Staff satisfaction and organisational performance: evidence from a longitudinal secondary analysis of the NHS staff survey and outcome data

Article history

The research reported in this issue of the journal was funded by the HS&DR programme or one of its proceeding programmes as project number 10/1011/11. The contractual start date was in January 2012. The final report began editorial review in July 2013 and was accepted for publication in December 2013. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HS&DR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

none

Copyright statement

© Queen’s Printer and Controller of HMSO 2014. This work was produced by Powell et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Introduction

It was noted in Chapter 1 that the search for causal links between HRM and organisational performance has involved a number of academic disciplines and a variety of terms;10,11,19,21 however, Macky and Boxall10 state that it broadly involves the following elements: a coherent and integrated ‘bundle’ of HR practices, a synergistic relationship between the practices, and an assumption of an underlying causal link flowing from HR practices via the responses of employees to organisational performance. Similarly, Patterson et al. 19 write that there is substantial variability between studies included in the model of high-involvement management or used to assess the link to performance. Some of these are terminological and reflect disciplinary biases or a desire to make their merchandise stand out, but others represent different foci and approaches.

Guest17 provides a chronological overview of the HRM performance field. The first phase (‘the beginnings’) began in the 1980s with work linking business strategy to HRM. The second phase (‘empiricism’) began in the 1990s when the empirical analyses of HRM and performance started to appear, with the seminal paper by Huselid65 and contributions by Arthur,66 Ichniowski et al. ,67 MacDuffie68 and Delery and Doty69 suggesting a positive relationship between HR practices and performance. The third phase (‘backlash and reflection’) focused on some key conceptual issues. For example, Dyer and Reeves70 and Becker and Gerhart71 showed that studies used many different dependent (performance) and independent variables (HR practices), raising questions over generalisability, differentiating between universalist (best practice), contingency and configurational perspectives. 69 A different type of backlash focused on the impact on employees. 72

Both sets of responses brought further streams of conceptual and empirical work, which were termed phases 4 and 5 by Guest. 17 The fourth phase (‘conceptual refinement’) stressed theoretical underpinning, with Guest16 arguing that we needed a better theory about HR practices, about outcomes and about the link between them. A number of authors discussed the ability, motivation, opportunity (AMO) model, while European authors, such as Paauwe,73 highlighted the importance of an institutional perspective. The fifth and overlapping phase (‘bringing the worker centre-stage’) pointed to the neglected impact on employees, suggesting the need to open the ‘black box’ that explored the process linking HRM and performance. The most recent phase (‘growing sophistication’) stressed the need for multilevel and longitudinal studies, including ‘big science’. 74

The field has some highly cited sources, e.g. Huselid65 (5559 Google Scholar citations; 1544 Web of Science citations; on 20 June 2013). It has also seen numerous special issues of international academic journals,75 narrative reviews,65 systematic reviews19 and meta-analyses. 76 In addition, there are reviews of the literature modelling the mediators and moderators of the HRM practice–performance relationship. 77 For example, Patterson et al. 19 identified six recent major reviews, while Purcell and Kinnie18 claim there have been ‘at least’ (p. 533) 11 review papers since 2000.

Reviews tend to find a broadly positive association, but include optimistic and pessimistic verdicts. For example, Combs et al. 76 claim that the results of their meta-analysis (that decrease the effects of sampling and measurement error) eliminate any doubt about the presence of a relationship as well as providing researchers with a baseline approximation of the extent of this relationship. They state that organisations can expand their performance by 0.20 of a standardised unit for each unit that HPWP use increases. On the other hand, Wall and Wood74 write that ‘it is premature to assume that HRM initiatives will inevitably result in performance gains . . . the existing evidence for a relationship between HRM and performance should be treated with caution.’

Commentators point to theoretical and methodological challenges. 18,74,78 Paauwe and Boselie79 state that empirical research provides evidence that ‘HRM does matter’, but the relationships are frequently weak in a statistical sense and the results are often unclear. Paauwe1 writes that although there is considerable evidence (at least regarding the number of studies), several authors still question HRM and particularly the HRM–performance relationship. Purcell and Kinnie18 state that numerous review papers have found this field of research often wanting in terms of method, theory and the specification of HR practices to be used when establishing a relationship with performance outcomes. According to Combs et al. ,76 the diversity of ‘sample characteristics, research designs, practices examined and organizational performance measures used’ has exasperated efforts to approximate the extent of the link between organisational performance and HPWP. Harris et al. 25 reported that the reviews all mention difficulties in identifying the theoretical perspective taken in each study, measuring HRM and performance consistently and drawing causal conclusions about the HRM–performance link owing to the predominantly cross-sectional nature of the research designs.

Performance/dependent variable

Human resource management/independent variable

Contextual perspectives on the NHS setting

As we saw above (see Fit/universalistic, configurational or contingency perspectives), there is a debate on the importance of context, with some authors differentiating between ‘best practice’ and ‘best fit’ approaches, while others suggest universalistic, configurational, or contingency approaches. At one level it is clear that context is important if only because many outcome measures used in studies of manufacturing such as profit are not appropriate for institutions such as the NHS. However, it is not fully clear which contextual features are most important. First, some studies point out that national context is important. According to Boselie et al. ,3 the question is whether or not the US-oriented models, however suitable for that country, can be used in relation to other countries and in other contexts. For example, the strict Dutch labour laws mean that some HPWP that vary widely elsewhere are required by law or regulated in the Netherlands. 86 Second, in a study carried out in the Netherlands, Boselie et al. 3 found evidence for mediating effects of institutionalisation. The effects on average duration of absence due to illness are weaker in a high-institutionalised context, such as hospitals and local authorities, than in a low-institutionalised context such as hotels. However, the discussion below focuses on three important but underexplored contextual elements relevant to the NHS setting of the study. The literature review of Hyde et al. 24 found few studies on services (3%), health care (2%) or the public sector (1%). It can be argued that important contextual features of a study on the NHS may be related to its setting within services, health care or the public sector.

Conclusion

The reviews tend to conclude that the HRM performance link is complex and unclear and that the literature still often suffers from theoretical and methodological problems. According to Patterson et al. ,19 the HRM performance relationship is ‘complex, multifaceted and multidirectional’. They stress the lack of ‘longitudinal research exploring the totality of any causal chain from HRM to intermediate outcomes and employee behaviours, to organisational performance’.

Moreover, while much research argues that context is important, there are few studies on health care in general and on the NHS in particular (see Chapter 3). 19,24–26 Patterson et al. 19 go on to state that while associations between HRM and performance have been found in many cross-sectional studies outside the health sector, and by a small number within the health sector, this does not demonstrate a causal link. Considerably more longitudinal research is needed on a wider range of variables to understand the impact of HRM practices on final outcomes in the NHS. According to Hyde et al. ,24 although the NHS is clearly different from other types of organisations, this does not imply that a new theory of the relationship between HRM and performance is needed, but rather that great care is required when introducing successful approaches or practices that have not already been applied in this context.

Structured summary

Introduction

Research has demonstrated that HPWS are linked to a wide range of important organisational and employee outcomes across various research settings, designs and industries. 8,138–142 Reviews and meta-analyses in the field have successfully confirmed the generalisability of these effects;71,76,138 nevertheless, there is strong evidence that context is an important factor that needs to be taken into consideration when studying the link between HRM and performance. This is because HR practices may differ between organisations in, for example, the public sector as compared with the private sector and in health care as compared with other industries. 3,122,128,132

With regards to the health-care sector in particular, two reviews that have recently been conducted give us a basis for drawing generalisable conclusions on the effects of HPWS on outcomes in health care. The first one, carried out by Etchegaray et al. ,136 is a narrative review that addresses issues of HPWS measurement, as well as links between HPWS and performance. The second one is a realist review of the field conducted by Garman et al. ,137 which combines research from health care with findings from other industries to develop and propose a comprehensive framework of HPWS and the mechanisms through which they affect outcomes, tailored specifically for the health-care sector. Surprisingly, the publications identified in the searches, as well as the main findings and conclusions the authors reach, differ markedly between the two reviews. Possible reasons for the discrepancies between the two reviews are discussed in this chapter.

The primary purpose of this chapter is, therefore, to explore in more depth the literature on HPWS and thus address the discrepancies found in the two reviews. With this general objective in mind, a systematic literature review was conducted in order to obtain our own collection of publications on HPWS in health care and compare this with the published reviews. This is followed by an in-depth critical discussion of the published articles that report empirical quantitative studies in terms of (1) HPWS definitions in relation to those commonly adopted in non-health-care research and publications, (2) the extent to which the primary characteristics associated with HPWS in general literature are reflected in the health-care literature, (3) the dominant theoretical frameworks used in linking HPWS with outcomes in health care, (4) the terminological choices and their appropriateness in the health-care literature on HPWS, (5) the evidence on the link between HPWS and outcomes in health care and (6) the various mechanisms through which, and conditions under which, HPWS have a positive effect on outcomes in health care. Finally, articles reporting qualitative studies and commentaries are discussed and integrated with the quantitative evidence. This is not designed to be a full literature review of the topics being covered by this report, but instead provides context for the wider study about what is known about the links between HPWS and outcomes in health care.

Method

A systematic literature search was conducted to identify publications, published up to and including 2012, dealing with HPWS in health care. The keywords were derived from past reviews and general literature (e.g. ‘high-performance work systems/practices/environment’, ‘high-commitment HRM practices’, ‘high-involvement HRM practices’, ‘HRM policy’, ‘HRM practice’, ‘human capital’ and ‘SHRM’). The search was restricted to publications containing any of the keywords and the word ‘health’. The databases searched (HMIC, MEDLINE, PsycINFO, SSCI, EBSCOhost) cover both managerial and health-care literatures.

The search yielded 126 publications. Our screening identified 27 publications both relevant and referring to the health-care sector, 47 non-health care, 31 non-HPWS, nine duplicates, 10 non-peer reviewed and two book reviews. We focused on the 27 articles that were both relevant and specific to health care. These were further broken down into qualitative (n = 5), quantitative (n = 13), mixed method (n = 2), qualitative reviews (n = 5), commentaries (n = 1) and theoretical analyses (n = 1). As the overlap between the two recently published systematic reviews141,142 was modest, to say the least, we decided to extend our search in order to create as complete a list as possible of relevant publications. As a first step, we compared our search outputs with those of the two published reviews, which yielded 11 publications that were added to our database. One of these, by Dawson et al. ,143 was not included in further analysis as it was a working paper, not published yet at the time of our review. Further manual search and search of the citations within the identified publications produced another five relevant publications that were added to our database as well. Therefore, the final collection of publications that was reviewed comprised 42 publications, of which 23 were quantitative empirical studies, seven were qualitative empirical studies, four were mixed-methods studies, five were reviews, two were commentaries and one was a theoretical article. The description of all publications that were included in the review is presented in Table 1, including the excluded working paper for reasons of comparison between our literature search and the searches by Garman et al. 137 and Etchegaray et al. 136 Drawing on the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2011 guidelines for systematic reviews, we present the search procedure using a four-stage flowchart (Figure 1).

FIGURE 1.

Flow diagram of literature search.

The publications reporting on empirical quantitative studies (including as part of a mixed-method approach) were coded based on 24 variables of interest: key words (if any reported in the original publication), source (whether the publication was identified in the systematic literature search or added manually through other searches or from citations in other publications), country, industry (if other non-health-care studies/samples are reported in the publication), type of health-care organisation (e.g. general practitioner practice, hospital, age-care sector, regional health service, etc.), sector (public, private, both), terminology used (e.g. HPWP, HPWS, HPM, etc.), appropriateness of terminology, conceptualisation of HPWS, theoretical framework underpinning the study, whether or not internal and external influence on HPWS have been assessed, internal and external fit of HPWS, the general approach to HPWS (‘best practice’, context-specific, organisation-specific approach), measurement of HPWS, level of analysis (e.g. individual, group, organisational, country, multilevel), method (e.g. survey, secondary data, multisource, etc.), methodology (e.g. cross-sectional, longitudinal, mixed methods, etc.), sample size, independent variables, dependent variables, control variables, moderators, mediators, and main findings.

Systematic review and critical evaluation of quantitative empirical studies: the ‘hard’ evidence

The reason a more detailed coding and analysis was conducted on the empirical studies was that these are most relevant in contextualising the findings of the analysis in the present report within the wider literature. The main factors addressed in the detailed review are summarised in Appendix 1, with Table 21 summarising the theoretical factors under consideration and Table 22 the methodological considerations and main findings. In this section, we provide a critical discussion of the quantitative empirical studies that were reviewed.

Bartram et al. 167 report on a survey study of hospital chief executive officers (CEOs), HR directors and senior managers in Australia regarding their views on SHRM and HR functions, as well as their effect on performance outcomes. They adopt the theoretical framework proposed by Bowen and Osrtoff168 to develop arguments on the link between SHRM and performance, with particular emphasis on communication and interaction among organisational members. The theory emphasises the importance of distinctiveness, consistency and consensus in SHRM as key factors that influence employees’ behaviour and organisational performance. They found significant links between perceptions of SHRM and perceived organisational performance. Additionally, their analysis unveiled interesting differences between managers at different levels in the organisation and in different functions regarding their perceptions of SHRM. In particular, they investigated the moderating effects of managerial role, organisation size, industry tenure, managerial tenure and gender. They found significant differences in perceptions of SHRM between CEOs and other managers, and no differences in perceptions between types of managers on perceptions of HR priorities on the overall sample. More detailed moderation analysis revealed that these differences were more prominent in large organisations. In relation to managers’ tenure in industry, tenure in organisation and gender there were no significant differences in perceptions among respondents. The authors conclude that ‘the strategic human management paradigm is “lost in translation”, particularly in large organisations, and consequently opportunities to understand and develop the link between people management practices and improved organisational outcomes may be missed’. 167 Overall, the study by Bartram et al. 167 has some very important advantages, such as a sample of senior managers, adequate measures of SHRM and other relevant variables and is theoretically robust in that it takes into account the strategic role of HRM in a holistic manner, as can be observed from the measures used in the study. Nevertheless, their findings cannot be classed as conclusive owing to the small sample size and common-method bias.

Berg and Frost120 published a paper on the topic of dignity at work for low-paid, low-skill workers using a sample from 15 hospitals in the USA. They used the main premises and assertions of the HPWS theory and available empirical evidence to link various work-related factors to three indicators of dignity at work, namely, fair treatment, intrinsically satisfying work and economic security. Although not explicitly stated, the assumption underlying the investigation is that the work-related factors form part of a HPWS. These include broadened jobs, participation in problem-solving teams, formal and informal training, union coverage, high-involvement union coverage, wages, staffing adequacy, resource adequacy and role overload. The main focus of the analysis was on the effects of enhancing workers’ jobs and having union representation on workers’ dignity at work. Although they did not find support for their overall model, they report some interesting associations: dignity at work was associated to higher pay, adequate levels of staffing and resources, and access to training. Some of the variables assessed were collected using surveys on the low-skilled workers sample, while others were collected from interviews with managers. Although the data are nested in organisations, the paper did not employ multilevel analysis; rather, the authors corrected for the effect of the organisation. Furthermore, rather limited information is provided regarding the psychometric properties of the scales used, which appear to be developed for the purpose of the reported study, rather than taken from other studies.

Bonias et al. 144 conducted an investigation in 2010 of the link between HPWS and organisational performance, measured as employees’ perceptions of the quality of patient care in their hospital. The study was conducted in a public sector regional health service organisation in Australia and had an initial sample of 541 responses across various occupational groups, both clinical and non-clinical. The study’s main finding was that HPWS are not directly associated to quality of patient care and that psychological empowerment fully mediates this relationship. Thus, the authors unveil one of the mechanisms through which HPWS have an effect on an important outcome in health-care settings, namely patient care. One of the concerns regarding this study is that the final sample used in testing the hypothesised relationships was markedly smaller, ranging from 319 to 329 responses. This is an indication of a large quantity of missing responses in the questionnaires and could raise concerns regarding the representativeness of the recorded responses, even though the authors tested for non-response bias and found that it was not an issue. Another concern is that the study was cross-sectional; therefore, we cannot infer causality in the reported relationships even though there was strong theoretical support for the direction of the relationships and the findings are potentially affected by common-source bias. With regard to the measurement of HPWS in health-care settings, Bonias et al. 144 initially used a general 42-item measure of HPWS that was developed by Zacharatos et al. 140 and used to measure HPWS in various industries. Nevertheless, on taking a closer look and analysing the scale, Bonias et al. 144 discovered that some of the dimensions from the general measure were inappropriate for their study. They omitted the ‘status distinctions’ dimensions owing to the prevalence of hierarchical structures in health care, the ‘management practices’ dimension owing to the working conditions being centrally determined through collective bargaining, and the ‘contingent compensation’ dimension as it is not employed in Australian public health-care organisations. This indicates that although general measures of HPWS have some benefits (e.g. they are reliable and validated), they may not always be appropriate for use in the health-care sector and in every country, owing to sector-specific and country-specific factors (e.g. culture, degree of centralisation, organisational structure, unionisation). The article by Bonias et al. 144 was not included in the two reviews of HPWS in health care; a possible reason is that the literature searches for these two reviews were most likely completed prior to the publication of the article.

Boselie et al. 3 make a series of important contributions to our understanding of HPWS on a theoretical level. Even though they do not use the HPWS framework directly and explicitly in their study, they do investigate a bundle of HR practices in terms of their effect on important organisational outcomes, namely absence rates, absence duration and turnover. Their findings indicate that the positive effect of HR practices, in terms of reduced absence rates and duration, is higher in non-institutionalised industries (hotels), than in institutionalised ones (local government and hospitals). They did not find any significant effects of HR systems on turnover in their sample. Therefore, the authors provide evidence for the role of context, environment and organisational characteristics with regard to the effect of HR on performance, indicating clearly that what is true for one industry, for instance in manufacturing, is not necessarily true in another, such as health care. The second important finding is that they revealed a two-factor structure of HR systems, contrary to what is theorised in the case of HPWS. 138 Their findings, although enlightening, should be interpreted with caution for three reasons: (1) their sample size is relatively small and in the hospital sector includes only nurses, (2) they did not analyse their data for hospitals separately to the data from other industries and (3) their statistical analysis is rather weak in terms of technique, where they claim to be testing mediating effect when in fact they are testing the moderating effect of institutionalisation on the relationship between HR systems and outcomes. The article by Boselie et al. 3 was included in the review by Etchegaray et al. 136 but not in the review by Garman et al. ,137 nor was it a hit in our own literature search. Possible reasons for this may be that the study is not framed around the HPWS theory, nor is it limited to the health-care sector and, therefore, did not match the key words used in the searches for this article nor that by Garman et al. 137

Boselie127 investigated the effect of HPWP in a public Dutch hospital on two individual level outcomes: affective commitment and organisational citizenship behaviour. The study is included in the present review with caution, as varying components of the study do not comply with the assumptions underlying the HPWS approach. In particular, HPWP was classified into three categories on the basis of the AMO model, namely practices enhancing abilities, motivation and opportunities. The study considers these as separate independent variables, therefore disregarding the ‘systemic effect’ assumption of HPWS, synergistic effects, and internal and external fit. However, the findings have important implications as they reveal that the ‘best practice’ approach is not always appropriate and that context-specific factors need to be taken into account. More specifically, it was found that abilities enhancing practices are linked to affective commitment, opportunities enhancing practices are linked to organisational citizenship behaviours, while motivation enhancing practices are not linked to any of the outcomes, contrary to what was hypothesised. As the motivation construct relates mostly to pay and rewards systems employed by the organisation and their degree of fairness, the author interprets the absence of statistically significant relationships to outcomes as a reflection of the contextual factors surrounding health-care organisations in the Netherlands. In particular, Boselie127 concludes that the lack of relationship could be due to the high institutionalisation of the sector that is characterised by collective agreements and legislative interventions. One potential criticism of Boselie’s study127 is that it does not test for the potential mediating effect of affective commitment in the relationship between HPWP and organisational citizenship behaviour. Such a hypothesis would be justified by both extant literature on the relationship between affective commitment and organisational citizenship behaviour169,170 and the high correlation between the two in Boselie’s sample (r = 0.34, p < 0.001). If this is the nature of the relationships, then a further indirect link between HPWP enhancing abilities and organisational citizenship behaviour might have been unveiled. The study by Boselie127 was not reviewed by Garman et al. 137 or Etchegaray et al. ,136 most likely because the article was not yet published when the literature searches for the two reviews took place.

Deshpande145 reports on a study looking at changes in HR practices and organisational performance following union elections in 101 hospitals in the USA. Although the data are methodologically strong, as they are collected from multiple sources and come from a wide range of HR practices and a relatively large number of organisations, the analysis that was conducted does not allow for any inferences to be made with regard to the link between HR practices and performance. However, some interesting differences in the use of HR practices and in performance are observed following union elections, both positive and negative regarding the HR practices, and generally negative regarding performance outcomes, with union certification having a negative effect and union rejection a positive one. Despite the fact that the author does not directly utilise the HPWS terminology and framework, the study can be classed as such, as it investigates a wide range of HR practices.

One of the more methodologically advanced studies in the health-care literature comes from Gittell146 who used multisource and conducted a multilevel analysis in order to investigate the effect ‘relational work systems’ on employees’ collective coping response (relational co-ordination) in nine hospitals in the USA. The author developed the ‘relational work systems’ theoretical framework drawing on the HPWS literature and proposed a series of practices, namely selection and training for cross-functional teamwork, the use of conflict resolution to build relationships between workers, feedback and rewards that are oriented towards contributions to shared goals, and information sharing or co-ordinating mechanisms (such as team meetings and boundary spanners) that will have a synergistic positive effect on employees’ resilience to external pressures. The results provide support for the hypothesised mediated model with environmental pressures being associated with perceived work pressures, which in turn are associated with collective coping response (relational co-ordination). Additionally, formal work practices (relational work systems) were found to be associated with collective coping response. Although this paper did not assess HPWS in a general sense, it provides a prime example of a theoretically and methodologically robust study that inspires confidence that targeted HPWS, when employed in health-care organisations, can produce targeted outcomes that are valuable to the organisation. Unfortunately, the direction of relationships and links to performance are hard to assert and further research should aim at filling these gaps.

Gittell et al. 8 followed up the above study with an extension using what appears to be the same sample of hospital employees as reported in the original Gittell146 paper discussed above. This article extends the study by adding performance outcomes collected from third sources. Theoretically, this study was framed around the HPWS paradigm; however, the conceptualisation remains focused on the relational aspect of HPWS and, using multilevel data analysis methods again, the authors find support for the proposed mediated model, with HPWS linking to relational co-ordination and this in turn linking to quality and efficiency outcomes. Therefore, this study fills one of the gaps identified in the previous study by showing that the proposed model links to organisational performance.

In a 2006 study, Gowen et al. 147 assessed the link between health-care error sources and error reduction barriers to quality management processes, SHRM and quality management practices, which were in turn assessed on the basis of the link with quality programme results and sustainable competitive advantage. The data used came from two sources (questionnaire surveys) and were analysed using regression analyses. The main findings show health-care error sources and error reduction barriers to be associated to quality management processes, quality management practices and SHRM. Quality management process, quality management practices and SHRM are related to quality programme results, and quality management practices and SHRM are related to sustainable competitive advantage. Owing to the way the variables and analysis are reported in the publication and the basic type of analysis that was conducted, it is difficult to assess the validity and generalisability of the reported findings.

Harley et al. 118 conducted a large-scale cross-sectional investigation of HPWS in the aged-care sector in Australia, across both public and private organisations. Although they looked at the effects of individual groupings of practices, rather than HPWS as a system or bundle, their study makes important theoretical contributions. First, it demonstrates the positive effects on individual-level outcomes (including autonomy, affective commitment, job satisfaction, psychological strain, turnover intention, and work effort) of a wide range of HR practices. This finding provides support for the theories that propose that HPWS will positively affect organisational performance through employees’ attitudes and behaviours and, in contrast to propositions under the RBV, that HPWS lead to work intensification and negative individual employee outcomes. Second, it demonstrates that HPWS are no less perceived by low-skilled than by high-skilled workers in the health-care sector. Third, it shows that there are no major differences in the nature of the relationships between HR practices and outcomes in the two occupational groups. Although these findings suggest that HPWS should be viewed as a ‘best practice’ because, according to the authors, they have positive effects across the board; these findings need to be interpreted with caution as several characteristics of HPWS were not investigated, such as synergistic effects, systemic effects, internal fit, external fit, external influences and link to performance.

Harmon et al. 4 report on an exceptionally large-scale study conducted in the USA among Veteran Health Administration (VHA) organisations, which are publicly funded and provide various types of health services. They used 10 items from a 1997 nationwide survey to measure HIWS (dependent variable) and two items from the same survey to measure employee satisfaction (mediator). The outcome data were obtained from a different source and are a measure of service cost. This study has two advantages over the majority of health-care research on HPWS. First, the authors used structural equation modelling (SEM) to analyse their data, a technique that is an advancement on the regression analysis employed by most other researchers as it compensates for measurement error. Second, the outcome data were obtained from a different source, thus avoiding common-source bias which plagues the vast majority of research in health care. The study’s findings indicate that job satisfaction partially mediates the relationship between HIWS and service cost. The authors demonstrated that, on average, a one standard deviation increase in the adoption of HIWS by an organisation in the sample equates to a $1.2M saving per annum.

The study by Harmon et al. 4 was followed up by two similar investigations in the USA VHA organisations. With a somewhat smaller sample size and using the same measure of HPWS, Scotti et al. 123 found links between HPWS and customer satisfaction, mediated by customer orientation, employee-perceived service quality and customer-perceived service quality. Their outcome data also came from a different source and the analysis was conducted using SEM. Scotti et al. 164 supplemented the existing VHA data with further data from the Veterans Benefits Association (VBA) organisations, which deal with benefits claims and are not directly involved in providing patient care. They tested the Scotti et al. 123 model and found that the effects stand for both occupational groups the high-contact service (VHA) and low-contact (VBA) ones. However, the effect sizes of the relationships were different between the two groups, with a stronger relationship between HPWS and employee-perceived service quality among low-contact employees, and stronger links between HPWS and customer orientation among high-contact employees. Interestingly, although Scotti et al. 123,164 use the same measure of HR practices as Harmon et al. ,4 they use different terminology, with Harmon et al. ,4 using HIWS and Scotti et al. 123,164 using HPWS. This a prime example of the ambiguities and inconsistencies in the terminology used in the HPWS literature.

Lammers et al. 150 report on a study that looked at the effects of commitment to quality improvement, quality councils, teams, budgets and training on perceptions of improvements as a result of total quality management programmes. This study was not framed around HPWS or relevant frameworks, but the measured variables could be argued to be loosely linked to HRM. The study showed some variation in the importance of level of commitment at different levels in the organisational hierarchy. Further, they report that the main factors explaining a large proportion of variance in numbers of teams, training intensity and total perceived improvement are the age of the quality council, overall organisational commitment to total quality management philosophy and physician commitment. Overall, the findings of the study need to be interpreted with caution owing to the small sample size and weak statistical analysis methods.

Laschinger et al. 151 conducted a study on a large sample of nurses in Canada to test a proposed mediated theoretical model. The model suggested that organisational characteristics (autonomy, control and physician relationships) are linked to trust in peers and managers, which in turn links to burnout (emotional exhaustion), which leads to poorer job satisfaction and assessments of quality of patient care and unit. Their data supported a modified model with both burnout and organisational trust mediating the relationship between organisational characteristics and outcomes. Although the study does not address HPWS or related theories directly, the large sample size and sophisticated methods of data analysis indicate and we can conclude with some confidence that HR-related factors are associated with perceived performance among nurses.

One of the more sophisticated studies looking at the link between HPWS and organisational performance in the health-care sector comes from Lee et al. 152 who investigated a complex mediational model to gain understanding of the mechanisms through which HPWS affect customer loyalty. Using multisource data and SEM, thus partly avoiding the problems of measurement error, they showed that HPWS predict employee reactions, which in turn predict service quality, which then predicts customer satisfaction, which finally links to customer loyalty. In spite of the relatively small sample size and the limited factors of HPWS measured, this study provides some strong evidence on the link between HPWS and performance with regard to customer satisfaction and loyalty.

Leggat et al. 153 conducted a cross-sectional survey investigation of the prevalence of various HRM related factors in hospitals of varying characteristics (12 metropolitan, 13 regional, 37 rural and district). In particular, they looked at HR priorities, performance management, training and development, employee participation and empowerment. The study revealed that there is insufficient emphasis in hospitals on practices that facilitate patient safety. Particular weaknesses of Australian hospitals were identified in the areas of performance management, lack of link between organisational performance indicators and staff/management performance indicators, and insufficient emphasis on training. Further, there was no significant differences in HR-related factors among the different hospital types. Although the study investigates a wide range of HR factors, it is not possible to infer any links to employee and performance outcomes.

Leggat et al. 6 also report on a series of studies, both qualitative and quantitative, the results of which have been published in detail in other articles. Their conclusions should therefore be interpreted with caution, as they might inflate the perception of readers and research users regarding particular findings. In terms of quantitative studies, they report on the findings from three surveys in Australian hospitals, without making references to the particulars of the statistical analysis that was conducted. They found that HPWS are associated with perceived quality of care and that HRM outcomes function as a mediator in this relationship. Furthermore, they observe that HPWS in Australian health-care organisations are generally deficient, in spite of the policies that encourage such systems. There is a difference in the identification of HPWS among various managers, with CEOs generally reporting higher levels than HR and other managers.

Lemmens et al. 155 report on a study of impressive design, conducted in the Netherlands. They measured various HR-related factors pre and post intervention, with a 1-year gap. The intervention was linked to changing the systems for delivering care to patients with chronic obstructive pulmonary disease. The factors that were assessed included culture, climate and quality improvement commitment. A change in the systems from pre to post intervention was observed in terms of self-management support, clinical information systems and delivery system design. The authors found associations between changes in processes of care, factors of organisation and professional commitment. Professional commitment and group culture appeared to be predictors of process implementation. As appealing as these findings are, we cannot draw conclusions regarding the effects of HPWS with regards to the interventions, as these were not directly assessed. A further limitation of the study is the small sample size (52 participants).

The study by Parkes et al. 159 was a large-scale longitudinal study in the UK, with data collected from both managers and employees at two time points. The main focus of the investigation was employee involvement. Unfortunately, statistical analysis and findings are not reported apart from the lack of link between employee involvement and organisational performance. However, without further details, it is difficult to draw any definite conclusions. The paper further reports on a series of case studies, which provide a rich insight into the potential relationships of employee involvement and outcomes.

Pas et al. 160 conducted a focused study that looked at family-friendly policies in a female sample of medical professionals in the Netherlands. Feminisation and collective labour agreements were found to have a positive effect on the offer of family-friendly policies. Offers of reduced participation arrangements had a negative effect on contracted working hours, while full participation arrangements had a positive one. Female doctors tend to work extra hours if they feel supported in improving their work–life balance, if they feel supported in achieving their career goals, and if they do not feel that their careers will be hindered. Reduced participation arrangements had a negative effect on contracted working hours, while full participation arrangements had a positive on effect. Family-friendly workforce philosophy was found to be a moderator in these relationships. Although limited in scope, this study provides some valuable insights with regard to the potential gender differences in terms of responses to various HPWS.

Preuss161 conducted a large-scale study of nurses in the Netherlands in order to investigate the mediating role of information quality in the relationship between HPWS and organisational outcomes. The author reports that the quality of information available for decision-making ‘partially mediates how employee knowledge, work design and total quality management systems affect organizational performance (measured as the inverse of medication error incidence)’. Although the paper reports significant results, it is ambiguous whether or not indeed the variables measured represent HPWS.

Rondeau and Wagar163 conducted a study of nurses in nursing homes in Canada in which they investigated the effect of ‘magnet’ status on nurse and patient satisfaction, participatory decision-making cultures and resources dedicated to job-related training. They found support for all the above proposed links. Interestingly, they did not find a significant association between HIWP and ‘magnet’ status of nursing homes.

In an earlier study on nursing homes in Canada, Rondeau and Wagar162 found that high-performance HRM practices and workplace climates that value employee participation, empowerment and accountability are linked to favourable organisational outcomes. Similarly, high-performing organisations are characterised by implementation of high-involvement practices and favourable climate. Although the study suffers from common-method variance, it provides some tentative evidence of the HPWS–performance link.

West et al. 126 conducted an extensive study of HR practices on CEOs and HR directors from 81 hospital trusts in the UK using patient mortality rates as the outcome measure. They found that all three predictors, namely sophistication of training policies, teamworking and sophisticated appraisal systems, were linked negatively to patient mortality rates, with the strongest relationship found for appraisal. This study was followed up with a similar investigation by West et al. 21 that extends these findings. They report links between a bundle of HR practices, including training, sophistication of performance appraisal system, staff participation, teamworking, employment security, Investors in People (IIP) status and patient mortality rate, even when controlling for previous patient mortality levels and other potentially confounding factors.

Finally, Young et al. 124 conducted a study of hospital employees in Australia and found that for managers what matters the most is the distinctiveness, consistency and consensus of HPWS. This finding provides evidence in support of the underlying theoretical principles of HPWS. Their findings further show that social identification facilitates the associations between HPWS and both affective commitment and job satisfaction.

Conclusion

The review found that a multiplicity of terminology, frameworks, settings and variables meant that overall conclusions were difficult, with little clear, consistent evidence for the link between HRM and performance in health care. This manifested itself in a number of ways, which we explain in more detail in the following paragraphs.

One of the limitations of the HPWS framework, which raises the difficulty of reviewing the literature and research evidence in the field, is the definitional ambiguity surrounding the term, which is reflected in the wide range of alternative terminologies used, such as HIWS, HCM and high-performance work environments, high-commitment work systems, high-performance management practices. 3,171 An example of how HPWS is defined from the health-care domain comes from Etchegaray et al. ,136 who do not limit their definition to HR practices, but rather extend the concept to encompass a wider range of practices, termed ‘work practices’. They further expand the implied effect of practices on outcomes beyond organisational performance to include employee attitudinal outcomes as well as outcomes at various levels. In particular, they define HPWS as ‘an integrated set of practices that result in engaged employees and positive individual-, unit-, or organizational-level outcomes’. 136 From our review, we can conclude that in terms of both terminology and approaches to the definition of HPWS, there is extensive variation among the publications.

The term ‘high performance’ implies that some systems or HR practice configurations will produce ‘low’ performance. 3 Therefore, in order to validate the HPWS theoretical underpinnings, research needs to show that not all system configurations lead to performance improvements and can be classified as HPWS, and to narrowly define the distinctive characteristics of high-performance compared with low-performance systems. When studying HR practice bundles that have the HPWS characteristics (e.g. synergistic effects), but are not assumed to produce high performance, then the terminology used should be adjusted accordingly. For instance, if the bundle of practices enhances involvement it should be termed ‘high-involvement’ work system, while, when the bundle is tailored to enhance abilities and competencies it should be termed ‘high-ability’ work system, and so on. Our review revealed that there are a considerable number of publications that do not take into account the particular point of reference of each term and rarely justify their selection of terms in light of their theoretical framework and study design.

In spite of the wide literature on the theoretical underpinnings of HPWS as a HRM theory, ambiguities regarding the characteristics of HPWS and what distinguishes them from HR practices in general were prevalent in the early work on HPWS and still remain. One major area that lacks clarity is the question of whether HPWS are a ‘best practice’ theory of HR, or whether it is context sensitive. Becker and Huselid138 put forward compelling arguments for the contingent nature of HPWS based on the notions of inimitability, internal (or horizontal) fit and external (or vertical) fit. Their arguments can even be interpreted to mean that a HPWS can only be characterised as such if it is unique to the organisation employing it, and uniquely aligned and fitted to the particular characteristics, strategy, culture, goals and environment of the organisation at hand. In this light, HPWS should be studied as organisation-specific configurations of practices and studies should go beyond identifying which practices are optimal for enhancing specific organisations’ performance to investigate how these practices are being applied and enacted to complement each other (thus creating synergies) and to fit the organisation’s strategy. Others, on the other hand, view HPWS as a universalistic ‘best practice’ approach and identify this as the main weakness of the HPWS approach. 3,16 This approach is reflected in the body of research that aims to identify a set of HR practices that are linked to high performance across organisations and contexts. Although this appears to be the dominant approach, if not always explicitly stated and recognised, we argue that it is fundamentally flawed as it ignores the salient role played by organisational and contextual characteristics. This can be demonstrated simply by looking at a specific HR practice across different contexts – performance-contingent rewards. Pay-for-performance is generally considered as one of the performance-enhancing practices and has been consistently included in generalist bundles of HPWS. 76,140 However, in the health-care sector, literature indicates that pay-for-performance is associated with various potential dangers and the evidence of the benefits of such practices is scarce and inconclusive. 172,173 It is encouraging to see that a large proportion of the reviewed studies take into account the peculiarities of the health-care context and reflect this in their theorising and study design. However, there is insufficient evidence of the proposed characteristics of HWPS in the health-care literature (i.e. lack of evidence of synergies, internal and external fit, link to productivity). There is a need for a clearer distinction between organisational-level and individual-level effects – what is good for the organisation is not always good for the employee and vice versa. 174 For example, an increase in performance might come at the cost of increased stress levels. Overall, it cannot be conclusively derived whether or not there is sufficient and appropriate evidence of the link between HPWS and performance in the health-care sector based on the reviewed papers; nevertheless, the reported findings provide some initial evidence of such links.

A wide range of theoretical frameworks have been utilised by the authors of the reviewed publications in order to provide the rationale behind HPWS. Common theoretical perspectives include the RBV,85 social exchange theory;175 AMO theory;83,127,176 structure, process and outcomes;141,142,177 attraction–selection–attrition model;142,178 motivation;179 configuration, contingency, universalistic;69 and sociotechnical systems (STS) design. 4,180 Harmon et al. 4 justify the selection of the STS design as the framework behind HIWS, as it is based on the same principle of alignment between human and technical factors. Although the theoretical variation is often well justified and aligned to terminological and measurement choices, the sheer range of theoretical frameworks used makes a systematic review and comparison challenging.

Overall, the primary gaps and limitations identified in the literature are (1) a lack of longitudinal studies that investigate causality; (2) various studies appear to report on the same data, thus possibly inflating the reported effects; (3) the country variation among the reported studies is limited, thus making it difficult to reach generalisable conclusions; and (4) the majority of studies investigate a limited range of HR practices, thus making it difficult to reach conclusions with regards to the effects of the HR system overall.

In the health-care literature, two recent reviews give us a basis for drawing generalisable conclusions on the effects of HPWS on outcomes. The first, by Etchegaray et al. ,136 is a narrative review that addresses issues of HPWS measurement in health care and their link to performance. The second is a realist review of the field conducted by Garman et al. ,142 which combines research from health care with findings from other industries to propose a comprehensive framework of HPWS and the mechanisms through which they affect outcomes, tailored specifically for health care. Surprisingly, the publications identified in the searches, and the main findings and conclusions, differ markedly between the two reviews. Owing to the conceptual and theoretical similarities among the various terms, the reviews that have been conducted thus far generally assess the concepts together. For example, in their review of HPWS in health care, Etchegaray et al. 136 included in their search the terms ‘high commitment’ and ‘high involvement’ as well. The present review aimed at overcoming this discrepancy by conducting a more thorough and inclusive literature search. Although it was not successful in providing clear, consistent evidence of the links between HPWS and outcomes, it does provide a firm basis for suggesting that more coherent research is needed. In addition, only three of the reported studies were conducted in the UK, suggesting a lack of evidence from within the NHS specifically. 21,126,159

Introduction

As noted earlier, the ‘business case’ that staff satisfaction leads to greater organisational performance has been accepted by government and the NHS. There are a series of reports by a number of bodies drawing on different, but connected, debates inside and outside the NHS. The generic business case has been carried out with reference to ‘Good Jobs’,27,28,181 work and well-being,29–33 and engagement. 34,100 Similarly, a series of reports from the Department of Health (DH) and other organisations have stressed the importance of staff involvement and engagement and health and well-being over a period of about 15 years. According to Hyde et al. ,182 the NHS presents a particularly interesting environment because of attempts, through national policies and legislation, to introduce HPWP throughout 2000–10. Atkinson and Hall5 report that, in line with HPWS theory, the NHS has adopted a range of HR practices as a means to enhance organisational performance. We present a chronological outline of the debates in general and for the NHS.

The generic business case

In a report for The Work Foundation, Coats and Max181 claim that studies demonstrate that better workplaces have better financial results. They argue that there is a compelling case for organisations of all sectors and sizes to move beyond the traditional health and safety agenda to embed health and well-being at their heart and to create an empowering and rewarding work environment for all employees. In particular, they focus on sickness absence, pointing out that the annual economic costs of sickness absence and worklessness associated with ill-health are over £100B a year – greater than the current annual NHS budget.

Waddell and Burton29 were commissioned to review the link between work, and health and well-being. As part of this review, PricewaterhouseCoopers183 were commissioned to consider the wider business case and specifically the economic case for employers to invest in wellness programmes for their staff. PricewaterhouseCoopers found considerable evidence from literature reviews and over 50 UK-based case studies that health and well-being programmes have a positive impact on intermediate and bottom-line benefits. Intermediate business benefits include reduced sickness absence, reduced staff turnover, reduced accidents and injuries, reduced resource allocation, increased employee satisfaction, a higher company profile and higher productivity. Waddell and Burton29 sum up that work is usually good for both mental and physical health as well as well-being, but it should be ‘good work’ which is healthy, safe and offers the individual some influence over how work is done and a sense of self-worth. They conclude that the message is clear: good health is good business.

The Scottish Executive100 considers that the literature finds measurable impacts of employee engagement and disengagement on the performance of the organisation. The level of employee engagement matters because it affects HR (e.g. recruitment and retention) as well as the bottom line for companies, although the links to these more distal outcomes tend to be more tenuous. Moreover, there is not an abundance of information on this in the literature and there is still discussion regarding quantifying the cost-effectiveness of commitment of an organisation to employee engagement.

In a report commissioned by government, Black33 writes that research found substantial evidence that economic benefits in all types of business could be offered by health and well-being programmes; good health allows for good business. However, employers do not adequately understand the information regarding investment in the health and well-being of employees. The government response184 welcomed the review, the evidence it presented, the conclusions drawn and the recommendations made.

Lekhi and Blaug27 produced a literature review for the Health and Safety Executive which argued that the existing literature focuses on associations, saying little about causation. They argue that job satisfaction is not a useful measure of job quality or a good job, as satisfaction can sometimes reflect individuals getting used to anything, which suggests that job satisfaction per se may be a poor measure of organisational commitment to good jobs.

MacLeod and Clarke34 were asked by ministers to explore how employee engagement can lead to organisational performance. They state that while the meaning of the term remains unclear, there is evidence that employers can increase engagement in a ‘win–win’ context (i.e. benefits for both employers and employees).

Writing for The Work Foundation, Constable et al. 28 report that considerable benefits can be achieved from increasing the number of jobs into ‘Good Jobs’ in the UK. These benefits can be reaped by government and other employers, and include greater labour productivity, higher workforce stability, a healthier workforce, and more engaged and committed employees. In particular, they stress the importance of impact on organisational performance by changing levels of sickness absence and presenteeism. They list the benefits of ‘Good Jobs’ for government departments such as the Treasury, the DH, the Department for Business, Innovation and Skills, and the Department for Work and Pensions. 28

Business in the Community30 argues that healthy people lead to healthy profits. It sets out 20 case studies that document business benefits that followed the introduction of health and well-being interventions in their workplaces. Similarly, Business in the Community31 argues that it is good business to have a happy, healthy and engaged workforce. It stresses that maximising the wellness and engagement of your employees is a win–win situation in that it benefits employees, customers, the organisational ‘bottom line’ as well as the wider society and nation.

In a Bupa report, Vaughan-Jones and Barham32 note the cost of sickness absence to the economy and to society. They examine more than 600 pieces of evidence regarding how effective a range of interventions are to find what is best for different employers. It states that evidence demonstrates that a variety of interventions benefit employers (by providing better productivity and a decreased number of absences) and employees (by providing earlier discovery of disease and better well-being), presenting employers government and society with an uncommon chance for a win–win.

In a report for IIP, Bevan185 argues that there is growing and convincing evidence that work is good for the vast majority of employees. He explores seven areas of business performance that are directly or indirectly linked to improvements in employee health: reduced work absenteeism, fewer work accidents, higher employee retention, greater employee commitment, increased labour productivity, enhanced employer brand and a higher level of employee resilience.

A report for the Chartered Institute of Personnel and Development186 states that although we have limited knowledge about employee engagement in theoretical, conceptual and empirical terms, the concept has positive associations at individual and organisational levels with a range of beneficial outcomes. The evidence suggests that the UK has relatively low levels of engagement; however, it tends to be higher in the public sector than in the private sector. The report lists the following drivers of engagement: voice, the ability to feed views upwards; senior management communication and vision; supportive work environment; person–job fit; line management style; and the work that is perceived to be meaningfulness.

Following the publication of the review by MacLeod and Clarke,34 the new government asked for additional evidence of the associations between employee performance and engagement. The Employee Engagement Task Force responded by calling for evidence of connections between employee engagement and organisational outcomes from UK-based organisations. Rayton et al. 97 report that the utter weight of the evidence should convince the most sceptical that employee engagement is not a weak topic, but an important issue that has an impact on success or on service outcomes. The authors regard employee engagement as something essential, not just desirable.

Rayton et al. 97 examine the consequences on business performance (such as innovation, customer measures and profits, efficiency) and people indicators (such as health and safety, attendance and welfare) in the public and private sectors. Multiple reports using meta-analysis have confirmed strong associations between employee engagement and improved efficiency, returns, beneficial discretionary effort, innovation, customer happiness and retention. Moreover, they point to a causal relationship from engagement to performance. In short, the evidence supports the existence of a strong longitudinal synergistic connection between employee performance and employees who feel engaged work better.

The business case for the NHS

The term ‘model employer’ has often been used for public services in general and the NHS in particular7,38,187 when government seeks to manage its employees along ‘best practice’ lines. As in the section The generic business case, we provide a largely chronological summary of the main documents. 36–39

The New Labour White Paper, ‘The New NHS’,42 acknowledged that staff involvement had not been a high priority, but pledged a new approach to appreciate staff more, spearheaded by a taskforce on staff involvement. 44 This was followed by the NHS HR Strategy, ‘Working Together’. 43 According to Ellins and Ham,36 this was one of the first documents produced by the DH that clearly linked better staff conditions with enhanced services. However, Bach and Kessler38 consider that the proposals were relatively modest in terms of staff involvement, but they still represented a significant departure for government because it was the first time that the NHS had set out a detailed approach to employee relations. However, it appeared to have a low priority; in a survey of 75 trusts, ‘reviewing staff involvement’ and ‘establishing a partnership agreement’ were the lowest priority in terms of progress on 13 HR goals.

The NHS Taskforce44 provided three key messages for ministers and for the NHS: staff involvement matters, works and can be made to work across the NHS. It issued 11 wide-ranging recommendations: encourage good leadership; promote good industrial relations; develop and use a self-assessment tool; develop a local statement of rights; provide support and advice; promote good practice on intelligence networks; improve communication; invest in personal development; monitor performance and progress; include in attitude surveys; and commission regular independent research.

Bach and Kessler38 state that the NHS Taskforce report ‘pointed to private sector best practice using almost evangelical language to persuade sceptical employers that staff involvement works’. However, the authors note that staff expressed scepticism about senior management’s interest in workforce perspectives despite the development of an impressive array of top-down communication mechanisms. For example, the NHS staff survey indicated limited change. The ‘What Matters’ research highlighted widespread frustration and commented that many staff regard the NHS as moving an inapproriate business agenda on finance and incorrect targets. 188

The NHS Plan40 makes a commitment to invest in NHS staff. It states that making the working lives of staff better partly results in improved patient care via staff retention and recruitment and because patients would rather be cared for by staff who are enthused. The way in which NHS Employers handle staff will in future be part of the central performance measures and related to the monetary resources provided.

This was set out in the Improving Working Lives Standard,41 which recognises the necessity for modern health services to be built on modern employment services. It sets out a series of targets to achieve annual improvements in the quality of working life for staff, and it expects that the Improving Working Lives Standard should be put into practice by all NHS employers by April 2003.

Shifting the Balance of Power45 states that a real shift in the balance of power will not occur unless staff are empowered to make the necessary changes, stressing that issues of cultural shift and staff ownership needed will in many ways. It lists the actions that will support this work at a national level, including mainstreaming staff involvement; publishing leadership competencies; development programmes; developing a Staff Involvement Toolkit; and establishing a joint forum for partnership and involvement. Local actions include appointment of a staff involvement leader who reports to a nominated non-executive and executive director; the reduction of hierarchies and development of self-managed teams; the preparation of a staff involvement plan; and ensuring that staff involvement is built into objectives for managers and into the arrangements for performance monitoring.

In 2002, the government launched the first comprehensive HR strategy for the NHS, HR in the NHS Plan. 189 The strategy described four pillars on which the goal of additional staff who are operating differently would be built: making the NHS a model employer by implementation of best policies, facilities and practices; ensuring the NHS provides a model career through the concept of the skills escalator, with lasting learning and development; increasing staff morale; and building people management skills, by developing the capacity and proficiencies of HR. 189

In 2003, the DH initiated resources to aid NHS organisations translate the idea of staff participation into reality,36 including a staff charter, partnership framework, staff forums, staff representation on committees, staff surveys and other feedback and communication tools. 47

The DH issued a national framework to support local workforce development that discussed staff engagement, enabling NHS organisations to provide high-quality services by using advanced employment methods and representing ‘model employers’. 46 The framework proposed 10 changes in HR practices, which evidence indicated would have the greatest benefit to delivering organisational goals, one of which was participation, staff involvement, and positive employee relationships. It also put forward a number of model employment practices including flexible working, good appraisal systems and staff participation policies as well as partnerships with staff-side organisations. 46

NHS Employers has drawn attention to many aspects of staff satisfaction in a series of reports. 55,57,62 Bullying and harassment are not normally considered in HPWS (and so was not considered in Chapter 3), but can clearly be an important factor in organisational performance. For example, Woodrow and Guest190 state that both physical violence from members of the public and non-physical harassment from colleagues are highly prevalent in the health-care workforce. They note that while policy has tended to focus on the more visible problem of public violence, it is not clear which of the two behaviours is the most damaging. They compared the consequences of public violence and staff harassment for well-being in two large samples of English nurses. The results showed that, while both types of aggression were associated with decreased levels of staff well-being, staff harassment had a stronger negative association with well-being than public violence. The relationships between each of the types of aggression and some aspects of well-being were moderated by perceived supervisory support, such that the negative effects on well-being were greater for those with higher levels of support, although the effect sizes were very small. This is in contrast to previous research showing that support (although not specifically from supervisors) can buffer against the effects of aggression. They conclude that the major implication of the study is that health-care organisations must pay more attention to the prevention of staff harassment in the workplace.

NHS National Workforce Projects191 points to research that shows that an engaged workforce is more productive, with better recruitment and retention rates. It sets out top tips developed from lessons learned in this study.

The DH commissioned the ‘What Matters’ research programme to develop understanding of the improvement of NHS values and how staff experience relays to care of patients. 188 Qualitative and quantitative research concluded that it is particularly important that the NHS aims to improve the following elements for staff:

• I understand my role and where it fits in.

• Senior managers are involved with our work.

• I have the opportunity to develop my potential. 188

The final Darzi Report192 notes the importance of empowering staff and supporting NHS staff to provide high-quality care. It pointed to two national issues of high-quality places of work and high-quality training and education. In the same way that patients should have high-quality care, NHS staff should have high-quality work. It proposed a NHS constitution that would reflect NHS values as well as valuing and empowering staff.

Staff engagement was also identified as a major priority by Clare Chapman, then NHS Director General of Workforce. 36 This was reiterated in the NHS Operating Framework for 2009/10, which challenged all NHS organisations to increase staff involvement. 193

The interim Boorman report48 argues the case for investing in improving staff health and well-being services as this will result in benefits to individual staff, patients and employers. It reported research from commissioned reports that shows that there is a positive relationship between staff health and well-being and key performance issues. It sets out the business case for improving staff health and well-being. For example, it calculated that reducing current sickness absence levels by one-third could lead to efficiency savings of some £555M. However, it points to a widely held view that staff health and well-being was not seen as a priority either at organisational or local management level. Finally, the report made a number of recommendations for action at both national and local level to deliver change.

The final Boorman report49 made further recommendations that the NHS Operating Framework should require staff health and well-being to be included in national and local governance; form part of standards and targets for the Care Quality Commission’s annual assessment of NHS; and to be considered as part of Monitor’s assessment process for foundation trust status and in annual monitoring arrangements. At the local level, it recommends that a staff health and well-being strategy should be developed by all NHS organisations. In short, all NHS organisations should be seen as exemplar employers that need to invest in their workforce’s health and well-being in order to deliver sustainable, high-quality services.

The DH50 accepted Dr Boorman’s recommendations, agreeing that this attitude must change and that all NHS staff and managers must give priority to staff health and well-being. In the foreword to the report, Secretary of State, Andy Burnham, stated that he was convinced by the business case presented in the report and accepted all suggestions. The document accepted the central case that good staff health and well-being is vital for ensuring that the NHS can meet the quality and productivity challenge is well made, and that the NHS must be an exemplar employer. 50

In the July 2010 health White Paper ‘Equity and Excellence: Liberating the NHS’,51 the coalition government committed to continuing to implement the recommendations from Dr Boorman’s49 report on NHS health and well-being. It stated that staff who are engaged, empowered and are supported provide better care of patients. The Coalition Government will therefore encourage staff engagement as well as partnership working and the initiation of Dr Boorman’s improvements to staff health and well-being.

The Operating Framework for the NHS in England 2011/1252 stated that the NHS is dedicated to developing and protecting staff health and well-being as well as decreasing the level of unnecessary sickness leave, as discussed in Dr Boorman’s review of NHS health and well-being. 48 It adds that substantial staff engagement will help to provide the productivity and quality challenges faced by NHS organisations and will lead to better patient outcomes and financial management.

Briefing 78: Health, Work and Well-being in the NHS57 recognises that the improvements to staff health and well-being recommended in the Boorman review48 contribute towards meeting the staff pledge in the NHS Constitution and delivering the four elements of the QIPP programme. Moreover, the 2010–11 Operating Framework194 requires all NHS organisations to set up a health and well-being strategy for their staff. It argues that evidence from the Black33 and Boorman49 reviews in addition to earlier research show the close links between staff health and well-being and engagement, and that high-performing NHS organisations tend to have good staff engagement polices. The document notes a set of five high-impact actions (leadership, evidence-based plan, management capacity, staff engagement, occupational health service) from the DH’s Well-being Delivery Group. It also notes that West et al. 21 find evidence of a clear association between reducing patient morbidity and effective HR, occupational health and health and safety services.

Briefing 78: Health, Work and Well-being in the NHS57 focuses on some local experience of staff engagement in the NHS. It argues that the challenges faced by the NHS such as reducing costs, increasing productivity and implementing the organisational changes associated with the NHS White Paper are linked with staff engagement as high levels of engagement are associated with positive outcomes for patients and for staff.

The Operating Framework for the NHS in England 2012/1353 argues that staff continue to be the most fundamental resource of the service. It suggested that all organisations should continuously progress staff involvement and services to patients by drawing on the NHS staff survey, and suggested models and frameworks for improvement, which will help to achieve the Boorman ambition of reducing the level of sickness absence towards 3% and towards meeting the QIPP challenge.

In a report for the DH, West et al. 106 state that effective NHS staff management results in better care, happier patients and reduced mortality. In more detail, engagement, the number of staff receiving good appraisals, working in successful teams, receiving supportive training and management are associated with several trust outcomes.

Generating Savings by Improving Health and Well-Being55 notes the high level of sickness absence in the NHS (10.7 days a year, compared with 9.7 days in the public sector as a whole and 6.4 days in the private sector). It states that evidence submitted to the Francis inquiry suggests that staff disengagement can damage care quality. NHS Employers56 has produced a series of factsheets on the staff engagement challenge, which provide evidence from the commercial sectors in the UK and USA and from the NHS of the positive association between staff and organisational performance.

The Francis report15 said little on staff engagement, but stressed that staff must be valued and front-line staff must be empowered with the capability and accountability to deliver safe care. However, the report has cast a long shadow in terms of engagement. Engaging Your Staff: The NHS Staff Engagement Resource62 states that the importance of staff engagement has never been higher as the NHS faces the biggest reforms since its inception and begins to change the poor cultures highlighted in the Francis report. The document sets out a pledge to work to improve the health and well-being of health-care staff.

The government’s response to the Francis report60 includes a number of relevant issues. It stresses the importance of the ‘Friends and Family Test’. However, as staff are asked this question only annually, the NHS Commissioning Board aims to ensure that this type of staff feedback becomes more frequent. It states that there is already good evidence to show organisations that treating their staff well will deliver better outcomes for patients. It adds that the Northumbria Healthcare NHS Foundation Trust has some of the most satisfied staff and patients in England and firmly believes the two must go hand in hand for a healthy organisational culture.

The business plan for NHS England195 states that as the main touchstones of success are patients recommending their local NHS care and individual NHS staff members having faith in the service they are contributing towards, an 11-point scorecard will set out progress against the key measures of success of satisfied patients and staff who feel positive about what they are doing.

The government’s mandate to Health Education England196 include excellent education; competent and capable staff; flexible workforce, receptive to research and innovation; NHS values and behaviours; and widening participation. One of the ‘longer-term objectives’ includes continual improvement supporting efforts to deliver a continual improvement in proportion of staff, patients and the public who recommend friends and family by ensuring an adequate supply of suitably qualified staff.

The NHS Constitution was refreshed in 201360 with more details being given in the Handbook to the NHS Constitution. 61 It draws on the ‘What Matters’188 research, which identified four themes, which are now reflected in the NHS staff survey and were also used to inform the NHS Constitution’s values: the resources to deliver quality care for patients; the support they need to do a good job; a worthwhile job with chances to develop; and the opportunity to improve the way they work. To really embrace the full and challenging definition of quality set out in ‘High Quality Care For All’,192 it must be recognised that high-quality care requires high-quality workplaces, with commissioners and providers aiming to be employers of choice. In addition to legal rights, there are a number of pledges, which represent a commitment by the NHS to provide high-quality working environments for staff:

• to provide a positive working environment for staff and to promote supportive, open cultures that help staff do their job to the best of their ability

• to provide all staff with clear roles and responsibilities and rewarding jobs for teams and individuals that make a difference to patients, their families and carers and communities

• to provide all staff with personal development, access to appropriate education and training for their jobs, and line management support to enable them to fulfil their potential

• to provide support and opportunities for staff to maintain their health, well-being and safety

• to engage staff in decisions that affect them and the services they provide, individually, through representative organisations and through local partnership working arrangements. All staff will be empowered to put forward ways to deliver better and safer services for patients and their families.

It is argued that a positive working environment not only has benefits in terms of the experience of staff, it is also linked to positive outcomes for patients and that several studies have shown clear evidence of the link between good staff experience and good patient experience. 39,106 An open and supportive culture has been identified by the Mid Staffordshire Foundation Trust Public Enquiry (‘Francis inquiry’) as a key element in successful organisations. It notes that there are already a considerable number of initiatives at all levels. The DH, NHS Protect, NHS Employers, NHS Plus and others are actively supporting programmes to provide a healthy working environment, improve the health and well-being of NHS staff and tackle violence, bullying, harassment and stress in the workplace. Finally, the NHS staff survey will continue to be an important benchmark, encouraging organisations to engage with their staff.

NHS Employers62 has produced the staff engagement toolkit. It presents the ‘Staff Engagement Star’ of:

• great management and leadership

• a healthy and safe work environment

• ensuring every role counts

• supporting personal development

• enabling involvement in decision-making.

Engaging Your Staff: The NHS Staff Engagement Resource62 points out that the Operating Framework for the NHS in England 2012/1353 refers to the need to improve staff experience and take account of staff survey results.

The King’s Fund59 report on patient-centred leadership argues that a change in management, systems and the culture of organisations in the NHS is required if the recommendations of the Francis Inquiry15 are to be noted and implemented. The Francis Inquiry identified a culture that was dangerous and unhealthy, including detachment by medical leaders, low staff confidence and bullying, as a cause of the problems at Mid Staffordshire NHS Foundation Trust. The King’s Fund report discusses the problems associated with disengagement between managers, staff and patients. It argues that a supportive and positive environment should be created for staff and that without this, staff may not attain good levels of employee engagement, and it has been shown that, patient happiness is higher and patient mortality is lower when they have been dealing with staff who are engaged. 95,106 While organisational climate (or culture) played a role in staff well-being, the local work climate – the ward – was key. 39 In short, cultures of positivity, compassion, engagement, thoughtfulness and respect for staff and patients as well as the public delivers the perfect environment for caring for the nation’s health. If staff are well cared for then they will be able to supply better patient care. 95

In short, this policy review has shown that issues such as staff engagement and health and well-being have been on the generic national and NHS agendas for a long time, although most of the focus has been on the topic of involvement or engagement. However, much of the discussion in the policy documents can be argued to be either too broad or too narrow. At one level, there are fairly vague assertions that ‘staff engagement’ will lead to better performance without consideration of issues such as cost, context, causality or mutual gains. At another level, the case study material reports that engagement interventions lead to reduced levels of absenteeism, but there is little consideration of whether or not they would work in different contexts. First, most studies report ‘benefits’ without any consideration of cost, making a ‘business case’, presumably based on assessing costs and benefits, difficult to sustain. Second, it has been shown (see Chapter 2, Fit/universalistic, configurational or contingency perspectives) that there are major debates over best practice compared with best fit. For commentators who favour a contingency perspective, it is difficult to argue that simple transfer of evidence from other countries, sectors/industries will produce enhanced organisational performance, and there is little evidence on the NHS (see Chapter 3). Third, most documents report cross-sectional correlations, making it difficult to establish causality (see Chapter 2, Methodological issues). For example, it is difficult to rule out reverse causality when high organisational performance causes staff satisfaction (rather than staff satisfaction causing organisational performance). Fourth, employer perspectives tend to get much more attention than employee perspectives. This lack of evidence makes it difficult to appraise the ‘mutual gains’ and ‘conflicting outcomes’ approaches (see Chapter 2, Employer and employee outcomes) and making it difficult to determine if high-performance work practices are ‘win–win’ or ‘win–lose’.

Moreover, Ellins and Ham36 note that despite many policy initiatives having been launched since 1998 to increase staff involvement, relatively few staff state that they are involved in important decisions, are consulted about changes that affect them, feel encouraged to suggest ideas for improving services or feel that their organisation values their work. Finally, they conclude that as there are strong similarities between recent DH initiatives and policy documents from the late 1990s, exhortation and guidance alone appear insufficient to convert policy into practice.

Conclusion

The business case for staff engagement and health and well-being has been recognised by a variety of bodies both inside and outside the NHS over a period of many years. However, a number of untested optimistic assumptions, ignoring costs, transferring evidence from contexts such as the USA and from for-profit industry, causality, and ‘win–win’ have been largely taken for granted. Moreover, implementation has been rather variable and patchy. It is possible that renewed emphasis may be placed on this case in the ‘post-Stafford’ era.

However, it is clear that there is little evidence on HPWS in the NHS. There is insufficient evidence on the applicability of HPWS concepts to the NHS in terms of its contextual setting of being located in England, in the service sector, as a public service organisation and in the health-care sector (see Chapter 2). There are few empirical studies on health care in general and on the NHS in particular (see Chapter 3). Finally, the policy review highlights a rather broad and vague ‘business case’ based on a number of untested optimistic assumptions (see Chapter 4). All these factors suggest that the empirical study of Chapters 6–8 is worthwhile.

Introduction

This chapter gives a detailed description of the quantitative analytical methods used to answer the research questions, and describes the data sources, the variables and (when appropriate) the samples used. As a reminder, the research questions were:

1. What are the links between individual staff experiences (e.g. satisfaction, engagement, turnover intentions) and intermediate staff outcomes (e.g. staff absenteeism, actual turnover)?

2. How do these link with organisational performance (e.g. patient satisfaction, mortality)?

3. Do these measures and relationships differ by occupational, demographic groups, trust types and geographical areas and, if so, what is the relative change for each group?

Analytical methods used

Data from the NHS national staff survey

Outcome data used

Outcome variables were selected based on the following criteria: (1) variables that clearly reflected either intermediate outcomes, patient outcomes or organisational performance, (2) variables that were published for all three years of the study and (3) variables that had a clear direction, i.e. in general terms ‘more’ is either better or worse. Because of criterion (3), we did not use financial performance as there is no clear consensus about an indicator that would uniformly reflect performance (e.g. having too much surplus at the end of a year is more likely to represent poor use of resources rather than good management). Other variables that were suggested either by the researchers or by members of the advisory group [e.g. Commissioning for Quality and Innovation indicators (CQUINs)] were not used because either they did not provide comparable data for all trusts, or they were not available for the three years of the study. The variables that were used were absenteeism, turnover, patient satisfaction, patient mortality and infection rates.

Absenteeism was measured via the Electronic Staff Record and obtained via the NHS Information Centre website (www.hscic.gov.uk). 208 It is measured as the total proportion of working time lost to sickness absence in each of the three NHS years (April 2009–March 2010; April 2010–March 2011; and April 2011–March 2012). Because it was measured by the Electronic Staff Record – the official NHS HR information system – data should be comparable across trusts and across years, although there are some doubts about the fidelity of reporting absences particularly among senior medical staff.

Turnover was measured via the stability index (the proportion of staff working on 31 March of a given year still working on 31 March the following year; it excludes bank staff, locums and trainee doctors). It also came from the NHS Information Centre website. Although this was always intended as an outcome variable, it was compromised somewhat by the structural changes affecting the NHS over the study period, which meant that more staff may have left their organisations for reasons to do with restructuring rather than the more common reasons for turnover.

Patient satisfaction was measured via one question from the NHS acute inpatient survey, which is conducted each year in a similar fashion to the NHS staff survey, but is limited to acute trusts. The question asks ‘Overall, how would you rate the care you received?’, and response options are ‘Excellent’ (scored as 100), ‘Very good’ (scored as 75), ‘Good’ (scored as 50), ‘Fair’ (scored as 25), and ‘Poor’ (scored as 0). Data were gathered from the UK Data Service (http://ukdataservice.ac.uk/)209 and were aggregated from all patients in each trust to the trust level. Details of the numbers of participants in each trust in each year are shown in Table 5.

Patient mortality was measured using two different indices: the Hospital Standardised Mortality Ratio (HSMR) for 2009/10 and previous years, and the Standardised Hospital Mortality Index (SHMI) for 2010/11 and 2011/12. The use of two different indices was forced on us by a change in policy during the study period. Although both indicators use similar data to give a ratio of actual to expected deaths (controlling for a variety of demographic and diagnostic data), a change in the formula used – and the number of conditions coded – between the two indicators means that the use of both in the same analysis presents a limitation that we cannot easily overcome. These data were gathered from Dr Foster^® – a provider of health-care variation analysis – (www.drfoster.org.uk)210 (HSMR) and the NHS Information Centre website (SHMI).

Infection rates for both methicillin-resistant Staphylococcus aureus (MRSA)211 and Clostridium difficile212 were gathered from the Health Protection Agency website (www.hpa.org.uk/). 211,212 Specifically, the MRSA rate measures the annual rates of trust apportioned cases of MRSA bacteraemia, while the C. difficile rate measures the rate per 100,000 bed-days for specimens taken from patients aged ≥ 2 years (trust-apportioned cases).

Other variables used

When indicated, we controlled for the following variables:

• trust type: acute, acute specialist, PCT, mental health/learning disability or ambulance

• teaching status (for acute trusts)

• foundation status

• location (whether or not the trust is in London)

• doctors per bed (ratio gathered from Dr Foster, www.drfoster.org.uk)210

• trust size (log of number of employees gathered from NHS staff survey advice centre).

These are all variables that have been shown in previous research to be linked to one or more of the outcomes. In addition, we examined the effect of the trust chief executive’s tenure on outcomes. This was gathered by using public records (including websites) and telephone calls to trusts when information was not available.

Chapter summary

This chapter gives the findings from the analysis relating to the first research question, which was ‘What are the links between individual staff experiences (e.g. satisfaction, engagement, turnover intentions) and immediate staff outcomes (e.g. staff absenteeism, turnover)?’. This included three main types of analysis: individual multilevel analysis, latent growth curve modelling and cross-lagged correlation tests. These three are reported separately, but findings are then brought together to examine common threads, so that overall conclusions can be drawn.

The main findings from this research question are that there are some very clear associations between staff experience and individual outcomes (e.g. job satisfaction, intention to leave jobs, well-being) and staff absenteeism, all in the expected direction, but less clear effects on turnover. Longitudinal analysis suggested that poorer staff experience is likely to lead to lower subsequent absence, rather than vice versa. However, there were a number of contradictory or counterintuitive longitudinal results involving turnover, which may reflect the complex restructuring of the NHS over recent years more than any truly causal effects.

Chapter structure

The analysis conducted, which was described fully in Chapter 5, was very extensive, and thus the full results of each analysis are not reproduced in the main body of the report. Therefore, full tables of results can be found in Appendices 2–4. Summary tables, with enough information to show the findings of primary interest, are instead given in the main body of the report. Within this chapter, the multilevel analysis is presented, with staff experiences predicting individual staff outcomes (measured from the NHS staff survey); the latent growth curve analysis is presented with staff experiences predicting intermediate outcomes (absenteeism and turnover), and the cross-lagged correlation analysis is presented, involving absenteeism and turnover. Summary of results identifies common themes between the different types of analysis and what can be concluded with appropriate levels of confidence from the findings.

Multilevel analysis

For each of six individual outcomes, we conducted multilevel analysis with each of the 21 staff survey variables identified in Chapter 5 (as well as chief executive tenure) predicting the outcome, controlling for gender, age, managerial status, tenure, working hours, occupational group, disabled status, ethnic background, trust location, trust type, trust teaching status, foundation status, trust size, and ratio of doctors per bed (acute trusts). The outcome variables in question were:

• impact of health and well-being on ability to perform work or daily activities

• work-related stress in previous 12 months

• job satisfaction

• presenteeism (feeling pressure to attend work when feeling unwell)

• intention to leave job

• advocacy (recommending trust as a place to work or receive treatment).

Although we included all control variables in each analysis (i.e. with each separate predictor), we do not report the coefficients for each – if we did, the set of tables would run to several hundred pages. Instead, we report the effect of the control variables for each outcome without predictors and then give a table showing the effect of each predictor (separately) after taking the controls into account (see Appendix 2). Predictors are entered in separate analyses owing to the very large correlations between some of them, which would make estimates unstable.

The vast majority of effects were statistically significant, which is unsurprising given the large sample size and the possible common method variance due to the shared source of predictors and outcomes. Therefore, we have identified the most important effects – those with the largest effect size for each of the six outcomes – using unstandardised effect sizes, so that the effects shown are the effect of the presence of an experience for the binary variables, or a one-point change for the scale variables. These are shown in Table 6, with estimates and 95% confidence intervals (CIs) of effect sizes included. Note that because the direction of causality is unclear, we have included the outcomes as predictors of each other outcome also. We would wish to make clear that this does not equate to causal relationships in either direction – it is impossible to detect this from such analysis. Therefore, although we report them in the tables here, we do not discuss them in this summary.

The key findings were that most of the predictors behaved entirely as expected – those that represented positive experiences at work (e.g. good job design features, good people management practices, work engagement) were associated with better outcomes for staff and those that represented negative experiences (e.g. violence and harassment, work pressure, discrimination) were associated with poorer outcomes. It is somewhat instructive, therefore, to look at the size of the effects. As noted previously, the effect sizes shown are the expected change in the dependent variable given the presence of the experience (e.g. appraisal) for the binary variables, or a one-point change for the scale variables (usually equivalent to about a standard deviation, or a little more).

In predicting the impact of health and well-being on ability to perform work or daily activities, what are particularly noteworthy as predictors are the violence and harassment variables. Although there is a reasonable effect of violence or harassment from patients or their relatives, there is a far greater effect of that coming from colleagues. Thus aggression from colleagues appears particularly harmful to individual health. Similarly, large effects are found for experience of discrimination, or the belief that the trust does not present equal opportunities to all in terms of career progression/promotion.

Similar findings for these predictors arise with the other outcomes, notably presenteeism. For work-related stress, this is joined by a moderately large effect from work pressure, but also effects from staff motivation (work engagement), quality of job design and having a good-quality appraisal (these are not apparent in Table 6, but can be seen in Appendix 2). For job satisfaction as the outcome, a number of people management and job design factors are particularly important: good-quality appraisals, opportunities to develop, good communication and good incident reporting procedures are the most substantial. For intention to leave as an outcome, the key predictors were almost identical (although in reverse, of course, and, if anything, the effects tend to be even bigger). For example, if an individual believes that the trust does not provide equal opportunities to staff, they are likely to be around a standard deviation higher in terms of their intention to leave. Staff are most likely to recommend their trust as a place to work or to receive treatment when they have good opportunities to develop, they are satisfied with the quality of the work they deliver, there are fair and effective procedures for reporting incidents and near misses, there is good communication, and (again) there are equal opportunities for staff to progress with their careers. Note that the ‘overall engagement’ predictor should not be interpreted strongly because part of this indicator includes the outcome itself.

Overall, the really notable effects are those of negative experiences that adversely affect all of the outcomes, in particular, violence and harassment from colleagues (rather than from patients), perceptions of unequal treatment by the organisation, and experiencing discrimination come out consistently as big predictors. Being treated badly by other employees may not be an everyday experience (even if it is more common than it should be), but when it does happen, it is particularly damaging.

Latent growth curve analysis

Two stages of latent growth curve analysis were completed, and each is used for a slightly different interpretation. Both stages predicted levels, and changes, in intermediate outcomes from 2009/10 to 2011/12; the first stage used 2009 staff experience variables as predictors, whereas the second stage used differences from 2009 to 2010 (denoted with a ‘D’ suffix in Appendix 3, Tables 35–38), to examine whether or not there was any evidence of change in staff experience affecting longer-term change in intermediate outcomes.

Owing to the complexity of the latent growth curve analysis procedure, there were occasionally statistical problems preventing the estimates being achieved, which was a common problem with latent variable procedures. In order to circumvent this, in some cases we had to omit control variables from the models to get estimates. These cases are clearly indicated in the relevant tables.

Table 7 shows the significant relationships between staff survey variables from 2009 and the starting level (intercept) in absenteeism and turnover. These indicate when there are significant cross-sectional relationships between aggregate staff experience and behaviour in terms of absenteeism or leaving jobs. Note that the outcome variable for turnover is actually the stability index and, therefore, a positive relationship for this suggests a negative result for turnover. The tables report what was found in the analysis in terms of stability (for the sake of accuracy), but the text reports these findings in terms of turnover instead.

In summary, absenteeism is lower in trusts for which:

• a higher proportion of staff report working extra hours

• a higher percentage of staff report feeling valued by colleagues

• staff have well-designed jobs

• a higher proportion of staff work in a well-structured team environment

• a lower percentage of staff report experiencing physical violence from other staff

• a lower percentage of staff report experiencing harassment, bullying or abuse from patients or their relatives

• a higher percentage of staff report good communication between management and staff

• a higher percentage of staff report that they are able to contribute towards improvements at work

• staff report that they are willing to recommend their trust as a place to work or receive treatment

• staff report a higher level of motivation at work

• staff report higher overall work engagement.

Almost all of these suggest that better experiences equate to lower absence. The only dubious finding is that when more staff report working extra hours, absenteeism is lower; however, this makes sense because, if more staff were absent, there would be less opportunity to work extra hours.

Turnover is lower in trusts for which:

• fewer staff work extra hours

• a higher percentage of staff receive any type of training and development (health and safety training in particular)

• staff report lower levels of work pressure

• a higher percentage of staff are appraised, or have agreed a personal development plan

• fewer staff experience harassment, bullying or abuse from other staff

• staff perceive that effective action is taken from the employer towards violence and harassment

• staff have lower intentions to leave their job

• fewer staff experience discrimination at work

• availability of hand-washing materials is higher.

The ‘extra hours’ finding makes more sense here because, even though absenteeism may be lower when staff work extra hours, turnover is higher. All other findings are in the direction that suggests turnover is lower when staff experience is more positive.

Table 8 shows the significant relationships between staff survey variables from 2009 and the change (slope) in absenteeism and turnover. These indicate where starting levels of staff experience are associated with subsequent changes in absenteeism and turnover. These are more difficult to interpret because a drop in absenteeism (for example) may be due to a very high starting level – in other words, regression to the mean. Therefore, we recommend not interpreting these results particularly strongly, but instead focusing on the (far stronger) results in later tables. However, they are included for the sake of completeness.

A much stronger form of the analysis is using changes in staff experience (i.e. differences in staff survey variables between 2009 and 2010) as predictors of the change in intermediate outcomes (slopes). Table 9 shows the significant results from this analysis. In summary:

• An increase in staff agreeing that their role makes a difference to patients is associated with a decrease in turnover in subsequent years.

• An increase in the percentage of staff feeling that there are good opportunities to develop their potential at work is associated with a decrease in turnover in subsequent years.

• An increase in the percentage of staff suffering work-related injuries or illness is associated with a decrease in turnover in subsequent years.

• An increase in the percentage of staff experiencing harassment, bullying or abuse from other staff is associated with an increase in turnover in subsequent years.

• In trusts for which staff report an increase in their level of willingness to recommend the trust as a place to work or receive treatment, there is a decrease in turnover in subsequent years.

• There were no significant findings with absenteeism as the outcome.

This clearly indicates that when the number of staff having meaningful jobs increases, when there are decreases in aggression from other staff and when belief in their employer as both a place to work and a place to receive treatment increases, then turnover tends to decrease over the 3-year period in question.

Cross-lagged correlations

Cross-lagged correlations compare the relationship between two variables in subsequent years, testing whether or not there is a stronger effect in one direction than the other. Again, full results are in Appendix 2, but Tables 10 and 11 show the significantly different cross-lagged correlations between staff experience and intermediate outcomes. All differences are for the two most recent years, i.e. staff survey variables from 2010 and 2011, and outcomes from 2010/11 and 2011/12.

These findings reveal some unexpected results, particularly with regard to turnover. For example, the first finding in the table shows that the quality of work–life balance has a negative relationship with stability the next year, whereas the converse finding (stability in 2010/11 and quality of work–life balance in 2011) is that of a relationship close to zero. This suggests that in trusts for which the climate for work–life balance is better, turnover tends to subsequently increase.

Owing to the large number of separate tests here, and some seemingly contradictory or unexpected results, it makes most sense to concentrate on those for which there are obvious patterns. For this purpose, we define a pattern for which two distinct but theoretically similar staff survey variables have the same pattern of results with the same outcome, or one staff survey variable has the same pattern of results with both outcomes (i.e. higher stability/lower absenteeism, or vice versa). With this in mind, the patterns can be summarised as follows:

• Quality of work–life balance and opportunities for flexible working are both more closely related with higher subsequent turnover than vice versa. However, they are both also more closely related with lower subsequent absenteeism than vice versa, so there may be a contextual effect on turnover due to changes in the NHS.

• When staff agree that their role makes a difference to patients, or when they feel satisfied with the quality of care delivered, or would recommend their trust as a place to work or receive treatment, this is more negatively associated with subsequent turnover than vice versa.

• When staff say their health negatively impacts their ability to do their job or when they suffer from work-related stress, this is more positively associated with subsequent turnover than vice versa.

• Job satisfaction, support from supervisors and working in well-structure teams is more positively associated with subsequent turnover than vice versa.

• Absenteeism results are far more straightforward. In addition to those already mentioned, having good communication, less discrimination, fewer errors, less extra-hour working, less work pressure, less harassment and violence from patients, fewer injuries, higher job satisfaction and ability to contribute towards improvements, lower turnover intentions, less pressure to attend work when feeling unwell, better support from supervisors, and more well-structured teamworking are all associated with lower subsequent absenteeism than vice versa.

Summary of results

Overall, the results presented from all three sets of analysis give some clear messages, although for those involving turnover, the messages are sometimes far less clear. In general, there is a clear pattern that better staff experiences are associated with better health and behavioural outcomes for the employees concerned; the results from the individual (multilevel) analysis confirmed what had been expected here. In particular, the effects of staff believing there were equal opportunities for career progression and promotion on individual outcomes were especially strong and also the negative effects of aggression (particularly from colleagues) and discrimination were telling. Negative experiences, particularly negative treatment from colleagues, were far more damaging to staff well-being than the positive effect of positive experiences.

Organisational-level analysis with absenteeism is probably the most instructive and clear set of findings from this chapter. The cross-lagged correlations suggest that there is clear evidence for the direction of the effect between absenteeism and over half of the staff survey variables: it is much more likely that good staff experience leads to lower absenteeism than vice versa. These effects are particularly strong for negative experiences such as violence and harassment, but are also very strong for the positive experiences of staff being able to contribute towards improvements at work and when there is good communication between management and staff. Combined with the latent growth curve analysis that gave similar results, this presents a very clear and unambiguous set of findings about the nature of NHS staff jobs and absenteeism.

Results involving turnover were the most equivocal. Even though there was some strong latent growth curve analysis results suggesting that improvements in the number of staff having meaningful jobs increases, when there are decreases in aggression from other staff, and when belief in their employer as both a place to work and a place to receive treatment increases, then turnover tends to decrease over subsequent years. Many of the other results, particularly the cross-lagged correlations, gave inconsistent or counterintuitive findings. This has to be placed in the context of major changes over the NHS over the study period, including many large reorganisations of services, necessitating more movement of staff between trusts (and, in some cases, redundancies) than would normally be expected. Therefore, although objective data are usually better to use than subjective, it is probably more instructive to look at the patterns of results with self-reported turnover intentions (from the multilevel analysis), when the findings met with expectations, than the more surprising results using the stability index. Because of this, there is little that can be learned from the longitudinal analysis with turnover.

Chapter summary

This chapter gives the findings from the analysis relating to the second research question, which was ‘How do staff experience and intermediate outcomes link with organisational performance (e.g. patient satisfaction, mortality, infection rates), and is there a mediated link from experiences to performance via intermediate outcomes?’. This again included three main types of analysis: latent growth curve modelling, cross-lagged correlation tests and mediated regression. As in Chapter 6, these three are reported separately but findings are brought together.

The main findings from this research question are that the relationships with organisational performance are complex. There is clear evidence of significant (and often strong) links between staff experience and patient satisfaction, although this does not appear to be mediated by intermediate outcomes. The longitudinal effects are much less clear. There were some links between changes in staff experiences and subsequent improvements in patient outcomes, but this was not consistently found across all predictors and all outcomes. The cross-lagged correlations failed to reveal a consistent pattern of results to provide evidence for causal relationships.

In terms of the mediation, a striking finding was that although many staff experiences were associated with absenteeism and with patient satisfaction, there were not any mediated effects here. That is, the reason for staff experiences affecting absenteeism appears completely separate from the reason they affect patient satisfaction. Given that both are important for trusts for different reasons, this points to an even greater importance of staff attitudes and experience.

Chapter structure

The analysis conducted for this chapter was extensive and, thus, the full results of each analysis are not reproduced in the main body of the report. The full tables of results can be found in Appendices 5 and 6. Summary tables, with enough information to show the findings of primary interest, are instead given in the main body of the report. Within this chapter, the latent growth curve analysis is presented, with staff experiences and intermediate outcomes predicting trust outcomes (patient satisfaction, mortality, and two forms of infection rates), and the equivalent cross-lagged correlation analysis is also presented, as are results of mediated regression analysis. Summary of results identifies common themes between the different types of analysis and what can be concluded with appropriate levels of confidence from the findings.

Because of the nature of these trust outcomes, they apply only to acute trusts. Therefore, the analysis in this chapter is for acute trusts only.

Latent growth curve analysis

There were two stages of latent growth curve analysis completed and each is used for a slightly different interpretation. Both stages predicted levels, and changes, in intermediate outcomes from 2009/10 to 2011/12; the first stage used 2009 staff experience variables as predictors, whereas the second stage used differences from 2009 to 2010 (denoted with a ‘D’ suffix in Appendix 5, Tables 41–48), to examine whether or not there was any evidence of change in staff experience affecting longer-term change in intermediate outcomes.

Owing to the complexity of the latent growth curve analysis procedure, there were occasionally statistical problems preventing the estimates being achieved, which is a common problem with latent variable procedures. In order to circumvent this, in some cases we had to omit control variables from the models to get estimates. These cases are clearly indicated in the relevant tables.

Table 12 shows the significant relationships between staff survey variables from 2009, intermediate outcomes from the same year, and the starting level (intercept) in patient satisfaction, mortality, and infection rates for both MRSA and C. difficile. These indicate where there are significant cross-sectional relationships between aggregate staff experience and outcomes.

In summary, we can see that for patient satisfaction, there were many significant relationships and satisfaction was higher when:

• fewer staff work extra hours

• more staff have received any training and development and, in particular, health and safety training

• more staff feel valued by their colleagues

• staff report lower work pressure

• a higher percentage of staff have appraisals and personal development plans

• fewer staff report experiencing violence, harassment, bulling and abuse from patients and their relatives; it is also higher where staff perceive that action taken from the employer towards violence and harassment is effective

• the perceived fairness and effectiveness of incidence reporting is high

• more staff feel that they are able to contribute towards improvements at work

• there are high levels of job satisfaction among staff and lower intentions to leave jobs

• staff report that they would be more likely to recommend their trust as a place to work or receive treatment

• where staff believe that the trust provides equal opportunities for career progression or promotion and where fewer staff report experiencing discrimination at work

• there is a higher availability of hand-washing materials

• staff report high levels of engagement

• there are lower objective turnover rates (i.e. higher stability)

• there were fewer significant associations with patient mortality, but still some important (and theoretically expected) significant findings. In particular, mortality was lower when:

  • more staff report that they understand their role and where it fits in

  • more staff feel able to contribute towards improvements at work

  • staff are more likely to recommend the trust as a place to work or receive treatment

  • staff report higher overall work engagement.

However, mortality was also lower when more staff experienced physical violence from patients or their relatives. This is unexpected and also appears to be an anomaly because when compared with other types of analysis and different years, this was not replicated. Therefore, it is most likely to be a type I error.

For infection rates, one finding was that higher levels of harassment, bullying and abuse from patients or their relatives was associated with higher MRSA rates. If there is a genuine link between these two variables, then causality in either direction (or both) is perhaps reasonable. However, this is somewhat contradicted by the finding that such higher rates (as well as higher rates of work pressure and of shift working) are associated with lower C. difficile rates. This suggests that links with infection rates in general may not be very understandable.

Table 13 shows the significant relationships between staff survey variables from 2009 and the change (slope) in trust outcomes. These indicate where starting levels of staff experience are associated with subsequent changes in outcomes. These are more difficult to interpret because a drop in patient mortality (for example) may be due to a very high starting level – in other words, regression to the mean. Therefore, we recommend not interpreting these results particularly strongly, but instead focusing on the (far stronger) results in later tables. However, they are included for the sake of completeness. Indeed, some results (notably the positive links between work pressure and percentage staff experiencing harassment, bullying and abuse from patients or their relatives and changes in C. difficile rates) may partially explain the contradictory results in the previous table; when such negative experiences are high, there may also be a low starting value of infection rates, but these rates then increase over time.

A much stronger form of the analysis is using changes in staff experience (i.e. differences in staff survey variables between 2009 and 2010) as predictors of the change in intermediate outcomes (slopes). Table 14 shows the significant results from this analysis. In summary:

• An increase in the reported negative impact of health and well-being on employees’ ability to perform their work and daily activities is associated with a decrease in patient satisfaction.

• An increase in shift working is associated with an increase in patient mortality rates.

• An increase in perceptions of effective action from employer towards violence and harassment is associated with a decrease in patient mortality rates.

• An increase in line manager support is associated with a subsequent drop in MRSA rates.

• An increase in staff feeling there are good opportunities to develop potential at work and an increase in the fairness and effectiveness of incident reporting procedures are associated with a subsequent drop in C. difficile rates.

• However, an increase in shift working and in experiencing of harassment, bullying or abuse from patients or their relatives, is also associated with a drop in C. difficile rates.

Cross-lagged correlations

Cross-lagged correlations compare the relationship between two variables in subsequent years, testing whether or not there is a stronger effect in one direction than the other. Full results are in Appendix 2, but Table 15 shows the significantly different cross-lagged correlations between staff experience and intermediate outcomes, and also between staff experience and trust outcomes. Because of the changes in some outcomes (particularly mortality) in 2010 and 2011, we also give the year of measurement in the table because in this table only the interpretation of the values changes depending on which year is which.

These findings reveal a relatively small number of significant effects, but some are not as expected. In summary:

• When absenteeism is higher, this tends to lead to higher subsequent mortality, rather than vice versa.

• When turnover is lower, this is associated with greater mortality in the subsequent year.

• High infection rates of C. difficile are associated with higher subsequent staff absence than vice versa.

• When there are more opportunities for flexible working, mortality tends to be higher the following year rather than vice versa.

• Worryingly, when there is lower patient satisfaction, this is associated with higher subsequent discrimination (rather than vice versa).

• Higher patient satisfaction is associated with more flexible working subsequently and better subsequent job design, rather than vice versa.

• When more staff recommend the trust and fewer experience harassment, bullying or abuse from patients, this is associated with higher subsequent MRSA rates.

• A number of good job design factors are associated with lower subsequent C. difficile rates, rather than vice versa.

As in Chapter 6, it is dangerous to read too much into these results, particularly for those that stand alone and/or are contrary to the direction expected from theory. However, it is clear that the picture of how staff experience and trust outcomes are linked is not straightforward and the relationship is certainly not a simple causal one. Staff experiences are likely to be affected by trust outcomes as well and it appears that this may not always be a positive thing, but it is impossible to say exactly how these effects occur.

Mediation

We tested for whether or not there were significant indirect (mediated) effects of the staff experience variables on trust outcomes via absenteeism or turnover. This analysis controlled for the usual control variables for acute trusts, and used data from 2011/12 only.

Results suggested that there was not, on the whole, evidence of any mediated effects. Those that were significant are shown in Table 16. For patient mortality there were no significant indirect effects at all, as was the case for C. difficile rates. For patient satisfaction, there was a single indirect effect: that of the proportion of staff working extra hours, mediated by absenteeism. This is difficult to interpret because the indirect effect is positive, but small; the more staff working extra hours, the higher patient satisfaction is, but only very slightly. Given the singular nature of this effect, the fact it only just reaches statistical significance and the number of effects tested, it is quite possibly a type I error and, therefore, we do not attach any particular significance to it.

However, for MRSA infection rates, there were a large number of significant indirect effects, again all via absenteeism. Most of the effects are actually for job design factors, suggesting that, when jobs are better designed and staff experiences are better, absence rates are likely to be lower and, as a result, MRSA rates are likely to be lower. However, we need to temper the interpretation of this with the results from the previous section, which cast some doubt over the direction of relationships between staff experience and infection rates. Therefore, we cannot assume there is indeed such a causal relationship and although a set of consistent and interesting results, it is too much of a step to say that this proves such a mediated link.

Summary of results

Overall, there is a real mix of results presented here. By far the most consistent and clear finding is the link between staff experiences and levels of patient satisfaction, replicating previous work examining these constructs. 106 These reveal that there are clear associations, at least cross-sectionally, between many staff experiences (across most domains) and patient satisfaction.

However, the longitudinal effects are much less clear. Cross-lagged correlations did reveal some patterns suggesting directional effects, for example that absenteeism in one year is more closely associated with mortality in the subsequent year than vice versa; however, others (particularly those involving infection rates) were much less convincing. There were some links between changes in staff experiences (particularly those relating to the quality of job design) and subsequent improvements in patient outcomes, but this was not consistently found across all predictors and all outcomes. This reveals the limitations of the analysis: looking at year-on-year changes may not be sensitive enough to the variables in question (particularly across whole trusts) to be able to detect time-lagged effects that could help provide more evidence for causality. Overall, it would be dangerous to conclude anything substantial from the cross-lagged correlations. The latent growth curve model results were not much clearer, as the stronger design (modelling change in staff experiences) revealed only a few statistically significant results, some of which could have been false positive findings. However, experiences linked to violence, harassment and actions dealing with it were linked (in the expected direction) with a number of different outcomes, suggesting that this may have not only impact on the staff immediate outcomes, but also directly on patients too.

In terms of the mediation, a striking finding was that, although may staff experiences were associated with absenteeism and patient satisfaction, there were no mediated effects here. That is, the reason for staff experiences affecting absenteeism appears completely separate from the reason they affect patient satisfaction. Given that both are important for trusts for different reasons, this points to an even greater importance of staff attitudes and experience.

Chapter summary

This chapter gives the findings from the analysis relating to the third research question, which was ‘Do these relationships [between staff variables and outcomes] differ by occupational, demographic groups, trust types and geographical areas, and if so what is the relative change for each group?’. This involved a series of many regression analyses, examining data aggregated by each particular staff breakdown, as well as some separating out trusts by geographical region.

The main findings from this research question are that few clear patterns emerge and some of those that do are not at all surprising. The most effects (and largest differentials) are for predictors of absenteeism, with nursing staff generally had the strongest effects of all the occupational groups followed by medical/dental staff. Most other differences by groups of staff were fairly inconsistent and, thus, more difficult to interpret reliably.

In terms of geographic regions, absenteeism was most readily predicted – by most staff survey variables – in the West Midlands, while the health of workers in Yorkshire had the strongest effect on patient satisfaction and work pressure in the South Central region was a stronger predictor of turnover than in other regions. Aside from the West Midlands results, which were consistent across many findings, these may be one-off results with no clear patterns emerging.

Introduction

This was, by its nature, a far more exploratory piece of work; we did not have a priori theoretical expectations about any particular group of staff – whether work-related or demographic groups – having more important relationships than any others (however, for some staff survey variables in particular it would seem likely that there could be some differences by occupational group). Because of this, and the fact that there were bound to be some differences between groups simply by chance, we report all of the analysis in Appendix 7, but only report a small subset of that analysis in this chapter. As described in Chapter 5, the criteria we used were that the standardised coefficients for two groups should have differences of at least 0.20 for them to be considered differential effects and at least one group’s effect should have a p-value of < 0.01. We did this separately for the analysis that controlled for the prior year’s outcome (a very strong control) and that which did not.

Analysis by staff groups

Six breakdowns met the criteria controlling for the outcomes of prior year and 39 met the criteria without this control included. These breakdowns are shown in Tables 17 and 18, respectively.

The differential breakdown for the effect of turnover intentions on turnover is particularly interesting, as this suggests that some groups are more likely to carry out their intentions than others. Examination of the results in Appendix 7 reveals that there are larger effects for medical/dental staff (β = –0.124, p = 0.008) and AHPs (β = –0.110, p = 0.021), but not for nurses (β = 0.023, p = 0.627). It is noteworthy that these effects (negative because the actual outcome is stability rather than turnover) may seem small, but because the outcome refers to the whole of the trust and not just that staff group, they are still important differences.

For the effect of work pressure on turnover, it was the AHPs who had the strongest effect by far (β = –0.165, p = 0.001). The other effects shown here are more difficult to interpret theoretically, but they suggest that the strongest link between turnover intentions and C. difficile rates is for general managers, the strongest links between general health and well-being and MRSA rates are for those who have been in post for between 6 and 15 years, the strongest links between general health and well-being and patient satisfaction are for women, and the links between line manager support and patient mortality are strongest for those in the 41–50 years age band.

There are many separate findings here and, as with Table 17, quite a few may be difficult to interpret. Therefore, we discuss those that are most likely to be understood theoretically: differential effects by occupational group and other effects on absenteeism.

For the various differential effects of occupational group on absenteeism, it is the nurses that have the strongest effect in each case, followed by medical/dental staff. The nurses’ effects are easily understood by the fact that they are the largest constituent group in each trust and, therefore, it is completely reasonable that they would have the largest single effect on trust absence rates. The medical/dental staff findings are not so easily understood by this same explanation, as they do not usually form the second highest proportion of staff. Rather, it may be that the influence of medical staff in trusts is such that their attitudes and behaviours affect other staff to a great enough degree to have an impact on absenteeism. These results are clear and consistent and, given the criteria used, this is one set of findings that can be stated with confidence.

However, it is the turnover intentions of the AHPs and perceptions of work pressure that are the strongest predictors of actual staff turnover. As predictors of patient satisfaction, all main clinical groups as well as administrative/clerical staff had large effects – more so than the other non-clinical groups (for the effect of work pressure, it was nurses who had the greatest effect). The effect of turnover intentions on C. difficile mirrored that in Table 17 and it was general managers who had the strongest effect. Of course, this may represent an inverse effect – if general managers are aware that infection rates are high, then they may be more likely to create workforce changes.

For the differential effects by age group on absenteeism, in the case of two of the three predictors (job satisfaction and staff involvement), it is the ‘average’ category of 41- to 50-year-olds who have the strongest effect on absenteeism. This suggests there is no particularly strong effect of young or old employees. However, for line manager support, the effect was strongest among 51- to 65-year-olds. This suggests that sickness absence is most likely to be affected by line manager support among older workers.

The effect of several variables on absenteeism differed significantly by ethnic group. In most cases, it was white workers’ effects that were the largest, again, perhaps not surprising given they constitute the majority of employees in most trusts. However, for the effects of both work pressure and advocacy, it was the Asian staff who had the greatest effect. It is not clear whether this difference is meaningful or just down to statistical chance.

Analysis by region

Seven breakdowns met the criteria controlling for the outcomes of the prior year and 10 met the criteria without this control included. These breakdowns are shown in Tables 19 and 20, respectively.

It can clearly be seen that in the West Midlands there are much stronger predictors of absenteeism than in any other regions (in each case here, one region had a far stronger predictor than the other regions). This is not due to an unusual distribution of absenteeism or outliers; in fact, the range of absenteeism figures for 2010/11 was lowest in the West Midlands of all 10 geographical regions.

Once again, it is the West Midlands where absenteeism is most strongly predicted and these results essentially mirror those in Table 20. The same is true for the effect of general health on patient satisfaction in Yorkshire and the Humber, and for the effect of work pressure on turnover in the South Central region. Interestingly, London did not come out with the strongest effects in any of this analysis (although if the criteria were relaxed, that could change).

Summary of results

The sheer number of separate analyses presented in Appendix 7, combined with the variability of results and lack of consistent patterns for many of them, mean that interpretation of most results is impossible (or, at least, not sensible). This, in fact, can be considered a finding in its own right. For the most part, there is not a single group of staff (or geographical region) for which staff experiences are the most important; despite this, there are some patterns that become evident when studying the findings in more detail.

Unsurprisingly, given the theoretical proximity as an outcome, there are the most effects (and largest differentials) for predictors of absenteeism. Nursing staff generally had the strongest effects of all the occupational groups, perhaps unsurprising given that they form the largest group of staff. However, medical/dental staff also had substantial effects for most predictors. The turnover intentions of AHPs and perceptions of work pressure were the strongest predictors of actual staff turnover and all the main clinical groups, as well as administrative/clerical staff, had large effects as predictors of patient satisfaction. White employees’ experiences had larger effects as predictors of absenteeism than those of other groups, mainly because they formed the vast majority of the workforce. There were no other easily explainable differential effects by ethnic group.

In terms of geographic regions, absenteeism was most readily predicted – by most staff survey variables – in the West Midlands, while the health of workers in Yorkshire had the strongest effect on patient satisfaction, and work pressure in the South Central region was a stronger predictor of turnover than in other regions. However, aside from the West Midlands, these may be one-off results, with no clear patterns emerging. It is not absolutely clear why these differences should emerge, but the West Midlands has been a region with significant levels of uncertainty in some trusts such as The Mid Staffordshire NHS Foundation Trust15 and a further 3 of the 14 hospitals included in the Keogh Review. 213

Introduction

We implemented two ALSs, partly to help ground the statistical findings in the ‘real world’ and partly to help start dissemination of findings. First, we wished to discover whether or not the statistical results from the study made sense to those dealing with the issues ‘on the ground’ and to hear the real ‘stories’ behind the statistical associations. However, this is not to say that we regarded the ALSs as a formal qualitative method to generate and analyse data such as interviews and focus groups. From the outset, it was made clear that the ALSs would provide support to the quantitative research but would not be part of formal data collection and, therefore, that only the general nature of discussions would be recorded. This means that it was not necessary to record detail as there was no intention to ‘code’ and analyse material.

The second reason was more important. While interviews or focus groups may have been a better way to generate and analyse data, they would, in our view, have been weaker in terms of participant involvement (or ‘buy in’) and dissemination, and so less tempting to those involved as they would not contain the element of shared learning on issues identified by the group. The ALSs were seen as an important element in dissemination to a group of over 40 NHS managers and staff, and public and patient representatives who had an interest in staff and patient experience, and gave them a personal stake in the research which was invaluable in local dissemination. However, our version of action learning differs in one major way from ‘pure’ ALSs (see Rationale for Action Learning Sets). Rather than only set members learning from the theories and results of the study (and from each other), the team aimed to learn from set members. Put another way, there was greater reciprocity in the process. In this way, participants were active subjects (rather than passive objects) and, in some ways, coproducers in the study.

It is generally argued that action learning is difficult to define and can take a variety of meanings in practice. 214 Dilworth215 states that it is difficult to define action learning as it takes a variety of forms. Most commentators state that action learning dates back more than 50 years and has much in common with action research, a concept and term originated by the German psychologist, Kurt Lewin, in the 1940s. However, the term ‘action learning’ itself is generally associated with Reginald Revans, who is seen as the ‘father’ of action learning. 215

Throughout his various writing, Revans avoids defining ‘action learning’, arguing that definition was counterproductive,214 as reported by Dilworth. 215 However, Revans216 suggests that learning is derived in two ways – through both programmed instruction (which he calls ‘P’) and questioning insight (‘Q’). However, by definition, the ‘P’ is all based in the past. Therefore, he suggests beginning with questioning insight (‘Q’) rather than by using past knowledge, which can highlight areas that require the creation of new knowledge (new ‘P’). 215 Revans216 (p. 3) sets out his action learning formula of ‘L (learning) = P (programmed knowledge) + Q (questioning)’. However, some authors also add R for reflection – for which the questioning insight is more important than knowledge acquisition in action learning. For example, Cho and Egan217 stress that reflection is important to balance action and learning in the action learning process. Through reflection, action learning teams can convert tacit knowledge into explicit knowledge and improve their thinking and solutions to challenges. Dilworth215 stresses the importance of bringing people together for reasons other than problem resolution, adding that the learning that occurs is regarded as the primary value.

Action learning has seen a significant growth and is now widely used. Park et al. 218 report a content analysis of 127 articles (case studies and case reports included) published in the journal Action Learning: Research and Practice between 2004 and 2012. Cho and Egan217 point out that special issues on action learning have been included in the following journals: Performance Improvement Quarterly (1998), Advances in Developing Human Resources (1999 and 2010), Journal of Workplace Learning (2000), Management Learning (2001), Learning Organization (2002), Journal of Asia Entrepreneurship and Sustainability (2006), Public Administration Quarterly (2008) and International Journal of Human Resources Development and Management (2012).

Action learning is being used across diverse contexts. In their survey of recent 127 cases, Park et al. 218 found that health accounts for 18% of cases (with some of Revans’ early work being in hospitals),215 while organisation development make up 25% of total cases. Action learning is eclectic and can take a variety of forms. In a variant that is closely interwoven with other organisational interventions. It is closely related to organisation development, management development, team building and transformative learning. The flexibility of action learning in promoting learning and elevating organisational performance can be highly attractive215 and it has been linked to HR development. Action learning has been used for effective communication, work climate, co-operation, shared vision and development at individual and organisational levels. 218 When used appropriately in organisational contexts, action learning can be a powerful approach to HR development. 219 Rigg and Trehan214 write that action learning is a mode of inquiry that has particular value in situations in which people both want to change something about their situation and gain greater insight into the issue as well as their own practice. This leads to action learning being ‘employed for a variety of individual and organisational development purposes as well as to address broad systemic and societal problems’.

Rationale for Action Learning Sets

Weinstein220 discusses the debate in the UK regarding whether some variants of action learning remain acceptable versions or are travesties, concluding that ‘debates involving definitional purity rigid sets of principles and processes are unhelpful’ and any effort to limit the use are unsafe as they hinder research and learning and ‘privilege the ideas of the past and downgrade experience’. Revans himself has said that action learners are ‘. . . always having to re-invent their own ways of putting the basic ideas into practice. This inventing element is what maintains the life in action learning’. 220

‘Pure’ ALSs generally last between 6 months and 1 year, with sets meeting around once a month and set size is generally small, usually no more than six people. 215,220 Our sets were larger (albeit with smaller subset discussions) and met less frequently (see Action Learning Sets participants). We chose diverse sets (see Action Learning Sets participants). Weinstein220 argues that although ‘horizontal’ sets (participants with the same levels of responsibility and authority) are more common than ‘vertical’ or ‘diagonal’ sets (participants from different levels of responsibility and authority, either within the same function or across different ones), and ‘vertical’ or ‘diagonal’ sets often expose a range of issues from different perspectives that are missing in horizontal sets. Similarly, team members with diverse backgrounds are highly desired because participants of diversity can generate innovative ideas and explore different solutions. 216,218

However, we argue that we have kept to the spirit or ethos of action learning. According to Weinstein,220 a ‘true action learning program’ must incorporate the four ‘P’s of action learning. First, it should achieve the two end Products, a task accomplished and implemented, and learning gained; second, it must adhere to the Procedures – the set, the processes in the set and the set adviser; third, it has to value the underpinning Philosophy – honesty, respect for others and the taking of responsibility; fourth, it should explore Programmed ‘knowledge’.

Similarly, we are in accordance with Dilworth’s215 summary of action learning fundamentals that includes the starting point of questioning insight, tackling real problems, strategic learning, the importance of reflection and the primacy of learning.

Moreover, our aim was in line with a key perception of the power of action learning and its benefits that was taken from interviews with a number of current practitioners and ‘users’: to resolve real business problems (‘Action learning is a bridge between analysis and implementation’). 220

In short, our version of action learning made two major contributions to the project. First, it explored whether or not the statistical results from the study made sense to those dealing with the issues ‘on the ground’ and allowed the statistical associations to be compared with some real ‘stories’. Second, it was associated with local involvement and dissemination, in which a stress on reciprocal learning made participants active subjects (rather than passive objects) or (in some ways) coproducers in the study and led to larger stake in dissemination.

Action Learning Sets participants

Given the topical and practical nature of the research, ALSs were used at three stages during the work to provide soundings with the current context of the NHS. First, ALSs were used at the commencement of the research in order to gauge the current preoccupations of managers and the challenges they face about issues connected with staff satisfaction and organisational performance. Second, they were used part-way through the work to hear the reflections of managers, and patient and public involvement (PPI) representatives on the initial findings. Finally, they were used in the closing stages of the research to test how strongly findings resonated with this group of individuals and as a part of local dissemination.

The NHS managers and staff representatives who participated in the ALSs had an interest in the area of study, working in trusts that were taking action to support staff in shifting attitude and behaviour, in the belief that such action would have a positive impact on patient outcomes and other organisational performance measures. Participating members of the public and patient representatives were foundation trust governors, and Local Involvement Networks (LINk) and Healthwatch members, self selected from open invitations to the groups to which they belonged. All participants had links with, or worked for, a range of NHS trusts in the then West Midlands and the East Midlands and from national NHS organisations (a list of participants and their job roles in included in the Acknowledgements). Given their individual interests, there was no need to convince participants of the importance of the area of study.

The first meetings involved only managers and staff side representatives. Invitees were asked to select one of two dates and the same format was used for each session. The approach to the meetings was flexible and adapted to suit participants and discussion area. In those first meetings (June 2012), traditional set methodology was used focusing on the experiences and challenges faced by set members. In the second round of meetings (January and February 2013), PPI representatives were also invited and the methodology shifted into more of an inquiry set exploring the views and experiences of those present and focusing on themes emerging at this stage of the research. All who had participated were invited to a final workshop (June 2013) that discussed provisional findings from the research and commented on dissemination proposals. Notes were taken during meetings and circulated after the meeting to check accuracy. The summaries presented here are constructed from those notes.

Reflections from the Action Learning Sets

Reflections from the first set meetings were grouped into three areas of discussion: an initial exploration of issues perceived as important by NHS managers and staff; an exploration of challenges presented; and possible areas for action. With the frequently rehearsed caveats (response level, interpretation of questions, etc.), it was agreed that the staff survey did provide much useful information and the experience of participants was that NHS trusts did provide a serious focus on the action plan developed annually to improve survey ratings.

There was general agreement that there was a positive relationship between what can be termed staff satisfaction and outcomes and, therefore, that finding ways to improve satisfaction is important, but there was some discussion about terminology such as satisfaction, happiness, engagement, well-being or experience. It was broadly agreed that ‘discretionary effort’ given by the individual member of staff such as smiles, reassurances and personal touches can make a big difference to patient satisfaction (experience) and ultimately to organisational outcomes.

However, this link is complex as it was pointed out that professional cultures are often stronger than organisational ones and that sometimes there will be good engagement in some parts of an organisation and not in others. This reinforces the importance of the ‘microsystem’39 and that action at the organisational level does not permeate down to every ward.

The importance of the pivotal role of leadership, line management, the quantity and quality of appraisal and managing poor performance was pointed out. It was suggested that sometimes line managers tended to avoid ‘difficult conversations’, perhaps as there can be a thin line between fair criticism, and bullying and harassment.

At their second meeting, set participants were joined by a number of members of the public and the discussion was focused on the four factors that seemed (at this stage of the research) to be the most important indicators of staff satisfaction and organisational outcomes: quality of job design, work pressure felt, work–life balance and support from supervisor.

Members tended to agree that understanding roles and having clear goals and objectives in one’s job was vital. This is related to the quality and quantity of feedback that one receives and the extent to which one is involved in decisions regarding changes to one’s job, team or department. Members tended to recognise the differences in the staff survey between having an appraisal in the last 12 months, a well-structured appraisal and a personal development plan resulting from that appraisal. Work pressure involved time pressure, lack of sufficient staff to complete the allocated work and inability to maintain a desirable standard of work in terms of quality. Members pointed to the pressure associated with the organisational turbulence of ‘change, change and more change’. Work–life balance referred to the extent to which the trust and the line manager are committed to offering opportunities for flexible working and how much they promote a work–life balance. Finally, support from supervisor referred to the extent to which the line manager encourages teamworking, provides help and feedback, encourages participation in decision-making and shows individualised interest.

There is much anecdotal evidence about a bullying and harassment culture in the NHS, and bullying, because of pressure to meet targets, was noted as a possible side effect of poor job design. Members pointed out that such managerial behaviour can result in a spiral of consequences including sickness and absenteeism. Finally, it was noted that agency staff are often disconnected from management systems (including the staff satisfaction survey) and can reduce productivity. It is possible for whole shifts on wards to be agency staff, making it very difficult for supervisors to know the skills and weaknesses of those they are supervising.

A final ALS was held in June 2013 to discuss the emerging findings from the research examining the links between staff satisfaction and organisational performance and to comment on policy implications. There were three areas for discussion: appraisal, teamworking and differences linked to gender and occupational group. There was strong agreement between the two groups of participants (NHS staff and PPI members) in seeing the value of a satisfactory and supportive appraisal process, but it was suggested that appraisal needs to be more of an ongoing process than an annual ‘event’, although the notion of ring fenced appraisal time was regarded as important. Although teamworking was seen as important, it was noted that the definition of a ‘team’ is often problematic and the notion of working in a team is sometimes hidden in many questions about such areas as ‘feeling involved’ or ‘contributing to developments’. Finally, managers were particularly interested in the disaggregated results as they felt that there was little previously available research that raised questions regarding individually specific approaches and needs. For example, they recognised that there may be a range of approaches to promotion and to pressures external to work (e.g. women disproportionately affected by child care and other caring roles), and the question of possible increasing gender and occupation differences as people work longer was raised.

Conclusions

The second and third ALSs included a mixture of individuals with diverse backgrounds and experiences but sharing a common intention – to look for ideas within the unfolding research that would provide insights to enable NHS organisations to enhance their abilities to become high-performing organisations. The NHS managers found that working with members of the PPI representatives was valuable and challenging. Similarly, those members of the public who are part of external groups or serving as governors in foundation trusts were keen to take back discussions to their organisations and have some influence from a position of authority given by additional information gained.

Both groups recognised the importance of the issues discussed and noted that there were complex relationships between many contributing factors. Discussions frequently returned to the impact of appraisals, with participants stressing that it would be unwise to focus purely on the annual and possibly ritualistic process that brings together manager and member of staff to talk about progress and intentions. All participants emphasised the ongoing relationship between manager and staff as the key factor and one that would ensure a supportive and challenging environment for work.

Those managers who participated were well versed in using the information from the annual staff survey to tackle areas of concern within their organisation. They were also interested in exploring the tensions that the survey illuminates but cannot answer (for instance the ‘right’ way for middle managers to manage weaker members of staff) and were keen to see research results as headlines for exploration.

The ALSs involved over 40 managers, PPI representatives and national policy influencers in total, who added a valuable element to the quantitative research in three main ways. First, it has added an active dimension of staff and PPI involvement. Second, it provided some ‘validation’ for the statistics in checking findings against the real life stories and experiences of the set members, which feeds into implications for practice (see Chapter 10). Third, it gave set members a personal stake in the research, which is a valuable component of local dissemination.

Introduction

This chapter aims to summarise and explore the main themes arising from the quantitative analysis. It also relates these to the ALSs, described in Chapter 9, which put some ‘flesh on the bones’ of the results in order to make them meaningful for NHS managers.

The main aim of the project was to use secondary data to test part of the overall model that hypothesises a positive link between HRM and organisational performance in the English NHS. In broad terms, HRM practices (e.g. training and development, appraisal/performance management) are associated with intermediate outcomes, including staff attitudes (e.g. staff satisfaction, turnover intentions, absenteeism) and final outcomes (e.g. patient satisfaction, mortality). This may be conveniently seen as two ‘chains’: between HR practices and intermediate outcomes, and between intermediate outcomes and final outcomes. This leads to the main research questions that are the focus of the main empirical chapters:

Q1 (see Chapter 6): What are the links between individual staff experiences (e.g. satisfaction, engagement, turnover intentions) and intermediate staff outcomes (e.g. staff absenteeism, actual turnover)?

Q2 (see Chapter 7): How do these link with organisational performance (e.g. patient satisfaction, mortality)?

Q3 (see Chapter 8): Do these measures and relationships differ by occupational, demographic groups, trust types and geographical areas and, if so, what is the relative change for each group?

It should be noted that, although we had originally framed the questions in terms of staff satisfaction and attitudes, this was broadened somewhat to explore staff ‘experiences’. Such experiences (as measured in the NHS National Staff Survey) included a variety of attitudinal and well-being scores, but also experiences of negative events at work (e.g. violence, harassment) and other features that are closer to the HR practices end of the chain (e.g. appraisal, training, a variety of job design features). Some of these measures could be seen to cut across different elements of the model (e.g. satisfaction with the ability to provide care for patients, which is an attitude in relation to a job design feature), and, therefore, we included all staff experience variables from the NHS staff survey for the first two research questions.

Summary of literature reviews

Owing to the complex nature of the study, we did not complete one single literature review, but instead conducted three separate reviews. The first examined the HRM performance literature in general terms, the second was a systematic review of this relationship in health care and the third studied policy literature relating to the topics.

Chapter 2 provides a critical review of the theoretical and methodological challenges associated with the broad field of the HRM–performance relationship. This was loosely structured around Guest’s16 argument that the field requires a better theory about HR practices, outcomes and the link between them. We discussed HRM/independent variables in terms of single practices compared with bundles, and fit/universalistic, configurational or contingency perspectives. The main issue here is whether ‘one size fits all’ in all situations or whether best practices vary in different contexts of countries and industries.

We discussed outcomes/dependent variables in terms of organisational and employee perspectives. Many commentators point out that most studies focus on organisational outcomes, while fewer studies focus on employee outcomes. However, the relationship between them is far from clear, with two competing views of the ‘mutual gains’, ‘optimistic’ or ‘win–win’ perspective and the ‘conflicting outcomes’, ‘pessimistic’ or ‘sceptical’, ‘win–lose’ or ‘lose–lose’, or ‘counteracting effects’ perspective.

Many researchers note the ‘black box problem’ in linking HR practices with outcomes. Although the theoretical basis of many studies is often implicit, commentators have noted some focus around AMO theory. Methodological problems include the dominance of cross-sectional rather than longitudinal designs, which makes it difficult to say anything meaningful about causality.

We argue in favour of the best fit rather than the universal best practice approach. However, while at one level it is clear that context is important if only because many outcome measures used in studies of manufacturing such as profit are not appropriate for institutions such as the NHS, it is not fully clear which contextual features are most important. We discuss three contexts from the literature that may justify a study of the NHS, associated with its setting within services, health care and the public sector. Commentators argue that there are few studies on health care in general and on the NHS in particular,19,24–26 which suggest that reviews focusing on health care and studies of the NHS are useful.

Our argument of both the limited information on HPWS in health care and importance of context suggests that a review of health care is valuable (see Chapter 3). Moreover, we found that some existing reviews Etchegary et al. 136 and Garman et al. 137 cover rather different ground and come to rather different conclusions from each other. This justified conducting our own systematic literature review in order to explore (1) HPWS definitions in relation to those commonly adopted in non-health-care research and publications, (2) the extent to which the primary characteristics associated with HPWS in general literature are reflected in the health-care literature, (3) the dominant theoretical frameworks used in linking HPWS with outcomes in health care, (4) the terminological choices and their appropriateness in the health-care literature on HPWS, (5) the evidence on the link between HPWS and outcomes in health care, and (6) the various mechanisms through which, and conditions under which, HPWS have a positive effect on outcomes in health care. We initially identified 27 publications and added a further 15 publications through snowballing to yield 42 publications (23 quantitative empirical studies, seven qualitative empirical studies, four mixed-methods studies, five reviews, two commentaries and one theoretical article). Our main conclusions are that there is a lack of longitudinal studies that investigate causality. Various studies appear to report on the same data, thus possibly inflating the reported effects. The country variation among the reported studies is limited, thus making it difficult to reach generalisable conclusions. Finally, the majority of studies investigate a limited range of HR practices, thus making it difficult to reach conclusions with regards to the effects of the HR system overall.

Chapter 4 reviewed the policy literature of documents from government, business and public bodies of the ‘business case’ that staff satisfaction leads to greater organisational performance. There are a series of reports by a number of bodies drawing on different, but connected, debates inside and outside the NHS. The generic business case has been carried out with reference to ‘Good Jobs’, work and well-being, and engagement. Similarly, a series of reports from the DH and other organisations have stressed the importance of staff involvement and engagement and health and well-being over a period of about 15 years.

We concluded that issues such as staff engagement, and health and well-being have been on the generic national and NHS agendas for a long time, although most of the focus has been on the topic of involvement or engagement. However, at one level much of the discussion in the policy documents is too broad, consisting of fairly vague assertions that ‘staff engagement’ will lead to better performance without consideration of issues such as cost, context, causality or mutual gains. Moreover, there is an element of a continuing and sometimes recycled policy debate with variable implementation into practice. In short, our literature reviews suggest that there is limited evidence on the applicability of HPWS concepts to the NHS (English; service sector, public service; health care) setting (see Chapter 2); few empirical studies on health care in general and on the NHS in particular (see Chapter 3); and a rather broad and vague ‘business case’ based on a number of untested optimistic assumptions, which has seen a rather patchy pattern of implementation (see Chapter 4). All these factors suggest that the empirical study of the following chapters is needed.

Summary and discussion of quantitative results

The analyses in Chapters 6 and 7 addressed research questions 1 and 2, which concerned the links between staff experience, intermediate outcomes and organisational performance. This included both cross-sectional (individual level and trust level) analysis and longitudinal analysis. Although previous work has examined cross-sectional individual data, this is the first study to do so in such a large NHS sample in a systematic way. There have been far fewer longitudinal studies and we believe that this is also the first to examine such longitudinal data on hundreds of health-care organisations in a systematic way. Thus, we believe that our analysis makes a significant contribution to our knowledge about substantial parts of the HR model and how it operates in the English NHS.

However, the nature of this contribution in terms of the findings is varied. Although many of the cross-sectional results were in line with expectations and demonstrate the more important staff experiences in determining outcomes, some of the longitudinal results actually demonstrated that the picture of how parts of the overall model (in terms of causality) is a lot less clear than might be anticipated.

Many of the individual level results are largely as expected and tend to confirm results found in earlier studies (in particular those discussed in Chapter 2), with some added insight in some cases. In general, there are highly significant links between positive experiences (particularly well-designed jobs, meaningful roles, lower work pressure) and individual outcomes including higher job satisfaction and advocacy, lower stress, lower presenteeism, fewer adverse effects of health and lower turnover intentions. These results are somewhat expected owing to large sample size and possible common method variance (i.e. outcomes being measured by the same people who have the experiences), so should not be overinterpreted; however, there were some strikingly substantial effects.

For example, individual outcomes (such as turnover intentions, well-being and satisfaction) are strongly affected by negative experiences – not just aggression from patients and (even more so) colleagues, but particularly by not believing that their employer offers equal opportunities for career progression and promotion. On the other hand, these outcomes were particularly enhanced by staff engagement, both in terms of affective work engagement and other job design factors that allow the contributions of employees to be clearly made, for example being able to provide care to a level that staff find satisfactory and being able to develop potential. It seems clear that the motivation of NHS staff to provide a good service is an important factor in their individual well-being.

In fact, the importance of discrimination and perceptions of equal opportunities was a feature of much of the analysis. This mirrors the earlier findings on an NHS data set221 that showed links between discrimination on the basis of ethnic background and overall job satisfaction (not just for those being discriminated against); these authors also found that the level of diversity training in an organisation predicted the extent of discrimination. Taken in conjunction with the findings of this report, it appears that diversity training and other measures to prevent discrimination – whether on the basis of ethnic background, age, gender, disability or other characteristics – should be important not just to fulfil mandatory requirements, but to ensure the well-being and health of the workforce in general.

Absenteeism is lower when staff job design is better, including good-quality teamworking; when motivation is higher; when communication is better; and when there are fewer instances of violence and harassment, both from patients and from other staff. Absenteeism is actually lower when more staff work extra hours, but this may be explained by the fact that staff cannot both be working more hours and be absent. However, when more staff work extra hours, turnover is higher, possibly indicating that the negative effect of working too much becomes clear in terms of employees wanting a change of direction. The cross-lagged correlations suggest that when staff work in a more supportive environment, they are less likely to be absent (rather than the other way round).

In fact, the organisational-level analysis with absenteeism is probably the most instructive and clear set of findings from this analysis. The cross-lagged correlations suggest that there is clear evidence for the direction of the effect between absenteeism and over half of the staff survey variables: it is much more likely that good staff experience leads to lower absenteeism than vice versa. These effects are particularly strong for negative experiences such as violence and harassment, but are also very strong for the positive experiences of staff being able to contribute towards improvements at work and when there is good communication between management and staff. Combined with the latent growth curve analysis that gave similar results, this presents a very clear and unambiguous set of findings about the nature of NHS staff jobs and absenteeism, mirroring and extending previous research on the same variables. 48,50,106

We found that turnover is lower when work pressure is lower, training is more widespread, appraisals happen more frequently and effective action towards violence and harassment is perceived to take place. Improvements in job design and decreases in aggression from other staff are associated with subsequent decreases in turnover. However, results from the longitudinal analysis involving turnover are far more difficult to interpret; there is certainly some evidence of direction of causality from turnover to these outcomes being quite confused. For example, while there is an increase in turnover in subsequent years in trusts for which a higher percentage of staff are experiencing physical violence from patients or their relatives, it is more difficult to explain the link with the availability of hand-washing materials. There is also an increase in turnover in subsequent years in trusts for which staff receive more training; have more opportunities for flexible working; and higher levels of appraisal. This may partly be linked with the training paradox: that a better trained workforce may be better placed to compete for posts and perhaps promotions in other organisations, but it is difficult to explain the link with flexible working.

There was some strong latent growth curve analysis results, suggesting that when the number of staff having meaningful jobs increases, when there are decreases in aggression from other staff, and when belief in their employer as both a place to work and a place to receive treatment increases, then turnover tends to decrease over subsequent years. However, many of the other results – particularly the cross-lagged correlations – gave inconsistent or counterintuitive findings. In some cases, it appears that trusts may react to high turnover by, for example, increasing opportunities for flexible working. Of course, these findings have to be placed in the context of major changes over the NHS over the study period, including many large reorganisations of services, necessitating more movement of staff between trusts (and, in some cases, redundancies) than would normally be expected. Therefore, although objective data is usually better to use than subjective, it is probably more instructive to look at the patterns of results with self-reported turnover intentions (from the multilevel analysis), where the findings met with expectations, than the more surprising results using the stability index. Because of this, there is far less that can be learned from the longitudinal analysis with turnover than would be the case under less turbulent circumstances.

Patient satisfaction is the one organisational performance variable (final outcome) that showed consistent results. As found in a less rigorous study of NHS staff and patient surveys,222 a large number of staff experience variables – in particular when staff are engaged, not under particularly high work pressure, and do not experience discrimination – result in patients in those trusts being likely to rate the care they received more highly. Patient mortality, which had been shown in other studies126,223 to be related to HR variables, was not so clearly related in this instance.

However, it is more difficult to interpret some of the longitudinal analyses with many of the outcomes. Staff experiencing physical violence from patients or their relatives is associated with higher patient mortality rates in subsequent years. However, feeling valued by colleagues is associated with higher mortality rates in subsequent years and the percentage of staff suffering work-related injuries or illness is associated with lower patient mortality rates in subsequent years. At least some of these counterintuitive results may partly be methodological artefacts – the latent growth models employed can pick up on ‘regression to the mean’, for example if a trust had an unusually low level of patient mortality one year, and in this same year it had good staff experiences, then the mortality rate may naturally rise back to its expected level in subsequent years, making it appear as though good staff experiences are followed by a rise in mortality rates. For this reason, we do not recommend interpreting these findings without much clearer attention to the other analyses, particularly those looking at changes in staff survey variables and subsequent changes in outcomes.

There are some links between changes in staff experience from 2009 to 2010 and changes in outcomes over the 2009–11 period. A decrease in turnover in subsequent years is associated with an increase in staff agreeing that their role makes a difference to patients, an increase in the percentage of staff feeling that there are good opportunities to develop their potential at work, and an increase in their level of willingness to recommend the trust as a place to work or receive treatment. An increase in the percentage of staff experiencing harassment, bullying or abuse from other staff is associated with an increase in turnover in subsequent years. However, an increase in the percentage of staff suffering work-related injuries or illness is associated with a decrease in turnover in subsequent years. An increase in perceptions of effective action from employer towards violence and harassment is associated with a decrease in patient mortality rates, but an increase in shift-working is associated with an increase in patient mortality rates. Finally, an increase in the reported impact of health and well-being on employees’ ability to perform their work and daily activities is associated with a decrease in patient satisfaction.

We also looked at cross-lagged correlations to examine whether or not there was evidence of directional relationship between staff variables and outcomes. This set of analyses was expected to be a great strength of the study; the availability of data from so many organisations in consecutive years allowed a large number of tests with relatively high statistical power. Although such cross-lagged correlation tests, which examine whether the relationship between two lagged variables is stronger in one direction than another, cannot say anything definitive about causal relationships, if there is a causal relationship between two variables then we would normally expect the correlation to be stronger in that direction. Therefore, if there were a consistent set of causal relationships, we would expect to see a clear pattern of results emerging for similar variables.

When looking at absenteeism as the (intermediate) outcome, there was a clear pattern that there was a stronger link between staff experiences and subsequent absenteeism (i.e. staff absence in the year following the measure of experience) than vice versa, i.e. there is evidence of a causal link between what staff experience and their subsequent levels of absence. This is entirely in line with theoretical expectations and represents a significant contribution above previous studies that have shown cross-sectional links between these variables, for example the Boorman review. 48

However, in general, for other outcomes this was not the case. When examining turnover, patient satisfaction, patient mortality and infection rates, these correlations were not always in the expected direction; for example, patient satisfaction was more closely related to subsequent levels of flexible working than vice versa. For some outcomes (turnover and mortality) there were methodological limitations that may contribute towards this. Infection rates as an outcome are more troublesome; the links between staff experience and infections are more distal theoretically and there is the danger of reverse causality (i.e. when infections abound, this may affect staff well-being both directly and indirectly). However, for patient satisfaction as an outcome, neither of these problems should be the case. Therefore, the mixture of direction in links between staff experiences and patient satisfaction suggests that we can be less certain about the causality here. It may be, in fact, that there are more complex relationships at play; not only do staff experiences affect outcomes, but outcomes (including patient satisfaction) also have an impact on staff attitudes. Such reciprocal relationships are difficult to capture, particularly when the measurements are relatively blunt (i.e. measured for entire, large organisations, and with annual frequency).

There were some clearer indications about the links between intermediate and final outcomes. In particular, mortality is lower following lower levels of absenteeism – the direction of causality is fairly clear here. However, we see that higher infection rates are usually associated with subsequent increases in staff absenteeism, as the infections can affect staff as well as patients.

Despite these clear effects, there is little or no evidence for the links between staff experience and organisational performance being mediated by intermediate outcomes. That is, although the experiences of staff contribute directly to levels of absenteeism (and maybe turnover), and in many cases to outcomes (particularly patient satisfaction), these appear to be separate mechanisms. Patients are not less satisfied because of greater absence among staff or a greater turnover among staff. These two intermediate outcomes have a great financial cost for trusts48 and decreasing them is not only in the interests of general good management, as staff experiences and attitudes have a more direct effect on patient views.

We need to acknowledge that it is likely that some of these results are type I errors, that is, they are among the 1 in 20 non-relationships that are found to be statistically significant by chance alone. When testing as many relationships as we do in this report, it is inevitable that some type I errors will occur and it is impossible to know which these are. As a general rule, relationships that are predicted by theory and have been demonstrated in other samples are more likely to be believable. Patterns of similar findings arising from different variables, even if not replicating results found before, are also more likely to be representing genuine effects; one of the reasons for conducting this study is that many of these things had not previously been looked at in large health-care samples, and, therefore, the generation of new knowledge – or at least findings which may be confirmed by subsequent studies – is an important part of that. However, when a significant result does not conform to expectations, and stands alone (without similar patterns), these are most likely to be the type I errors. Although we would not discard such findings completely, we urge further examination before any firm conclusions are drawn.

The analyses in Chapter 8 address research question 3, which concerns the differential effects by groups of staff and different geographical regions. It is possible that relationships may be contingent rather than universal. For example, it may be expected that the relationship between turnover intention and actual turnover may vary depending on the nature of the labour market in different parts of the country. Similarly, there may be different relationships between engagement and patient satisfaction for doctors and nurses, for example. If associations are differential rather than universal, this may suggest focusing on different issues in different contexts and exploring different ‘policy levers’.

Few previous studies have examined this disaggregate level and so this should be considered exploratory – we did not begin with a priori expectations about which groups would differ from which. Previous reports106 have shown there are differences in the experiences between groups of staff; these are also shown in the annual publication of NHS staff survey results at www.nhsstaffsurveys.com. 206 Our interest was not in replicating results from these studies, but in examining relationships between experiences and outcomes.

A very large number of tests (corresponding to the large number of breakdowns and variables) were conducted and, especially given the lack of particular expectations, this would likely lead to a large number of type I errors if we were simply to consider the results at face value. Instead, we imposed criteria to select out those for which there was the clearest difference between groups or regions. Given the theoretical proximity as an outcome, it was not surprising that there are the most effects (and largest differentials) for predictors of absenteeism. Nursing staff generally had the strongest effects of all the occupational groups, which makes sense given that they form the largest group of staff and can, therefore, have the largest single effect on trust-level absence rates. However, medical/dental staff also had substantial effects for most predictors. The turnover intentions and perceptions of work pressure of AHPs were the strongest predictors of actual staff turnover, and all the main clinical groups as well as administrative/clerical staff had large effects as predictors of patient satisfaction. These results point to the conclusion that no single group of staff (clinical or non-clinical) has a monopoly on outcomes and, although it may be tempting to conclude that nurses’ experiences are the most important for some, this may simply be because nurses are the largest constituent of the workforce.

For other breakdowns by staff, there were few results meeting the criteria that were immediately explainable. One exception was that white employees’ experiences had larger effects as predictors of absenteeism than did the experiences of other groups; however, this is likely, again, to be because white employees formed the vast majority of employees in most trusts. Overall, these results again point to the conclusion that it is the experiences of all staff, wherever they are in the organisation, that are important.

In terms of geographic regions, absenteeism was most readily predicted, by most staff survey variables, in the West Midlands, while the health of workers in Yorkshire had the strongest effect on patient satisfaction and work pressure in the South Central region was a stronger predictor of turnover than in other regions. Although regional differences are potentially interesting, it is difficult to attribute reason for these findings. In particular, it is tempting to interpret the West Midlands result in the light of the Francis report15 and the documented issues not just with Mid Staffordshire NHS Foundation Trust, but with others in the region as well (the West Midlands Strategic Health Authority being criticised for its use of, and reaction to, data within the report); however, it is not immediately clear why experiences should predict absenteeism more in this region than in others. Indeed, these may be one-off results, with no clear patterns emerging. Given the exploratory nature of the analysis, the difficulty of interpreting all these results, and the absence of many clear patterns, any conclusions must be tentative.

Discussion of Action Learning Sets

Action Learning Sets were used for two main reasons: to ground the statistical findings in participants’ experiences and as a contribution towards local involvement and dissemination. The first meetings (June 2012) involved NHS managers and staff, while the second meetings (January and February 2013) and final workshop (June 2013) also involved PPI members.

Reflections from the first set meetings were grouped into three areas of discussion: an initial exploration of issues perceived as important by NHS managers and staff, an exploration of challenges presented and possible areas for action.

At their second meeting, set participants were joined by a number of members of the public. It was broadly agreed that the four factors that seemed (at this stage of the research) to be the most important indicators of staff satisfaction and organisational outcomes – quality of job design, work pressure felt, work–life balance and support from supervisor – made sense to participants.

A final ALS was held in June 2013 to discuss the emerging findings from the research examining the links between staff satisfaction and organisational performance and to comment on policy implications. There were three areas for discussion: appraisal, teamworking, and the differences linked to gender and occupational group. Although participants largely agreed on the importance of appraisal and teamworking (which was in line with previous studies), it was recognised that the implications of disaggregated results were much less clear as they posed new questions on issues with little previously available research.

Integration of study elements

Just as the HPWS stresses the additionality of elements into synergistic bundles, links between the different elements are intended to increase the synergy of this study. The literature reviews provided the conceptual and policy foundations for the empirical study. The conceptual review (see Chapter 2) explored the relevance of the HPWS literature. It found that the literature suffers from some conceptual and methodological problems. In particular, the ‘best fit’ literature argued that conclusions from other settings (countries, sectors) cannot easily be transferred to the NHS. This justified a literature review of health-care settings (see Chapter 3) and the subsequent empirical study of the NHS (see Chapters 6–8). In addition, it suggested that a major problem of the literature was the lack of longitudinal research, which justified the research design that analysed data from three data periods.

The literature review of HPWS in health care (see Chapter 3) found that many of the issues from the broader literature also occurred in this setting. In particular, we found that there were few studies on the NHS and very few that used longitudinal research designs.

The policy review (see Chapter 4) showed that the ‘business case’ linked staff satisfaction and organisational outcomes have been broadly accepted by government, public and other bodies both generally and within the NHS. However, a long history of policy documents have not clearly resulted in effective local action in all trusts.

There were some obvious links between the quantitative study and the ALSs. First, the ALSs were used to ‘validate’ the quantitative study in the sense that the relationships found were confirmed by set members as ‘real’ rather than statistical artefacts and that the data sources were seen as broadly relevant measures of the constructs. Second, emerging results from the quantitative study were generally ‘validated’ by set members in that they fitted into their experiences and this provided some experiential backing for findings and implications based on issues such as staff appraisal. The findings from the ALSs also help further interpret the findings from the quantitative analysis, including the analysis not presented in the ALSs (either because it had not been conducted by the time of the ALS in question, or because it was necessary to focus on a subsection of the findings only due to time constraints). For example, the ALSs participants described the subcultures that could exist within trusts and microsystems within organisations where things operated better (or worse) than elsewhere in an organisation. When such variation exists, the overall trust-level results (for either staff experience or outcomes) become less meaningful, as they represent an average of disparate departments or teams rather than a coherent organisation-wide culture. This is to be expected, of course, but the reports from the ALSs that individual line managers can vary significantly suggests that results around staff experiences where line managers are important (e.g. appraisal, well-structured teamworking, support from immediate managers, work pressure, opportunities for flexible working) may be particularly compromised by such variation and these relationships may benefit from more finely grained analysis on individual teams or departments.

The viewing of appraisal as more of an ongoing process rather than an annual event also suggests that the survey questions around appraisal may not be ideal for examining what goes on and this may be why fewer significant results involving appraisal were found than might be expected.

Practical implications

It is difficult to draw clear practical implications as we considered only the second link in the chain of the model (staff satisfaction and organisational outcomes) and did not focus on the first link between HR practices and staff satisfaction. This means that drawing implications for HR practices is problematic. However, some elements that appear in the staff survey (such as appraisal) can be linked, while others can be inferred (e.g. if reported harassment is strongly linked to organisational outcomes, then this suggests a focus on anti-harassment measures). Our broad conclusions support existing policy and other work in the field, e.g. by Maben et al. ,39 that individual employee satisfaction is best seen as a precursor rather than a result of the performance of patient care and so it is important to encourage staff satisfaction not just for its own sake but to enable the delivery of patient care that is of a high quality.

Many of the individual-level results are largely as expected and tend to confirm results found in earlier studies. For example, intermediate outcomes (e.g. higher job satisfaction and advocacy, lower stress, lower presenteeism, fewer adverse effects of health and lower turnover intentions) are positively linked with experiences (e.g. staff engagement, well-designed jobs, meaningful roles, lower work pressure) and negatively linked with aggression from patients and colleagues, and not believing that their employer offers equal opportunities for career progression and promotion, for example. Similarly, we found that turnover is lower when work pressure is lower, when training is more widespread, when appraisals occur more frequently and when effective action towards violence and harassment is perceived to take place. We found evidence of some causal relationships in some cases. For example, the cross-lagged correlations suggest that when staff work in a more supportive environment, they are less likely to be absent (rather than the other way round).

Some of these measures showed clear links to the organisational performance variable (final outcome) of patient satisfaction. In particular, when staff are engaged, are not under particularly high work pressure and do not experience discrimination, then patients in those trusts are likely to rate the care they received more highly.

There are some links between changes in staff experience from 2009–10 and changes in outcomes over the 2009–11 period; a decrease in turnover in subsequent years is associated with an increase in staff agreeing that their role makes a difference to patients, an increase in the percentage of staff feeling that there are good opportunities to develop their potential at work, an increase in their level of willingness to recommend the trust as a place to work or receive treatment. An increase in the percentage of staff experiencing harassment, bullying or abuse from other staff is associated with an increase in turnover in subsequent years.

We also looked at cross-lagged correlations to examine whether or not there was evidence of directional relationship between staff variables and outcomes. For example, there was a clear pattern of a stronger link between staff experiences and subsequent absenteeism (i.e. staff absence in the year following the measure of experience) than vice versa, i.e. there is evidence of a causal link between what staff experience and their subsequent levels of absence. This is in line with theoretical expectations and represents a significant contribution above previous studies that have shown cross-sectional links between these variables, e.g. the Boorman review. 48

The strongest evidence is derived from results associated with different methods and research questions, ‘validated’ by our ALSs and in line with prior evidence. On this basis, the most obvious implications for practice appear to be:

• Set clear guidelines and take effective action on harassment, bullying or abuse from other staff (as stated in Woodrow and Guest190).

• Ensure that appraisals are conducted effectively, not just as a ‘box-ticking’ exercise: with clear objectives and personal development plans agreed, with appraisers trained to conduct these appropriately (i.e. see also Powell et al. 224).

• Ensure that teams are constructed to meet the needs of the task and the patients, and that these teams have clear objectives, with clear interdependent roles of team members but with opportunities to reflect on performance (i.e. see also West et al. 21,126 and Maben et al. 39).

• Invest in unit-level leadership and supervisor support, including appropriate training for both new and existing supervisors, so that clinicians promoted to management positions have the appropriate skills to deal with matters such as bullying and harassment, time management of staff, and monitoring health and well-being (as well as conducting appropriate appraisals and team leadership) (i.e. see also Maben et al. 39).

Limitations and directions for future research

The research conducted had many strengths, including the use of large-scale data sets in extensive analysis without the need for expensive primary data collection. It made use of 3 years’ data across nearly 400 organisations [longitudinal data analysis on staff experiences in organisations is seldom (if ever) seen on such a scale in any industry and any part of the world] and so we believe that this is ground-breaking in that respect. It also made use of relatively sophisticated methods (latent growth curve modelling and cross-lagged correlation analysis, as well as multilevel regression analysis) to conduct the tests. Much of the analysis was informed by a clear theoretical model and, thus, we were able to test relationships that had been clearly hypothesised by previous researchers.

However, there were also several limitations. Most of these related to the measures available to us. First, the NHS staff survey – although providing excellent coverage of many issues – does not include everything that is part of the HR model.

Second, some of the outcome measures did not perfectly capture what we would want. The mortality index that is published and used within the NHS changed in the middle of our survey period, which makes it difficult to interpret longitudinal findings involving mortality. Turnover (or, more precisely, stability) is also difficult to assess longitudinally, owing to the changing NHS environment over the study period (a mixture of uncertainty about the future, cutbacks in many areas and reconfiguration of trusts and services) means that turnover cannot be assumed to be a result of individual decisions.

Third, the organisational performance variables (trust outcomes) were all measured in acute trusts only. We had hoped to include some performance measures that could be applied to all NHS trusts; however, despite our examination of a variety of potential sources (including measures relating to CQUIN), no suitable quantitative measures for the whole of the NHS were found. Previous research, for example West et al. ,106 had examined ratings produced for the Healthcare Commission’s Annual Health Check as outcomes; however, these have not been produced since 2009. It is impossible to say whether or not relationships with patient satisfaction, for example, would also be applicable across other types of trust, although there is no particular reason to believe that they would not.

Fourth, owing to the nature of the data we were constrained to study entire trusts with annual measurements. In reality, this may not be sufficiently sensitive to pick up all the effects of staff experiences. Many may become a lot clearer if measured in smaller units and with appropriate outcomes measured a suitable length of time afterwards. Although it is somewhat reasonable to expect that the overall experiences of staff in a large hospital in 2010 may affect overall absence rates between April 2011 and March 2012, it is far more likely that the experiences of staff in a single team or department at a given point in time would be reflected in absenteeism over the next few weeks or months.

Fifth, the analytical methods themselves are often insufficient to detect causal relationships. This is partly due to that mentioned in the previous paragraph: the design of the study was too blunt to pick up more finely grained effects. However, it is partly because the actual relationships may be more complex than such methods can model, at least with the extent of data available to us. There may be reciprocal causal relationships between staff experiences and outcomes that we were unable to account for fully. In addition, the techniques of cross-lagged correlation analysis and latent growth curve modelling are known to be insufficient to capture complex, multivariable relationships with absolute accuracy. 197,201,202 However, short of a fully randomised control trial, few methods are able to assign causality in a very clear way.

Finally, the staff experience variables themselves are often closely related and it is not always possible to distinguish between the effects of different variables. We chose not to include all staff experience variables in the same analysis and, for the most part, this would not allow sufficient degrees of freedom to conduct the tests adequately. Even if it would (which would be possible in the individual-level analysis), it would result in far too much multicollinearity between predictors for the results to be interpretable. Therefore, although we find many links between different staff experiences and certain outcomes, it may well be that it is the same overall effect that we observe in multiple ways.

All of these point to some interesting possible directions for future research. There is still much scope for detecting exactly how staff experiences and outcomes are linked, but in our view there are some priorities for further research.

First, there is a need to explore the link between HRM and performance more fully in a health-care setting, exploring the full model rather than our focus on the second ‘chain’ between staff satisfaction and organisational outcomes. This might involve the careful evaluation of interventions designed to improve staff experience. The use of appropriate designs (e.g. randomised control trials or stepped-wedge designs at individual, group or department levels) could identify when such interventions actually do have an effect on staff experiences and patient outcomes. Many such interventions take place in trusts up and down the country, but most are evaluated in a far less rigorous way, if at all, meaning that the evidence base is rather thin. This could be connected to the HPWS literature to examine whether or not ‘bundles’ are more effective than individual practices. Moreover, the vast majority of work focuses on effectiveness or the benefit side with little or no consideration of efficiency, which also includes the cost side. Some work on cost-effectiveness of bundles or practices may suggest priorities for investment.

Second, most secondary data such as those used in this study focus on the organisational level. It is important to examine links between staff satisfaction and performance at the micro (e.g. ward) level, as it is clear that there may be some highly performing wards in poorly performing trusts (and vice versa). 39 Therefore, qualitative work exploring local leadership and microclimates are necessary to complement organisational-level quantitative work.

Third, there is a need to explore HPWS in settings beyond acute care. Many existing data are more clearly suited to acute settings, but there is much public, policy and professional concern over long-term care. It is unclear if broad conclusions that are largely based on acute care can be easily transferred to long-term care, with a very different pattern of staffing.

Fourth, we suggest continued longitudinal examination of the links between staff satisfaction and organisation performance. It is possible that links between satisfaction and performance may change from historical patterns owing to the change to less generous funding after the financial crisis (the ‘Nicholson challenge’) and the ‘external shock’ of the reorganisation associated with the coalition health reforms.

Fifth, we would recommend the continued use of secondary data sources, such as those used in this report, to answer more specific, theoretically driven questions. Such research is relatively inexpensive and can make good use of data that have already been collected. In some cases, this can be complemented by further data collection to expand the possibilities. For example, if outcome variables that could be applied in non-acute trusts were to be developed or collected, this would allow a far greater set of analyses that could be of use to the NHS more widely. The NHS staff survey itself could assist this process in a number of ways by collecting improved data on trust leadership, by asking more detailed questions about the ongoing support from line managers for staff (i.e. not just the annual appraisal), and by allowing identification of subsections of trusts, such as individual departments, localities or teams.

Finally, we would urge more longitudinal data to be collected for individual staff members because this way a far more sensitive analysis could be conducted. Although this would not allow examination of all of the outcomes, for some (e.g. absenteeism, turnover, patient satisfaction) careful design would allow linkages to be drawn, particularly if data were collected more frequently than once a year, which shed far greater light on the causal mechanisms behind the data.

Conclusions

Overall, this research using sophisticated analytical methods (including extensive use of longitudinal data) gives a mixture of clear answers and further questions. Clear answers include those that suggest staff experience is clearly linked to outcomes, especially intermediate outcomes such as absenteeism. Building on previous research, this has shown that negative experiences such as discrimination, violence and harassment are most detrimental to outcomes, while staff engagement and the design of jobs so that, for example, staff feel they are clearly able to make a difference to patients, are most beneficial. These links also apply clearly to patient satisfaction as a cross-sectional outcome, although less clearly to other organisational performance measurements. However, although there is some clear evidence for causal links between staff experiences and absenteeism, other causal relationships are much more equivocal and in many cases it is not possible to say whether or not there is a causal relationship in either direction.

Given that there are relatively few empirical studies in the NHS, and we have demonstrated that it is not sensible to transfer findings from other contexts or countries, this represents a significant advance on our knowledge about how staff management and experiences play an important role in health care.

Contributions of authors

The authors are listed in the order of principal investigator (Martin Powell), coinvestigator (Jeremy Dawson), research fellow (Anna Topakas) and Health Services Management Centre associates (Joan Durose and Chris Fewtrell). The contribution of each author is as follows:

Martin Powell was the principal investigator and he drafted Chapters 1, 2 and 4.

Jeremy Dawson was the coinvestigator. He was responsible to overseeing the quantitative analysis. He drafted Chapters 5, 6, 7, 8 and 10.

Anna Topakas was the research fellow and she carried out the systematic review for health care and the quantitative analysis, and drafted Chapter 3.

Joan Durose is an associate at Health Services Management Centre and was jointly responsible for the ALSs. She drafted Chapter 9.

Chris Fewtrell is an associate at Health Services Management Centre and was jointly responsible for the ALSs.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health.

Dependent variable: impact of health and well-being on ability to perform work or daily activities

Dependent variable: suffering work-related stress in previous 12 months

Dependent variable: job satisfaction

Dependent variable: presenteeism (feeling pressure to attend work when feeling unwell)

Dependent variable: intention to leave job

Dependent variable: advocacy (recommend trust as a place to work or receive treatment)

NHS staff survey variables as predictors of absenteeism