A multifaceted intervention to reduce antimicrobial prescribing in care homes: a non-randomised feasibility study and process evaluation

Article history

The research reported in this issue of the journal was funded by the HS&DR programme or one of its preceding programmes as project number 13/97/12. The contractual start date was in April 2016. The final report began editorial review in May 2018 and was accepted for publication in November 2018. 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

Carmel Hughes is a member of the Health Services and Delivery Research Commissioned Panel. Mark Loeb has worked for the World Health Organization as a consultant to develop antibiotics for an essential list of medicines and algorithms for appropriate antibiotic use. Martin Underwood is a member of National Institute for Health Research (NIHR) Journals Library Editors Group. He was chairperson of the National Institute for Health and Care Excellence accreditation advisory committee from 2013 until March 2017, for which he received a fee. He is chief investigator or co-investigator on multiple previous and current research grants from NIHR and Arthritis Research UK and is a co-investigator on grants funded by Arthritis Australia, Australian National Health and the Medical Research Council. He has received travel expenses for speaking at conferences from the professional organisations hosting the conferences. He is a director and shareholder of Clinvivo Ltd (Tenterden, UK), which provides electronic data collection for health services research. He is part of an academic partnership with Serco Ltd (Hook, UK) related to return-to-work initiatives. He is an editor of the NIHR journal series, for which he receives a fee. He has accepted an honorarium for advice on Research Excellence Framework submission from Queen Mary University of London. He is co-investigator on an Efficacy and Mechanism Evaluation grant, receiving support in kind from Orthospace Ltd (Caesarea, Israel).

Copyright statement

© Queen’s Printer and Controller of HMSO 2020. This work was produced by Hughes et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. 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.

2020 Queen’s Printer and Controller of HMSO

Introduction

Care homes (with or without nursing) provide care for older people who can no longer live independently. The most frequent acute health-care intervention that care home residents receive is the prescribing of medications. 1 There are serious concerns about the quality of prescribing for care home residents generally, and in particular the prescribing of antimicrobials (antibiotics, antifungals and antivirals). 1 Care homes have been identified as important ‘reservoirs’ of antimicrobial resistance (AMR). 2,3 This has important implications for individual residents, and has broader public health implications due to the development of widespread AMR. A number of prescribing decisions (not just antimicrobials) for care home residents may be made over the telephone,4 and this can lead to medicines-management problems, with erratic medication reviews and prescribing errors.

We have previously shown that Northern Ireland (NI) nursing homes have the highest levels of, and greatest variation in, antimicrobial prescribing among facilities in 20 other European countries/jurisdictions. 5 England was ranked fourth in terms of overall prescribing. 5 Similar findings were reported for residential homes (those facilities that are not required to have qualified nursing staff). 6 Indeed, antimicrobial prescribing in care homes is seen as a global problem, contributing to increasing resistance. 1 There are several policy-level reports that have highlighted this issue, such as Infections and the Rise of Antimicrobial Resistance from the Chief Medical Officer (CMO) in England. 7 The ageing population and increasing requirements for high-quality long-term care are important considerations for the NHS8 and were recognised in the CMO’s report on AMR. 7 This report highlighted ‘the older adult’, with an acknowledgement of this population’s greater vulnerability to infection, which can be exacerbated by living with other older people with risk factors for infection in care homes. The report stated that infections could be managed better, with the appropriate prescribing of antimicrobials being highlighted as an aspect of health care that needed to be tackled in the context of AMR. This theme was echoed in the UK Five Year Antimicrobial Strategy 2013–20189 and in the earlier NI Strategy for Tackling Antimicrobial Resistance (STAR) 2012–2017. 10 These reports emphasised the importance of better stewardship of antimicrobials, which encompasses optimising therapy for individual patients, prevention of overuse, misuse and abuse of antimicrobials and the subsequent minimisation of resistance at both patient and community levels. Education of the health-care workforce was seen as an essential element to draw attention to AMR and appropriate antimicrobial stewardship. 7,9,10

The CMO’s report7 indicated that ‘there is a need to take an international view of this problem [AMR] and work with other nations’ to tackle it. Indeed, there have been a number of international commentaries and studies focusing on antimicrobial prescribing in the care home setting, particularly from North America. Morrill et al. 11 described AMR as a national security threat to the USA and issued a ‘call to action’ to address antimicrobial stewardship in care homes. Thompson et al. 12 noted a high prevalence of antimicrobial prescribing in care homes with nursing in four states in the USA and that there was little documentation to support such prescribing. Scales et al. 13 reported that nurse leaders within care homes and associated medical staff could act as champions for antibiotic stewardship, and Kistler et al. 14 noted that health-care providers may need more support to guide the decision-making process to reduce the excessive use of antimicrobials. Crnich et al. 15 reported that structured assessment, communication between care home staff and prescribers and education about AMR were important facets of interventions that may be effective in supporting antimicrobial stewardship.

Therefore, a more ‘whole-systems’ approach, involving education, diagnosis, treatment and feedback, may help improve practice more broadly. One such study that incorporated a number of these components was conducted by Loeb et al. 16 and focused on prescribing for urinary tract infections (UTIs). In this study, 12 nursing homes in Ontario (Canada) and Idaho (USA) were randomised to receive a multifaceted intervention (based on education and the use of a structured approach to the management of infections) and 12 care homes were allocated to usual care. The intervention consisted of the application of diagnostic (signs and symptoms) and treatment algorithms for UTIs at nursing home level, supported by small group educational interactive sessions for staff, videotapes, written material, outreach visits and face-to-face sessions with physicians. Findings indicated that fewer courses of antimicrobials were prescribed for suspected UTIs in the intervention care homes than in the usual care care homes. No significant differences were found between the intervention and control sites in terms of the total numbers of antimicrobials, admissions to hospitals and mortality.

Rationale for the research

We ran a short feasibility study, based on the original Canadian study,16 in two nursing homes in NI, using some of the same intervention components,17 such as interactive sessions, written material, outreach visits to care homes and educational sessions with general practitioners (GPs), along with the use of algorithms. The intervention was well received by staff and GPs and provided confidence that we could extend this approach on a greater scale. However, this feasibility work was conducted in only two care homes in one region of the UK and focused on one infection (i.e. UTIs), and therefore the relevance of our findings was extremely limited. To extend this research, we considered that it was important to undertake a more comprehensive piece of work that would adapt the intervention that was originally developed for the Canadian care home setting for use in two UK geographic regions in a non-randomised feasibility study, extending the focus to infections common in care homes (including respiratory and skin infections). Importantly, it was also considered essential to address aspects of intervention adaptation, implementation and acceptability by undertaking a process evaluation, which would overarch the conduct of the feasibility study.

At the time of writing this report, a search of trial registries revealed no ongoing intervention studies of this specific topic, although we are aware of one US-based study that is currently recruiting to a trial that will seek to reduce antimicrobial use in nursing home residents with Alzheimer’s disease and other dementias and that will focus on UTIs and lower respiratory tract infections. 18 This study is quite specific in terms of the target population and target infections. The study described in this report is broader in its scope and context (i.e. including care homes with and without nursing care) and is more relevant to the UK setting.

Aims and objectives

Our aim was to evaluate the feasibility and acceptability of a multifaceted intervention on rational prescribing for infections in a non-randomised feasibility study in care homes. The intervention consisted of an educational and management approach, supported by discussion of residents cases. The objectives of the study were as follows:

• to recruit six care homes – three in NI and three in the West Midlands, England

• to adapt and develop an intervention (a decision-making algorithm and small group interactive training) originally developed and implemented in Canadian care homes

• to deliver training in respect of the intervention in the care homes and associated general practices

• to implement the intervention in the six feasibility care homes

• to undertake a process evaluation of the non-randomised feasibility phase and test data-collection procedures.

The outcomes that we were interested in for this feasibility study were primarily related to the process evaluation; these included the acceptability of the intervention in terms of recruitment and delivery of training, the feasibility of data collection from a variety of sources, the feasibility of measuring appropriateness of prescribing and a comprehensive overview of the implementation of the intervention. Patient and public involvement (PPI) was embedded in all aspects of the study through our care home representative on the research team and the role of participating care home staff and family members (see Patient and public involvement).

Structure of the report

• Chapter 1 describes the background to and rationale for the study.

• Chapter 2 gives an overview of the research approach and methods, providing detail about the study design, data collection and analysis.

• Chapter 3 focuses on the recruitment of care homes.

• Chapter 4 describes the adaptation of the intervention, covering the adaptation of the decision-making algorithm and of the small group interactive training materials.

• Chapter 5 focuses on the implementation of the intervention and the collection and analysis of the associated data.

• Chapter 6 details the process evaluation that was conducted over the course of the study.

• Chapter 7 focuses on a survey that was posted to care homes in NI and the West Midlands to assess interest in participating in a larger study.

• Chapter 8 discusses the key findings from the study, its strengths and limitations, proposals for future research and final conclusions.

Study organisation and oversight

Patient and public involvement

For this feasibility study, we convened an Advisory Group to provide PPI perspectives as well as to contribute to the study design and development. This group was convened with the assistance of the Independent Health & Care Providers, a member organisation that consists of those providing care to vulnerable and older adults. The group was composed of staff and resident family members and it contributed to the drafting of information sheets and related documentation prior to the start of the study. As part of our research team, we had Mr Robert (Bob) Stafford, who is Head of Care and Compliance at Orchard Care Homes. Mr Stafford has responsibility for care compliance across the organisation, which consists of over 100 care homes across the UK. As someone who has direct experience of managing and overseeing care homes, his perspective has been invaluable. He actively participated in all Feasibility Study Management Group meetings and advised on implementation and troubleshooting as and when required. In addition, care home staff contributed to aspects of study conduct, particularly in respect of refinement of data-collection forms, insight into the research process and how the study affected workload.

Ethics considerations and approval

We considered the potential ethics issues for this study very carefully and took advice from a number of organisations. We were advised that the data required for the proposed primary outcome (drug dispensing data) could be obtained without requiring individual resident consent as the data would be available at care home level from community pharmacies and we would not be able to link this back to individual residents. We also needed to collect data from care homes in respect of the resident population and hospitalisations and mortality. In this case, data were extracted, anonymised and/or aggregated by the direct care team (care home staff). We consulted with the Health Research Authority, the Office of Research Ethics Committees Northern Ireland and the Privacy Advisory Committee in NI, which advised that our general approach was acceptable.

The research team have considerable experience of carrying out research within care homes. The team is very aware that a care home is the ‘home’ for each and every resident within it and that care homes are also complex workplaces for the staff. The team liaised closely with the care homes’ managers to ensure minimum disruption to the day-to-day running of the care homes. When possible, researcher visits to the care homes were pre-arranged and the researchers had appropriate training and approvals. Researcher visits were an important part of this study, during which the researcher acted as an ‘observer’. In a setting such as this, non-participant observation is almost impossible as residents and staff may want to interact. The researchers were respectful of residents’ wishes and space and remained in public areas of the care home.

The interviews undertaken with the various stakeholders were held at a time and place to suit participants. In this study, all interviews and focus groups with care home staff took place in the care homes and all interviews with GPs took place in GP surgeries.

The REACH (REduce Antimicrobial prescribing in Care Homes) study was reviewed by Office of Research Ethics Committees Northern Ireland Health and Social Care Research Ethics Committee B and given a favourable opinion (reference 16/NI/0003).

Summary

Care homes for older people are viewed as reservoirs of AMR. Previous research has shown high levels of antimicrobial prescribing in care homes in NI and England. Several national policies have advocated for better antimicrobial stewardship and looking to international examples of how best to manage this issue. A more ‘whole-systems’ approach, involving education, diagnosis, treatment and feedback, may help improve practice more broadly. This informed the conduct of the feasibility study that is described in this report, and which was based on Canadian research that demonstrated that the use of a decision-making algorithm and training sessions with staff led to fewer courses of antimicrobials being prescribed for suspected UTIs. Our study was set up to develop and adapt the Canadian approach to encompass respiratory and soft tissue infections, and addressed aspects of implementation and acceptability in a mixed-methods process evaluation. The study was managed by members of the research team, it had external oversight through a SSC, it had PPI input and all necessary approvals were in place prior to the start of the research.

Introduction

This chapter provides an overview of the research approach and methods, describing the study design, theoretical underpinning, and data collection, management and analysis.

Overview of methods

The REACH study is a non-randomised feasibility study that employed a mixed-methods design. A feasibility study is a piece of research that assesses whether a future main study can be carried out. 19 As a feasibility study, our focus was on the facilitators of or obstacles to implementation of our intervention and to test and adapt our processes to inform a possible larger study. Therefore, our methods and theories were chosen to reflect the focus on our ability to implement the intervention. This study was grounded within the Medical Research Council’s framework for the development of complex interventions. 20

In this section, we present an overview of the methods used within this study and the theoretical frameworks that underpin these. The study consisted of a recruitment phase, adaptation phase, implementation phase and process evaluation, as outlined in Figure 1. PPI was embedded throughout the study via the contributions of Mr Bob Stafford, who was a member of our research team.

FIGURE 1.

Overview of the REACH study.

Data management

All data collected during the study were handled and stored according to relevant legislation and standard operating procedures utilised by Queen’s University Belfast, the University of Warwick and the NICTU. Data were stored on secure servers and access to these data was restricted to authorised personnel. Any data transfer was in accordance with standard operating procedures and required data-sharing agreements to be in place. Study-related documents were made available for internal monitoring and audit activities; this was highlighted in participant information sheets. Study documentation and data will be archived for at least 5 years after completion of the study in accordance with the standard operating procedures of Queen’s University Belfast and the NICTU.

All CRF data returned to the NICTU were dealt with in accordance with its standard operating procedures and accessed only by authorised personnel. A member of the research team checked the data that were entered into a study-specific database designed by the NICTU. After all data had been entered into the database, the original CRFs were securely stored in archiving facilities.

Safety and adverse event management

An adverse event is defined as any untoward medical occurrence in a participant that does not necessarily have a causal relationship with the research treatment/intervention. We did not expect any adverse events related to the intervention in this feasibility study. However, we did expect a large number of reports to be made by staff owing to the nature of this population, but did not attempt to monitor these in real time. Adverse event forms were reviewed when received. Any such events were dealt with in accordance with the NICTU standard operating procedures for safety reporting. We also collected data on hospitalisations and mortality as reported by the care homes through other methods of data collection and monitored these data very carefully. In the original Canadian trial on which this feasibility study is based,16 no differences in admissions to hospitals or mortality between the intervention and control arms were found.

Summary

We have provided a brief overview of the methodologies and theoretical underpinning related to this study. Specific methods for particular aspects of this study will be described in the following chapters.

Introduction

In this chapter, we report on the recruitment of care homes. We gave careful consideration to the number and type of care homes required for this feasibility study. Residential care homes are staffed 24 hours a day and provide meals and help with personal care (activities such as washing, dressing and going to the toilet). Nursing homes also provide personal care, but in addition provide specialised nursing care for those who are sick, injured or require regular monitoring, and are required to employ at least one qualified nurse for 24 hours a day. Residential and nursing homes in the UK are collectively known as care homes. We planned to recruit both nursing and residential facilities to this study.

Aim and objectives

Our aim in this phase of the study was to recruit six care homes for participation in this study: three in NI and three in the West Midlands, England.

Methods

The sample size was informed by the research team’s previous experience in care home studies regarding what was considered acceptable for a feasibility study, regarding what would provide the type and quality of data required and to allow us to understand the process and implementation challenges. 4,28,29 We planned to recruit a purposive sample of six care homes: three in NI and three in the West Midlands, with two nursing homes and one residential home in each area. This was to reflect the broad breakdown of nursing home versus residential home numbers in the overall care home sector. The basic inclusion criteria were homes:

• with/without nursing care, providing 24-hour care for residents aged ≥ 65 years

• with a minimum of 20 (permanent) residents

• associated with a small number of general practices (up to four per home providing care for a minimum of 80% of residents within a home)

• with an exclusive arrangement with one pharmacy for dispensing medications (we also required that the pharmacy used specific dispensing software).

The recruitment process was conducted during April to June 2016. Our recruitment method differed slightly in NI and the West Midlands owing to particular contextual features of each research site. In NI, we compiled a list of care homes from the website of the Regulation and Quality Improvement Authority, the independent body responsible for monitoring and inspecting the availability and quality of health and social care services in NI. We then applied the following inclusion criteria to the Regulation and Quality Improvement Authority list to include care homes:

• within a 20-mile radius of Belfast (to ensure proximity to Queen’s University Belfast)

• with ≥ 20 permanent residents

• providing care for residents aged ≥ 65 years regardless of disability

• that were not dual registered (i.e. providing nursing and residential services)

• not owned by health trusts

• that were members of the Independent Health and Care Providers and any home identified as interested in research.

In the West Midlands, many care homes are part of the NIHR Clinical Research Network (CRN) Enabling Research In Care Homes (ENRICH) programme. 30 The ENRICH programme aims to bring together care home staff, residents and researchers in order to facilitate the design and delivery of research. We asked the CRN to distribute a flyer to these homes, which sought their interest in participating in the REACH study based on the inclusion criteria for the study.

This process identified refined samples of homes in NI and the West Midlands. The research fellows at Queen’s University Belfast and the University of Warwick then randomised their respective lists of nursing and residential homes and, beginning with number 1 on each randomised list, contacted the manager to confirm the initial inclusion criteria at each site. In NI, the survey also applied additional inclusion criteria to identify those homes with four or fewer general practices providing care for a minimum of 80% of residents and with an exclusive arrangement with one pharmacy for dispensing medications. For those homes in each area that met these initial inclusion criteria, we asked the home manager if they were interested in receiving information about how their home could participate in the REACH study.

For those homes that were willing and eligible to participate in the study, we asked the managers to provide contact details of the pharmacy that dispensed medications to their homes. We then contacted these pharmacies to ascertain the type of dispensing software used to ensure that the anonymised dispensing data we required for the study could be downloaded from these systems. We continued contacting care homes and pharmacies until we obtained the requisite number of homes that were fully eligible for the study.

We telephoned the managers of each care home that was eligible and willing to take part in our study, and made arrangements to visit on a mutually convenient date. At this visit, we verbally provided more detail about the study and answered any immediate queries that the managers had. We also provided written information, including an invitation letter, a participant information sheet and a consent form. The participant information sheet informed each home manager of the requirement to identify up to two members of staff (to account for different shifts within the homes) who could act as ‘REACH Champions’: individuals who would be responsible for delivering training to staff unable to attend the original REACH training session. Some managers agreed to take part in the study at this point and completed the consent form. Two homes in NI were part of a group of homes and so the managers of these homes advised that permission to take part in the study should be sought from the management group. The research fellow at Queen’s University Belfast contacted the management group and permission was obtained. The remaining homes were revisited within 1 week and written consent was obtained. The research fellows in each site maintained regular contact with the homes between taking consent and the start of the research processes.

Results

Following screening using the initial sampling criteria, 49 nursing homes and 16 residential homes in NI and seven nursing homes and eight residential homes in the West Midlands were identified (Table 1). This list of homes was randomised, from which the final sample for the feasibility study was recruited, is shown in Table 1.

We recruited three homes (two nursing and one residential) in both NI and the West Midlands (Table 2).

The number of beds ranged from 32 to 62, with occupancy at almost 100% in all homes apart from home A. In NI, more general practices provided care to the homes, whereas in the English homes, each participating home was served by one practice. Homes varied in ownership, with three being part of a chain and the remaining three being owned by single proprietors. All homes were each associated with one pharmacy that supplied all medication.

Summary

We successfully recruited six homes that met the inclusion criteria for participation in the study. The approach taken in the two geographic sites was largely comparable, but also took into account some contextual differences, such as the role of ENRICH in England. We noted the key characteristics pertaining to homes, such as number of beds and number of general practices associated with the homes.

The six recruited homes then progressed to the adaptation phase of the study, which is described in Chapter 4.

Introduction

In this chapter, we report on the adaptation of the intervention to reduce antimicrobial prescribing. The Canadian study16 had provided ‘proof of concept’ that antimicrobial prescribing could be influenced by a multifaceted intervention consisting of a decision-making algorithm and small group interactive training. However, context in research is important;31 in this case, the difference is between the Canadian care home context and that of the UK. Transposing the intervention from Canada to the UK without any modification was unlikely to be successful. Furthermore, we anticipated that the evidence on management of infections in older people would have developed since the Canadian study was undertaken (the last follow-up was in 2003). 16

Aims and objectives

The aim of this phase of the research was to adapt and develop an intervention (a decision-making algorithm and small group interactive training) originally developed and implemented in Canadian care homes. The objectives were as follows:

• to undertake a series of rapid reviews encompassing the most up-to-date literature on the management of the three target infections

• to conduct a consensus meeting, using the nominal group technique, with health-care professionals, to adapt the decision-making algorithm

• to undertake focus groups and interviews with key stakeholders, including care home staff, family members of residents and GPs, to adapt the decision-making algorithm

• to develop a training programme outlining the use of the decision-making algorithm, communication techniques between health-care professionals and aspects of the study process

• to produce an updated and refined decision-making algorithm, which, in conjunction with the training programme, would be used in the implementation phase of the study.

We report on the adaptation of the decision-making algorithm and follow this with a description of the adaptation of the small group interactive training programme.

Methods

Adaptation of the decision-making algorithm

The approaches developed by Loeb et al. in 200132 and 200516 were updated and adapted for UK use through a series of iterative steps: production of a rapid scoping literature review, focus groups with care home staff and family members of residents and semistructured interviews with GPs. As part of discussions with other members of the Feasibility Study Management Group, it was also agreed that a consensus group meeting with health-care professionals would provide further additional input into the development and adaptation of the intervention. The process was also informed by continual iterative internal review and analysis within the research team.

Production of rapid reviews

We undertook a rapid scoping review of the literature to obtain the most up-to-date evidence for the management and diagnosis of UTIs, respiratory tract infections (RTIs) and skin and soft tissue infections (SSTIs) in older people living in care homes. We considered systematic reviews, guidelines, reports, review articles and randomised controlled clinical trials published between 2000 and 2016.

Search methods for identification of studies

We searched the following electronic databases for primary studies: the Cochrane Central Register of Controlled Trials (via The Cochrane Library, 2016, issue 4), MEDLINE (via Ovid, 1946 to May week 1 2016), EMBASE (via Ovid, 1980 to week 41 2016), CINAHL plus EBSCOhost (1980 to May 2016), PubMed (1996 to May 2016) and SCOPUS (1983 to May 2016). We supplemented this with forward citation tracking. In addition, we contacted experts in the field of antimicrobial and geriatric medicine for advice on further potential studies. We conducted a ‘grey’ literature search of NICE, the European Centre for Disease Protection and Control, the Infectious Disease Society of America, the Society for Healthcare Epidemiology of America and NHS Evidence, using the appropriate terminology as applicable. This search focused on identifying guidelines relating to the management, stewardship, initiation, diagnosis and treatment of infections in the care home setting (see Appendix 1).

Screening and review of literature

We downloaded all titles and abstracts retrieved by electronic searching to the reference management database RefWorks (Pro Quest, LLC, Ann Arbor, MI, USA) and removed duplicates. Two review authors (AC and CS) independently examined the remaining references. A reference was included if it was:

• relevant to the key question ‘what is the most up-to-date evidence for the management of urinary tract, respiratory tract and skin infections and the stewardship of antimicrobials?’

• a review, guideline or report from a professional organisation or clinical trial

• published between 2000 and 2016

• related to the care home setting

• written in English.

We excluded those articles that clearly did not address the management and diagnosis of the key infections in the care home setting and obtained copies of the full text of potentially relevant references. At least two review authors (CS and AC) independently assessed the eligibility of the full-text papers.

Data extraction and management

Three reviewers (AC, DE and CS) independently extracted data from relevant publications using a predefined extraction table containing the following elements: author and year, setting (i.e. nursing or residential care home), type of publication [i.e. review article, randomised controlled trial (RCT), NICE guideline], population (i.e. older people), infection (i.e. urinary, respiratory or skin/soft tissue infection), objectives, relevant data relating to the target infections from each article and comments (from the reviewers AC, DE and CS), such as whether or not the paper was particularly useful or if it cited the original algorithm by Loeb et al. 16 An article was thought to be particularly useful if it informed the reviewers about new evidence relating to urinary, respiratory or skin/soft tissue infection. This process took place during a 3-day face-to-face meeting between Anne Campbell, David Ellard and Catherine Shaw to discuss and agree discrepancies in their interpretation; those discrepancies that could not be agreed were resolved by an additional author (CH).

An additional meeting was held between Catherine Shaw, Carmel Hughes and Michael Tunney to further discuss the findings from the evidence tables. This involved individual examination of the extracted data alongside group discussion, with the aim of reaching a final conclusion as to which articles should be used to update the algorithm.

The data extracted during this process were used to update the algorithm for the consensus meeting (see the following section) and informed ongoing iterative discussions within the research team during the process of adaptation.

Consensus group meeting

Ethics approval was granted by the School of Pharmacy, Queen’s University Belfast Ethics Committee, to conduct a consensus group meeting using the nominal group technique (reference 022PMY2017). Potential participants were identified through personal networks of members of the research team in Belfast, and were approached by e-mail and given brief details about the consensus group meeting. Participation was voluntary and required written informed consent. All participants received an honorarium for their time, as noted in the relevant information sheet.

Prior to the meeting, each participant was provided with a draft version of the decision-making algorithm as shown in Figure 2, alongside supporting documents (comprising key papers that provided the research team with updated evidence in relation to urinary and respiratory infections in care homes that had been retrieved from the literature search) to allow them to familiarise themselves with the material. No new evidence was found relating to SSTIs (see Adaptation of the decision-making algorithm).

FIGURE 2.

Draft algorithm presented to the consensus group. b.p.m., breaths per minute; COPD, chronic obstructive pulmonary disease.

Nominal group technique

The consensus group meeting format was based around the nominal group technique. 33 Potential benefits of nominal group technique include significant idea generation that may happen through face-to-face discussion and debate, even though it may be in a limited and prestructured format. 34 The research team formulated three nominal questions: (1) how do we best manage UTIs in older people?, (2) how do we best manage RTIs in older people? and (3) how do we best manage SSTIs in older people? These questions were presented to the participants during the face-to-face meeting (Figure 3). Initially, each participant recorded his or her views regarding the nominal questions independently and privately.

FIGURE 3.

The REACH consensus group nominal group technique. Reproduced from Hughes et al. 35 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Reproduced from Hughes et al. 35 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

The group members were asked to write down the five most common symptoms they associated with UTIs, RTIs and SSTIs in the older population, and to study the algorithm (see Figure 3, session 1) and supporting evidence. This allowed the participants to refamiliarise themselves with the material. They were then encouraged to share the five most common symptoms they associated with each infection with the rest of the group (see Figure 3, session 2). The facilitator noted these symptoms and any comments on a whiteboard for each member to see. The next step involved group discussion of the symptoms and comments made, and members of the group had the opportunity to seek further explanation from one another if required (see Figure 3, session 3). The symptoms included within the algorithm were discussed. Participants were then asked to rank the top five symptoms of each infection, derived from their personal experience (most common symptoms listed previously) and those included within the algorithm (from the literature), based on their clinical expertise. Participants were provided with a separate table for each infection to guide the ranking of the symptoms and were asked to complete the table by selecting a numerical rating for each one, with the most important receiving a rank of 5 and the least important receiving a rank of 1 (see Figure 3, session 4).

When each participant had ranked their top five symptoms of each infection, the data were entered onto a Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) spreadsheet by the facilitator. The ratings for each symptom were summed and item totals were ranked. The facilitator compared the ranked symptoms with those included in the algorithm as presented. If there were any discrepancies between the top-ranked symptoms and those included within the algorithm, the algorithm was amended accordingly and presented back to the group. Further discussion ensued (see Figure 3, session 5) and this was followed by another round of voting and ranking of the symptoms to be included in the algorithm (see Figure 3, session 6), which was amended accordingly again.

Pre-implementation focus groups and interviews

Sampling and recruitment

We aimed to recruit care home staff as users of the intervention and family members of residents, given that family members may be influential in decision-making in relation to the prescribing of antimicrobials. 4 To recruit care home staff and family members of residents prior to the implementation of the intervention, the research fellows at Queen’s University Belfast and the University of Warwick asked the manager in each participating care home for assistance. They provided managers with individual participant information packs for care home staff and family members. Each participant information pack included the following documents: an invitation letter, a participant information sheet and a consent form. The research fellows asked the managers to distribute this information to senior and junior care home staff and family members. The research fellows were available in person, by telephone or by e-mail to provide further explanation. We aimed to recruit 4–12 participants for each focus group36 and to include senior and junior staff in the care home focus groups. The managers of the care homes were not invited to take part in the focus groups so as to prevent the management relationship directing or constraining the group discussion. 36 To recruit GPs, the research fellows approached the practice manager in each general practice associated with the participating care homes. They provided the practice manager with a verbal overview of the study and asked that an invitation letter, information sheet and consent form be sent to a named GP. A follow-up telephone call was made to the GP. All focus group and interview participants received an honorarium for their time, as noted in the relevant information sheet. Participation was voluntary and written consent was obtained from all participants.

Interview and focus group discussion guides

Even though this stage of the study focused on developing and adapting the intervention and did not involve its implementation, we considered the first three constructs of normalization process theory – coherence, cognitive participation and collective action – as these were useful in alerting us to factors that may have an impact on its implementation; hence these constructs were used in the development of topic guides. The fourth construct, reflective appraisal, was not considered useful at this stage given that the intervention was not yet implemented. We used these three constructs to create the following broad working definitions so that normalization process theory could be usefully applied within our study:23–25

1. Making sense (coherence) – how do participants understand the issue of AMR and what is their usual practice?

2. Engagement and commitment (cognitive participation) – what do participants see as necessary to engage staff in the new practice?

3. Facilitating the use of the intervention (collective action) – how do participants envisage the intervention working and what are the factors that may facilitate or inhibit its use?

These three constructs of normalization process theory were used to shape the questions within the discussion guides for focus groups (see Appendix 2 for care home staff and Appendix 3 for family members) and individual interviews (see Appendix 4) that were devised and agreed by members of the research team. For example, both guides included questions about usual practice when a resident was suspected to have an infection.

Conduct of the interviews and focus groups

At the beginning of a focus group or interview, the facilitator assured participants regarding confidentiality, gave a brief background to the study and offered participants an opportunity to ask any questions. We then used questions from the discussion guides, with appropriate prompts, to ask participants about their usual practice when a resident was suspected to have an infection. The participants in the care home staff and family member focus groups were also asked about their understanding of AMR. Next, participants were shown a version of the decision-making algorithm and, using the ‘think–pair–share’ approach,37 were asked to reflect on it for a few moments. We then asked the participants, using appropriate prompts, questions that included how easy the algorithm was to follow; what they perceived to be missing, not needed or confusing; and any concerns they had about the use of the algorithm. This process is described in Box 1.

BOX 1 - How the think–pair–share approach was used during focus groups and interviews

Focus groups and interviews

We distributed the decision-making algorithm to participants and asked them to think briefly about it on their own (1–2 minutes). We then asked them to imagine that they had a resident with a suspected infection and to consider how well they thought the decision-making algorithm would actually work in practice.

Focus groups

We asked participants to form groups of two or three people to discuss the algorithm (3–4 minutes) and to think about at least one question they would like to ask or one comment they would like to make about it. We then asked each small group of participants about their questions or comments.

Focus groups and interviews

We proceeded to use the questions and appropriate prompts from the discussion guides to stimulate discussion (25 minutes).

Data analysis

All interviews and focus group recordings were digitally recorded and transcribed verbatim by an external organisation. The transcripts were checked against the recording by the research fellow in each area to ensure accuracy and anonymisation. The transcribed data were uploaded into NVivo^® and repeatedly read to increase familiarity with the data. Data analysis was based on the constant comparison method. 27 A selection of focus group and interview transcripts were first open coded inductively, with codes created from the patterns and themes emerging from the data, and an initial coding frame was developed. This thematic coding frame was then applied to subsequent transcripts and iteratively refined as new codes were defined (see Appendix 5 for care home staff and family member interviews and Appendix 6 for GP interviews). We used the framework matrix facility within NVivo^® to assist the analytic process. These matrices enabled the research fellows to summarise the text associated with a theme in order to develop a narrative for each of them. These themes were then structured in accordance with key aspects of the decision-making algorithm. A COREQ (consolidated criteria for reporting qualitative research) checklist was completed for this aspect of the study and can be found in Appendix 7.

Internal review by the Feasibility Study Management Group

In addition to the activities outlined in the previous section, the algorithm was also reviewed and refined by the research team through an iterative process. Via the monthly Feasibility Study Management Group meetings that involved all members of the research team, aspects of each draft of the algorithm were discussed and debated. Changes were made by the research team based on results from the literature review, consensus meeting, focus groups and interviews. The final algorithm was agreed on by all members of the research team.

Results

Adaptation of the decision-making algorithm

Production of rapid reviews

Our searches retrieved 1905 articles. Just under 1500 articles (1487) remained after removal of duplicates, with the addition of another 24 records identified from other sources (hand-searching of other papers), of which 118 were carried through to the next stage. References were included if they met the pre-established criteria outlined in Introduction. The main reason for exclusion of articles at this point was that they did not relate to the older population or the care home setting. Following the 3-day meeting, 38 articles and two guidelines were included and data from these documents were extracted into evidence tables. The main reason for exclusion at this stage was that the article did not specifically relate to the management of urinary, respiratory or skin infection in care home residents. Following a further review of the evidence tables by members of the research team (CS, CH and MT), a further 34 articles were excluded because they did not provide any updated evidence in relation to UTI, RTI or SSTI.

An overview of screening and assessment of all papers/resources is provided in Figure 4.

FIGURE 4.

The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram outlining the review process for identification of new evidence. Reproduced from Hughes et al. 35 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Reproduced from Hughes et al. 35 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

The extracted evidence from the remaining six papers is summarised in Table 4.

TABLE 4 - Data extracted from the articles contributing to updating of the algorithm

SIGN, Scottish Intercollegiate Guidelines Network.

Reproduced from Hughes et al. 35 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Article	Setting	Design	Population and condition	Objective	Relevant information for updating/refining algorithm	Group agreement for updating algorithm
Falcone et al.41	Community and hospital (includes care home setting)	Review	Older people, pneumonia (community-acquired pneumonia, health-care-associated pneumonia and hospital-acquired pneumonia)	This review sought to produce a summary of therapeutic recommendations on the basis of the most up-to-date clinical and pharmacological data	Signs and symptoms most commonly associated with pneumonia: cough, fever, chills, pleuritic chest pain. Extrapulmonary symptoms such as nausea, vomiting, alternation to sensory stimuli or diarrhoea may also be present. It is important to remember that pneumonia in older patients tends to occur more often with extrapulmonary manifestations. For example, the appearance of a delirium or acute confusion is found in approximately 45% of elderly patients with pneumonia	Agreed to add in extrapulmonary symptoms
Juthani-Mehta et al.42	Nursing home	Prospective observational cohort study	Older people, UTI	To identify, among non-catheterised nursing home residents with clinically suspected UTI, clinical features associated with bacteriuria plus pyuria	The most commonly reported clinical features for suspected UTI in this cohort were change in mental status (39%), change in behaviour (19%), change in character of the urine (i.e. gross haematuria and change in the colour or odour of urine, 15.5%), fever or chills (12.8%) and change in gait or a fall (8.8%). Dysuria, change in character of urine and change in mental status were significantly associated with the combined outcome of bacteriuria plus pyuria. Absence of these clinical features identified residents at low risk of having bacteriuria plus pyuria (25%), and presence of dysuria plus one or both of the other clinical features identified residents at high risk of having bacteriuria plus pyuria (63%)	Change in character of urine (i.e. gross haematuria and change in the colour or odour of urine) was considered but not supported by more recent guidelines
SIGN 8843	All settings	Guideline	Older people, UTI	To provide guidance in the diagnosis and management of suspected UTI in older people	Algorithm to be used in people/residents with fever defined as a temperature of > 37.9 °C or > 1.5 °C above baseline First stage of algorithm used to differentiate between UTI, RTI, SSTI and gastrointestinal infection Advises against using dipstick test in diagnosis of infection. Provides supportive care advice	Agreed to add supportive care advice to algorithm
Stone et al.44 – updated McGeer et al.45	Long-term care	Position paper	Older people, infection (general)	To update the 1991 McGeer criteria (infection surveillance definitions for long-term care facilities) using an evidence-based structured review of the literature in addition to consensus opinions from industry leaders including infectious diseases physicians and epidemiologists, infection control specialists, geriatricians, and public health officials	Acute swelling of the testes, epididymis and prostate should be included in surveillance definitions for UTIs as these symptoms are a common complication of UTI in both catheterised and non-catheterised males	Agreed to add acute swelling of testes, epididymis and prostate
D’Agata et al.46	Nursing home	Prospective study	Older people, UTI	To describe the presentation of suspected UTI in nursing home residents with advanced dementia and how they align with minimum criteria to justify antimicrobial treatment	In long-term care residents with dementia, the most common reason for suspected UTI was a change in mental status (44.3%)	Agreed to subdivide the UTI element of the algorithm into two sections to account for two populations within care homes – those with and without dementia; changed in later iterations
Rowe47	Nursing home	Review	Older people, UTI	This review sought to provide an overview of the prevalence, diagnosis and diagnostic challenges, management and prevention of UTI and asymptomatic bacteriuria in older adults	The most commonly reported clinical features for suspected UTI in this cohort were change in mental status (39%), change in behaviour (19%), change in character of the urine (i.e. gross haematuria and change in the colour or odour of urine, 15.5%), fever or chills (12.8%) and change in gait or a fall (8.8%) (Juthani-Mehta et al.42)	Change in character of urine (i.e. gross haematuria and change in the colour or odour of urine) was considered but not supported by UK SIGN guidelines

Following the data-extraction process and analysis, three separate algorithms were developed: one for each infection of interest (i.e. UTI, RTI and SSTI). The UTI algorithm was developed using the original algorithm by Loeb et al. 16 as a starting point, and updated in accordance with the new evidence identified during the literature review (see Table 4). Thus, the UTI algorithm was based on data from Loeb et al. ,16 Juthani-Mehta et al. ,42 Scottish Intercollegiate Guidelines Network (SIGN) 88 guidelines,43 Stone et al. ,44 and D’Agata et al. 46 The RTI algorithm was informed by the original study undertaken by Loeb et al. 32 (preliminary work leading up to the study by Loeb et al. 16) alongside research by Falcone et al. 41

Because no new evidence was identified from the literature regarding SSTI, it was agreed that the minimum criteria for initiation of antimicrobial therapy for suspected SSTI in long-term care facilities by Loeb et al. 32 should be used. Each decision-making algorithm is described in more detail in the following sections.

Changes made to the urinary tract infection decision-making algorithm

In the first instance, the updated UTI decision-making algorithm was redrafted by the research team into two distinct pathways to account for two populations within care homes: those with and without dementia. This was based on findings from the literature (D’Agata et al. 46) that previous studies did not consider cognitive decline and the effect that this may have on residents with dementia and a suspected UTI. These residents may not experience any specific urinary symptoms, but rather a more pronounced decline in cognition and function. However, following ongoing discussions within the research team, it was agreed that, for the sake of simplicity, only one pathway encompassing residents with or without dementia would be presented.

The updated algorithm was designed to be used in residents who had a raised temperature (> 37.9 °C or > 1.5 °C above baseline) as per Loeb et al. 32 and SIGN 88 guidelines. 43 Swelling/tenderness of testes, epididymis and prostate were additional symptoms to those reported by Loeb et al. ,32 following data published by Stone et al. 44

Changes made to the respiratory tract infection decision-making algorithm

The RTI decision-making algorithm largely mirrored that of Loeb et al. ,32 with the addition of extrapulmonary symptoms as per Falcone et al. 41 Nausea, vomiting, altered response to sensory stimuli and diarrhoea were included in the algorithm as a separate step, provided that these symptoms presented in combination with a cough and a raised temperature. 41

Changes made to the skin and soft tissue infection decision-making algorithm

No changes were made to the approach for SSTI management previously published by Loeb et al. 32 owing to the absence of new evidence.

A combined decision-making algorithm

At this point, the management group decided that it was more practical to combine all three algorithms into a single decision-making algorithm with three ‘paths’ – one for each infection. It was considered that this would be more acceptable to care home staff and was more likely to be used. A single algorithm, encompassing all three infections, updated from evidence extracted from the literature, was then presented to the consensus group (see Figure 2). In all three infections, a supportive care step was added as per the SIGN 88 guidelines. 43 This step instructs the user to continue to monitor the resident, to provide pain relief if necessary and to encourage fluids.

Consensus group meeting

Four participants were approached by e-mail and invited to take part in the study. Three agreed to join the group. The remaining invitee was unable to attend on the day of the consensus group, but suggested an alternative participant who was able to attend. The participants consisted of a hospital consultant (respiratory medicine) with experience of prescribing in older people, a GP, a geriatrician and an expert in microbiology. All had the necessary clinical expertise to comment on the findings from the literature review and the updating and adaptation of the algorithm. The consensus group meeting was convened on 12 September 2016 and was facilitated by a member of the research team (CS).

Following sessions 1 and 3, as outlined in Figure 3, the most common symptoms identified for the three target infections by the consensus group are summarised in Box 2.

BOX 2 - Consensus group results: the most common symptoms for target infections

UTI symptoms

• Urinary frequency.

• Urinary incontinence.

• Suprapubic pain.

• Pain when urinating.

• Blood in urine.

• Non-specific feeling of being unwell.

• Feeling ‘off’.

• Loin pain.

• Swelling of testes.

RTI symptoms

• Cough.

• Sputum.

• Breathlessness.

• Chest pain.

• Delirium (not otherwise delirious in normal circumstances).

• Pain with inspiration.

• Confusion.

• Shortness of breath.

• Rapidly breathing.

• Fast respiratory rate.

• Hypoactive delirium.

• Change in confusion status – more confused than normal.

SSTI symptoms

• Swelling.

• Erythema/redness.

• Heat.

• Pain.

• Drainage from area.

• Hot to touch.

• Change in colour of area affected.

• Pus.

The consensus group participants viewed the presented algorithm positively and felt that it had a place within the care home setting, particularly where there was nursing support. Their specific comments on each infection, and how each would be managed using the decision-making algorithm, are summarised in the following sections, reflecting discussions in session 3, as outlined in Figure 3.

Urinary tract infection

• There were concerns that the ‘two or more’ symptoms requirement suggested that some residents who needed to be seen by a GP would not have a referral. Strong views were expressed by one participant that a resident with one symptom, such as new or increased urinary frequency, would need to be seen.

• There was discussion regarding whether or not ‘anecdotal evidence’, such as change in character of urine (i.e. smell or colour), could be considered a valid indicator of infection.

• Confusion was discussed as a non-specific symptom of UTI. All participants felt that confusion should be included within the algorithm as a non-specific sign of any infection (UTI/RTI/SSTI), but it was not necessarily specific to a urinary infection. This is expanded on in the following section when considering RTIs.

Respiratory tract infection

• Concerns were expressed regarding the use of the algorithm in those with chronic respiratory disease. Such patients may, at any one time, display several symptoms included within the RTI section of Box 2, yet not have an infection.

• As with the UTI section of Box 2, concerns were expressed by one participant that the ‘two or more’ symptoms requirement meant that some residents would not see a GP.

• Confusion and delirium were emphasised by most participants as lesser known symptoms of a lower RTI.

• Chest pain/pain when breathing/pain on inspiration could be renamed dyspnoea.

Skin and soft tissue infection

• The presence of pus should be included as a standalone symptom.

Following discussion of the algorithm (see Figure 3, session 3), participants continued to debate other aspects of infection management in older care home residents. All participants stated that they would not rely on temperature alone when assessing an older person, particularly in a care home setting.

Most participants considered assessment of pulse/blood pressure/respiratory rate/colour of resident (e.g. pallor) to be more valuable indicators of infection. Rigors was considered to be a very serious presentation and would require medical intervention, and it was recommended that this should be acknowledged in the algorithm. There was extensive discussion regarding the management of infections in residents with specific conditions such as dementia, type 2 diabetes or chronic respiratory disease and whether the algorithm should be applied in such cases or if there should be a specific algorithm for those residents with dementia. However, it was concluded that excluding such residents would render the algorithm essentially redundant as these residents constituted the majority of a care home population. There was also discussion regarding the different types of staff within care homes (with and without nursing) and how non-nursing staff would use the algorithm. However, having a single algorithm was considered to be the most practical option, and feasibility of its application by a range of staff would be tested in the study.

Ranking of the symptoms (see Figure 3, session 4) by participants resulted in the findings that are summarised in Table 5.

TABLE 5 - Round 1 ranking of symptoms by consensus group participants

Infection	Symptoms
	Accepted (ranked score)	Rejected (ranked scores)
UTI	Urgency (19) Frequency (12) New or increased incontinence (10) Suprapubic pain (9) Dysuria (6) Change in smell or colour of urine (4)	Swelling of the testes (0) Loin pain (0) Blood in urine (0)
RTI	New or increased shortness of breath/difficulty breathing/rapid breathing/wheeze (19) Rapid breathing (10) Confusion or delirium/feeling confused/change in confusion status (10) New or worsening cough (7) Sputum (6) Respiratory rate > 25 breaths per minute (3) Dyspnoea (3)	None
SSTI	Redness (15) Tenderness/pain (12) Warmth (11) Swelling (11) Pus (as a standalone symptom) (11)	None

These findings informed the next iteration of the decision-making algorithm, which was presented to the participants for further comment (see Figure 3, session 5).

Urinary tract infection

It was agreed that symptoms such as urgency, frequency and incontinence should be prefixed by ‘new or worsening’ and that suprapubic pain should be replaced by lower abdominal pain.

Respiratory tract infection

One participant felt that ‘Respiratory rate > 25 (shortness of breath/difficulty breathing/rapid breathing)’ should be presented separately as ‘Respiratory rate > 25 and Shortness of breath/difficulty breathing/rapid breathing’ within the decision-making aid as respiratory rate is a standalone symptom within other algorithms such as those prepared by the British Thoracic Society (e.g. British Guideline on the Management of Asthma48). All participants were in agreement given this additional information.

It was suggested that the term ‘dyspnoea’ might not be understood by staff in residential homes, so it was agreed that this should be replaced by ‘difficulty breathing’.

Skin and soft tissue infection

It was agreed that all symptoms should be prefixed with ‘New or worsening’.

More generally, participants felt that a text box containing the following statement should be added to the overall algorithm:

Please note:

Residents with dementia may not show the usual signs of urinary, chest or skin infection at first. Their symptoms may be more general such as a change in behaviour or functional decline. This may be seen as hyperactivity such as aggression and agitation, or hypoactivity such as appearing withdrawn and loss of appetite.

There was also further discussion regarding the steps preceding the consideration of specific symptoms, and participants suggested the inclusion of observation of other vital signs or symptoms of concern (e.g. rigors) that would highlight the need for medical intervention.

The final round of voting and ranking of symptoms by participants (see Figure 3, session 6) resulted in the findings summarised in Table 6.

TABLE 6 - Round 2 ranking of symptoms by consensus group participants

Infection	Symptoms
	Accepted (ranked score)	Rejected (ranked scores)
UTI	New or increased urgency (19) New or increased frequency (12) New or increased incontinence (10) Lower abdominal pain (9) Dysuria (6) Change in smell or colour of urine (4)	None
RTI	New or increased shortness of breath (19) New or worsening cough (7) Sputum (6) Rapid breathing (10) Confusion or delirium/feeling confused/change in confusion status (10) Respiratory rate > 25 breaths per minute (3) Difficulty breathing (3)	Dyspnoea (0)
SSTI	New or increased redness (15) New or increased tenderness/pain (12) New or increased warmth (11) New or increased swelling (11) Pus (as a standalone symptom) (11)	None

These rankings informed the final stage of refinement following on from the consensus group meeting.

Pre-implementation focus groups and interviews

We conducted semistructured face-to-face focus group interviews with care home staff and relatives of residents in a total of six care homes in NI and the West Midlands during September to October 2016. There were 12 focus groups in total: six with care home staff and six with families of residents, with one care home staff and one family focus group in each of the care homes. We conducted semistructured one-to-one interviews with eight GPs during January to March 2017 (five in NI and three in the West Midlands). The interviews in NI were carried out by Anne Campbell and Catherine Shaw, and the interviews in England were carried out by Rachel Potter. All participants gave written informed consent. The numbers participating in each focus group and interview are shown in Table 7.

TABLE 7 - Number of participants in pre-implementation focus groups and interviews

Home	Number of participants
	Care home staff focus groups	Family focus groups	GP interviews
A	4	4	1
B	4	5	2
C	8	4	2
D	9	5	1
E	8	8	1
F	8	5	1
Total	41	28	8

The focus groups were held in a suitable room in the care homes and participants were provided with refreshments. The duration of these meetings ranged from 50 to 76 minutes with care home staff and from 46 to 71 minutes with family members of residents. The interviews with GPs were conducted in the practice room of the participating GP. The duration of these interviews ranged from 16 to 31 minutes. In the presentation of our findings, all quotations have been given an anonymised label (e.g. A–F represents each participating home) (see Chapter 6, Care homes, for a description of each home). ‘Staff, pre implementation’ indicates that the interview or focus group was conducted with staff pre implementation of the intervention. Other participant groups are family members (‘family’) and GPs.

Here we present the main findings identified from the pre-implementation focus groups with care home staff and relatives of residents and GP interviews that relate to, and informed, the adaptation of the algorithm. The themes that were identified from the analysis were structured in accordance with three key aspects of the decision-making algorithm: initial assessment of the resident, observation of the resident and action by care home staff.

Initial assessment of resident

Care home staff described many examples of new or worsening non-specific symptoms that they thought may indicate that a resident had an infection. Most related to the way in which care home staff observed change in behaviour, such as reduced mobility, increased confusion, agitation or aggression, poor appetite, lethargy, changes to fluid intake and output, not recognising their relative or just ‘not being right’:

Well they’d still be, you’d be going by their mobility, because they’d be off their feet, they’d be shaky, they’d be clammy confused, you know a lot of those symptoms. (. . .) We see a lot of changes in mobility or increased falls when somebody has an infection.

C: staff, pre implementation

They manifest like agitation, anxiety or different from what they are. For example, they are always pleasant to the staff and other patients and their families, they can be different when they come in and then the staff go ‘there’s something wrong, something’s not right’.

A: staff, pre implementation

Some care home participants also felt that the decision-making algorithm should distinguish changes in behaviour for those with and without dementia as many of the examples provided in the algorithm were already present in residents with dementia, therefore making it difficult to notice change. However, the study team felt that it would be difficult to provide a comprehensive list of all examples of change in behaviour suggested by staff as the decision-making algorithm would become illegible. We therefore decided that the examples already provided in the algorithm would remain unchanged, but additional suggestions from the participants would be included and discussed in the training session and accompanying study handbook.

Observation of the resident

Assessing the resident’s temperature

Some care home staff considered 37.9 °C too high a threshold and 4 hours too long before contacting the GP. They expressed concern that this may put the resident at risk and suggested a temperature range of 37.3–37.6 °C as being more appropriate to initiate GP contact. However, without exception, staff used temperature to alert them to a possible infection in a resident, in combination with other indicators such as behaviour change, prior knowledge of the resident and vital signs, and sought to reduce temperature with supportive care. The decision-making algorithm already directed staff to look for additional symptoms that may have indicated infection if a resident’s temperature was within this range.

Staff in one nursing home expressed concern about stipulating a time scale before contacting a GP:

I wonder if it will not be safer not to put [the number of] hours because you know if you have this instrument and you want people to follow it and something happens then they said we’ve followed this and it’s your [care home staff] fault.

E: staff, pre implementation

Similarly, family members expressed concern about a 4-hour wait between temperature observations. Family members did not consider temperature a sufficient indicator of infection in older people and were concerned that the algorithm appeared to be over-reliant on temperature as a means of assessment:

I think I would be worried about the 1 to 4 hours I would worry about the time, if the answer is no, I think it should be a shorter period of time. (. . .) Usually an increase in temperature is a good sign that there’s something going on?

A: family, pre implementation

The GPs reported that although having a measurement of temperature as a guide was useful, some older patients may be quite unwell without having a raised temperature and it was not always a reliable indication of infection. One GP expressed concern that staff may feel compelled to follow the algorithm strictly and would wait 12 hours to see if a temperature rise was sustained before contacting the GP. Specific time requirements may be confusing and may lead to accusations of staff not adhering to guidance. Another GP felt that it was important to consider the time at which the temperature was taken, and suggested that this would have implications for requests for out-of-hours visits:

You see there, the over-65 [year-old] with COPD [chronic obstructive pulmonary disease] and delirium is much more clinically urgent than someone with increased frequency of their urine. And yet the end result of that algorithm is phone GP. Now if you phoned the out-of-hours service at 3 o’clock in the morning [for resident with chronic obstructive pulmonary disease] that’s reasonable. If you call the out-of-hours service at 3 o’clock in the morning like that [for resident with increased frequency], that’s unreasonable.

D: GP, pre implementation

Staff and relatives at the residential homes reported that residential home staff were not allowed to measure (‘take’) a resident’s temperature using a thermometer, as it was deemed to be a ‘nursing task’. There is no requirement for nurses to be present in residential homes. Participants reported that they would monitor a resident’s temperature in a number of ways, including feeling the head of the resident, noticing whether or not the resident was sweating or flushed, or whether or not the resident wished to remove clothing to cool down:

See where it says take the resident’s temperature we obviously can’t do that, step 1. (. . .) That’s why we rely wholly on behaviour and that because we don’t have a lot of tools that we are allowed to use. (. . .) We would love to be able to take temperatures and things like that there but company frowns upon it.

C: staff, pre implementation

Although family member participants from residential homes recognised that staff were not clinically trained, they expressed surprise that staff were not able to assess temperature using a thermometer. They queried the possibility of staff doing so, even though it was not part of their normal role:

It would be something to look into, if (. . .) they needed to take the temperature, that they would be able to do it. But you know they are not, they are carers, they are not medical people you know? So whether they would want to take that on is as their everyday work and job?

C: family, pre implementation

Initially, the research team considered changing ‘take resident’s temperature’ to ‘assess resident’s temperature’ to accommodate this practice. However, it was later decided that it would be more useful to train staff to use thermometers in the residential homes.

Additional signs and symptoms

Care home staff from nursing homes described how they routinely recorded other signs and symptoms alongside temperature if they suspected that a resident might have an infection, such as by measuring pulse, blood pressure and respiration rate, listening to the chest and measuring oxygen saturation levels. Participants in both nursing and residential homes discussed not relying solely on temperature as an indicator of infection and how they considered other signs and symptoms as being important to their decision-making:

We would notice a difference in their mood or the way they are, or confusion would be a big thing with elderly people (. . .) temperature would nearly be the last thing I would take. I would look at all the other things first and then I would take temperature, the GPs will always ask for the temperature.

B: staff, pre implementation

Participants from residential homes described how they would also look for non-clinical indicators of infection, including pallor, tiredness and struggling to breathe.

Again, the study team felt that it would not be useful to include all suggestions as they were not essential criteria in order to move on to the next stage of the decision-making algorithm.

Urinary tract infections

Participants expressed concern that it could be challenging to assess new or increased urinary urgency, frequency or incontinence, blood in urine and lower abdominal pain in care home residents, particularly in those who are incontinent or have dementia. They also noted that they would place more importance on some urinary symptoms than others (e.g. staff would be more concerned about evidence of blood in the urine than about increased urgency or frequency). More common indicators of infection that were suggested included change in smell or colour of urine and dehydrated skin:

Yeah and if someone is incontinent you don’t always know about the urgency of it because with dementia, not everyone can tell you when you need to go so it makes it quite difficult. And with the lower abdominal pain not everyone will tell you if they are in pain.

F: staff, pre implementation

I don’t think anyone ever tells us that they’ve got that it’s burning. Because a lot of them are already incontinent, they are wearing incontinent pads, you’re not going to see the increased urgency or frequency or increased incontinence, so it’s rare that we actually see blood in the pad when they’ve had a urinary infection. A lot of them can’t tell you if they’ve got a lower abdominal pain OK, you could see the shaking and the rigors but that’s not a symptom that we see often.

D: staff, pre implementation

Because there was insufficient evidence from the literature to support inclusion of a change in smell or colour of urine as an indicator of infection, this was not changed.

Participants who were registered nurses described how they would not expect to observe shaking or rigors in residents with a temperature below 37.9 °C. Evidence from the literature supported a 1.5 °C increase in baseline temperature for UTIs, which may not necessarily be greater than 37.9 °C,32,43 so, again, no changes were made based on the feedback.

Respiratory tract infection

Staff in nursing homes described aspects of the algorithm that gave them concern when assessing a RTI, for example that a respiratory rate of > 25 breaths per minute would often warrant emergency assistance (ambulance). There was also concern that residents with RTIs can deteriorate very quickly and waiting 4 hours before contacting GPs would be too long. Staff described how, in their experience, green/yellow sputum may indicate infection and recommended that this should be included in the algorithm. However, evidence did not support this inclusion, so no changes were made.

Staff described how a new or worsening chesty cough was likely to be accompanied by a high temperature, green or blood-tinged sputum, a change in pallor, clamminess and mood. If a resident was exhibiting these signs and symptoms, staff were likely to contact the GP irrespective of what was contained within the algorithm because a resident’s condition – especially a resident with a prior history of RTI – could decline very quickly:

See it’s not mentioning here the sputum, the colour of the sputum because COPD [chronic obstructive pulmonary disease], every patient of COPD has sputum (. . .) you can see a lot from the colour. When it’s infection it’s yellowish, greenish.

A: staff, pre implementation

Staff in residential homes thought that the algorithm largely reflected their usual practice apart from taking temperature and respiratory rate. For the latter, staff discussed how they would be alerted to the resident being more breathless than usual, having a change in colour, experiencing tiredness, having blue lips, struggling to breathe and having reduced mobility. Family member participants reported that although they considered that residential care home staff were competent to observe and test for UTIs, they felt that they were not qualified to do so in the case of RTIs, which would require a GP to assess:

The staff do all the vital signs which is the main things, and as I say some things they can do for UTIs is take a dip, but the like of chest infections or suspected chest infections they have to get a doctor out, they are not qualified to do that, so some of it’s OK and some of it’s not, you have to have a doctor.

A: family, pre implementation

Skin and soft tissue infection

Care home staff from nursing homes described how they would routinely swab when there was pus draining from a wound and that the results from the swab test would be sent to the GP, who would then prescribe an antibiotic. Conversely, GP respondents differed in their views of pus; one thought that it could be potentially confusing to ask those using the decision-making algorithm to contact the GP if pus was draining from a wound, as an antibiotic was not usually required:

If they have pus draining from a wound, we always swab it and send the swab. Always. We would never leave that. (. . .) And then like one was done the other day and it goes back to the GP, the results and then the GP contacts us and then with the antibiotic and then it comes from the pharmacy, so we always swab a pus-y wound.

B: staff, pre implementation

If we get someone with an abscess and we let the, once you let the pus out you don’t usually have to give the antibiotic cover, so it’s more the antibiotics needed here rather than whenever there actually is pus draining.

B: GP, pre implementation

Participants also discussed whether or not they would wait until there were two or more of the other symptoms listed in the algorithm for SSTIs before contacting the GP. Staff in residential homes reported that they would contact a district nurse if they had concerns about a SSTI and the district nurse would contact the GP if needed:

But with ours being residential, if we have any sort of concerns with any of the residents’ skin, we would contact the district nurse to come out, and the district nurse would sort of be the one then to say, if it’s maybe an open wound they would take a swab, or if it’s the likes of say, if it’s the redness, the hot to touch, and they would go back and voice concerns to the GP, and then issue an antibiotic, so the GP would.

C: staff, pre implementation

Well, if there’s a wound, we would do that if it’s localised redness heat or swelling or anything like that where there’s no abrasion or no wound then we go through the GP – if there’s a sore or the skin like moisture lesions we see. Or if it looks like breaking, then we go to through the district nurses.

F: staff, pre implementation

Action by care home staff

Care home staff participants described how they felt that the algorithm generally reflected their usual practice. When a resident had a temperature > 37.9 °C, staff provided supportive care in the form of fluids and paracetamol and rechecked the temperature approximately 2 hours before contacting the GP:

If they are not taking the paracetamol and not taking a drink, I’d be more inclined to contact the GP, because I know I will not get that down, if I can get that down with paracetamol and lots of drinks I would wait, usually.

B: staff, pre implementation

Format of the decision-making algorithm

Participants also discussed aspects of the format of the decision-making algorithm. They suggested that the algorithm should be laminated and preferred A4 size for personal use and A3 size for display. Participants proposed that the decision-making tool could be made more visually appealing and easy to follow with the addition of colour, and this suggestion was incorporated into the final version of the decision-making algorithm. Generally, staff, family members and GPs were happy with the format of the algorithm, but not everyone understood the ‘≤’ symbol used in the algorithm and this was removed in the final version.

Internal review by the Feasibility Study Management Group

As part of the iterative process in the development and adaptation, there were ongoing discussions within the research team. The literature review, the consensus group, focus groups and interviews contributed to these discussions. A further two additional papers49,50 were identified, which informed our thinking regarding the use of baseline temperature to initiate the use of the algorithm. The final version of the decision-making algorithm was structured in accordance with the three distinct stages that had emerged from the findings from the qualitative data collection earlier and provided a step-wise approach to the management of residents. These stages were initial assessment of the resident, observation of the resident and action by care home staff. Each stage is described in the following sections.

Initial assessment of the resident

The research team concluded that the most appropriate way to begin the algorithm was by providing a list of various non-specific signs and symptoms of infection (including change in behaviour of the resident), together with specific signs and symptoms of each of the three infections (see Figure 7). This was because of the emphasis on non-specific indicators as signs of infection in older people within the literature,14,41,46,51,52 expert opinion from the consensus meeting and the opinions of care home staff and family members of care home residents. This approach would ensure that the majority of residents who may be unwell were identified in the first instance (Box 3).

BOX 3 - Initial assessment of the resident

Resident has one or more new/worsening symptoms making them unwell

Non-specific: suspected fever, change in behaviour (e.g. delirium, confusion/agitation, unco-operative, reduced mobility/’off legs’, loss of appetite, withdrawn).

AND/OR

Urinary/chest/skin symptoms.

Observation of the resident

It was agreed that temperature should not be used as a standalone criterion in the initial assessment of the resident (as was the case in the original algorithm by Loeb et al. 16) as older people do not always present with an increase in temperature when they are unwell. 49,50 In addition, an increase of 1.5 °C above the baseline temperature (as denoted in the original algorithm16) was not applicable to the UK setting as baseline temperatures are not routinely recorded in care homes. However, assessment of temperature was included in the observation step to identify residents who may be extremely unwell, and so temperature was included as an additional step, with a value of > 37.9 °C denoting the requirement of additional monitoring. This value was agreed within the research team, in consensus group meetings and focus groups as a temperature that would cause concern and require additional monitoring if presented by a resident. Figure 5 shows the observation section of the algorithm.

FIGURE 5.

Observation of the resident.

Following on from measurement of the resident’s temperature, this stage also incorporated the specific signs and symptoms of each infection. Following internal review by all those contributing to the updating and adaptation of the algorithm, and using a UTI as an example, the final signs and symptoms are outlined in Box 4.

BOX 4 - The UTI signs and symptoms

Urinary symptoms

Resident (without catheter) has burning on urination or two or more of the following:

• new or increased urgency

• new or increased frequency

• new or increased incontinence

• blood in urine

• lower abdominal pain

• shaking/rigors.

Action by care home staff

An action stage was added to the end of the algorithm to instruct care home staff on how to proceed, depending on the symptoms presented by the resident. For example, if the resident fulfilled the criteria for a suspected UTI (i.e. two or more symptoms from the list or dysuria alone), the care home staff were instructed to telephone the GP. If the minimum number of symptoms were not present, an additional step instructed the staff member to monitor residents with a temperature between 37.3 °C and 37.9 °C. In this scenario, the algorithm instructed the user to repeat step 1 (taking the resident’s temperature) after 6 hours. Figure 6 shows the ‘action’ stage of the algorithm relating to a UTI.

FIGURE 6.

Action by care home staff.

A copy of the final decision-making algorithm is shown in Figure 7.

FIGURE 7.

Final REACH decision-making algorithm. b.p.m., breaths per minute; COPD, chronic obstructive pulmonary disease. Reproduced from Hughes et al. 35 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Reproduced from Hughes et al. 35 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Summary

In this phase of the study, we set out to adapt and update the decision-making algorithm and training materials developed for the original Canadian study by Loeb et al. 16 for implementation in the feasibility study. This was achieved through a series of rapid reviews of the literature, focusing on identifying new evidence that had been published since the original Canadian study. This fed into the development and adaptation of the decision-making algorithm to encompass management of RTIs and SSTIs in addition to UTIs. A draft version of the decision-making algorithm was presented to, and discussed with, a range of health-care professionals, all of whom had experience of the management of infectious diseases and/or care of older people, as part of a consensus exercise. Further versions were reviewed and discussed in a series of focus groups and interviews with key stakeholders (i.e. care home staff, resident family members and GPs). Continuous iterative review was also undertaken by the research team while the other activities were taking place.

The final agreed version of the decision-making algorithm was included in the training material that was developed to support intervention implementation. The material consisted of a Microsoft PowerPoint presentation and study handbook. The presentation was also available in a number of other formats to facilitate accessibility of training for care home staff.

Introduction

This phase of work focused on the implementation of the intervention (training and use of the decision-making algorithm) in the six care homes. Aspects regarding the feasibility of collecting data are discussed in detail in Chapter 6.

Aims and objectives

The aim of this phase of the research was to implement the intervention in the six recruited care homes. The objectives were as follows:

• to deliver the training programme to care home staff in the participating homes

• to collect relevant data from a variety of sources to assess aspects of feasibility, including the use of the decision-making algorithm and data pertaining to the dispensing of antimicrobials

• to measure the resource use and costs associated with the training and implementation.

Methods

Data analysis

Results

Evaluation of data-collection feasibility

Care home data

Data were collected on the characteristics of the six participating care homes (Table 11). The number of beds ranged from 32 to 62 at baseline and occupancy ranged from 58% to 100%. Care home A was undergoing refurbishment over the course of the study and one floor within the home was not available, hence the reduced occupancy. Over three-quarters of all residents were female. The age range for male residents was 63–96 years, whereas for females the range was 57–103 years. The majority of all staff in both types of care home were junior carers. Half of care homes belonged to national chains/groups, and only one home reported having a policy relating to AMR.

TABLE 11 - Characteristics of the participating care homes at the start of implementation

Characteristics	NI			England			Range	Total
	Care home			Care home
	A	B	C	D	E	F
Type of home	Nursing	Nursing	Residential	Nursing	Nursing	Residential	–	–
Registered to provide care for residents with dementia?	Yes	No	Yes	Yes	Yes	Yes
Number of beds in home (n)	62	32	36	56	51	40	32–62	277
Reported total occupancy at baseline, n (%)	36 (58)	28 (88)	36 (100)	42 (75)	51 (100)	37 (93)	26–51	230 (83)
Males, n (%)	12 (33)	8 (36)	5 (14)	12 (29)	13 (25)	5 (16)	5–13	55 (24)
Females, n (%)	24 (67)	18 (64)	31 (86)	30 (71)	38 (75)	31 (84)	18–38	172 (76)
Age of residents (years)
Males (median)	75.5	83	80	82	86	82		–
Males (range)	65–89	73–95	71–93	63–96	71–95	74–89	63–96	–
Females (median)	85.5	88	87	92	89.5	85
Females (range)	67–99	70–97	75–103	68–99	57–98	66–97	57–103
Total reported staffing at baseline (n)	67	39	18	53	38	50		265
Manager/deputy manager	2	1	2	1	1	2		9
Nurses	15	8	0	9	10	0		42
Senior carer	2	3	3	3	5	9		25
Junior carer	30	18	7	27	22	39		143
Ancillary staff/others	18	9	6	13	0	0		46
Home is part of a chain or national group?	Yes	No	Yes	No	No	Yes
Home has a policy relating to AMR?	No	No	Yes	No	No	No

The decision-making algorithm and its use in the care homes

Five of the six care homes recorded using the decision-making algorithm on a study-specific CRF; the sixth home was unable to complete these forms primarily owing to staffing shortages and a safeguarding issue within the home that prevented them from engaging fully with the study. The decision-making algorithm form was completed 135 times across five care homes during the 6-month implementation phase of the study. On 81 out of 135 completed CRFs (60%), staff reported using the decision-making algorithm (Table 12). We do not have data on frequency of use of the algorithm, or where training might have affected decision-making, when a CRF was not completed. Data in this section are likely to represent only an unknown proportion of actual activity.

TABLE 12 - Summary of the use of the decision-making algorithm

Items completed on the decision-making algorithm form	n (%)
Use of decision-making algorithm (N = 135 forms completed)
Yes	81 (60)
Missing (neither box ticked)	1 (1)
No	53 (39)
Type of suspected infection (N = 135)
Urinary tract	66 (49)
Respiratory tract	38 (28)
Skin and soft tissue	11 (8)
Both urinary and respiratory tract	4 (3)
Do not know	10 (7)
Other specified	6 (4)
Biliary sepsis	1
Broad spectrum between mouth and urine infection	1
Eye infection	1
Swollen gum inspection	1
Vaginal thrush	1
Resident blood test indicates an infection, source not known, GP prescribed antibiotic	1

Urinary tract infections were the commonest suspected infection in residents (49%), with RTIs ranked second (28%). The most frequently reported reasons for not using the tool were that other tests were carried out (14/53) or the resident was too unwell (13/53). Of note, staff reported that in five cases (5/53) relatives either intervened or became involved in the decision-making process.

Table 13 shows that from the 135 forms completed by staff, 274 actions were initiated. The most commonly reported action was to contact the GP (39%), followed by continuing to monitor the resident or provide supportive care (33%). Conducting additional tests was the most reported ‘other’ action (21/60); of note, junior staff reported their findings to senior staff on seven occasions (7/60). There were 186 actions taken by the GP when contacted. In 24% of these cases, the GP visited the resident. In 52% of such cases, GPs prescribed an antibiotic. Of the 25 ‘other actions’ taken by the GP, eight related to residents being sent to the emergency department. There were 101 records reporting on outcomes at 2 weeks post suspected infection. In most cases, the resident had recovered (59%) and 15% were admitted to hospital. Of the ‘other’ outcomes at 2 weeks, five out of 20 residents were missing data, four out of 20 residents were still in hospital and four out of 20 residents were still on antibiotics.

TABLE 13 - Records of actions taken resulting from the use of the decision-making algorithm

If yes, the GP later prescribed antibiotics.

Records	n (%)
Actions taken by care home staff (N = 274)
Continued to monitor/provide supportive care	89 (32.5)
Contacted the GP	106 (38.7)
Over-ruled the decision aid	18 (6.6)
No action taken	1 (0.4)
Other	60 (21.9)
Actions taken by GP (N = 186)
The GP visited the resident at the care home	44 (23.7)
The GP prescribed antibiotics	97 (52.2)
The GP advised to continue to monitor and/or provide supportive carea	20 (10.8)
Other	25 (13.4)
Outcome 2 weeks after suspected infection (N = 101)
Resident recovered	60 (59.4)
Resident was prescribed further antibiotics by the GP	5 (5)
Resident admitted to hospital	15 (14.8)
Resident died	1 (1)
Other	20 (19.8)

Figure 8 displays the pathway from the decision-making algorithm being used, GP contacted, antibiotics prescribed to the outcome after 2 weeks. The figure illustrates the decision-making process as it was recorded by the staff. At each step, there is a ‘yes/no’ pathway until an outcome is achieved. If the extreme left-hand pathway is followed from when the decision-making algorithm was used, an infection was suspected and the GP was contacted (64 times), an antimicrobial was prescribed in most cases (57 times). If a pathway was followed in which the decision-making algorithm was used, and the GP was not contacted (17 times), an antimicrobial was less likely to have been prescribed (prescribed twice).

FIGURE 8.

Flow chart of the use of the decision-making algorithm to an outcome after 2 weeks.

The effect of the decision-making algorithm is borne out further by the data presented in Table 14. From the 81 cases in which the algorithm was used, 27% (n = 22) did not receive an antimicrobial, compared with 73% (n = 59) who did (see Table 14). Even with these small numbers, the trend was in the expected direction [i.e. when the decision-making algorithm was used, there were fewer cases of antimicrobial prescribing (73%) than when the algorithm was not used (76%)].

TABLE 14 - Antimicrobial prescribing following the use of the decision-making algorithm

	Use of tool?		Total
	No	Yes
Antimicrobial prescribed?
No
Cases (n)	13	22	35
% of use of tool	24.5	27.2	26.1
Yes
Cases (n)	40	59	99
% of use of tool	75.5	72.8	73.9
Total
Cases (n)	53	81	134
% of use of tool	100.0	100.0	100.0

Use of health and social services data

The use of health and social care services by residents over the 6-month implementation phase was collected using two forms for completion by care home staff: the use of hospital services form and the contacts with health and social care professionals form. The function of this exercise was to inform the design of a full cost-effectiveness analysis in a future definitive trial. We wanted to establish if the forms were fit for purpose (i.e. did they collect the information that we needed?) and if they were easily completed by care home staff or if they were burdensome. We also wanted to establish the type of resources used most frequently by residents so that future versions of the data-collection forms could be refined. Comments from staff on data collection are provided in Chapter 6.

Hospital services

In total, 127 hospital events involving the use of 151 services were recorded on the use of hospital services form from five out of the six homes. No data were collected from the English residential home (F) as they were reported to be too difficult/inaccurate to collect retrospectively; in addition, this home had faced a number of difficulties during the study. In only 22 of the hospital events an infection was suspected, and half of these were RTIs (Table 15).

TABLE 15 - Suspected infections reported

Suspected infection	Frequency
RTI	11
UTI	5
Sepsis	2
Unspecified	4
Total	22

The majority of hospital services were outpatient appointments (84/151); however, none of these or the day procedures were due to a suspected infection (Table 16). In total, 34 hospital admissions (three following an outpatient appointment, 21 from accident and emergency visits and 10 direct admissions) were reported and only one of these was a planned admission. Most of the accident and emergency admissions were associated with an infection (18/21). The outcome from the hospital admission was recorded in 33 out of 34 cases, with the majority (25/33) returning to the care home. The mean length of stay associated with a hospital admission was 6.8 days; for those admissions in which an infection was suspected, it was longer (7.9 days). One resident who was admitted to hospital with a suspected infection (sepsis) later died in hospital.

TABLE 16 - Use of hospital services and suspected infection

CI, confidence interval.

Only one admission was planned.

One patient remained in hospital; one was a respite patient and returned to the home.

Hospital service used	Number of residents (n = 151)	Cases in which infection was suspected (n = 22)
Outpatient appointment	81	0
Outpatient appointment leading to admission	3	0
Day procedures	5	0
Accident and emergency visit	7	1
Accident and emergency visit leading to admission	21	18
Direct admissionsa	10	3
Outcome from all admissions (n = 34)
Returned to care	25	16
Moved elsewhere	3	3
Died	3	1
Otherb	2	1
Missing	1	0
Mean length of stay (days) (95% CI)	6.8 (4.8 to 8.7)	7.9 (5.4 to 10.3)

Contacts with health and social care professionals

In total, 1540 contacts with health-care professionals were recorded on the contacts with health and social care professionals form from all six care homes. In only 145 cases, this was due to a suspected infection (Table 17). A total of 2220 residents were involved in these contacts. Nurses were the most frequently contacted health-care professional, with 584 contacts recorded, of which 544 involved a visit to the care home. However, in only seven of these contacts was an infection implicated.

TABLE 17 - Contacts with health and social care professionals over the study period

Number of residents involved not reported in one case.

Number of residents involved not reported in three cases.

Resource used	Number of contacts	Cases in which infection was suspected	Number of out-of-hours contacts	Total number of residents involved	Mean number of residents involved per contact (range)
GP telephone call	257	68	24	261	1.0 (1–3)
GP visit at homea	210	65	22	601	2.9 (1–29)
GP clinic visit	9	0	0	9	1.0 (1–1)
Nurse telephone call	37	3	0	39	1.1 (1–2)
Nurse visit at homeb	544	4	5	599	1.1 (1–5)
Nurse clinic visit	5	0	0	5	1.0 (1–1)
Other telephone call	178	2	5	216	1.2 (1–36)
Other visit at home	279	3	2	466	1.7 (1–30)
Other clinic visit	21	0	0	24	1.2 (1–2)
Total	1540	145	58	2220

There were fewer contacts with GPs (n = 476) than with nurses, but GPs tended to see multiple patients per visit, with 210 home visits involving a total of 610 patients, and these visits were more frequently associated with a suspected infection (133/476).

The type of nurse varied considerably (Table 18), with the vast majority of nurse contacts made by district nurses. There were also nurse contacts recorded that were associated with NHS continuing health-care assessments; this is an organised visit related to the provision of resident health care and not connected to a specific health event.

TABLE 18 - Types of nurse contacts specified

Nurse type	Number of contacts
District	469
Continuing health-care assessment	29
Tissue viability	25
Diabetic	13
Mental health	13
Practice	11
Palliative	8
Not specified	5
Incontinence	3
Respiratory	3
Clinical nurse facilitator	2
Care home nursing team	2
Test results	2
Renal	1
Total	584

There were 478 contacts recorded with other health-care professionals (Table 19), but an infection was suspected in just five of these contacts. Again, some contacts could be regarded as inappropriately recorded (e.g. telephone calls to obtain test results or medication and contacts with hospital and ambulance services that should have been recorded on the hospital services use form).

TABLE 19 - Other health-care professional contacts specified

Other health-care professional	Total number
Social worker	173
Speech and language therapist	75
Chiropodist/podiatrist	57
Dietitian	56
Physiotherapist	30
Psychiatrist/psychologist	26
Optician	23
Pharmacist	10
Ambulance	9
Audiologist	7
Test results/medication orders	5
Care home on-call manager	4
Intermediate care team	1
Hospital	1
Unspecified	1
Total	478

Community pharmacy dispensing data

Based on the data supplied by the community pharmacies, the number of antimicrobial items dispensed for residents in the participating care homes was summarised, together with a calculation of the total number of DDDs of antimicrobials dispensed, as shown in Table 20. The most frequently prescribed antimicrobials were nitrofurantoin (n = 84 items pre implementation; n = 82 items post implementation), trimethoprim (n = 71 items pre implementation; n = 46 items post implementation) and amoxicillin (n = 59 items pre implementation; n = 56 items post implementation). Comparing the antimicrobial prescribing against the NI Management of Infection Guidelines for Primary and Secondary Care (2016)65 and the Community Antibiotic Guidelines for Common Infections in Adults (2017) produced by the Coventry and Warwickshire Area Prescribing Committee66 suggests that the prescribing patterns were broadly in line with recommendations for UTIs and RTIs.

TABLE 20 - Antimicrobial prescribing trends pre and post implementation of the intervention

12 months divided by 2 and rounded to the nearest whole number.

Antimicrobial agent	Total number of prescriptions			Total number of DDDs
	Before intervention (6 monthsa)	After intervention	Change (%)	Before intervention (6 monthsa)	After intervention	Change (%)
Nitrofurantoin	84	82	–2 (2)	544	531	–13 (2)
Trimethoprim	71	46	–25 (35)	401	244	–157 (39)
Amoxicillin	59	56	–3 (5)	501	518	+17 (3)
Doxycycline	53	45	–8 (15)	612	445	–167 (27)
Flucloxacillin	35	23	–12 (34)	223	142	–81 (36)
Cefalexin	20	13	–7 (35)	76	53	–23 (30)
Co-amoxiclav	18	13	–5 (28)	149	95	–54 (36)
Clarithromycin	15	25	+10 (67)	161	328	+167 (104)
Ciprofloxacin	8	8	0 (0)	35	40	+5 (14)
Erythromycin (includes ethyl succinate)	4	4	0 (0)	31	33	+2 (6)
Pivmecillinam	11	12	+1 (9)	41	48	+7 (17)
Others	5	7	–2 (40)	74	82	+8 (11)
Total	383	334	–49 (13)	2848	2559	–289 (10)

The total number of prescriptions dispensed in the 6 months prior to the intervention was 383, whereas 334 prescriptions were dispensed post intervention, equating to a reduction of 49 prescriptions (13% reduction). In terms of DDDs, the pre-implementation total was 2848, whereas the post-implementation total was 2559, a reduction of 289 DDDs (10% reduction). As highlighted in Table 20, there was a reduction in the prescribing of trimethoprim, flucloxacillin and cephalexin. The prescribing of nitrofurantoin remained relatively unchanged, whereas clarithromycin prescribing increased post implementation. These changes were somewhat reflected in the DDDs prescribed for each drug. From the data, we could see that approximately 30 nitrofurantoin prescriptions pre intervention had been written for a quantity of 30 tablets; in the case of trimethoprim, there were prescriptions written for 30 × 100-mg tablets. These prescribing courses may suggest prophylaxis of UTIs rather than treatment.

Table 21 summarises data at the care home level. In three homes, there was a reduction in both the number of antimicrobial items and the number of DDDs. In two care homes, there was a slight increase in the number of items (homes A and E). The home designated F demonstrated a larger increase (an increase of 20 prescriptions post intervention), but this home had experienced a number of difficulties over the course of the study and we have concerns about the completeness of the data.

TABLE 21 - Antimicrobial prescribing trends in the participating care homes pre and post implementation of the intervention

CI, confidence interval; SD, standard deviation.

12 months divided by 2 and rounded to the nearest whole number.

Home	Total number of prescriptions			Total number of DDDs
	Before intervention (6 monthsa)	After intervention	Change (%)	Before intervention (6 monthsa)	After intervention	Change (%)
A	77	78	+2 (2)	548	595	+48 (9)
B	69	49	–20 (28)	691	479	–212 (31)
C	68	44	–24 (35)	433	276	–156 (36)
D	68	34	–34 (50)	464	280	–184 (40)
E	82	89	+7 (9)	572	640	+68 (12)
F	21	40	+20 (95)	142	289	+147 (103)
Total	383	334	–49 (13)	2848	2559	–289 (10)
Mean	63.8 (SD 22.0)	55.7 (SD 22.4)	–8.2 (95% CI –29.7 to 13.4)	474.7 (SD 186.6)	426.5 (SD 167.1)	–48.2 (95% CI –209.2 to 112.8)

Following removal of all non-antibiotic products [in addition to unusual drugs such as rifaximin (Targaxan^®)], analysis at the resident level, as shown in Table 22, reveals the number of residents who were prescribed antimicrobials 12 months prior to the intervention being implemented and 6 months post implementation.

TABLE 22 - Number of residents prescribed antimicrobials pre and post implementation of the intervention

This home had 62 beds in total, but 22 were closed owing to refurbishment from December 2016 to December 2017.

Home	Number of residents prescribed antimicrobials (average occupancy)
	12 months pre implementation	6 months post implementation
A	52 (45)a	30 (35)
B	32 (32)	21 (32)
C	33 (36)	16 (36)
D	40 (53.5)	19 (53.5)
E	45 (51)	19 (51)
F	21 (37)	24 (37)
Total	223 (254.5)	125 (244.5)

Analysis at the resident level, as shown in Table 23 [again, with the removal of non-antibiotic products and rifaximin (Targaxan)] illustrates the number of prescriptions issued per resident pre and post implementation. We were unable to collect data pertaining to the number of residents pre and post implementation who had not been prescribed any antimicrobial.

TABLE 23 - Number of prescriptions for antimicrobials issued per resident

Number of antimicrobial prescriptions	Residents in receipt of an antimicrobial prescription, n (%)
	Pre implementation to 12 months	Post implementation to 6 months
1	81 (36)	52 (42)
2	47 (21)	33 (26)
3	20 (9)	12 (10)
4	21 (9)	11 (9)
5	14 (6)	3 (2)
6	6 (3)	3 (2)
7	8 (4)	4 (3)
8	5 (2)	0 (0)
9	4 (2)	2 (2)
≥ 10	17 (8)	5 (4)
Total	223 (100)	125 (100)

Calculation of the ICC revealed that it was 0.11 [95% confidence interval (CI) 0.00 to 0.24] at baseline, 0.05 (95% CI 0.00 to 0.13) post implementation and 0.09 (95% CI 0.00 to 0.24) overall.

Administrative data

English data

In England, we approached NHS Digital to discuss if we could obtain anonymised HES data for the residents in the three care homes in England. At the outset of this project, NHS Digital was going through considerable organisational changes and this initially proved to be a barrier to engagement. Our enquiries focused on identifying an approach to obtain HES data on all of the residents in the three English care homes covering the period of the study. As our study did not have resident-level consent, we had hoped that the NHS database could be searched against the addresses and postcodes of the care homes. This would give three sets of data relating to each of the homes. However, NHS Digital informed us that a search based on a postcode and address was not possible on their system. Our discussions have been ongoing and it now seems that we may be able to obtain pooled anonymous data through using the residents’ NHS Number. As noted, we did not have ethics approval for individual consent so we would have to seek permissions from the Confidentiality Advisory Group. Resident NHS Numbers can be found in two places in the study: the first is within the records maintained by the care homes and the second is at the pharmacy/dispensary that supplies the home with its medicines. We would require someone (e.g. care home manager) who has the appropriate permissions to access the residents’ NHS Numbers, collate them and send these numbers to NHS Digital, who would then supply the anonymised data for that particular care home.

Northern Irish data

The algorithms developed within the patient registration system by the Health and Social Care BSO were used to identify the number of residents in each of the three care homes in NI at the start of the study. These are presented in Table 24 along with the number of residents ascertained by data collection undertaken by the research fellows and staff within the homes.

TABLE 24 - A comparison of the residents in the three care homes in NI at the start of the study, as ascertained from the study and from administrative data sources

From the patient registration system.

Includes two respite residents.

Care home	Number of residents
	Data collected during study	Data collected from administrative dataa
A	36	35
B	28b	24
C	36	34

As can be seen, there is a reasonably close correlation between the numbers of residents with data derived from the administrative data systems for two care homes and those with data derived from data collection over the course of the study. It appears, however, that the algorithm tended to under-record residents, perhaps because it erred on the side of caution/selectivity rather than sensitivity. According to the data collected directly during the study, there was one additional resident recorded in care home A, four additional residents in care home B and two additional residents in care home C, compared with data collected from administrative sources; it should be noted, however, that care home B included two respite residents who would not have been detected by the administrative data system, which registers only permanent residents as temporary addresses are not recorded. Some further discrepancies would be expected as a result of slight differences in the timing of the dates of the data collection during the study and extraction from the registration systems. The patient registration system, which depends on changes of address at the GP surgery, will always lag behind ‘real-time data’.

Table 25 shows some of the additional information garnered from the administrative data systems. This includes a tally of all residents entering the three care homes in NI during the 6 months of implementation, producing the total number of residents for the study period; the total number of deaths (all causes) of residents (n = 15) in these homes was extracted, again during the same period.

TABLE 25 - Other data abstracted from the patient registration system

Care home	Total number of residents (1 May to 31 October 2017)	Number of deaths (1 May to 31 October 2017) (all causes)
A	39	6
B	33	7
C	42	2

Summary

The intervention (training and use of the decision-making algorithm) was implemented in the six participating care homes. Data were collected from a number of sources. Analysis of the data provided insight into the use of the decision-making algorithm, use of hospital services and contact with other health-care professionals. Dispensing data were summarised to provide the number of antimicrobial items dispensed over different time periods and converted to DDDs, and then used in subsequent analysis. Very few administrative data were available owing to logistical difficulties. Resource use and costs associated with the intervention were calculated.

Introduction

In this chapter, we present the process evaluation. Our process evaluation was adapted from and based around a Medical Research Council framework67 for such evaluations and was inclusive of a number of key components for process evaluations proposed by Steckler and Linnan. 68 The components chosen for this study reflect that this was a feasibility study and that our main interest was in evaluating the feasibility of a number of aspects of the study, for example the ability to recruit, deliver (implement) and evaluate an intervention in care homes in NI and England.

As this was a feasibility study, it was not appropriate for us to consider fidelity as a component of this process evaluation. Should a larger trial be conducted in the future, a full evaluation of the fidelity of the intervention would be undertaken.

As outlined in Chapter 2, normalization process theory was the underpinning theory within this study. Normalization process theory is a middle-range theory that can underpin process evaluations of complex interventions in health care. 22 Middle-range theory is an approach to sociological theorising aimed at integrating theory and empirical research. 69 We have used normalization process theory to evaluate the development and implementation of the intervention. Normalization process theory is not a methodology or how-to-do guide to research and it is not prescriptive or rigid; rather, it is a theoretical device that is meant to be used in a flexible and active way, allowing us to both direct and explain the everyday and critical course of the project. 22,70,71 In the description of the methods, we outline how normalization process theory was used within this process evaluation.

Aims and objectives

The aims of the process evaluation were:

• to comprehensively describe the implementation of this intervention, including the facilitators of and barriers to implementation

• to develop a set of recommendations regarding the intervention to inform its implementation on a wider scale.

The objectives were:

• to monitor implementation processes (e.g. recruitment, development of the intervention, delivery of the intervention and acceptability/use of the intervention in practice)

• to undertake ethnographic observations in the care homes to understand current practice and to explore possible changes due to the intervention

• to carry out in-depth interviews/focus groups with a sample of care home staff, care managers and other stakeholders (e.g. GPs).

Methods

We used a mixed-methods approach, combining qualitative and quantitative data72,73 to facilitate exploration of apparent discrepancies between findings. 74,75 The principal data-collection method was qualitative (e.g. interviews, focus groups and observational field notes), complemented and illuminated by the quantitative data (e.g. recruitment to focus groups, training sessions provided and attendances at sessions), providing an in-depth and breadth of understanding of the study processes and the implementation of the intervention into practice. We used normalization process theory (see Chapter 2, Normalization process theory) as the theoretical underpinning to our methodological and analytical approach in this study. As this was a feasibility study, all six care homes were included in the process evaluation. We have outlined the components of process evaluation we explored in the study (Table 27).

In the following sections, we outline the procedures used to gather process evaluation data.

Data analysis

Results

Summary

The process evaluation explored the implementation of the intervention presented in Chapter 5. As this was a feasibility study, the process evaluation provided valuable information about how the intervention was implemented and, indeed, to what extent it was implemented.

Our results have demonstrated that we could recruit care homes and train staff to use the decision-making algorithm, and there was some evidence of a level of implementation of the intervention. The process evaluation has identified a number of challenges and suggestions that should be considered in a future study.

Introduction

In this study, we explored the views of care home managers, or the individuals who they designated, on the REACH training programme and intervention and the likelihood of recruitment to a future trial by conducting a short postal survey with all care home managers in NI and the West Midlands, England.

Aim and objectives

Our aim was to assess the likelihood of care home participation in a future trial evaluating the effectiveness of the REACH intervention to reduce antimicrobial prescribing in care home residents and to gain care home managers’ views on the REACH training programme and intervention.

Ethics approval

This study was approved by members of the School of Pharmacy Ethics Committee at Queen’s University Belfast (reference 022PMY2017).

Methods

In order to fulfil the aims and objectives of this study, quantitative methodology, employing a brief postal questionnaire survey, was used. Using surveys as a research tool is advantageous when the desired outcome is to gather information from a large sample within a relatively short time frame. 77 In addition, respondents had the opportunity to provide free-text responses.

Data analysis

Responses were entered into IBM SPSS^® Statistics version 20 (IBM Corporation, Armonk, NY, USA) statistical analysis software. All respondents were given a unique identifier to ensure anonymity when reporting any free-text responses provided. The content of such responses were imported into NVivo^® for analysis of recurring themes. Descriptive statistical analysis (e.g. frequencies, means and medians) were used to report findings from closed and multiple-response questions. As with all studies using questionnaires, a level of missing data was to be expected; however, it was hoped that this would be minimised through the process of piloting and refining the questionnaires. Where there were missing response data, they were coded as such and omitted from the final analysis.

Results

In total, 167 care homes (23%) responded to the survey; the response rate for care homes in NI was 47% (n = 107) and the response rate for English care homes was 12% (n = 60). Over half of responders (n = 89) provided additional free-text comments in response to one or more of the questions.

From those who responded, 83% (80% of care homes in NI and 88% of English care homes) indicated that they would welcome such a study and 6% (6% of care homes in NI and 7% of English care homes) indicated that they would not (11% did not know; 14% of care homes in NI and 5% of English care homes). Free-text comments suggested that respondents thought it an important area to research, although managers felt that some staff may be more interested than others. Some respondents expressed reservations that care homes already use a range of decision-making aids and guidelines and had limited time to commit to research.

There was good support for an education and training programme on antimicrobial prescribing (91%; 91% of care homes in NI and 90% of English care homes), and this was reflected by free-text comments advocating the benefits of any training for staff and increasing the awareness of the subject area:

Most care staff including myself as the home manager would benefit from more training as we have little knowledge about this subject.

There were also a number of comments expressing concern that care staff have limited control over what GPs prescribe and that either the education and training programme should be aimed at GPs rather than care staff or GPs should be included in the training programme. Similarly, there was concern that some families insist that GPs are contacted in order to prescribe antibiotics and, therefore, respondents thought that families should also be offered the programme:

I feel it would be beneficial to have GPs on board to avoid prescribing antibiotics when patients are asymptomatic.

Some GPs are reluctant to give [antibiotics] as client does not require, but families insist so they need the training more as they are unwilling to listen.

Most of the respondents (90%; 91% of care homes in NI and 87% of English care homes) reported that 2 hours was a reasonable period of time for staff to attend such a programme. Among those who did not think that this was reasonable and provided free-text comments, most suggested that a 1-hour training session would be more appropriate and that care staff could have shorter sessions than nursing staff.

Approximately three-quarters of respondents (73% of care homes in NI and 80% of English care homes) indicated that it was feasible for staff to attend an education and training programme if it were provided at different times of the day, but 19% (20% of care homes in NI and 17% of English care homes) did not think that it was feasible for all staff in the home to attend. Free-text comments suggested that shift patterns, annual leave, sickness, difficulty taking staff away from direct care and relying heavily on agency staff made it difficult for all staff to attend. Some respondents suggested that it was necessary only for senior staff and managers to attend the training. Respondents reported that they currently provided training sessions at 14.00–16.00 and 06.00–08.00 to allow both day and night staff to attend, and suggested timetabling sessions weekly for a few weeks to maximise attendance:

Ideally I’d like to say yes, however experience proves that lots of things can happen throughout the day, and therefore limit staff availability.

In terms of using a ‘decision-making aid’ to help staff decide when to contact the GP if they suspected that a resident had an infection, 88% of respondents thought that it would be useful (87% of care homes in NI and 88% of English care homes) and 9% (10% of care homes in NI and 9% of English care homes) did not think that it would be useful; 3% (3% of care homes in NI and 3% of English care homes) did not know. Comments regarding the usefulness of the ‘decision-making aid’ included support for the use of evidence-based tools to give staff confidence in decision-making:

The care staff would report concerns to the deputy manager. I think it would be great for them to take more responsibility and take action in a resident’s best interest.

A number of managers reported that staff already used tools for the assessment of UTIs and considered additional tools unnecessary:

My own nurses express concern about the number of tools already in place. Paperwork has become excessive. The practical, observant and experienced nurses suggest tool overload.

The most common comment was that the decision-making tool would be useful only if GPs and families were also supportive:

If the GPs are also in agreement as many GPs are still reluctant to listen to qualified nurses who know the patients.

One manager expressed concern based on participation in a previous study that decision-making aids can be too simplistic for residents in care homes who have complex conditions and that they do not take into consideration the knowledge nurses use to assess change in residents with dementia who are unable to verbalise their symptoms.

Overall, 79% of care homes (78% of care homes in NI and 82% of English care homes) were, in principle, prepared to participate in a study such as that described in the participant information sheet that was provided with the questionnaire. Free-text comments suggested that managers were keen to be involved in research, felt the subject area to be important and welcomed training for staff:

I believe it is an essential area of study and the home and myself would be happy to participate.

Some respondents, however, commented that despite the undoubted benefits of research, current staffing levels would prevent the care homes from taking part:

Depending on how much time it would involve as staff are already very busy.

Summary

A survey was conducted in a sample of care homes in NI and the West Midlands to gauge interest in a larger study. The overall response rate was 23%. There was interest and support for a larger study and recognition of the importance of AMR, and the opportunity for training was welcomed. Concerns were expressed regarding time commitment and the need to involve GPs and family members.

Introduction

This chapter summarises and discusses the findings from each phase of the study in order to directly address the aims and objectives of the research. The overall aim was to evaluate the feasibility and acceptability of a multifaceted intervention on rational prescribing for infections in a non-randomised feasibility study in care homes. The objectives related to recruitment of homes, adapting and developing the intervention (a decision-making algorithm and small group interactive training), implementing the intervention, undertaking a detailed process evaluation of the non-randomised feasibility phase and testing data-collection procedures. We discuss these issues in subsequent sections in this chapter, in addition to the findings of a survey in care homes that was undertaken in order to gauge interest in a larger study.

Recruitment

We successfully recruited six care homes that met the inclusion criteria. The approach taken in the two geographic sites was largely comparable, but also took into account some contextual differences, notably the role of ENRICH in England; there is no equivalent organisation in NI. In NI, we also restricted the geographic area from which care homes were recruited to reduce the amount of travelling that would be required to be undertaken by research staff. It was at this point in the sampling approach that many homes were excluded in NI. Therefore, a greater sample of care homes would be available from which to recruit if the requirement of geographic proximity to Belfast was lifted. We had also excluded care homes that were dual registered (i.e. providing nursing and residential services) as we wished to explore the feasibility of this intervention in the residential setting compared with the nursing home environment. Dual registration may have confounded this assessment. It had also been decided, a priori, to exclude trust-owned homes; almost all nursing homes in NI are privately owned, but this is not the case with residential facilities. Applying this criterion ensured a degree of comparability between nursing and residential homes. Interest in research was gauged by a care home being a member of an advocacy organisation in NI or a member of ENRICH in England. Further attrition was noted at this point in the NI sample.

The number of beds in the care homes ranged from 32 to 86, with almost 100% occupancy in all of the homes apart from one in NI. In the case of this one care home, one floor containing 22 beds was closed for refurbishment from December 2016 to December 2017.

The number of general practices providing care to each home reflects the organisation of primary care services in care homes in different parts of the UK. In NI, residents will often remain registered with the general practice that they attended prior to admission to a care home. In England, residents, on admission to a care home, will often register with a practice that has an exclusive arrangement with the home. Hence, care homes in NI tend to be associated with more general practices, and this was reflected in the characteristics of the participating homes. 79 All care homes were associated with only one community pharmacy, which also reflects usual practice. 80

The sampling and recruitment process achieved the aim as we recruited the requisite number of homes. The inclusion criteria do not appear to have been unduly restrictive, apart from geographic proximity to Belfast in NI. This would be removed in the case of a larger study, providing a larger sample from which to recruit. We could also consider focusing on care homes that have an interest in research. Clearly, this can be an advantage, but may also introduce an element of bias in selection. However, if appropriate support could be provided to care homes in a larger study, this may also increase the sample available for recruitment. Findings from the survey reported in Chapter 7 suggested that there was interest and support from care home managers for a larger study.

Adaptation and development of the intervention

In this phase of the study, we set out to adapt and update the decision-making algorithm and training material developed for the original Canadian study by Loeb et al. 16 for implementation in the feasibility study. This was achieved through a series of rapid reviews of the literature, a consensus exercise, a series of focus groups and interviews, and continuous iterative review undertaken by the research team. This comprehensive approach ensured that a robust and rigorous approach was taken to updating of all material, while also navigating through the tension between published evidence and ingrained clinical practice. This phase also generated a huge number of data, which represented a challenge in terms of synthesis and producing an updated decision-making algorithm that would be practical and feasible to use in a busy care home environment. Box 5 illustrates the key components of the original Canadian intervention compared with the components of REACH.

In terms of the rapid reviews, it was striking that there were relatively few new publications in the field of managing the three key infections. The reason for the exclusion of most papers was that they did not relate to the older population or the care home setting, or that they did not offer any new evidence regarding management of the three target infections. Six papers that met the inclusion criteria were consulted and used in the updating of the decision-making algorithm. Most focused on UTIs, which may not be surprising as this is the most common infection type in care home residents. 4 There was no new evidence regarding the management of SSTIs. The most recent papers had been published in 2013. 46,47 The lack of more recent papers may be due to there being a relative paucity of research in the care home population and the difficulty in establishing definitive diagnostic criteria for infection in this population. Three papers referred to change in mental status as being important in the identification of UTIs;42,46,47 however, this is difficult in a population in which cognitive impairment is highly prevalent. This led to discussion within the research team regarding whether or not two management ‘pathways’ should be presented for UTIs (i.e. one for those with dementia and one for those without dementia). However, it was agreed that it would be more straightforward and practical for staff for a single ‘pathway’ to be presented. This was also the view in respect of three separate algorithms for the three target infections. Again, to facilitate ease of use, it was agreed that a combined algorithm, with a common starting point, followed by three ‘pathways’ for each of the infections would be the preferred form of presentation.

The consensus exercise was not part of the original application submitted to NIHR. The decision to conduct such an exercise arose through discussions within the research team, and was seen as an additional method to assist us in deciding on the content of the updated decision-making algorithm. Separate ethics approval for the consensus exercise was obtained and granted. The participants came from a range of backgrounds, but all had experience of the care of older people and/or management of infectious diseases. The consensus approach allowed for a consideration of the key symptoms that could be included when updating the algorithm. Temperature was identified as important, but its interpretation was seen as problematic in the care home population as a rise in temperature was not always indicative of infection. 49,50 As had been identified in the literature, confusion and cognitive impairment were recognised as difficult in care home residents, and the participants in the consensus exercise recommended that staff should be aware of changes in behaviour in residents as these may be suggestive of infection rather that confusion per se. It was also the view of the participants that a single decision-making algorithm with three ‘pathways’ for the three target infections was the most practical approach for use in the implementation phase.

The qualitative aspects of this phase (focus groups and interviews) generated rich and complex data, which were managed through careful analysis. The participants included key stakeholders: a range of staff from care homes, GPs and family members. All could see the value in the use of the decision-making algorithm, and, in many cases, care home staff reported that it reflected their usual practice. The findings also highlighted aspects that had been previously raised through the rapid reviews and the consensus exercise: the challenges presented by confusion in this population and concerns regarding the interpretation of temperature. There was the additional challenge in that staff in residential care homes did not usually measure temperature as this was viewed as a ‘nursing’ task. However, this was overcome by providing the necessary training to these staff, as outlined in Chapter 5. Many of the participants also suggested a range of other symptoms that they thought should be included in the decision-making algorithm. However, the inclusion of all suggestions would have produced an illegible and probably unmanageable tool to use. There were also interesting contrasts between evidence and usual practice informed by experience. Care home staff frequently reported on the smell of urine as being indicative of infection, but SIGN, in particular, had not included this within its most recent publication. 43 Other publications (albeit ones that did not meet the criteria for inclusion in the rapid review) have highlighted this symptom as being problematic in terms of evidence for a UTI. For example, Midthun et al. 81 evaluated freshly voided urine from care home residents for the presence of odour compared with urine analysis and culture results. The positive predictive value of odour was 54% for bacteriuria and only 28% for situations in which pyuria was present with bacteriuria.

Involving key stakeholders was challenging as it elicited a huge body of data and views that contradicted published evidence, which is discussed further later in this section. However, it may have engendered a sense of ownership of the decision-making algorithm and encouraged its use during the implementation phase of the study. A number of publications have highlighted the importance of the role of co-creation in terms of promoting practice change and generating ownership of proposed changes. Goeman et al. 82 showed the value of a co-creation approach in developing a model of care for dementia support in culturally and linguistically diverse communities. This required input from community aged-care services, consumer advocacy organisations and ethnic community group representatives to develop and refine the dementia model of care. In contrast, Beck et al. 83 reported that nursing home managers did not view advance care planning as part of their role, with lack of ownership having an impact on current practice behaviours. Similarly, a broad range of health-care professionals who were interviewed regarding delivering appropriate care to patients with multimorbidity and polypharmacy did not use or were unaware of structured approaches, such as decision aids, to support activities in improving care. This was partly attributed to the participants avoiding ownership of multimorbidity management. 84

The rapid reviews, consensus exercise, focus groups and interviews contributed to the ongoing discussions and work of the research team, which was refining and updating the decision-making algorithm on an iterative basis. This was not a linear process, as a number of the activities were overlapping. The 3-day meeting to undertake data extraction from papers and to identify relevant papers took place in June 2016; the consensus exercise took place in September 2016 and the interviews and focus groups took place during September 2016 to March 2017. As data emerged from these activities, the decision-making algorithm was refined, altered and refined again until the final version, as shown in Figure 7, was produced. This was a culmination of discussion within the research team, as it attempted to produce a tool that would have relevance in the unique context of a care home and reflect practice as far as possible while incorporating the most up-to-date evidence. This was particularly challenging as data emerging, notably from the staff focus groups, reflected practice that was often not supported by evidence, as previously highlighted. This has been seen in other examples of practice in care homes (e.g. person-centred interventions in care homes). Fossey et al. 85 found that despite a range of evidence-based support materials to promote person-centred care, many interventions that were being employed did not meet recognised quality standards, and few had been evaluated in trials. In the management of residents with dementia who displayed symptoms of agitation, the role of non-pharmacological interventions was advocated based on evidence, but for consistent and long-term implementation, staff training was required. 86 Training has been identified as a means by which care home staff can become empowered in their roles, to facilitate clinical reasoning and critical thinking. 87 Therefore, as part of this phase of the project, we developed a comprehensive training programme that would suit a range of staff and provide a number of modes of delivery.

Although conventional in format (a presentation and study manual), we tried to make the material as interactive as possible. Evidence underpinning the content of the algorithm was incorporated into slides and text, highlighting the key points and where referral to a GP would be required. Case scenarios were also included and would provide staff with an opportunity to compare and contrast practice with and without the use of the decision-making algorithm, and with and without the use of the SBAR tool. In order to facilitate as many staff as possible, other modes of delivery of the presentation were also made available. Access to up-to-date and easily accessible information was found to be important in supporting hospital-based nurses in antimicrobial stewardship programmes,4,88 and required the development of a specific app (application). We supplied the training material in a number of formats to facilitate access, and staff were also given a copy of the study handbook.

In the original submission to NIHR, we had planned to involve GPs in intervention implementation. However, based on the advice of the independent members of the SSC, we did not proceed with this. The SSC were of the view that training GPs and care home staff would have been overly complicated in terms of intervention delivery. We consider that this was an appropriate course of action for this study. The focus of the study was on the care home and its staff, as staff instigate contact with a practice when an infection was suspected. 4 We are aware that GPs currently issue most prescriptions for residents within the care home settings and would be logical targets for an intervention. GPs did have an opportunity to contribute their views on the decision-making algorithm, and were informed that the study would be taking place in homes to which they provided care. However, as reported in the process evaluation and the survey to care homes (described in Chapters 6 and 7, respectively), staff thought that GPs and, indeed, family members should be involved in the training aspect of a future study. How this would be implemented would need to be carefully considered, in view of the scale of a larger project and the time and resources required.

Implementation

The implementation of the intervention included monitoring training attendance and the use of the decision-making algorithm. In addition, we also set out to collect data pertaining to a number of key outcomes, notably the acceptability of the intervention in terms of recruitment and delivery of training, feasibility of data collection from a variety of sources, the feasibility of measuring appropriateness of prescribing and a comprehensive overview of the implementation of the intervention. We achieved many of these outcomes, but there is still uncertainty regarding data collection from a variety of sources, notably from administrative databases, and the feasibility of measuring appropriateness of prescribing.

The training was well attended, with a total of 87 staff from the six care homes receiving training from the REACH team, which delivered 21 training sessions over 35 hours. Providing the training in the alternative format of a DVD was useful for those staff who were unable to attend scheduled sessions in the homes, and this was overseen by the REACH Champions.

Process evaluation

The process evaluation aims were to comprehensively describe the implementation of this intervention, including facilitators and barriers, and to develop a set of transferable principles regarding the intervention to inform its implementation on a wider scale. Our purpose in this feasibility study was to generate data that would help us decide if we should proceed to a larger randomised study. Indicators for success or failure in this regard related to our ability to recruit care homes (discussed in Chapter 3), engage with and train staff in the homes, implement an intervention and collect data. Process evaluation results, in general, reveal that we were successful in meeting these indicators. In Chapter 3, we demonstrated that we recruited the requisite number of care homes in both NI and England. The care homes were in various settings, both rural and urban, with residents reflecting the general mix of residents in care homes across NI and England. 96

We also delivered a training package to a substantial number of the key staff in these homes and provided information about AMR and possible antimicrobial overuse in an older person care setting. This also provided an opportunity to introduce them to the updated and refined decision-making algorithm, which was then used in the implementation phase of the study. Feedback on the training shows that it was generally well received. A larger study of longer duration may require more frequent training to ensure continued engagement with the use of the decision-making algorithm.

The results of the normalization process theory analysis gave us an interesting insight into the implementation of the decision-making algorithm into practice over a 6-month period.

In the construct ‘making sense’ (coherence), the findings demonstrated a range of views. Pre-intervention interviews with relatives and some of the care staff reflect the findings of a number of recent publications, with some participants not understanding the problem (AMR) and others being very knowledgeable. 97 This suggests that more work may need to be done within the homes and with relatives to raise awareness and knowledge. Interviews post intervention revealed that staff were more aware of the issues around AMR and could see that there was a need to ensure that GPs prescribed an antimicrobial only when it was needed. They also noted that the decision-making algorithm was useful in the care home but they were unsure if it would change how GPs prescribed.

The analysis revealed that ‘engagement and commitment’ (cognitive participation) was generally high. Care home managers felt that being involved helped to empower the staff to increase their knowledge for the benefit of the residents. Empowerment is the process of enabling others to do something, to make them feel free to act on their own judgement and to trust their own decisions. 98 Empowerment contributes to each carer’s sense of worth and inspires greater aspirations. 99,100 A number of previous studies have been undertaken on ‘structural empowerment’, defined as the presence of social structures in the workplace that enable people to accomplish their work in meaningful ways. 98,101 It appears from our results that the intervention conferred a sense of empowerment among the carers and nurses. Evidence shows that a sense of empowerment can facilitate the implementation of beneficial practices. 102–104

Appointing a REACH Champion in each care home was a challenge for two of our six care homes, but, generally, our analysis suggested that REACH Champions were very important in helping to ensure the engagement and commitment of the staff in the homes. Little is specifically reported about the effectiveness of such champions in supporting the implementation of study interventions, but they are seen as invaluable in some research settings. 105 Problems were also reported regarding having time to engage in the research activities (e.g. completing paperwork) when staff were very busy with day-to-day care activities. This burden will need to be reviewed for any future study. Our training was seen to be very useful for most staff in terms of securing engagement, but some nurses did suggest that they were aware of all of the issues and others felt that we should not have included junior care staff in our training. Although acknowledging these views, we stand by our inclusive approach as we clearly demonstrated the need for increased knowledge of AMR. Including junior staff who cared for residents in the training appeared to raise their awareness to alert a senior member of staff if they observed a significant change in a resident’s status, and this was demonstrated in the implementation phase. Training in care homes is challenging but can bring about changes to practice. 106

In ‘facilitating the use of the REACH intervention’ (collective action), our results were somewhat inconsistent. There was evidence that many staff were implementing the decision-making algorithm but others were not. In some cases, staff forgot to refer to it; in others, staff reported not having time to use it. In the case of the latter, this was partly related to staff having to complete additional forms if they used the decision-making algorithm. It seems that REACH provided access to learning for the teams within the care homes, but implementation was mediated by the context of working in a care home environment, as seen in other studies. 107 Finding the time to participate in an intervention and complete associated paperwork is a challenge that is encountered in many studies; other studies have also found that such tasks conflict with day-to-day resident care activities. 108

The staff were very willing to provide feedback on the decision-making algorithm, particularly with regard to some of the symptoms that had been included following the adaptation and development phase. Issues included not seeing or being able to identify particular symptoms that we had identified as evidence based. Some staff also noted that their own knowledge of the resident (through close contact from delivering care) was important and, therefore, they did not use the decision-making algorithm. Interestingly, a study undertaken in a nursing home setting109 concurs with our finding that staff knowledge of residents was the main driver for decisions about their care; research evidence did not feature highly. If, as some of the findings suggest, our evidence-based tool was being ignored because it did not reflect staff’s everyday practice, we may need to consider how best to embed and convey evidence-based practice in training. The issue is complex as staff do know their residents well and this is invaluable in recognising the onset of new symptoms, but this should not preclude the use of the decision-making algorithm in making an informed decision. We also have to be cognisant of the comments provided by staff relating to the evidence-based symptoms within the decision-making algorithm that were seldom seen or difficult to observe in a care home population.

Some staff seemed to be a little confused regarding how the decision-making algorithm was to be used. There appeared to be a number of staff who did not realise that any member of staff could be involved with the recognition that a resident was unwell (through the use of the ‘trigger’ symptoms at the start of the decision-making algorithm). Instead, they considered that subsequent actions were dependent on their role, with the more junior staff relaying findings to a senior member of staff. In all of the care homes in this study, if a GP contact was indicated, this was initiated by the senior staff (nurse, carer or manager). Some staff noted that working through the decision-making algorithm helped them when communicating with relatives as it was a clear pathway documenting their decision process. However, some staff felt that relatives would overrule decisions they had made. Medical situations involving decisions with relatives about ill residents are very challenging. Trust and confidence can be seriously challenged and conflicts may arise. 110,111 Van Keer et al. 112 suggest that families may lack understanding of a prognosis and have unrealistic expectations, perhaps based on media misinformation. This is a complex issue as conflict between staff and relatives is likely to have an impact on the care of the resident. In terms of testing the REACH intervention further, engaging with family members on the topic of AMR and the use of the decision-making algorithm may improve knowledge and acceptance. For staff, training on how to handle discussions with relatives may also be useful.

Our normalization process theory analysis of the ‘value of the intervention’ (reflexive monitoring) reflected a more negative outcome than in the other constructs. Although most participants believed that it was a good idea, operationalising it was more problematic. Some reported that junior staff lacked the skills and competence to use the tool and some nursing staff considered that it was beyond the accepted role of some levels of staff. Again, the workload issue of time associated with intervention implementation and documentation was highlighted. There was, however, evidence that some staff fully understood how to implement the decision-making algorithm and how and when to refer to a senior member of staff. The problems of task orientation and a strict division of labour are increasing concerns for many workplaces, including nursing homes. 113 Task orientation can be defined as work that is highly focused and prioritised on the completion of tasks, without adequately considering the outcomes of the tasks completed. 113 For example, in a care home, dining might be completed at a designated time and baths are given when time allows, but these do not necessarily take place in reference to the individual needs of the residents. Rather, these activities take place in accordance with a work schedule that breaks down care into discreet tasks. 114 Task orientation is in marked contrast to relational work. Relational work involves co-operative task sharing, more flexible regulations, limited restrictions on work and increased work autonomy. 115 Relational work is beneficial to workers through stress reduction and beneficial to residents through improvements in their care. 115 Therefore, we need to consider and address these issues before trying to implement REACH on a larger scale.

When presenting normalization process theory, May and Finch21 note that there is an interplay and non-linear relationship between the four constructs. Studies using normalization process theory, when identifying challenges or crucial drivers in the implementation, highlight just one of the constructs. 116–118 Something we did not see in our study was a lack of coherence, which is often cited as an important challenge in the implementation (i.e. the intervention does not make sense or is met with conflicting attitudes). 116,117 In this study, reflexive monitoring highlighted a number of challenges; although REACH was valued by staff, it was not fully implemented. In a feasibility study, this may be expected as the implementation period is short, whereas in other studies, such periods have been longer, allowing for a new practice to be normalised. 97 It has been argued that for practices to become accepted, integrated, and sustained in day-to-day work, they must be experienced as dealing effectively with real everyday problems. 119

Limitations to this study are addressed elsewhere in the report; however, it is important to note here that only five of the six homes fully embraced the implementation of the intervention. The residential home in England experienced a number of problems during the study period, which made it difficult to achieve complete engagement. Staff were trained but they did not record the use of the decision-making algorithm, in contrast to the other homes.

A future study should consider the items in Table 33, which are based on our findings, presented against components of normalization process theory, to help overcome some of the issues identified from this feasibility study. These could help to ensure the success of implementation on a larger scale.

Care home survey

A postal survey was undertaken with a sample of care homes in NI and the West Midlands to assess the likelihood of care home participation in a future trial evaluating the effectiveness of the REACH intervention. The overall response rate was 23%, but there was a marked difference in the response rates between the NI and West Midlands samples, which were 47% and 12%, respectively. It is unclear why there was this differential as the questionnaire was quite short (seven questions), requiring ‘yes’ or ‘no’ responses, with free-text answers being optional. Therefore, responses should be considered in the light of this overall response rate, recognising that there will be an element of response bias, and that those who did respond are likely to be those who are interested in this topic. Moreover, to some extent, the positive results would support this. It was encouraging to note that those who did respond were generally supportive of a larger study, recognising the importance of the topic. Reflecting some of the findings from the process evaluation, a number of respondents suggested that the training should be extended to GPs and, indeed, family members. Most agreed that the duration of training (2 hours) was a reasonable period of time but highlighted the importance of flexibility in scheduling of training to maximise attendance. Respondents also considered that a ‘decision-making aid’ was useful but highlighted the range of tools that were already being employed within care homes. Overall, 79% of respondents were, in principle, prepared to participate in a future study. This is encouraging, and provides some evidence that we would have a sample from which to recruit to a larger study.

Reflections on patient and public involvement

This was a challenging study and PPI helped us navigate through some of the difficulties, particularly in respect of data collection and its impact on staff.

A group that had been convened in advance of the study advised on the content of study documentation before submission for ethics approval. The input received ensured that the language was clear and unambiguous. We had hoped to be able to involve residents in this process, but this could not be achieved as the Independent Health and Care Providers were unable to identify anyone willing to take part.

Mr Bob Stafford was a member of the research team and actively participated in all meetings. His insight into care home practice helped us approach issues with staff as to how to balance the demands of the study with the immediate demands of everyday work.

In many ways, the participating care home staff acted as research partners, particularly in the adaptation of the intervention and its implementation. It was a challenge in respect of the development of the decision-making algorithm as the evidence did not reflect usual practice. As outlined in the section Process evaluation, data collection was seen as a burden for staff, but some did actively engage in helping to refine the data-collection forms or devised systems that would make aspects of the study more practical for them. These types of innovations are useful when considering how best to engage with care home staff in future studies.

Strengths and limitations of the research

The study has a number of strengths that should be acknowledged:

• We have developed and updated a decision-making algorithm to help guide care home staff in the management of the three most common infections observed in this environment. The algorithm was generally well received by staff, as was the accompanying training programme.

• Stakeholder involvement, particularly in the development and updating of the algorithm, was comprehensive and generated rich data and important insights, particularly from care home staff who would use the decision-making algorithm.

• Staff appeared to recognise the importance of AMR in the context of their everyday work and were enthusiastic about implementation of the intervention.

• The process evaluation generated a rich and wide-ranging understanding of the facilitators of and barriers to implementation in a busy care home environment.

• We were able to confirm many aspects of feasibility relating to recruitment, data collection and implementation.

• The results of the survey to a sample of care homes in NI and England suggests support for a larger study.

However, there are a number of limitations that need to be taken into account when considering our findings:

• This was a feasibility study and no assessment of efficacy of the intervention can be made.

• The feasibility study and process evaluation took place in six care homes, with three homes in each jurisdiction. Therefore, all findings must be interpreted in the light of this small sample. It is also likely that a number of these homes were highly motivated to participate, and so cannot be considered representative of UK facilities.

• Although the process that we undertook to develop and adapt the intervention (decision-making algorithm and training) was extensive and comprehensive, the number of participants in the consensus exercise was small (n = 4). The focus groups and interviews were extremely valuable in generating rich data to contribute to the adaptation process, but caveats associated with qualitative work must be considered (i.e. findings may not be generalisable and, as noted, the participants were drawn from a small number of homes and practices). However, careful consideration was given to reflexivity and standard approaches to data analysis and interpretation.

• Not all staff attended the training sessions provided in their respective care homes. However, the availability of other modes of delivery, such as the DVD, was useful.

• Data collection was challenging. The lack of a unique patient identifier prevented the tracking of individual residents and collection of resident-level data. It also prevented us from accessing administrative data, notably in the English context. It was also noted that data collection was perceived to be a burden by staff. We attempted to streamline this as far as possible. We cannot be confident that we have a complete record of the use of the decision-making algorithm. We wanted to collect these data as a measure of process, but in any future larger study, it is unlikely that we would collect such data.

• Staff struggled with reconciling the evidence presented in the training and the guidance within the decision-making algorithm with usual practice, which led to the algorithm being over-ruled in a number of cases. This applied more often in the management of UTIs. However, staff also reported that the decision-making algorithm broadly reflected practice.

• General practitioners did not participate in any post-implementation interviews, so we are unable to comment on how they perceived the intervention. They were also not the focus of the intervention (following advice from the independent SSC). We recognise the role that GPs play in the prescribing of antimicrobials, but previous research has shown that care home staff initiate the prescribing process through contact with GPs.

• The response rate to the survey to gauge interest in a larger study was 23%. Therefore, the findings need to be interpreted carefully.

Implications for practice

This was a feasibility study to assess various elements of research methodology and possible progression to a larger trial, so implications for practice at this stage are somewhat limited. However, the following points are worthy of consideration:

• Training for care home staff was an important aspect of this feasibility study and would also be a key part of a definitive trial. Being able to integrate training into everyday practice and shift patterns was a challenge in the study, and would also appear to be difficult outside a research context. More generally, care home organisations should consider how best to provide and facilitate training events and opportunities for their staff to ensure that their practice is up to date and evidence based.

• It was accepted practice in care homes without nursing not to measure temperature; this would have been challenging for the implementation of the intervention. However, we obtained agreement from the management of such care homes to allow us to train staff to undertake this task during the course of the study. Allowing this to be part of everyday practice in care homes without nursing would be beneficial for staff (and indeed residents) outside the research context.

Recommendations for future research

As a result of these findings, we consider that we have achieved many of the objectives that were initially established for this study. We have demonstrated that we can recruit homes, oversee implementation and collect data. However, there are a number of key issues that need to be highlighted to allow a future study to proceed:

• We need to consider how best to minimise the data-collection burden on care home staff. Over the course of the data collection, forms were refined and it may be possible to further refine, or indeed omit, data collection for some variables. It may also be possible for research staff to undertake data collection (with the appropriate approvals to do so) in consultation with care home staff.

• There is a need to obtain resident-level data from care homes and other sources. Consideration regarding obtaining individual consent or employing an ‘opt-out’ approach may be the best course of action to obtain the data that would be needed. General use of administrative data sources is being advocated by research funding bodies, but this was very difficult to achieve in this study, as has been previously described.

• The content and focus of the intervention may need to be reviewed in the light of antimicrobial use for prophylaxis in the case of UTIs, and how evidence-based management of UTIs can be best promoted in UK care homes.

• Further consideration and guidance should be produced in respect of the most appropriate outcome measure to assess the effects of antimicrobial stewardship interventions, with a focus on a ‘prescribing outcome’.

• Progressing to a full RCT could be considered, but only with the inclusion of an internal randomised pilot study and the application of robust progression rules. Further consideration should also be given to whether or not GPs should be targets for the intervention and how best to engage all care home staff in the implementation of the intervention approach.

Final conclusions

Based on our findings, we draw the following conclusions:

• We have demonstrated feasibility in respect of recruitment, data collection and implementation of the intervention, although challenges remain with respect to accessing centralised administrative data.

• Stakeholder involvement in the adaptation and development of the intervention was challenging, but it was also valuable as it provided an important perspective and may have engendered a sense of ownership of the intervention, particularly among care home staff.

• We were able to collect dispensing data and health economic data, all of which would be critical for a larger study.

• The intervention appeared to be broadly acceptable to care home staff, and could be integrated into everyday practice; measurement of temperature in care homes without nursing would need to be considered if a larger study was to be undertaken.

• A larger definitive trial, using a cluster randomised design, with inclusion of an internal pilot study and appropriate progression criteria may be considered. However, individual resident consent may be the best option in order to obtain the data required.

Contributions of authors

Carmel Hughes (Professor of Primary Care Pharmacy, Queen’s University Belfast) was the chief investigator, led the study design, oversaw the whole study and is lead author of the report.

David Ellard (Principal Research Fellow, Warwick Clinical Trials Unit, University of Warwick) was principal investigator for the study in England. He was involved in the conception and development of the study, was responsible for oversight for the English part of the study and for the process evaluation, contributed significantly to the development of the protocol and the study’s processes and paperwork and was involved in analyses and interpretation of results. He contributed to the writing of the final report.

Anne Campbell (formerly Research Fellow, Queen’s University Belfast, during the course of the study, subsequently Research Associate, Imperial College London) had day-to-day responsibility for project management in NI and was involved in all aspects of the study. This included recruitment of care homes, recruitment for and conducting of pre- and post-intervention focus groups and interviews, rapid review of the literature, organisation of training of care home staff, observation visits and support of care home staff with all aspects of research activity during implementation, data analysis and administration of the care home survey. She was a co-author of the report.

Rachel Potter (Senior Research Fellow, University of Warwick) had day-to-day responsibility for management of the study in the West Midlands and was involved in all aspects of the study, including conducting the focus groups and interviews, organisation of the training of care home staff, observation visits, support of care home staff with all aspects of research activity, data collection pre, post and throughout the implementation of the intervention, administration of the care home survey and analysis She was a co-author of the report.

Catherine Shaw (Research Fellow, Queen’s University Belfast) conducted rapid review of the literature, updated and expanded the algorithm, developed and delivered training material and was a co-author of the report.

Evie Gardner (Head of Statistics, NI Clinical Trials Unit) was a co-applicant, conducted the statistical analysis and prepared the results for publication, and was a co-author of the report.

Ashley Agus (Health Economist, NI Clinical Trials Unit) was a co-applicant, conducted the economic analysis, the analysis of health service use and prepared the results for publication, and was a co-author of the report.

Dermot O’Reilly (Senior Clinical Lecturer, Queen’s University Belfast) led on the aspects of the study related to the potential use of administrative data in the monitoring of study outcomes and reviewed the final report for critical academic content.

Martin Underwood (Professor, Warwick Clinical Trials Unit, University of Warwick) was part of the team that developed the proposal, he contributed to data collection and analysis and he reviewed the final report for critical academic content.

Mark Loeb (Professor, Infectious Diseases, McMaster University) advised on the conduct of the study.

Bob Stafford (Group Head of Care, Orchard Care Homes) was a co-applicant, provided PPI input from a care home operator’s perspective and was a co-author of the report.

Michael Tunney (Professor of Clinical Pharmacy, Queen’s University Belfast) was a co-applicant, he advised on the conduct of the study and he reviewed the final report for critical academic content.

Publications

Potter R, Campbell A, Ellard DR, Shaw C, Gardner E, Agus A, et al. A multifaceted intervention to reduce antimicrobial prescribing in care homes: a process evaluation of a non-randomised feasibility study. BMJ Open 2019;9:e032185.

Hughes CM, Ellard DR, Campbell A, Potter R, Shaw C, Gardner E, et al. Developing evidence-based guidance for assessment of suspected infections in care home residents. BMC Geriatr 2020;20:59.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

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 and Social Care. 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 and Social Care.

Database: Ovid MEDLINE(R)

Date range searched: 1946 to April week 3 2016.

Date searched: 19 April 2016.

Housekeeping (15 minutes)

• Collect consent and arrange for copies to be given to participants (if not already done).

• Get contact details of participants so that payment can be made (if not already done).

• Introduce ground rules for focus group regarding (1) practicalities (need to record session, role of researcher in guiding discussion but allowing participants to take up the discussion in their own terms, importance of not talking over one another and encouraging people to voice their opinions) and (2) ethics considerations (seeking agreement regarding voicing disagreements in a reasonable way and maintaining confidentiality outside the group).

Introduction (5 minutes)

• Introduce study and give very brief overview of decision-making aid and data-collection forms that will be used by staff in the study (mention the training they will receive). Make clear that the decision aid is used to help care home staff decide whether or not to contact the GP when they think a resident might have an infection.

• Check whether participants have any questions before beginning focus group and recording.

• Begin recording.

Questions (approximately 70 minutes)

Housekeeping (15 minutes)

• Collect consent and arrange for copies to be given to participants (if not already done).

• Get contact details of participants so that payment can be made (if not already done).

• Introduce ground rules for focus group regarding (1) practicalities (need to record session, role of researcher in guiding discussion but allowing participants to take up the discussion in their own terms, importance of not talking over one another and encouraging people to voice their opinions) and (2) ethics considerations (seeking agreement regarding voicing disagreements in a reasonable way and maintaining confidentiality outside the group).

Introduction (5 minutes)

• Introduce study background, making clear that the decision aid is used to help care home staff decide whether or not to contact the GP when they think a resident might have an infection.

• Check whether participants have any questions before beginning focus group and recording.

• Begin recording.

Questions (70 minutes)

Decision aid (approximately 30 minutes)

1. Usual practice

1.1. How do you usually find out about a resident of a care home who may have an infection?

1.2. How do you think care home staff currently assess whether a resident has an infection that needs an antibiotic? [Prompt: what signs or symptoms do they prioritise? (e.g. temperature, behaviour change, etc.) What signs or symptoms would you like them to prioritise?]

1.3. How do care home staff usually convey their assessment of the resident to you? [Prompt: do they use any particular method of communicating this assessment to you e.g. SBAR? (Situation: Background: Assessment: Recommendation.)]

1.4. What influence do care home staff or family members have on any decision you might make regarding prescribing antibiotics for residents of care homes?

1.5. What else influences any decision you make regarding prescribing antibiotics for residents of care homes? (Prompt: knowledge of the resident and their history of infection; whether you are able to make a visit to the resident.)

2. Using the decision aid (show aid and let think briefly about it on own)

2.1. Imagine the care home staff have a resident with a suspected infection. How do you think the decision aid will actually work in practice?

2.2. How easy or difficult is it to follow?

2.3. In which ways do you think using the decision aid will help staff decide when to contact the GP?

2.4. In which ways do you think using the decision aid will help staff convey their assessment of the resident to you?

2.5. Which staff do you think would use the aid? (e.g. nurses, senior care assistants, junior care assistants).

2.6. Are there particular circumstances in which you think it will be easier or more difficult to use the tool than others (e.g. with particular residents, infections, times of day/year)? Why?

2.7. Do you foresee any problems in using this tool during an outbreak of a respiratory virus? What changes could we make to the aid to prevent this?

2.8. Is there anything missing? (Prompt: symptoms/signs. Anything confusing?)

2.9. What concerns do you have about care home staff using it?

2.10. How do you think using the decision aid will change how care home staff usually assess whether a resident has an infection that needs an antibiotic?

2.11. Is there anything else you would like to say about the decision-making tool?