Health literacy interventions for reducing the use of primary and emergency services for minor health problems: a systematic review

Article history

The research reported in this issue of the journal was funded by the HSDR programme or one of its preceding programmes as project number NIHR131238. The contractual start date was in November 2020. The final report began editorial review in January 2022 and was accepted for publication in May 2022. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HSDR 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.

Copyright statement

Copyright © 2022 O’Cathain et al. This work was produced by O’Cathain et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaption in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2022 O’Cathain et al.

Demand for primary and emergency care in the UK

There is considerable pressure on general practices, emergency departments (EDs) and emergency ambulances in the UK. In 2019, the Royal College of General Practitioners highlighted intense resource and workforce pressure on general practice out-of-hours services and expressed concerns about decreases in the number of general practitioners (GPs) per head of population, despite rising demand for general practice. In 2021, the Royal College of General Practitioners highlighted the lack of GPs to meet demand. 1 In 2019, policy-makers reported on the NHS England website that there were 110 million urgent patient contacts in England each year, with around 85 million being same-day GP appointments, and the rest attendances at EDs and minor injury units. NHS England estimated that between 1.5 million and 3 million attendances at EDs could be managed by other parts of the urgent care system. Emergency ambulance services deal with around 11 million calls annually, and around one in eight are dealt with through advice from a clinician by telephone. 2 Patients having these ‘hear-and-treat’ calls, that is calls that do not result in an ambulance being despatched, are offered self-care advice or are recommended to contact lower-acuity services. NHS England is currently encouraging people to use NHS 111 to help them make decisions about which service, if any, to use, including self-care and pharmacy. Demand for NHS 111 also outstrips supply, and policy-makers have attempted to ease demand for this service by encouraging people to use NHS 111 Online to help them make decisions about where in the NHS to seek care. 3

There is a need to consider ways of reducing pressure on emergency and primary care services. Although primary care is a setting that provides a first point of contact in the NHS for patients, and includes general practices, community pharmacies and dentists, a focus on general practice is important because of the current pressure on this service.

Minor health problems

Some people need to use emergency and urgent care, and it is important to maintain access to services to meet their needs. However, some people do not need the clinical resources or acuity level of primary or emergency care services. Some patients could contact their GP instead of calling for an emergency ambulance or attending an ED. Other patients could self-care or seek advice from a pharmacist instead of contacting general practice, the ED or the ambulance service. A variety of terms are used to describe this type of service use, such as ‘clinically unnecessary’,4 ‘primary care sensitive problems’ in emergency care,5 non-urgent or low acuity. The prevalence of this type of use varies depending on the definition used, but it is likely to be high. For example, 15% of attendances were classed as non-urgent in an analysis of routine data from EDs in the UK. 6

Some of these contacts in primary and emergency care are for minor health problems, injuries or ailments that could be managed through self-care or contact with a pharmacist. Examples of minor health problems include raised temperature in a child, musculoskeletal pain, eye discomfort, gastrointestinal disturbance and upper respiratory tract-related symptoms. 7 Research has shown that people attend EDs, general practice or a pharmacy with these minor health problems. When contacts with different services for similar minor health problems are compared, they have similar resolution rates, but the cost per consultation is higher for EDs and GPs than for pharmacy: £147 and £82, compared with £29, respectively. 7 People may also attend general practice out-of-hours services, as well as daytime general practice, for minor health problems. 8

Reasons why people attend primary and emergency care with minor health problems

There are a wide range of reasons why people contact services for minor health problems. A recent study of ‘clinically unnecessary’ use of ambulance services, EDs and GPs in the UK explored this issue using a realist review, qualitative interviews with patients and a national population survey. 4 Ten reasons were identified in the realist review9 and qualitative interviews:10

1. uncertainty about symptoms, causing anxiety

2. heightened awareness of risk as a result of experience or knowledge of traumatic health events, leading to anxiety

3. fear of consequences when responsible for others such as children or elderly people

4. inability to get on with daily life, such as going to work or looking after children

5. need for immediate pain relief

6. waited long enough for things to improve

7. stressful lives so cannot cope with any health problem

8. following advice of trusted others from NHS staff, family, friends, colleagues

9. perceptions or prior experiences of services, for example positive experiences of EDs lead to further use of EDs

10. perceived poor access to a GP in the time frame wanted.

These issues were measured in a national population survey of the tendency to contact primary and emergency care for minor health problems. Other issues that were less commonly identified in the literature were also measured, including awareness of alternative services, beliefs about how busy emergency and primary care services are and health literacy levels. The effect of people’s health literacy levels on their decision-making was measured because researchers in the USA found that people who attended EDs with non-urgent problems had lower levels of health literacy. 11 It was also possible that poor health literacy might underlie some of the reasons for using primary and emergency care for minor health problems. For example, people with low levels of health literacy might be more anxious about minor health problems because they do not understand what symptoms are associated with serious problems, they may be less likely to know how to manage pain, they may be more likely to look to others for advice, they may not know how long it takes for some minor problems to resolve and they may be unaware of alternative services. When a range of variables were tested in a population survey to explain tendency to contact primary and emergency care for minor health problems, lower health literacy levels were associated with this use of emergency ambulances and general practice, even when all other variables were controlled for. 9,12

The importance of health literacy

Given the role of health literacy in explaining the tendency to use emergency ambulances and general practice, and its relationship with the use of EDs for non-urgent problems, it is important to understand what it is, how it is measured, its relationship with use of primary and emergency care, and how it can be addressed to reduce the use of primary and emergency care for minor health problems.

NHS initiatives to address demand for primary and emergency care

Given that there is a need for multiple interventions to address demand for primary and emergency care for minor health problems, it is useful to consider the range of interventions that policy-makers, commissioners and service providers have introduced. They tend to focus largely on service and workforce reconfigurations that offer alternatives to the three pressurised services of general practice, EDs and emergency ambulances, or on initiatives to manage minor problems efficiently within each of these three services. Examples of these initiatives include provision of a Community Pharmacy Minor Ailment Service, the provision of walk-in centres, the extension of the role of pharmacists, GPs working in EDs, paramedics working in general practices and redirection of people attending EDs. Initiatives also include the national telephone helpline NHS 111 for urgent care whereby non-clinical call-takers direct people to the most appropriate service for their need, or pass callers to clinicians who assess more difficult cases and offer self-care advice or referral to services.

Intervention further back in the pathway of help-seeking behaviour may offer a complementary approach to these initiatives. That is, it may be possible to reduce population need to seek advice from health professionals in pharmacy or NHS 111 for minor health problems and thereby reduce demand for pressurised services. This will be useful at any time, and especially during pandemics, when some people stop attending primary care and emergency services for minor health problems and need to self-care or be able to distinguish between minor and serious problems.

Reviews in related areas

Prior to applying for funding for this systematic review, we undertook a scoping exercise and found no similar systematic reviews. We continued to search for other reviews during the study. We found a number of reviews in related areas, but no review of the same territory.

The need for this review

Although a number of reviews have been undertaken in related areas, and some reviews addressed parts of the territory covered in this review, there was a research gap for a review of health literacy interventions aiming to reduce the use of primary and emergency care for minor health problems.

Aim and objectives

The aim was to identify interventions that reduce the use of primary and emergency care for minor health problems by improving health literacy.

The objectives were to:

• identify the effectiveness, cost-effectiveness and safety of interventions

• identify the contexts in which different types of interventions are effective

• describe interventions

• identify the mechanisms of action of interventions

• construct a typology of interventions relevant to service use reduction

• describe the development process and its impact on effectiveness

• describe the literacy levels of information-based interventions and their impact on effectiveness

• identify patient, carer and staff views on the feasibility and acceptability of interventions

• consider how UK stakeholders could operationalise the evidence in terms of identifying a set of interventions that they would find feasible, appropriate, meaningful and effective within their localities.

These objectives are slightly different from those in the proposal. For example, we originally intended to consider the literacy levels of effective interventions only, but we subsequently decided to do this for all the interventions we could find because we were interested in whether or not the readability or ease of use of an intervention had an impact on its effectiveness.

Design

We undertook a systematic review with stakeholder involvement. First we held a stakeholder event to inform the systematic review, then we conducted a systematic review to identify and synthesise the evidence and, finally, we held another stakeholder event to consider how to implement the evidence. We attended to guidance on good practice when involving stakeholders in systematic reviews. We drew on the Guidance for Reporting Involvement of Patients and the Public 2 (GRIPP2) framework to reflect on our patient and public involvement (PPI). 46 See Report Supplementary Material 1 for the GRIPP2 reporting guidance checklist.

Ethics approval

Some researchers may use research methods for stakeholder involvement, such as Delphi exercises,47 and this requires research ethics approval. Our stakeholder involvement was akin to PPI in that we held workshops to encourage research users to shape and interpret the systematic review. We did not require ethics approval for the stakeholder involvement or for the systematic review.

Patient and public involvement

Background

In the first month of the study (November 2020), we held a virtual stakeholder event. It had to be virtual because of COVID-19. We aimed to involve representatives from the following groups: patients, carers and members of the public; Clinical Commissioning Groups (CCGs); NHS England Emergency and Urgent Care directorate; the NHS England Health Literacy Lead; general practice; EDs; and emergency ambulance services. The aim was to elicit stakeholders’ perceptions of key issues that might further shape the review, such as suggestions about the key types of interventions to include. The virtual format facilitated people from different parts of the UK to attend the event.

Stakeholders attending the event

Fourteen stakeholders attended the event, along with research team members:

• seven PPI representatives (three others provided input outside the event)

• one urgent care commissioner from a CCG

• one representative from NHS England

• one ED consultant

• one GP

• one academic pharmacist

• one representative from NHS Digital

• one representative from the third sector.

We designed the event with our PPI co-applicant Margaret Ogden to ensure that PPI representatives were supported to engage in the event. Our stakeholder lead on the team (AF) contacted all of the PPI representatives by telephone or e-mail before the stakeholder event to discuss the review and to support people to attend by explaining how the virtual platform worked. Alexis Foster also facilitated input from three PPI representatives who could not attend the event. Two did not feel confident with the technology and one person could not make the date. Alexis Foster sent them the notes from the event and then had a telephone call with each of them to discuss the event and get their input. We ensured that we included their viewpoints in the write-up of the event. This additional involvement was important because we did not want to exclude people from the event because of lack of technical skills.

The event

The event was held on 10 November 2020 and lasted 2 hours. It was a mixture of presentations and discussions. Alicia O’Cathain delivered two short presentations (each < 10 minutes) to explain the review, to justify why the research was needed and to discuss how the research fits within research on help-seeking behaviour in emergency and primary care. Jonathan Berry, the health literacy lead from NHS England, also presented on initiatives being delivered within the NHS to increase health literacy.

The group was split into two smaller groups to facilitate discussion, with each small group facilitated by a member of the research team and including a mix of stakeholders. In the discussion we focused on potential factors we needed to consider in the review. We also discussed potentially relevant interventions. Attendees gave us potential sources of grey literature such as work undertaken by the World Health Organization Healthy Cities initiatives, and by specific CCGs in England.

The issues raised and actions taken

After the event we recorded the points made by stakeholders and our comments on how we would address each point in the review (Table 1). We circulated our write-up to the stakeholders so they could add to the document. Comments from people who had been unable to attend the meeting were added at this stage (this is indicated by ‘post meeting’ in Table 1).

Ongoing communication

Following the event, we circulated newsletters to keep stakeholders informed about the progress of the review and to give people the opportunity to provide further feedback.

Design

We undertook a multicomponent review49 of quantitative and qualitative research. We registered the review on PROSPERO (registration number CRD42020214206). The funders played no role in the review.

Methods

Synthesis

Presentation of the studies

One reviewer (CC) reported the included studies in alphabetical order in four tables: (1) context/characteristics, (2) description of the intervention (including mechanisms and typology), (3) service use outcomes and (4) other outcomes. A second reviewer (AOC) read the papers independently and checked the presentation of each study.

Synthesis of outcomes

We used the typology of interventions to identify common types and then we presented a narrative synthesis of service use outcomes for each common type, with data summarised in tables. We ordered studies in these tables in alphabetical order, or, if there was a subset of the type, we ordered by this subset and then alphabetically.

We could not use meta-analysis because of the heterogeneity in how service use outcomes were measured. We did not measure heterogeneity in reported effects because it was clear from the data extraction that there was considerable variation in the way the outcomes were measured and the time periods over which they were measured. We planned to use the typology to synthesise other outcomes and processes, but numbers were too small. Instead we undertook a narrative synthesis of all the studies for each outcome or process. We followed the synthesis without meta-analysis (SWiM) reporting guidelines. 69 The SWiM checklist can be found in Report Supplementary Material 2. When drawing conclusions from the synthesis, we gave weight to studies that used RCTs in preference to other designs, measured the impact on service use using routine data/records rather than self-reported service use, were assessed as being of higher quality, and had greater relevance to the UK because the research was undertaken in the UK or a country with a similar health-care system to that of the UK.

Costs and cost-effectiveness

We knew from scoping the evidence base that there would be few economic evaluations for synthesis. We intended to follow the ‘brief economic commentary’ for synthesis, rather than the ‘integrated full systematic review of economic evidence’ as set out in the Cochrane Handbook. 70 We planned to follow methods used by Angus et al. 71 to summarise any directly reported costs associated with interventions as reported in the literature and convert them to Great British pounds for a specific year. We planned to aggregate these data across intervention types to report costs for comparison and narratively synthesise cost-effectiveness and report the relationship between cost-effectiveness and context. Unfortunately, the cost data and cost-effectiveness analyses were often poorly undertaken or reported. Instead we undertook a simple narrative synthesis.

Qualitative research

We intended to analyse qualitative research using qualitative synthesis methods. 72 There were only four articles reporting qualitative research, focused mainly on digital interventions, so synthesis was limited.

Explaining variation in impact on service use

Ideally we would have identified size of effect of each intervention within a meta-analysis, but, as explained previously, this was not possible because of heterogeneity in outcome measurement. We might also have considered a meta-regression to explain size of effect. 73 Instead, we identified whether or not studies that measured service use in a full evaluation (as opposed to pilot studies, which are underpowered statistically) had statistically significant reductions for impact on use of different services. Then we used IBM SPSS Statistics 27 (IBM Corporation, Armonk, NY, USA) to test whether or not different characteristics of the study, context and intervention explained reductions in service use.

An alternative way of explaining variation in an outcome is to consider the relationship between context, mechanisms and outcomes by identifying context–mechanism–outcome chains associated with understanding messy complex interventions. 74 We did this by considering mechanisms within the narrative synthesis by type of intervention.

Articulating a programme theory

We constructed a logic model to present a programme theory of how characteristics and mechanisms of interventions might produce short- and intermediate-term outcomes that lead to long-term outcomes. We did this to facilitate interpretation of the synthesis (Figure 1).

FIGURE 1.

Logic model of health literacy interventions to reduce use of primary and emergency care.

Results

Included articles

We identified 67 articles that met our inclusion criteria (Figure 2). The decision about inclusion was challenging when studies met some, but not all, of the inclusion criteria. We excluded 38 articles at the full-text stage (see Appendix 3). Articles about frequent users of EDs tended to be excluded at the full-text stage because these studies focused on a population of people identified as having serious problems such as homelessness and substance abuse, rather than minor health problems. Articles about self-management of chronic conditions such as asthma were excluded when they were about prevention of acute severe problems, rather than management of minor health problems. Case management interventions tended to be excluded because they did not focus on minor health problems, but on serious problems that resulted in hospitalisations.

FIGURE 2.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram.

Synthesis by type

Service use

We have summarised the proportions of articles showing a reduction in ED (Figure 3) or GP use (Figure 4) by type of intervention. The numbers in each intervention type are small, so it is not possible to say whether or not some types were more likely to show a reduction in service use. Although navigation tools look like they may be more likely to show reductions, there were a number of poor-quality studies in this type.

FIGURE 3.

Numbers of studies showing a reduction or no change in ED attendances, by type.

FIGURE 4.

Numbers of studies showing a reduction or no change in GP consultations, by type.

1. Navigation types (A.I, B.I and C.I)

Size and quality of the evidence base

Seven studies testing eight interventions aimed to guide people to the most appropriate service (including self-care) for their minor health problem. They included written navigation tools (A.1),91,110 person-delivered navigation tools (B.I),87,91,92,130 digital navigation tools (C.I)123 and a multicomponent written and person-delivered navigation tool (A.I and B.I). 86

The studies focused on interventions that persuaded patients to use primary care rather than EDs. All the studies were undertaken in the USA, where funding systems for health care (e.g. an insurance system) can save money directly if their patients use primary care rather than EDs. The studies mainly used strong designs for testing impact: either RCTs or controlled before-and-after studies. Only one study87 was rated as being of high quality using a formal assessment tool. We also had concerns about the quality of some studies in terms of the analyses undertaken,92,130 clerical errors potentially affecting the data collection,91 lack of intention-to-treat analyses91 and the potential that patients attended a different ED from the one being measured. 87 On a positive note, the studies measured service use using routine records, although they used different time periods for measurement.

Intervention description and mechanisms

These interventions tended to use the mechanisms of ‘raising awareness’ and ‘persuasion’. The person-delivered interventions took an active and targeted approach by communicating the message via ‘patient navigators’87 or health professionals91,92,130 to people who had used an ED for a minor health problem.

Impact on service use

In Table 5, ‘+’ indicates a statistically significant reduction in service use, ‘–’ indicates a null result and ‘±’ indicates that results were mixed. Five of the six articles measuring impact on EDs showed a reduction, although, in two studies,91,92 this was not a simple overall reduction. Racine et al. 92 showed no reduction overall in ED attendances, but then tested the data for types of service use, which showed a statistically significant reduction. Patel et al. 91 showed a reduction when they looked at subgroups, rather than for the sample overall. So strictly, a statistically significant reduction in ED attendances was found in three out of six studies. The studies that measured primary care use had an intention to increase use of this service because the aim of the intervention was to shift users from EDs to primary care. Given that increasing use of primary care was not a relevant aim of our study, we have not indicated that results were + or – in Table 5. There was some indication that different age groups might be persuaded using different delivery approaches: a postal approach for younger patients and a telephone approach for older patients. 91

TABLE 5 - Impact of navigation tools on service use

BA, before and after; CBA, controlled before and after; CI, confidence interval; OR, odds ratio; PCP, primary care provider; PED, paediatric emergency department.

Note

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed.

Study	Design	Quality	Mechanisms	Time period of measurement	Data source	Impact on ED	Impact on GP consultations
A.I Written
Adesara et al.,110 USA	CBA study	Weak	Awareness, persuasion	6 months	Records	All: (+) ED visits attributable to intervention patients – pre = 6.83%; during = 6.22%; post = 6.15%. Approximately 40 ED visits saved per month. Internal medicine clinic: pre = 1.34%; during = 1.69% (difference pre–during vs. intervention p < 0.001); post = 1.43% (difference pre–post vs. intervention p < 0.001)
Patel et al.,91 USA	RCT with three arms	High risk of bias	Awareness, tailoring	6 months	Records	All: (±) no change in the whole sample, but change in ED use differed by age group and intervention type, with reductions for older people getting a telephone call and younger people being mailed information. Telephone intervention vs. matched control outcome (model): 0.92 (0.77–1.10); p = 0.36 [≥ 65 years: 0.78 (0.62–0.99); p = 0.04]. Mail intervention vs. matched control outcome: 1.07 (0.92–1.23); p = 0.40 [< 65 years: 0.73 (0.55–0.98); p = 0.03]	All: PCP call for telephone intervention vs. matched control outcome (model): 0.93 (0.68–1.27); p = 0.63 Mail intervention vs. matched control outcome: 0.83 (0.60–1.13); p = 0.23
B.I Person-delivered
Chande and Kimes,130 USA	RCT	High bias	Awareness, education, persuasion and tailoring	12 months	Records, self-report	All: (–) ED – at study entry, 94 out of 102 in the intervention group and 87 out of 93 in the control group had made at least one visit to the ED (p = 0.79). At the 1-year follow-up, 84 out of 102 in the intervention group and 73 out of 93 in the control group had made at least one visit to the ED (p = 0.59). There was no significant difference in rates of use of the ED	All: at study entry, 95 out of 102 in the intervention group and 75 out of 93 in the control group had seen their PCP at least once (p = 0.59). At the 1-year follow-up, 81 out of 102 in the intervention group and 77 out of 93 in the control group made at least one visit to their PCP (p = 0.59); however, the total number of PCP visits by both groups increased by almost 50%. Although there was no difference between the control and intervention groups at the 1-year follow-up, the total increase in visits for both groups is significant (p < 0.001)
						Minor: (–) at study entry, 55 out of 102 (54%) in the intervention group and 55 out of 93 (59%) in the control group had made two or more ED visits for minor illness in the preceding year. At the 1-year follow-up, 61 out of 102 (60%) in the intervention group and 57 out of 93 (61%) in the control group had made two or more ED visits for minor illness (p = 0.95)
Enard and Ganelin,87 USA	CBA study	Strong	Awareness, persuasion, tailoring	12 months and 24 months	Records	All: (+) for people who did not use the ED frequently in the previous year, the mean number of primary care-related ED visits declined significantly in the corresponding post-observation period, compared with controls. The OR was 0.83 (95% CI 0.71 to 0.98) for people who had a baseline attendance of one or more visits at 12 months. At 24 months, this was 0.55 (95% CI 0.47 to 0.63)
Racine et al.,92 USA	RCT	Some concerns	Awareness, persuasion	12 months	Records	All: (±) both intervention and control groups were as likely to return to the PED after their initial PED visit (intervention, 38%; control, 39%; p = 0.26). But, looking at visits, the adjusted odds of returning to a PED rather than another service was 0.88 (95% CI 0.82 to 0.94). PED, n (%): 2145 (17.7) vs. 2329 (20.4); p < 0.001. Hospitalisation, n (%): 325 (2.7) vs. 302 (2.6); p = 0.96. Subspecialty, n (%): 2744 (22.6) vs. 2220 (19.5); p < 0.001	Primary care: 6905 (57.0) vs. 6546 (57.4); p = 0.68 All: 12,119 vs. 11,397; p < 0.001. The adjusted OR of a follow-up visit being to the PED by an intervention subject, compared with a control subject, was 0.88
A.I and B.I
Davis et al.,86 USA	BA study	Moderate/weak	Awareness, persuasion	3 and 12 months	Records	All: (+) 62 fewer PED visits than the same 3 months the previous year. The numbers continued to decrease at 12 months: a decrease of 284 in the total number of PED visits (29.8%). In 2015, the clinic was at 71.5 visits per 100 persons per year. Following interventions, in 2017, the rate changed to 41.9 visits per 100 persons per year	All: the proportions of paediatric patients using the (GP) clinic instead of the ED also shifted from 2015 to 2017: 47.5% (64 patients) used the clinic for upper respiratory infections in 2015; in 2017, that number increased to 78.2% (97 patients) who were seen in the clinic as opposed to the ED (p < 0.0001)
						Minor: (±) 41.1% decrease (from n = 82 to n = 48) in level-3 (urgent) visits to the ED and a 16.7% decrease (from n = 120 to n = 100) in level-4 (less urgent) visits in the initial 3 months. These changes were maintained when examining the data 1 year later. The number of level-5 (not acute) visits remained unchanged	Minor: Walk-in clinic visits during this time continued to increase. Similar shifts were seen in fever and otitis media; however, the sample size was not great enough to draw significant conclusions

Other outcomes

Satisfaction or helpfulness

Only one weak prospective cohort study of a navigation tool reported any findings on users’ satisfaction with the intervention: 90% of the respondents reported that they were satisfied with the triage website Strategy for Off-site Rapid Triage (SORT) for Kids. 123 The other six studies of this type reported no relevant outcome data on satisfaction. 86,87,91,92,110,130

Current examples in the NHS

The NHS in England makes use of navigation tools routinely to make the general public aware of the range of services they can use, and in what circumstances, so that they use EDs only when they need to. An example of this is the Choose Well page on the Sheffield Children’s NHS Foundation Trust website: www.sheffieldchildrens.nhs.uk/patients-and-parents/choose-well/ (accessed 7 September 2021). This type of information is also disseminated by social media. For example, Sheffield Teaching Hospitals NHS Foundation Trust tweeted a navigation tool in July 2021 under the title of ‘Injured or unwell?’ (Figure 5). These navigation tools present different service options and when to use them. They can use the mechanism of ‘positive persuasion’ in terms of the promise of quicker advice if people use alternatives to EDs. We also found an example of a person-delivered navigation process (with a leaflet) to persuade users of an urgent care centre with minor problems to look for alternative sources of care in their future health-seeking. 144 There are also videos available through YouTube (YouTube, LLC, San Bruno, CA, USA) that make people aware of new services in primary care that might reduce demand for face-to-face care. An example is produced by West Yorkshire Health and Care Partnership,145 which includes the use of British Sign Language and animation to facilitate accessibility.

FIGURE 5.

Sheffield Teaching Hospitals NHS Foundation Trust ‘navigation tool’ tweet.

Conclusion and relevance to English NHS

The evidence base identified that there was potential that people could be persuaded to use their primary care provider, rather than an ED. However, the quality of the studies was not high and the evidence base was likely to be context specific because it was generated in the USA where there are incentives to make cost savings within a health insurance agency context. In terms of relevance to England, the navigation tools in use in England are different from those used in the evidence base. Tools in England mainly attempt to disseminate information to potential service users, rather than actively communicate a message to a targeted group of recent ED users. The NHS tools use a light-touch persuasion technique, compared with the persuasive techniques used in the evidence base. So the evidence base is not relevant to the NHS tools in routine use. We would recommend that consideration is given to identifying how these NHS tools could be used in a more proactive and targeted way in England, and that this approach is evaluated.

Box 1 provides a plain English summary of this section.

BOX 1 - Plain English summary (Flesch Reading Ease: easy)

Some tools tried to persuade people to use their GP rather than go to A&E. Some of them worked. But the quality of the studies was not high. And the studies were done in the USA, where they have health insurance. This is different from the UK. The tools were different from ones used in the UK in other ways. They used nurses to persuade people who had gone to A&E with a small problem to go to their GP next time. In the UK we tend to have information on hospital websites and this is aimed at everyone. So the studies we found were not relevant to the NHS at the moment. We wonder what would happen if the NHS tried to target the information on their websites to the people who need it the most. Then we could test if this works.

A&E, accident and emergency.

2. Written information about how to care for minor ailments in leaflets, booklets, books and websites (A.II and C.II)

Size and quality of the evidence base

Seventeen articles focused on written education in paper format given at a specific time when patients sought care for a minor health problem,89,100,101,107,109,119 paper format given at a general time95–97,105,106,117,138 and a website given at a specific136 or general time. 90,124,129 Fourteen of these measured impact on service use. These are presented in Table 6 in the three groups of articles ‘at a specific time’, ‘at a general time’ and ‘digital’. There was one pilot study so this was not powered to detect a difference. Of the articles measuring impact, seven were undertaken in the UK, five in the USA, one in Denmark and one in the Netherlands. The study designs were strong, mainly RCTs; in fact, all the studies used controls. All the studies used records rather than self-report. The formal quality assessment identified only one high-quality study. Our more informal assessment identified some concerns with papers: limited results available because yet to be formally published,89 low statistical power,119 or poor analysis or presentation of the findings. 95,97,105,107,138 We felt that, given the difficulties of undertaking research in this area, and the harshness of the formal quality assessment, we identified some good-quality studies with strong designs undertaken in the UK. 84,89,96,100,101 We also considered the controlled before-and-after study undertaken in the Netherlands, which made use of a national database on GP consultations to measure outcomes, to be highly relevant to the UK and of good quality. 129

TABLE 6 - Impact of written education on self-care (e.g. leaflets, booklets, websites on service use)

CBA, controlled before and after; CI, confidence interval; IRR, incidence rate ratio; OR, odds ratio; SD, standard deviation; TLAS, time-limited acute symptom.

Notes

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. Pilot studies are indicated by italics.

Study	Design	Quality	Mechanisms	Time period of measurement	Data source	Impact on ED	Impact on GP consultations
Paper based, specific time
Holden,89 UK	RCT	Some concerns	Education	11 months	Records	Minor: (–) average number of non-urgent attendances (including reattendances):– intervention, mean 1.3 (SD 0.8); control, 1.3 (SD 0.7). Reattendance rate: intervention, 23.2%; control, 24.6%
Macfarlane et al.,101 UK	RCT	High risk of bias	Education	1 month	GP-recorded reconsultations		All: (+) for the whole group, patients who received the leaflet had significantly fewer reconsultations for the same symptoms over the following month (14.9%) than those who did not receive a leaflet (21.4%); (OR 1.56, 95% CI 1.11 to 2.19; p = 0.007). For the 723 patients who were treated with antibiotics, significantly fewer patients receiving a leaflet reconsulted (60/369, 16%), than those who did not receive the leaflet (81/354, 23%) (OR 1.53, 95% CI 1.03 to 2.26; p = 0.02)
Macfarlane et al.,100 UK	RCT	High risk of bias	Education	1 month	Records		All: (–) the reconsultation rates were similar for all patients: leaflet, 11/104; no leaflet, 14/105
Platts et al.,107 UK	RCT	High risk of bias	Education	12 months	Records	All: (–) the differences between the groups in the number of visits to EDs and in the numbers of telephone calls, home visits and nurse clinic visits were small and not statistically significant	All: (–) mean annual consultation rates, including routine GP consultations and emergency consultations with a GP or nurse practitioner, for the 12-month period of the study, classified by age group, sex and book group, did not, in general, differ between the three arms of the study. Use of NHS Direct telephone service: a total of 160 participants made an average of 1.19 calls to the NHS Direct telephone service. In a multivariable logistic regression model, the odds of calling NHS Direct were nearly doubled for being allocated to the NHS Direct group rather than to control or to the self-care book group (p < 0.01)
Steelman et al.,119 USA	Non-RCT	Weak	Education	4 months	Records	Minor: (–) intervention vs. control: appropriate ED and clinic visits: 5 vs. 9; inappropriate visits: 6 vs. 8; p > 0.99
Paper based, general time
Elsenhans et al.,138 USA	Non-RCT	Moderate/weak	Education	4 months	Records		All: (+) decrease in total visits: intervention, 21%; control, 8%. Decrease in total visits per member per month: intervention, 18%; control, 12% Minor: (+) decrease in TLAS visits: intervention, 33%; control, 9%. Decrease in TLAS visits per member per month: intervention, 31%; control, 13%. Decrease in TLAS total visits per member per month: intervention, 5.3; control, 0.31; p ≤ 0.05
Hansen,105 Denmark	RCT	High risk of bias	Education	6 months	Records		All: (+) families that received the booklet reported significantly more self-teatments, had significantly fewer consultations and had significantly fewer consultations in which worry was the main reason. Mean number of patient-initiated consultations: intervention 0.288 (2SD 0.3–0.2) vs. control 0.426 (2SD 0.5–0.4). For home visits: intervention 0.098 (2SD 0.1–0.7) vs. control 0.195 (2SD 0.2–0.2) (no p-values given). Worry was a reason for consultation: intervention 19% vs. control 31% (p = 0.0075). GP assessed appropriateness of consultation: intervention 88% vs. control 81%
Heaney et al.,96 UK	RCT	High risk of bias	Education	1 year pre intervention, 1 year post intervention	Records		All: (–) out-of-hours contacts: book – before vs. after: 0.13 vs. 0.11; Control – before vs. after: 0.11 vs. 0.13. Before–after difference –0.02 (95% CI –0.06 to 0.01); out-of-hours total contacts: book – before vs. after: 8.53 vs. 6.57; control – before vs. after: 9.02 vs. 6.65. Before – after difference 0.22 (95% CI –0.31 to 0.75) Minor: (–) book – before vs. after: 1.13 vs. 1.09; control – before vs. after: 1.06 vs. 1.07. Before–after difference 0.03 (95% CI –0.17 to 0.10). Total contacts for minor illness: book – before vs. after: 2.24 vs. 1.74; control – before vs. after: 2.43 vs. 1.84. Before–after difference 0.02 (95% CI –0.25 to 0.29)
Little et al.,106 UK	RCT	High risk of bias	Education	12 months	Records		Minor: (+) compared with the control group, fewer patients in the booklet and summary card groups attended frequently with minor illnesses – booklet: OR 0.81 (95% CI 0.67 to 0.99; p = 0.043); summary card: OR 0.83 (95% CI 0. 27 to 0.96; p = 0.011)
Usherwood,97 UK	RCT	High risk of bias	Education	12 months	Contact record completed by practice staff		Minor: (±) although there was no significant difference between booklet and control groups for daytime health centre contacts, there were differences for daytime home visits and for contacts out of hours. For daytime home visits, the mean frequency of initial contacts was substantially lower in the booklet group than in the control group for households with one or two children. For out-of-hours contacts for booklet symptoms, the mean frequencies of initial contacts were higher in the booklet group than in the control group for all sizes of household
Yoffe et al.,95 USA	CBA study	Weak/moderate	Education Anxiety	18 months	Records	Minor: (+) there was a dramatic drop in ED use rates by children seen in the clinic in which the booklet was distributed. The reductions ranged from 55% to 81%, compared with the same month of the previous year and compared with other geographies. Compared with the control groups, the decline in use rates by intervention patients was statistically significant (p < 0.001)
Digital
Ladley et al.,90 USA	RCT	Low risk of bias	Education	12 months	Records	All: (+) total number of visits – control 240 (mean 2.12, SD 2.26) vs. intervention 168 (mean 1.44, SD 1.65); p < 0.05. All ED visits: IRR 1.48 (95% CI 1.11 to 1.97; p < 0.01) Urgent ED visits: IRR 1.79 (95% CI 1.05 to 3.05; p = 0.03) Minor: (+) non-urgent ED visits: IRR 1.43 (95% CI 1.06 to 1.94; p = 0.02)
Lepley et al.,136 USA	RCT (pilot)	High risk of bias	Education	6 months	Records, self-report	All: (–) record review – no significant difference in rate of ED visits between the application group and the control group (14% more visits in the application group; IRR 1.14, 95% CI 0.56 to 2.34), between the book group and the control group (22% fewer visits in the application group; IRR 0.78, 95% CI 0.34 to 1.74) or between the book and application group and the control group (40% fewer visits in the application group; IRR 0.60, 95% CI 0.25 to 1.42)
Spoelman et al.,129 the Netherlands	CBA study	Moderate	Education	3 years	Records		All: (+) the consultation rate decreased by 6.2% in the first year and by 11.6% after 2 years. All consultations: 272.109/0.826 vs. –0.794/–1.620; p < 0.001; so the change in the slope of GP consultations before and after the website was introduced was –1.620 (p < 0.0001). The trend for the top 10 topics reduced whereas the rates for the reference group stayed the same Minor: (+) the decline was most distinct for those telephone consultations that usually provided non-essential care

Intervention description and mechanisms

Leaflets or booklets varied in size and in the range of minor conditions addressed. They were sometimes endorsed by health professionals within a slide presentation,119 a covering letter if a booklet was sent by post105,106,138 or by personally handing the leaflet/booklet to the patient. 95,101,107 The role of the health professionals was different from the role taken for person-delivered education interventions (B.II, discussed in later section) in that there was no discussion with patients of the content of the booklet/leaflet. In one study, patients had the option of calling a doctor with questions, but we did not consider this to be person-delivered because it was optional. 119 In another study, the intervention was a leaflet and verbal advice from the doctor, compared with verbal advice as the control, so the intervention being tested was a leaflet. 100 The common mechanism was ‘education’.

Impact on service use

In Table 6, (+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. Pilot studies are indicated by italics. Five articles of paper-based interventions measured the impact on Eds, and two of these identified a reduction. 90,95 We had concerns about the quality of one of these positive studies, and the formal quality assessment rated the other as having a low risk of bias, so results were not determined by the quality of the research. Of the null studies, one was very small119 and one was a good-quality UK study (according to our informal assessment), involving posting a leaflet to ED non-urgent users and measuring further attendances. 89 Again, research quality did not explain null results.

Of the nine studies measuring impact of written interventions on GP consultation rates, six showed a reduction. The two good-quality UK RCTs (based on our informal assessment) had conflicting results. 96,106 One study’s positive result relied on a subgroup analysis. 97 The single website intervention measuring GP consultations129 identified a reduction in GP consultations in the population for symptoms that were frequently accessed on the website, compared with consultations for conditions not frequently accessed on the website.

Other outcomes

Satisfaction or helpfulness

Six articles reported no relevant outcome data on satisfaction. 89,95,96,100,101,129 Eleven of the 17 articles reported one or more relevant outcomes (Table 7). Seven of the 17 articles reported findings on users’ satisfaction with the intervention. In all seven articles, users reported being satisfied or very satisfied with the intervention, regardless of whether the intervention was aimed at general populations,97,106,138 aimed at a specific group109 or was web based. 90,124,136 However, in some cases, participants might have been satisfied with the intervention, but also satisfied with the control intervention,109 or expressed preferences for one format of the intervention over another. 136 Six of these seven articles were assessed either as being weak (two non-RCTs) or at high risk of bias (four RCTs); only one RCT was judged to be at low risk of bias. 90

TABLE 7 - Other outcomes for written education

A Level, Advanced Level; CBA, controlled before and after; CI, confidence interval; mHealth, mobile health; SDI, standard discharge instructions; VDI, video discharge instructions.

p < 0.01 for comparison with the app group.

p < 0.001 for comparison with the app group.

Note

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed.

Study	Design	Quality	Mechanisms	Data collection time points	Data source	Impact on other outcomes
Paper based, specific
Platts et al.,107 UK	RCT	High risk of bias	Education	3 and 12 months	Questionnaire	(+) Helpfulness: comprehensiveness – participants’ perceptions of books: participants who had been allocated a book and had consulted it at least once were asked whether or not the book provided all of the information required. Responses were requested on a scale from 1 (all provided) to 5 (none provided). Those allocated the self-care book gave a mean response of 2.42 (SD 1.09), compared with 2.85 (SD 1.30) for those allocated the NHS Direct book (p < 0.001) (+) Easy to understand: similarly, although both books were found to be easy to understand by most participants, the self-care book was seen as better for understanding (p < 0.001). Participants whose educational qualifications were below A Level were more likely to report that they found the information in the self-care book ‘harder to understand’ than those whose qualifications were at or above A Level, at both 3 months (p = 0.004) and 12 months (p = 0.006) (+) Enablement: the self-care book was better for helping people know what to do (p < 0.001); the advice from the self-care book was followed more often (p = 0.002). Similarly, at 3 months, those whose educational qualifications were below A Level were more likely to report that they found the information in the NHS Direct book ‘useful for knowing what to do’ than those whose qualifications were at or above A Level (3 months, p = 0.012; 12 months, p = 0.009) (+) Intentions: using the self-care book was more often likely to lead to a change of mind on what to do (p = 0.003)
Steelman et al.,119 USA	Non-RCT	Weak	Education	Baseline and 2 and 4 months	Self-report instrument	(+) Knowledge: evaluation of knowledge on fever – average number of incorrect responses: Baseline – intervention 11.5 vs. control 11.4; p = 0.35 2 months – intervention 10.4 vs. control 11.8; p = 0.006 4 months – intervention 8.5 vs. control 10.3; p = 0.002
Wood et al.,109 USA	CBA study	Weak	Education	2 months	5-question survey	(±) Knowledge: significant improvements in knowledge scores were observed after SDI or VDI were provided (65% pre instructions vs. 75% post instructions; p < 0.001). Knowledge significantly improved post instructions for gastroenteritis (73% post vs. 57% pre; p = 0.005), fever (76% post vs. 69% pre; p < 0.001) and bronchiolitis (64% post vs. 49% pre; p = 0.025). Post-instructions knowledge improvement was significantly greater in the VDI group for all diagnoses combined (pre–post difference: 13% VDI vs. 6% SDI; p = 0.027). Caregivers demonstrated significantly greater knowledge improvement for gastroenteritis (pre–post difference: 23% VDI vs. 0% SDI; p = 0.027). Similar knowledge improvement trends were noted for other diagnosis groups, but group differences did not achieve statistical significance (fever pre–post difference: 9% VDI vs. 7% SDI; p = 0.47; bronchiolitis: 23% VDI vs. 11% SDI; p = 0.32). Knowledge level (percentage correct) was significantly higher for caregivers in the VDI group than in the SDI group for all diagnoses combined (82% VDI vs. 67% SDI; p < 0.001). When stratified by diagnosis group, caregivers of children with fever and bronchiolitis demonstrated significantly greater knowledge after receiving VDI than after SDI (fever: 84% VDI vs. 70% SDI; p < 0.001; bronchiolitis: 83% VDI vs. 53% SDI; p = 0.019). A similar trend was observed in the gastroenteritis group, but the difference did not reach statistical significance (75% VDI vs. 68% SDI; p = 0.41). Evaluation of question types revealed that caregivers were least knowledgeable regarding duration of disease (mean post-instructions knowledge for all participants: 81% diagnosis, 81% treatment, 43% disease duration, 78% seek care; p < 0.001). This pattern was similar across all diagnosis groups (p > 0.29) (±) Satisfaction/helpfulness: caregivers rated the discharge instructions favourably in the SDI group and in the VDI group. The highest VDI ratings were observed for the bronchiolitis discharge instructions (100% very/extremely helpful), and the lowest VDI ratings were for the fever instructions (63% very/extremely helpful). There were no significant differences in caregivers’ perceived helpfulness of SDI when compared with VDI (all p ≥ 0.13)
Paper based, general
Elsenhans et al.,138 USA	Non-RCT	Moderate/weak	Education	6 months	Telephone survey	(+) Satisfaction: there was general enthusiasm among members and staff and the majority recommended the continued use of self-care manual; no dissatisfaction was expressed
Hansen,105 Denmark	RCT	High risk of bias	Education	6 months	Self-report instrument (survey and interview)	(+) Enablement: self-management of episodes of illness – 51% in the intervention group and 36% in the control group reported some kind of self-treatment. This difference was statistically significant
Little et al.,106 UK	RCT	High risk of bias	Education	12 months	Survey	(+) Satisfaction: most respondents could remember receiving a leaflet or booklet (booklet 85%, card 70%, control 52%; p < 0.001) and found them useful (booklet 81%, card 78%, control 62%; p < 0.001)
Mullett and Hills,117 Canada	Prospective cohort study	Weak	Education	12 months	Diary	Helpfulness: examples from patient diaries describing how the handbook helped them make decisions and self-care, and also what was missing from the handbook
Usherwood,97 UK	RCT	High risk of bias	Education	12 months	Questionnaire	(+) Satisfaction: perceived usefulness of the booklet – frequency of response: Has the advice in the booklet changed anything that you would do for your child when he or she is ill? Yes definitely, 21; yes probably, 47; not sure, 17; no, probably not, 15; no definitely not, 1 Overall, has the booklet been useful to you? Yes definitely, 41; yes probably, 52; not sure, 7; no, probably not, 2; no definitely not, 0 Do you think that other families in the practice would find the booklet useful? Yes definitely, 71; yes probably, 55; not sure, 8; no, probably not, 0; no definitely not, 0
Digital
Ladley et al.,90 USA	RCT	Low risk of bias	Education	12 months	Survey	(+) Enablement: high levels of satisfaction and engagement at 12-month survey in intervention group: 91% (75/79 who completed the survey) felt more comfortable making decisions about when to take their baby to the ED
Lepley et al.,136 USA	RCT (pilot)	High risk of bias	Education	6 months	Telephone survey	(+) Satisfaction: modified intention to treat – combined app (n = 37) vs. combined book (n = 37) vs. control (n = 23): Used intervention – 35.1% vs. 73.0%a vs. 78.3%a Very/somewhat useful – 37.8% vs. 70.3%a vs. 78.3%a Very/somewhat understandable – 46.0% vs. 94.6%b vs. 100.0%b Recommend intervention to family or friends – 48.7% vs. 94.6%b vs. 100.0%a
						(–) Satisfaction: per-protocol analysis – combined app (n = 20), combined book (n = 37), combined control (n = 23): Used intervention – 60.0% vs. 73.0% vs. 78.3% Very/somewhat useful – 70.0% vs. 70.3% vs. 78.3% Very/somewhat understandable – 85.0% vs. 94.5% vs. 100.0% Would recommend to others – 90.0% vs. 94.5% vs. 100.0% In the per-protocol analysis, comparing groups of those who were followed up at least once, there was no difference in use (p = 0.530), understanding (p = 0.222), recommendations (p = 0.517), or usefulness (p = 0.983) between the app and the book. Of note, parents in the book + app group prefer the book over the app. When given a choice between the book and the app, the majority of parents would choose the book alone (61.1%) or both (27.8%); none of the parents would choose the app alone
						Authors note in the discussion that the mHealth app used a lot of medical jargon and was not good for people with lower health literacy levels
Murray et al.,124 UK	Survey	Weak	Education	At the time of using the website or visiting the GP	Survey	(+) Satisfaction: Online sample (n = 1559) – use of NHS Choices in relation to GP consultations: Very satisfied – n = 252, 18.1%, 95% CI 14.4% to 18.1% Satisfied – n = 794, 57.1%, 95% CI 48.4% to 53.4% Neither satisfied nor dissatisfied – n = 334, 24%, 95% CI 19.4% to 23.5% Dissatisfied – n = 7, 0.5%, 95% CI 0.2% to 0.9% Very dissatisfied – n = 4, 0.3%, 95% CI 0.1% to 0.7% Did not answer – n = 168, 10.8%, 95% CI 9.3% to 12.4% GP sample (n = 125) – use of NHS Choices if have used the website in relation to GP consultations: Very satisfied – n = 7, 5.6%, 95% CI 2.3% to 11.2% Satisfied – n = 39, 31.2%, 95% CI 23.2% to 40.1% Neither satisfied nor dissatisfied – n = 14, 11.2%, 95% CI 6.3% to 18.1% Dissatisfied – n = 0, 0.0%, 95% CI 0.0% to 2.9% Very dissatisfied – n = 1, 0.8%, 95% CI 0.0% to 4.4% Has not used – n = 34, 27.2%, 95% CI 19.6% to 35.9% Did not answer – n = 30, 24.0%, 95% CI 16.8% to 32.5%
						(+) Knowledge: online sample (n = 1559) – patients who agree/strongly agree that use of NHS Choices for their GP appointments means: I know more about my condition/illness – n = 1189, 76.3%, 95% CI 74.1% to 78.4% I know more about my treatment options – n = 108, 69.4%, 95% CI 67.0% to 71.7% I am more confident to ask questions – n = 984, 63.1%, 95% CI 60.7% to 65.5% I am confident to express my point of view – n = 870, 55.8%, 95% CI 53.3% to 58.3%

Knowledge about conditions or confidence in knowing what to do

Three studies of paper-based formats reported improved knowledge and understanding among patients. 107,109,119 The same finding was reported in a study of an NHS website,124 and improved confidence was found among users of a general paper-based intervention. 105

Intentions

People changed their minds about the actions they would take after using one of the interventions. 107

Current examples in the NHS

The NHS in the UK has a number of sources of web-based educational materials about how to care for minor ailments and when to seek health care. The NHS website [www.nhs.uk (accessed 31 August 2022)] has a section titled ‘Health A to Z’ where a minor symptom such as ‘sore throat’ can be searched for and self-care advice is offered in writing and a short video, along with the role a pharmacist could play. Figure 6 shows a screenshot of a page (accessed 7 September 2021). It is aimed at everyone. People find it by searching the internet.

FIGURE 6.

NHS website screenshot. Contains public sector information licensed under the Open Government Licence v3.0.

Contains public sector information licensed under the Open Government Licence v3.0.

There are also regional versions of web-based education to help parents of young children to self-care for minor ailments, advertised under the banner of ‘Healthier Together’. For example, a version of Healthier Together146 is available for people in the South Yorkshire and Bassetlaw area (Figure 7). The website link is emailed directly to parents and addresses common illnesses and conditions that affect children and young people aged 0–18 including rashes, fever, coughs, asthma, sickness, earache, conjunctivitis, head injuries, and limps. It offers a traffic light system of seriousness from ‘minor’ to ‘urgent’. A Consultant Paediatrician from local highly regarded children’s hospital signs the letter.

FIGURE 7.

Healthier Together screenshot.

Another example of a regional resource for parents is The Little Orange Book68 produced by Newcastle and Gateshead CCG for parents with young children (Figure 8).

FIGURE 8.

The Little Orange Book screenshot.

Conclusions and relevance to the NHS

There is a reasonable size of evidence base about sending booklets or leaflets to patients, but limited evidence about digitally available written information. There is a reasonable amount of evidence from the UK and the rest of Europe. The outcome of interest measured was more likely to be impact on GP consultations than ED attendances. Unfortunately, the results of different studies were not consistent. There was no apparent reason why some good-quality studies showed reductions in service use and others not. However, there were enough positive studies to consider this type of intervention as having potential. Given the availability of digital educational information in the NHS currently, it may be beneficial to undertake a high-quality evaluation of such information, specifically considering how any information is disseminated to relevant people. Our PPI colleagues were keen that ways of educating people other than through digital means should also be used because of a lack of digital health literacy in the UK population and the lack of access to technology. Some regional educational information is also available in paper format, for example The Little Orange Book. 68

Box 2 provides a plain English summary of this section.

BOX 2 - Plain English summary 2

Booklets, leaflets and websites that tell people how to look after minor problems have been tested. People find them helpful and think that they know more about what to do when they are ill. But some research says they reduce the use of a GP or A&E and some research says they do not. They have potential, but need to be tested more.

A&E, accident and emergency.

3. Person-delivered education (B.II, and multicomponent B.I and B.II)

Size and quality of the evidence base

Three interventions focused on a person delivering education about managing minor health problems,88,103,126 and two had an additional component of a navigation tool. 93,137 They are presented in Table 8 in that order, alphabetically for each group. All the studies were undertaken in the USA. Only one study was a full RCT. Two studies were pilots, and so did not have statistical power to detect changes in service use. 88,103 None of the studies was assessed as being of good quality in the formal assessment. In addition, we were concerned about the analysis or presentation of results for one study,137 and two studies relied on self-report. 103,126 So the evidence base was small and poor for this type of intervention.

TABLE 8 - Impact of person-delivered education (with or without navigation tool) on service impact

BA, before and after; CI, confidence interval; NS, not significant; PCP, primary care provider; PED, paediatric emergency department.

Notes

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. Pilot studies are indicated by italics.

Study	Design	Quality	Mechanisms	Time period of measurement	Data source	Impact on ED	Impact on GP consultations
B.II only
Fieldston et al.,88 USA	BA pilot	Moderate/weak	Education, empowerment	6 months	Records	All: (–) the mean number of ED visits for the index child reduced following the intervention, although the difference was not significant: 0.67 mean ED visits before, and 0.58 after (p = NS)	All: (–) the mean number of ambulatory visits for the index child reduced following the intervention, although the difference was not significant: 2.7 before vs. 2.3 after (NS). After-hours telephone calls to the PCP rose significantly from 0.33 per patient to 1.46 per patient ( p  = 0.047)
O’Neill-Murphy et al.,103 USA	RCT pilot	High risk of bias	Education, empowerment	2 weeks and 8 weeks	Self-report	All: (–) only one child, in the control group, had visited the ED in the time period, even though 30% (n = 11) of the control and 43% (n = 10) of the intervention had a fever in the previous 2 weeks
Stockwell et al.,126 USA	RCT	High risk of bias	Awareness, education, empowerment	5 months	Self-report	All: (+) children < 4 years of age in intervention families had fewer PED visits per illness episode than the standard curriculum group [18/205 (8.8%) vs. 29/191 (15.2%), respectively; p = 0.049]. The majority of this difference was for children aged 6–48 months [intervention families 16/194 (8.2%) vs. standard curriculum group 29/185 (15.7%) PED visits per illness episode; p = 0.025]. The relationship between intervention status and fewer PED visits was also significant on the family level (p = 0.03). For intervention families attending all three classes, 6.2% of illness episodes resulted in a PED visit for a child aged 6–48 months (5/80); for those attending two classes, this was 7.8% (8/102), and for those attending zero or one class, it was 25% (3/12) (p = 0.087)
B.I and B.II
DeSalvo et al.,137 USA	Non-RCT	Weak	Awareness, education, persuasion, tailoring	3 and 6 months	Records, self-report	All: (–) these patients had an average of 0.16 emergency room visits per patient in the 3 months subsequent to the index visit. In contrast, there were 0.29 subsequent visits per patient in the control group (p < 0.01). By the end of the first 3 months, the difference had disappeared (intervention 0.32 vs. control 0.28; p = 0.14). The addition of a social worker to the team in the second phase did not result in a change in emergency room use (0.3 visits per patient in both the intervention and control groups; p = 0.5)
Shnowske et al.,93 USA	Retrospective cohort study	Moderate/weak	Awareness, education, tailoring	3 months	Records	All: (+) mean number of ED visits per month before was 1.20 ± 0.66, and after was 0.71 ± 0.54. This represents approximately a 41% (95% CI 43.2% to 54.8%) decrease in ED visits per month Minor: The visit change analysis showed that visits for pain-related complaints decreased by 0.51 visits per month	All: mean number of PCP visits per month before was 0.31 ± 0.64, and after assignment was 0.28 ± 0.68 (p = 0.30). Had expected to see an increase

Intervention description and mechanisms

The interventions were delivered by nurses88,137 or community health workers126 and varied in terms of length of time spent on training patients and the number of minor illnesses addressed. For example, a paediatric ED nurse delivered a 90-minute training session to parents,88 and community health workers gave three 90-minute training sessions on minor illness. 126 The mechanisms included both education and empowerment, as patients were trained in self-management.

Impact on service use

In Table 8, (+) indicates a statistically significant reduction in service use, (–) indicates a null result, and (±) indicates that results were mixed. Pilot studies are indicated by italics. Removing the pilot studies, three studies measured impact on EDs and two of these showed a reduction in use. Only one full study measured impact on GP consultations; they expected an increase because they aimed to shift care from EDs.

Other outcomes

One RCT that was judged to be at high risk of bias, evaluating person-delivered education, reported high levels of satisfaction with the upper respiratory infection intervention. 126 Two studies judged to be weak or at high risk of bias reported improved knowledge scores following the intervention,88,126 but a third trial, also at high risk of bias, found that the intervention did not improve participants’ confidence when faced with fever in a child: anxiety was reduced more in the control group. 103 One study of this type reported no relevant outcome data on satisfaction or intentions at all (Table 9). 137

TABLE 9 - Other outcomes for person-delivered education

BA, before and after; URI, upper respiratory infection.

Note

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed.

Study	Design	Quality	Mechanisms	Data collection time points	Data source	Impact on other outcomes
Fieldston et al.,88 USA	BA pilot	Moderate/weak	Education, empowerment	6 months	Knowledge tests	(+) Knowledge: the intervention increased knowledge as measured by performance on the 19-item evaluation questionnaire, with mean scores increasing from 55% in the pre-intervention assessment to 77% (p < 0.001) immediately following the intervention. All three content areas (colds, fever and trauma) saw increase in knowledge, comparing the pre-intervention and post-intervention test results. Among the 20 participants, the mean score on the follow-up test declined from the value immediately after intervention (79% for those 20 participants to 71%; p = 0.031), but remained higher than before the intervention (61%; p = 0.015). For the same group, at follow-up, the mean scores for specific content areas were not significantly different from post-intervention scores, except for trauma, in a negative direction (74–62%; p = 0.044)
O’Neill-Murphy et al.,103 USA	RCT pilot	High risk of bias	Education, empowerment	2 weeks and 8 weeks	Telephone interview	(–) Anxiety levels: > 40% of parents in each group reported that they had moderate to high levels of anxiety related to their child’s fever on arrival to the ED. After either fever education programme, parents in both groups reported reduced levels of fever anxiety: control group, 82%; intervention group, 85%. At the 2-week follow-up, parents in both groups reported that they felt less anxious when their child has a fever (86% of the control group and 50% of the intervention group)
Stockwell et al.,126 USA	RCT	High risk of bias	Awareness, education, empowerment	5 months	Knowledge–Attitudes instrument	(+) Satisfaction: almost all intervention parents reported that the education sessions (97.3%) and URI kit (93.0%) were very useful; 87.3% used the kit at least a few times over the winter (+) Knowledge: URI knowledge/attitudes and home remedy safety attitudes – mean baseline Knowledge–Attitudes instrument scores were similar for intervention and standard curriculum families (4.2 vs. 4.6 out of 10, respectively; p = 0.27). Mean post-intervention scores increased to 5.5 for intervention families, but remained the same (4.7) for standard curriculum families (p = 0.011), and the mean difference between baseline and post-intervention scores for intervention families was significantly different than for standard curriculum families (1.3 vs. 0.097; p = 0.001). Home remedy beliefs at baseline were also similar (28.9% of intervention families believed that all home remedies were safe or were unsure vs. 37.2% of standard curriculum families; p = 0.28), but were different post intervention (17.8% vs. 38.9%; p = 0.005). The percentage of intervention families who had incorrect home remedy beliefs post intervention (52.0%) was also lower, but not significantly lower, than for standard curriculum families (75.7%; p = 0.053), when those who had both correct baseline and post-intervention beliefs were removed

Current examples in the NHS

We are aware that multiple types of health professional in the NHS educate patients informally about managing minor ailments when patients attend for care for that ailment, for example GPs, health visitors and pharmacists. We could find no examples of health professionals or trainers routinely offering formal education in managing minor health problems in the UK. The NHS has ‘care navigators’ who work in general practice and guide patients to the right clinician within a practice, and offer support for long-term conditions and social support. There are also social prescribers in primary care or volunteers who may offer some education and navigation.

Conclusions and relevance to the NHS

The evidence base for person-delivered education was small, context specific (USA only) and had some underpowered pilot studies, so few conclusions can be drawn. Although the NHS may have services similar to this type of intervention, we feel that there are no routinely available interventions currently in the NHS. The NHS would have to introduce this type of intervention and evaluate it. The feasibility of such an initiative delivered by health professionals is doubtful in times of staff shortage, and costs are likely to be prohibitive. Our PPI panel wondered about the use of volunteers to offer training, but also pointed out that patients need to trust the expertise of any trainers.

Box 3 provides a plain English summary of this section.

BOX 3 - Plain English summary 3

There was not much research on doctors or nurses educating patients about minor problems. The research was done in the USA. It may not be possible to use it in the UK because there is a shortage of doctors and nurses.

4. Combination of written education and person-delivered education (A.II and B.II)

Size and quality of the evidence base

Seventeen articles (15 studies) assessed leaflets/booklets and an interactive session with a person to educate people about minor health problems (Table 10). Ten were offered at a specific time in a patient’s care75–78,83,85,99,102,112,131 and seven at a general time. 81,82,120,121,127,134,135 They are presented in Table 10 in alphabetical order within these two groups. The evidence base for interventions given at a specific time was largely generated in Europe, whereas interventions offered at a general time were largely undertaken in the USA. There were seven RCTs and four controlled before-and-after studies that measured impact on service use. The formal quality assessment showed a mix of medium- to poor-quality studies. Seven articles relied on self-report and/or we had concerns about quality in terms of large loss to follow-up,83,127 the control group being very different from the intervention group,75 poor analysis with reliance on subgroup analysis78 or very short follow-up. 131

TABLE 10 - Impact of interventions combining educational information delivered in person and writing (e.g. leaflets) on service use

BA, before and after; CBA, controlled before and after; CI, confidence interval; OOH, out of hours; OR, odds ratio; PED, paediatric emergency department; PHCP, primary health-care provider.

Notes

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. Pilot studies are indicated by italics.

Study	Design	Quality	Mechanisms	Time period of measurement	Data source	Impact on ED	Impact on GP consultations
Specific
Chande et al.,85 USA	RCT	High risk of bias	Awareness Education Persuasion Tailoring	6 months	Records	All: (–) return to PED within 6 months: 21 patients (30%) from the intervention group and 16 (26%) from the control group (p = 0.68) Minor: (–) 17 patients (81%) of intervention group returnees to the PED had minor illness, as did 11 (69%) of control group returnees
de Bont et al.,102 the Netherlands	Cluster RCT	Some concerns	Education Tailoring	2 weeks (same illness) 6 months (OOH)	Records		Minor: (–) reconsultation with OOH service within 2 weeks: usual care group (N = 13,410): n = 861 (5.5%); access to booklet group (N = 11,945): n = 741 (5.4%); access to booklet, OR 0.95 (95% CI 0.83 to 1.09). Reconsultation with OOH service in 6 months: OR 0.99 (95% CI 0.84 to 1.18)
Francis et al.,99 UK	RCT	Some concerns	Education Anxiety Tailoring	2 weeks	Self-report		Minor: (–) there was no significant difference between the intervention and control groups in the odds of reconsulting in primary care during the 2 weeks after registration. Primary care reconsultation for same illness episode within the first 2 weeks: intervention 33/256 (12.9%) vs. control 44/272 (16.2%); OR 0.75 (95% CI 0.41 to 1.38). There was no significant intervention effect when telephone consultations were counted as reconsultations along with face-to-face primary care consultations (OR 0.81, 95% CI 0.47 to 1.42), or when consultations at EDs were included along with primary care consultations (OR 0.85, 95% CI 0.48 to 1.51)
Herman et al.,83 USA	BA study	Weak	Education	6 months	Self-report	All: (+) visited the ED – 73% pre vs. 43% post, p < 0.0001; visited the ED more than once: 36% pre vs. 26% post, p = 0.19
Plass et al.,75 the Netherlands	CBA study	Moderate	Education Tailoring	24 months	Records		Minor: (+) intervention group – a significant decrease in self-reported care-seeking behaviour concerning minor ailments and self-limiting health problems (p < 0.001). At baseline, the participants reported that, during the previous 6 months, they had consulted their GP 3.0 (SD 2.2) times because of 4.9 (SD 2.1) different minor illnesses. Six months later, they reported that they had consulted their GP 2.1 (SD 2.0) times for 5.1 (SD 3.2) different minor illnesses. One year after the intervention, they reported 1.7 (SD 1.8) consultations for 5.3 (SD 2.7) different minor illnesses. Control group: the number of consultations for minor illnesses was, on average, 2.4 (SD 2.3) for 5.4 (SD 3.7) different minor illnesses
Rutten et al.,77 the Netherlands	RCT	High risk of bias	Education	12 months	Not clear		All: (+) in both groups, the consulting behaviour changed. In the experimental practices, the GP was confronted less often with everyday cough (16 before vs. 11 after, compared with 28 before vs. 34 after in control practices) and more often with alarming symptoms (31 before vs. 47 after, compared with 16 before vs.19 after in control practices; p < 0.05)
Sustersic et al.,131 France	RCT	Some concerns	Education	2 weeks	Self-report		Minor: (+) 23.4% consulted for the same symptoms in the households of the intervention group, compared with 56.2% in the control group (p < 0.01)
General
Bertakis,134 USA	RCT	High risk of bias	Awareness Education Tailoring	12 months	Records	Minor: (–) appropriate visits: 0.2 ± 1.1 vs. 0.03 ± 0.9; p = 0.095. Appropriate visits: 75.3% ± 40.7% control vs. 84% ± 35.2% intervention; p = 0.072	Minor: (+) appropriate visits: 97.6% ± 11.2% control vs. 99.5% ± 4.8% intervention; p = 0.016
Herman and Mayer,82 USA	CBA pilot	Weak	Education Empowerment	6 months	Self-report	All: (+) 32 fewer visits to the ED (p < 0.01)	All: (+) visits to doctor or clinic – post intervention, there were 161 fewer visits to the doctor or clinic (p < 0.01) and 67 fewer calls to the doctor (p < 0.03)
Herman and Jackson,81 USA	BA	Weak	Education Empowerment	6 months	Self-report	All: (+) number of visits decreased from 0.79 per year to 0.33 i.e. by 58% (95% CI¼51%–66%)	All: (+) the pre-training mean number of clinic visits per child was 3.69 per year (0.3072 per child per month). The post-training mean was 2.1922 per year (0.1851 per child per month), showing a reduction of 1.47 visits per child per year (p < 0.001). Tracking survey data showed that doctor or clinic visits decreased by 42% (95% CI 33% to 46%)
McWilliams et al.,135 USA	CBA study	Moderate	Education Empowerment Anxiety Persuasion	12 months	Records	Minor: (+) intervention – 80% decrease in ED visits for ear pain [from mean 107 (SD 410) per 1000 to mean 21 (SD 144) per 1000]; after regression models: p = 0.009. Control: number of ED visits – from 48 (248) to 60 (295) (25% increase); p = 0.009 A 40% decrease in urgent care visits, but after regression models this was not statistically significant (p = 0.33): intervention – from 315 (658) to 188 (568); control – from 365 (835) to 263 (662), (28% decrease); p = 0.33	Minor: (–) 28% decrease in primary care visits for ear pain in the intervention sites [from 762 (1249) to 550 (1064) visits], compared with no change in control sites [from 667 (1491) to 639 (1269), 4% decrease], but after regression models, no change (p = 0.14)
Powell,127 USA	Prospective cohort study	Weak	Education	5 months	Self-report	All: (+) reduced ED visits by 52	All: (+) reduced GP visits by 126
Steinweg et al.,120 USA	Survey	Weak	Education Empowerment Tailoring	6 months	Self-report	Minor: (+) a total of 181 ED visits avoided, with 39.8% of the respondents indicating that they avoided at least one visit	Minor: (+) 72% of the respondents avoided at least one clinic visit. A total of 423 clinic visits were avoided during the 6-month study period
White et al.,121 UK	CBA study	Moderate	Education Empowerment Anxiety Persuasion	6 months	Records	All: (–) no statistically significant changes were seen in ED use when comparing the intervention group with the comparison (OR 1.64, 95% CI 0.84 to 3.32; p = 0.157), or in use of NHS Direct (OR 1.36, 95% CI 0.88 to 2.11; p = 0.169)	All: (–) the intervention did not make a detectable difference to the number of consultations with GPs or other PHCPs in the follow-up period. Expected number of PHCP consultations for a person in the intervention group was 1.17 times that of their counterpart in the comparison group (95% CI 0.93 to 1.46; p = 0.182). The proportion of people using services in the follow-up period was statistically significantly increased in the intervention group than in the comparison group for OOH services (OR 2.26, 95% CI 1.10 to 4.92; p = 0.031)

Intervention description and mechanisms

The interactive part of the intervention was provided by a researcher,85,134 a doctor in a consultation,75,78,99,102,112,131 a nurse135 or trainers/employers. 81,82,120,121,127 Because of the interactive component, there were additional mechanisms to the education mechanism in purely written educational interventions: tailoring, anxiety reduction and empowerment were common.

Impact on service use

In Table 10, (+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. Pilot studies are indicated by italics. Three articles did not report impact on service use. 76,78,112 Nine studies measured impact on ED attendances, but one was a pilot study (see Table 10). Five out of the eight full studies showed a reduction in any indicator of service use. Twelve measured impact on GP consultations, but one was a pilot study. Seven of the 11 full studies showed a reduction in the number of GP consultations.

In Table 10, one article showed an unexpected increase in the use of GP out-of-hours services and a non-statistically significant increase in the use of primary care generally,121 raising the issue of unintended consequences of trying to increase health literacy. The authors raise the concern that a lot of data were missing in their routine data set and their results should be treated with caution. Furthermore, they measured service use overall, rather than for minor conditions, so it is possible that any increase in service use was for major problems. Nonetheless, it is important to recognise the possibility that service use might be increased by these interventions.

Other outcomes

Eleven of the 17 articles reported relevant outcomes (Table 11). Four articles reported no relevant outcome data on these outcomes. 75,85,102,112,134

TABLE 11 - Other outcomes for interventions combining written education and person-delivered education

BA, before and after; CBA, controlled before and after; CI, confidence interval; ER, emergency room; N/A, not applicable; PIL, patient information leaflet; SD, standard deviation; WCV, well-child visit.

Note

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed.

Study	Design	Quality	Mechanisms	Data collection time points	Data source	Impact on other outcomes
Specific
Francis et al.,99 UK	RCT	Some concerns	Education Anxiety Tailoring	2 weeks	Telephone questionnaire	(–) Satisfaction: there were no significant differences in terms of satisfaction, level of reassurance, parental enablement, or the parent’s rating of the ‘usefulness of any information received in the consultation’
Herman et al.,83 USA	BA study	Weak	Education	6 months	Telephone survey	(+) Intention: hypothetical responses to low-acuity child medical problem scenarios: Temperature of 99.5 °F (37.5 °C) – pre intervention vs. post intervention: (1) do nothing and wait, 26% vs. 31%; (2) look in a health book, 5% vs. 27%; (3) talk to family or friends, 6% vs. 5%; (4) visit a physician’s office or clinic, 44% vs. 31%; (5) go to the ED, 16% vs. 7%; (6) do not know, 3% vs. 0% (p = 0.0002) Vomiting for 1 day – pre intervention vs. post intervention: (1) do nothing and wait, 14% vs. 10%; (2) look in a health book, 1% vs. 20%; (3) talk to family or friends, 7% vs. 10%; (4) visit a physician’s office or clinic, 65% vs. 55%; (5) go to the ED, 12% vs. 5%; (6) do not know, 1% vs. 0% (p = 0.0025) Earache – pre intervention vs. post intervention: (1) do nothing and wait, 4% vs. 7%; (2) look in a health book, 2% vs. 25%; (3) talk to family or friends, 3% vs. 3%; (4) visit a physician’s office or clinic, 82% vs. 61%; (5) go to the ED, 7% vs. 5%; (6) do not know, 1% vs. 0% (p = 0.0469) Cough – pre intervention vs. post intervention: (1) do nothing and wait, 20% vs. 34%; (2) look in a health book, 7% vs. 31%; (3) talk to family or friends, 7% vs. 3%; (4) visit a physician’s office or clinic, 61% vs. 29%; (5) go to the ED, 3% vs. 2%; (6) do not know, 1% vs. 0% (p = 0.0171) When asked again about the low-acuity child health scenarios, there was a reduction in the proportion who would visit a physician’s office or clinic for each complaint, and also a significant reduction in the proportion that would go to the ED for a low-grade fever and for vomiting for 1 day
Rutten et al.,78 the Netherlands	RCT	High risk of bias	Education	Unclear	Not clear	(+) Behaviour: changes in consulting behaviour – Initial behaviour did not follow guidelines: experimental (n = 122) vs. control (n = 232) – change of behaviour: desired change, 56% vs. 30%; undesired change, N/A vs. N/A; unchanged, 42% vs. 68%; undefinable, 2% vs. 2% (p < 0.001) Initial behaviour followed guidelines: experimental (n = 102) vs. control (n = 92) – desired change, N/A vs. N/A; undesired change, 30% vs. 67%; unchanged, 66% vs. 29%; undefinable, 4% vs. 3% (p < 0.001)
Sustersic et al.,131 France	RCT	Some concerns	Education	2 weeks	Telephone interview/survey	(+) Behaviour: for the whole population (adults and adults accompanying children), those in the PIL group significantly showed behaviour that was closer to that recommended by the PIL than those in the group that had not received a PIL (mean behaviour score 4.9 vs. 4.2; p < 0.01). This was confirmed by the alternative analytical approach, whereby the behaviour scores were dichotomised and used in univariate analysis (recommended behaviour 71.8% vs. 43.0%; p < 0.01)
						(+) Knowledge: likewise, those in the PIL group had a mean knowledge score that was significantly higher than that of those in the control group (mean knowledge score 4.2 vs. 3.6; p < 0.01).The adult patient subgroup showed behaviour that was closer to that recommended by the PIL than adult patients in the control group (mean behaviour score 4.9 vs. 4.0; p < 0.01). The adult and child group showed the same tendency, but did not reach significance (mean behaviour score 4.9 vs. 4.5; p = 0.11) For the adult patient subgroup, knowledge was significantly better in the group that received a PIL (mean knowledge score 4.2 vs. 3.5; p < 0.01), irrespective of the condition studied or of sociodemographic parameters (with the exception of the level of education, for which the difference was not significant)
General
Herman and Mayer,82 USA	CBA pilot study	Weak	Education Empowerment	6 months	Self-report instrument	(+) Satisfaction: most parents claimed to have used the book and had a positive experience with it. A total of 145 (96%) rated the book as ‘very easy to understand,’ with none reporting that it was ‘hard to understand’ and only 3% reporting that they had not used the book. A total of 122 parents (81%) found the book to be ‘very useful’ and 26 (17%) found it useful ‘sometimes’. Only 2% reported that they had not used the book in response to a question about the book’s usefulness (‘If you used this book, how useful was it?’) (±) In response to the question ‘If you used this book, what would make the book better?’, roughly 42% of the parents (63/151) thought that the book was ‘perfect the way it is’, and 32% (48/151) felt that it would be helpful to ‘add more information’. With 13% of parents (20/151) recommending that the authors ‘add more pictures’, more than half (51%) seemed curious to learn more, either by indicating their general desire for ‘more information’ and more pictures (a combined total of 45%) or by suggesting that the authors ‘make [the book] longer’ (6%). Seventy-one per cent of respondents (107/151) claimed to have used the book ‘frequently’, with 67% (101/151) rating the book ‘very well liked’. [One-third of parents (33%) found the book to be ‘okay’]
						(+) Behaviour: accessing health information – according to the survey, exposure to the self-care book or to the book with additional training affected the way many parents accessed their health information. Before the intervention, about half of the parents (52%) claimed to derive health information ‘from the doctor or clinic’. Following the intervention, however, only 18% claimed to access health information this way, a decrease of 34%. The effects of the training were evident in parents’ responses to the question ‘When your child is sick, where do you first go for help?’. In the control group (those who received the book without the additional training), 69% responded that they would ‘call [their] child’s regular doctor or health phone line’. In the intervention group, however, which received both the book and training in how to use it efficiently, 58% responded that they would ‘look in a book’, with only 28% reporting that they would ‘call [their] child’s regular doctor or health phone line’. (Only 1% of those in the control group responded that they would ‘look in a book’ first.) Eight per cent of those in the control group had noted that they would ‘take [their] child to the emergency room’, whereas only 3% of those in the intervention group claimed that they would take that route when a child was sick (17% in the control group would ‘call family or friends’, whereas only 7% in the intervention group chose that option). Overall, then, 6 months following the intervention, more parents claimed that they would turn to a book and fewer claimed that they would take a child to the clinic or ED in response to a perceived illness
						(+) Confidence: parents’ confidence levels seemed to be positively affected by the book and training. When asked if they felt confident caring for their child’s health-care needs after reading the book, 84% responded that they were ‘more confident after reading the book’ and 16% felt ‘the same after reading the book’
						(+) Intention: parent responses to mild conditions, including what they would do if their child had a fever of 99.5 °F, had an earache, was vomiting and had diarrhoea, or had a runny nose and a cough. In each case, more parents would look in a book and fewer would call 911, go to the ED or go to the doctor or a clinic Change in parent responses regarding method of treatment, from pre to post intervention:
Herman and Jackson,81 USA	BA study	Weak	Education Empowerment	6 months	Self-report instrument	(+) Behaviour (responses to illness): the proportion of parents who answered using the ER as a first source of help was 4% (369/9240) before the training and 1% (73/7281) after the training. Although 85% of parents reported in the pre assessment that they always can take care of their child, 90% reported getting worried to some degree when their child was sick, and 57% reported that they were sometimes unsure of what to do when their child was sick Results from the parental assessment showed a significant change in behaviour across all measures. When asked, ‘When your child is sick, where do you first go for help?’, responses that listed doctor visits as the first source for treatment decreased from 69% to 33%, and seeking treatment at an ER decreased from 8% to 2%. Parents also were asked how they would respond to specific common childhood illnesses, such as a fever of 99.5 °F. Possible responses ranged from using the health book provided in the training to taking the child to the doctor or the ER. After the training, the proportion of parents who stated that they would refer to a health book increased from 5% to 48%
						(+) Anxiety: after participating in the intervention and receiving training in the use of a health book, the percentage of parents who reported being ‘very worried’ when their child is sick decreased by one-third
						(+) Missed workdays: workdays missed by the primary caretaker per year decreased by 42% (95% CI 35% to 50%), and school days missed per year decreased by 29% (95% CI 23% to 35%). Changes in school days missed were confirmed through examination of school records
McWilliams et al.,135 USA	CBA study	Moderate	Education Empowerment Anxiety Persuasion	12 months	Questionnaire	(+) Satisfaction: using a scoring system from −10 to 10, parents at the 15-month WCV strongly endorsed this intervention, showing support for both aspects of the intervention (nurse teaching and access to prescription). Support for this intervention remained high for those seen at the 24-month WCV, even for families who had not experienced an episode of ear pain after the nursing intervention
						(+) Intentions: when surveyed at the 24-month WCV, 42.0% of parents indicated that their children had in fact experienced ear pain since the 15-month WCV. More than 80% of this subgroup believed that the nursing education helped them avoid an ED or after-hours visit: 54.3% ‘strongly agreed’, 31.4% ‘somewhat agreed’, 8.6% ‘were unsure’, 5.7% ‘somewhat disagreed’ and 0% ‘strongly disagreed’. Likewise, > 80% of these same parents indicated that access to antipyrine–benzocaine otic drops helped avoid an ED or after-hours visit: 62.9% ‘strongly agreed’, 25.7% ‘somewhat agreed’, 11.4% ‘were unsure’, 0% ‘somewhat disagreed’ and 0% ‘strongly disagreed’
Powell,127 USA	Prospective cohort study	Weak	Education	5 months	Questionnaire	(+) Intentions: (usefulness, understanding, etc.) > 90% for 607 employees who returned questionnaire; 70% felt that using the guide would help them visit the doctor less frequently
Steinweg et al.,120 USA	Survey	Weak	Education Empowerment Tailoring	6 months	Survey	(+) Enablement: programme outcome – confidence to treat minor illness: increased, 77%; no change, 23%; decreased, 0%
White et al.,121 UK	CBA study	Moderate	Education Empowerment Anxiety Persuasion	6 months	Questionnaire	(+) Enablement (locus of control) – change in scores from baseline: 6 months – Intervention: mean 20.02 (SD 4.40), n = 363 Control: mean 20.15 (SD 4.01), n = 266 12 months – Intervention: mean 0.72 (SD 4.09), n = 325 Control: mean 0.36 (SD 3.73), n = 251 There was a statistically significant positive effect of the intervention on recovery locus of control at 6 months, and small effects in favour of the intervention on recovery locus of control (0.88 points, 95% CI 0.27 to 1.49 points) and self-esteem (20.99 points, 95% CI 21.60 to 20.37 points) at 12 months
						(–) Knowledge: there was no statistically significant effect of being in the intervention group on perceived health status, levels of social support or knowledge about children’s cough, back pain or crying in babies at 6 and 12 months’ follow-up, controlling for baseline, gender and setting
						(+) Intentions: statistically significant positive effects of being in the intervention group were seen at 6 months in intention to use GP services less for minor ailments and, at 12 months in knowledge about back pain. Intention to use services: at baseline, > 90% of all participants stated an intention to use the family doctor, and > 80% stated an intention to use the pharmacist or family for support. At the 12-month follow-up, intention to use support from friends, hospital, the library, occupational health workers and social workers had increased in the intervention group, but not the comparison group. Intention to use support from health visitors and midwives had decreased in the comparison group, but not in the intervention group

Satisfaction

Only two of these articles reported on participants’ satisfaction with the interventions. 99,135 Both were of moderate quality (or were assessed as having some concerns); both were also for interventions targeted at parents: one found no significant differences in terms of satisfaction, level of reassurance, parental enablement or the parent’s rating of the ‘usefulness of any information received in the consultation’,99 whereas the other reported high levels of satisfaction with the education provided at well-child visits for infants. 135

Change in behaviour or intentions

Eight articles evaluated patients’ reported changes in behaviours and intentions to consult following an intervention, and each of these studies found a positive direction of change, regardless of whether the intervention was aimed at a specific group78,83,131 or populations more generally. 81,82,121,127,135 Five of the eight articles were judged to be weak or at high risk of bias; two non-RCTs were assessed as being of moderate quality121,135 and one RCT was judged to have some concerns. 131

Confidence

Two studies also reported increased confidence among participants to deal with minor illness120 or improved locus of control following the intervention. 121

Knowledge

Two studies compared knowledge between intervention and control groups: a moderate-quality trial of a targeted intervention reported a significant improvement in knowledge in the intervention group,131 whereas a moderate-quality controlled before-and-after study of a more general intervention found no such difference, despite improvements across other measures such as intentions and locus of control. 121

Current examples in the NHS

We could find no examples of this type of intervention in routine use in the NHS.

Conclusion and relevance to the NHS

Even though this type of intervention had additional mechanisms to written material only, the conclusions from the evidence base were similar to those drawn about written information only. There is potential for this type of intervention to affect service use, but there is no strong evidence base to say that it does.

Box 4 provides a plain English summary of this section.

BOX 4 - Plain English summary 4

A lot of research looked at a person training patients in how to look after minor problems and giving them leaflets or booklets too. The findings were the same as for leaflets and booklets by themselves.

5. Multicomponent interventions (A.II and B.II with mass media campaigns D)

Size and quality of the evidence base

Five articles (four studies) used additional components to the written information combined with a person-delivered approach described in the previous section, adding mass media advertisements79,80,115,128,139 (Table 12). They were based in North America, with one based in Taiwan. 139 None of the studies was a RCT. They were of moderate quality in our formal quality assessment, but in our informal assessment of quality, we identified concerns about a downwards trend in service use in the control group,79 and service use based on self-report. 80

Description of interventions and mechanisms

The mass media components included public education campaigns in a specific geographical area, or the use of posters to promote training workshops. The mechanisms tended to include empowerment as well as education.

Impact on service use

In Table 12, (+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. One study did not measure service use. 139 Three articles measured impact on ED attendances and three of these showed a reduction in use. Two measured impact on GP consultations, with one showing a reduction. Note that Hibbard et al. 79 and Wagner et al. 80 are different components of the same study, with the former using routine data to measure impact and the latter using a controlled before-and-after survey of self-reported use.

TABLE 12 - Impact of multicomponent intervention on service use

BA, before and after; CBA, controlled before and after; CI, confidence interval; NS, not significant; TLAS, time-limited acute symptom(s).

Note

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed.

Study	Design	Quality	Mechanism	Time period of measurement	Data source	Impact on ED	Impact on GP consultations
Barr et al.,115 Canada	BA study	Moderate	Education Persuasion	6 months	Records	Minor: (+) visits for non-disease-related crying: pre intervention: 724/20,394 (3.5%) of visits. The relative risk of a visit for crying or colic in the post-implementation period compared with the pre-implementation period was 0.705 (95% CI 0.59 to 0.85; p < 0.001), representing an overall 29.5% reduction in visits
Hibbard et al.,79 USA	Non-RCT	Moderate	Education Empowerment	12 and 24 months	Records	All: (+) emergency rooms visits began to decline in the intervention group following the intervention
						Minor: (±) TLAS visits showed a steady decline in the comparison communities. In the intervention communities, however, TLAS visits increased during the first follow-up year, but decreased during the second period. They remained higher during that period than at baseline and were higher in the intervention communities than in the other two comparison communities by a factor of nearly two. But they were lower in the second follow-up period than the first follow-up period. ED visits began to decline the intervention communities, whereas they increased in one comparison community
Powell et al.,128 USA	BA study	Moderate	Education Empowerment Tailoring	12 months	Records	All: (+) the frequency of ED visits decreased by 100 visits or 19.8%	All: (+) the frequency of health-care provider office visits decreased by 932 visits (18.4%). All categories of health-care provider office visits showed declines (freestanding laboratory, accidents, consultations, regular, all other). The frequency of all outpatient visits declined by 1032, or 18.5%. This averaged a little over 1 fewer visit per employee. All 938 subscribers (371 employees and 567 dependents): frequency of health-care provider office visits decreased by 522 visits, or 12.2%. The frequency of subscriber total outpatient visits declined by 495 visits, or 10.8%
Wagner et al.,80 USA	CBA study	Moderate/weak	Education Empowerment	12 months	Self-report		All: (–) seeing a doctor (0 vs. ≥ 1 visits) – at baseline, 84.5% of the intervention group had visited a doctor in the past year, and 86.1% of the control group had done so. At 1 year: 88.9% intervention vs. 86.8% control; difference in differences = 3.7%, or = 1.38 (95% CI 0.97 to 1.95); p = NS. Number of visits: baseline – 3.69 intervention vs. 3.84 control; 1 year: 3.73 intervention vs. 3.67 control; p = NS. After adjusting for observable characteristics, both seeing a doctor and number of visits were not statistically associated with the Healthwise Communities Project intervention

Other outcomes

Three of the five articles of the multicomponent type reported other outcomes (Table 13). Two studies of this type reported no relevant outcome data on satisfaction or intentions. 80,115

TABLE 13 - Other outcomes for multicomponent interventions

BA, before and after; ESCMS, easy-access self-care medical spot; MMI, non-prescription medications for minor illness.

Note

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed.

Study	Design	Quality	Mechanism	Data collection time points	Data source	Impact on other outcomes
Chiu et al.,139 Taiwan	Survey	Weak	Awareness Education Empowerment	9 months	Survey	(+) Satisfaction: 80% (n = 80) of the respondents were aware of the existence of the ESCMSs, and the MMI service was known to most residents and was used by most of them. Compared with respondents who had not used the MMI service (n = 30), those who had used the service (n = 50) considered the ESCMSs more helpful (90% vs. 77%; p < 0.05)
						(+) Intentions: compared with respondents who had not used the MMI service (n = 30), those who had used the service (n = 50) had greater willingness to seek help from the ESCMS in the future (88% vs. 57%; p < 0.01)
Hibbard et al.,79 USA	Non-RCT	Moderate	Education Empowerment	12 and 24 months	Survey	(+) Behaviour: the effects observed for manual use appeared to be greater in the intervention community than in the comparison communities. The intervention respondents were more likely to indicate that using a self-care manual helped them self-treat a symptom and saved them a visit to the doctor. The magnified effect observed in the intervention community for manual use does not occur for users of a nurse advice line or computer program. Thus, it appears that the intervention increased the use of manuals as well as the effects of using a manual
Powell et al.,128 USA	BA study	Moderate	Education Empowerment Tailoring	12 months	Survey	(+) Satisfaction: workshop evaluation – the results show a high level of satisfaction with the guide and the workshop: 97% of employees found the workshop helpful 96% of employees found the workshop enjoyable
						(+) Understanding: 89% of employees reported having a better understanding of when to see the health-care provider; 92% of employees feel that they are wiser health-care consumers
						(+) Intentions: 73% of employees think that they will need to visit their health-care provider less frequently; 66% of employees think that they will be absent from work less often

Satisfaction

Two articles reported on participants’ satisfaction: there were high levels of reported satisfaction both with an educational workshop128 and with the provision of a local minor illness service. 139

Behaviour and intentions

Three articles reported a positive impact. 79,128,139 Two of the cohort studies were of moderate quality79,128 and one was of weak quality. 139

Examples in the NHS

We could find no examples of this type of intervention in routine use in the NHS.

Conclusion and relevance to the NHS

The findings for this type of intervention were essentially the same as for the combination of written information with person-delivered education, and for written information by itself.

Box 5 provides a plain English summary of this section.

BOX 5 - Plain English summary 5

Not a lot of research has been done on using leaflets, people giving training in looking after minor health problems, and advertising in newspapers, etc. at the same time. Some of the research showed a reduction in use of A&E and GPs, but some did not.

A&E, accident and emergency.

6. Self-triage (C.III and A.III)

Size and quality of evidence base

Nine studies reported self-triage interventions: most of them were digital self-triage interventions,53,84,98,104,113,114,140 and two older studies report a paper-based symptom-scoring system akin to self-triage. 132,133 They are presented in Table 14 in alphabetical order within these digital and paper groups. The digital self-triage evidence base was highly relevant to the UK because three articles were undertaken in the UK and four in the Netherlands. Some did not measure outcomes. 104,114 Of the ones that did, some were RCTs,84,98 with another RCT acting as an exploratory study for one of these full RCTs. 53 The formal quality assessment did not identify these RCTs as being of high quality. Our own informal assessments identified concerns that only very educated people were included98 and that there was reliance on self-report for some outcomes on service use. 84 Yardley et al. 53 was an exploratory study/pilot, and so did not have the necessary statistical power to detect a change in GP consultations. The paper-based self-triage evidence base was small and limited to a scoring system for very young children.

TABLE 14 - Impact of self-triage tools on service use

CI, confidence interval; IRR, incidence rate ratio; OOH, out of hours; RR, risk ratio; RTI, respiratory tract infection.

Notes

(+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. Pilot studies are indicated by italics.

Study	Design	Quality	Mechanisms	Time period of measurement	Data source	GP use
Digital
Little et al.,84 UK	RCT	High risk of bias	Awareness Education Empowerment Tailoring	5 and 12 months	Records	All: (±) based on self-report, a modest increase in contacts for NHS Direct among those who had a RTI in the intervention group [37/1574 (2.4%) vs. 20/1661 (1.2%), multivariate RR 2.25, 95% CI 1.00 to 5.07; p = 0.048], but reduced contact with doctors [239/1574 (15.2%) vs. 304/1664 (18.3%), RR 0.71, 95% CI 0.52 to 0.98; p = 0.037]. There was no evidence that self-management advice resulted in delayed consultations for serious illnesses (e.g. lobar pneumonia, meningitis, septicaemia) and, hence, an increased number of hospitalisations; in fact, the number of hospitalisations reduced, albeit not statistically significantly, both in the shorter term (20 weeks) and longer term (1 year). Use of primary care records at 12 months showed no difference in consultations at the GP (RR 0.85, 95% CI 0.65 to 1.12; p = 0.259) and the hospitalisation rate was lower in intervention group, but this was not statistically significant (RR 0.35, 95% CI 0.11 to 1.10; p = 0.073)
Yardley et al.,53 UK	Exploratory RCT	High risk of bias	Education Empowerment Anxiety Tailoring	4 weeks	Self-report	All: of the people in the internet doctor group, 11 (11.6%) had consulted their doctor or used other health-care services (mainly NHS Direct) for their symptoms, compared with a substantially greater proportion ( n  = 21, 17.6%) in the control group, although this difference did not approach significance in this small sample ( p  = 0.22)
van der Gugten et al.98 the Netherlands	RCT	Some concerns	Education Anxiety Tailoring	12 months	Records	All: (–) intervention (N = 314) vs. control (N = 305): No visits: 156 (49.7%) vs. 150 (49.2%); IRR 0.96 (95% CI 0.85 to 1.09); p = 0.532. Record data only, similar for combined record and self-report data 1 visit: 27 (8.6%) vs. 27 (8.9%) 2 visits: 49 (15.6%) vs. 39 (12.8%) 3 visits: 32 (10.2%) vs. 37 (12.1%) > 3 visits: 50 (15.9%) vs. 52 (17.0%)
Paper
Thomson et al.,132 UK	RCT	Some concerns	Education Anxiety Tailoring	6 months	Records	All: (–) no differences were detected between groups in the use of primary care services, excluding number of child health surveillance and immunisation attendances: intervention, median 2; control, median 2 (p = 0.26). Use of OOH service did not differ either (86 vs. 85 consultations; p = 0.93)
Thornton et al.,133 UK	Prospective cohort study	Weak	Education Anxiety Tailoring	6 months	Self-report	All: only reports numbers of mothers reporting contact with health-care professionals. Contacts with their health visitor, midwife and doctor were mostly for minor complaints

Description of interventions and mechanisms

The digital interventions included eConsult104 and an especially developed digital triage intervention. 53,84 Three interventions focused on respiratory problems, either in the general population in the UK53,84 or among children in the Netherlands. 98 The paper-based system was the same in both studies, asking parents to score symptoms. The common mechanisms were education and tailoring, with some interventions also offering attention to anxiety reduction. One of the interventions adopted reinforcement techniques by sending e-mail reminders to users on an ongoing basis to maintain awareness of the intervention. 84

Impact on service use

In Table 14, (+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. Pilot studies are indicated by italics. Five studies measured impact on service use (see Table 14), focusing on GP use. One study was a pilot and one study offered such limited information about impact on service use that we could not interpret the findings. Two studies were of key interest because they were relatively recent digital tools. 84,98 Little et al. 84 offered a positive picture of digital self-triage, showing fewer GP consultations when this intervention was used. In Little et al. ,84 the proportions of GP consultations were 15% in the self-triage group vs. 18% in the control group, based on self-report of those who had a respiratory illness; a similar difference was evident when routine records were used, although this was not statistically significant because all consultations were included, not simply those for respiratory illness. In contrast, there was no difference in GP consultations for the European study. 98 Authors of this study point out that they tested the intervention on a highly educated sample and that their results may not be generalisable to those in lower socioeconomic groups.

Other outcomes

Seven of the nine included articles of the self-triage type reported relevant outcomes (Table 15). Of these seven articles, all but two were assessed as being weak or at high risk of bias; the exceptions being a moderate-quality qualitative study114 and a RCT with only some concerns. 98 Two studies of this type reported no relevant outcome data on satisfaction or intentions at all. 84,132

TABLE 15 - Other outcomes for self-triage interventions

CI, confidence interval; IQR, interquartile range; OOH, out of hours; OR, odds ratio; SD, standard deviation.

Study	Design	Quality	Mechanisms	Data collection time points	Data source	Impact on other outcomes
Digital
Nijland et al.,114 the Netherlands	Qualitative	Moderate	Education	6 months	Survey	(–) Satisfaction: user-friendliness – 106 (40.8%) out of the 260 times the 14 patients used the apps, they faced usability problems; quality of care – 146 (56%) times, they faced quality-of-care problems, such as comprehensiveness of information
Nijland et al.,113 the Netherlands	Survey	Weak	Education	54 weeks	Survey	(–) Enablement (self-care advice): 15% of patients were given self-care advice, so authors considered it to be a very cautious intervention
						(±) Intentions: patients used the triage system to gather information about their complaint (n = 72, 49%), more than for deciding whether or not a doctor has to be contacted (n = 38, 26%)
						Preceding the triage consultation, nearly half of patients (n = 73, 49%) had the intention to visit a GP for their complaint. Of the patients with an intention to visit (n = 72), most expected self-care advice (n = 30) and fewer expected to be advised to visit a GP (n = 22), or had no expectations at all (n = 20). All the patients who expected to be advised to visit a GP (n = 22) received such advice, whereas, of the patients who expected tailored self-care advice (n = 30), the majority (63%) received advice to visit a GP. In most cases, contact with a doctor (GP advice) was given (n = 54, 76%)
						A positive attitude towards advice was strongly related to the perceived usefulness of the advice, as reflected in understandability, reliability of information, applicability in daily life and effectiveness. It appeared that 57% actually had complied with the advice. Patients who complied with the advice were more highly educated (55%) and frequent users of medication (40%); furthermore, the received advice corresponded, in most cases (84%), with expected advice. Reasons for non-compliance were lack of confidence in the advice and fear to follow-up the provided advice
Yardley et al.,53 UK	Exploratory RCT	High risk of bias	Education Empowerment Anxiety Tailoring	4 weeks	Patient enablement instrument	(+) Enablement: the median patient enablement score was significantly greater in the internet doctor group than in the control group [median score of 3 vs. 2, respectively (IQR 0–5), for the whole sample; p = 0.03]
						(+) Satisfaction with web-delivered advice in the internet doctor vs. control groups (n = 332) (scale/item mean for each group) – Total scale (summed items divided by 3): 6.58 (SD 1.96) vs. 5.86 (SD 2.27); p = 0.002 The website gave me all the advice I needed: 6.40 (SD 2.05) vs. 5.63 (SD 2.51); p = 0.002 The website was helpful to me: 6.41 (SD 2.17) vs. 5.72 (SD 2.51); p = 0.007 I felt I could trust the website: 6.91 (SD 2.21) vs. 6.25 (SD 2.54); p = 0.01
						(+) Intentions to consult the doctor declined between baseline and the intermediate (48-hour) follow-up; although the decline was greater in the internet doctor group, this difference did not reach significance. Consultation necessity beliefs and emotional reactions to illness declined at follow-up to a similar extent in both groups. Poor understanding of illness declined in the internet doctor group, but slightly increased in the control group, resulting in a just significant interaction between time and group effects. Self-confidence to self-care remained stable, similar and high in both groups at both time points
van der Gugten et al.,98 the Netherlands	RCT	Some concerns	Education Anxiety Tailoring	12 months	Questionnaire	(+) Satisfaction: Clear information on programme (when applicable) – yes, 99.1% Possibility to find information that was needed (when applicable) – yes, 77.5%; no, 1.3%; partly, 21.3%
						(+) Behaviour: changed (when applicable) – yes, because of the information I went to the doctor, 3.8%; yes, because of the information I did not go to the doctor, 5.8%; no, I wanted to go and I did, 65.4%; no, I did not want to go and I did not go, 25.0%
Verzantvoort et al.,140 the Netherlands	Prospective cohort study	Weak	Education Anxiety	Within 24 hours of app use	Survey, telephone interview	(±) Satisfaction with the app (n = 4456): very dissatisfied, 3.3%; dissatisfied, 8.2%; neutral, 32.8%; satisfied, 46.5%; very satisfied, 9.2% Satisfaction with app (n = 126): 19/126 answered that they were dissatisfied with the app and provided a total of 21 explanations for this. The majority (62%) related to a perceived inability to enter the (complete) story of their illness. Other reasons for dissatisfaction related to the app’s advice (24%) and the structure of the app (14%). Of the participants who were telephoned, 39 did not intend to follow the app’s advice and gave 33 reasons for this. The three main reasons were feelings of being unable to tell their complete story (33%), already having contacted a doctor (27%) or trusting their own judgement better (27%). A total of 65 suggestions were communicated to further improve the app. These suggestions mainly related to the issues previously mentioned: 51% to enable better and more complete entry of all aspects of their illness into the app; 32% to improve the structure, speed and operation; 12% regarding the app’s advice; and 5% related to its layout
						(±) Intentions: follow the app’s advice? Yes, 65%. App’s advice: call doctor, 58%; GP, 15.6%; OOH clinic, 42.4%; do not call doctor, 42.1%; self-care advice, 33.8%; and wait and see, 8.3%. The intention to follow the app’s advice was greatest among participants receiving the advice to contact their GP during daytime (75%), and was 67% for those receiving self-care advice, 61% for contacting the OOH clinic and 56% for wait-and-see advice (p < 0.001) Furthermore, this intention was associated with satisfaction (OR 2.5, 95% CI 2.2 to 2.9; p < 0.001), age < 13 years (OR 1.8, 95% CI 1.3 to 2.3; p < 0.001) and male sex (OR 1.2, 95% CI 1.1 to 1.4; p = 0.045)
Paper
Thornton et al.,133 UK	Prospective cohort study	Weak	Education Anxiety Tailoring	6 months	Questionnaire	(+) Satisfaction: they all found Baby Check easy to use; between 68% and 81% found it useful, and 96% would recommend it to others

Satisfaction

Six of the nine articles of the self-triage type reported on participants’ satisfaction. Four articles reported clear satisfaction with the intervention,53,98,133 one reported that only slightly more than 50% of respondents were satisfied or very satisfied with the intervention app,140 and one study reported barriers to the acceptability of the website for self-triage. 114

Behaviour, enablement and intentions

The mixed findings on satisfaction for the app and website were also reflected in respondents’ intentions: satisfaction with the app or website was highest for those respondents who received advice commensurate with their prespecified intentions. 113,140 Two studies that reported a positive change in behaviour and intentions to consult also reported positive effects on either satisfaction53 or satisfaction and confidence. 98

Current examples in the NHS

The NHS in England has recently introduced NHS 111 Online, a digital self-triage service. Anyone can check a symptom, either to get advice about where and when to seek help, or to get general information about the symptom and how to manage it. Symptoms are checked only for patients aged > 5 years; people using the self-triage for young children are asked to contact NHS 111 by telephone. The service is available at https://111.nhs.uk (accessed 7 September 2022), and a page is shown in Figure 9.

FIGURE 9.

NHS 111 Online screenshot. Contains public sector information licensed under the Open Government Licence v3.0.

Contains public sector information licensed under the Open Government Licence v3.0.

We found that Buckinghamshire CCG had a similar service called ‘AskFirst’, that is a symptom checker that offers self-care advice and helps people find local services [www.buckinghamshireccg.nhs.uk/public/your-services/feeling-unwell/askfirst/ (accessed 18 November 2021)]. It can be downloaded onto a mobile phone.

A large number of general practices in the UK have signed up to eConsult where patients can get advice about self-managing symptoms and direction to services such as pharmacy [https://econsult.net/nhs-patients/how-to-use-econsult (accessed 23 September 2021)]. It appears that a patient makes a query online, and a GP triages that query, rather than self-triage by the patient. We included the Cowie et al. 104 study in our review, which used eConsult. In Cowie et al. ,104 eConsult was described as offering self-care advice, so we included it in our review. However, descriptions of its use routinely in the NHS imply that a GP makes decisions, not the patient.

Conclusion and relevance to the NHS

This type of intervention has potential, but the evidence base needs strengthening for impact on primary care, and does not exist for the impact on emergency care. NHS 111 Online is a relatively new service in England that offers digital self-triage. A RCT of the active use of NHS 111 Online (potentially with reminders to use it), with measurement of its impact on EDs as well as GP consultations, would add considerably to the evidence base.

Box 6 provides a plain English summary of this section.

BOX 6 - Plain English summary 6

There is not a lot of research on computer sites that check patients’ symptoms. This type of service is available in the NHS, so it is important to test if it helps to reduce the use of A&E or a GP for minor problems.

A&E, accident and emergency.

7. Other types of interventions

Size and quality of the evidence base

Seven articles did not fit the six common types of interventions. Six of these were a rare type of intervention94,108,111,116,122,125 and one was a qualitative study focusing on five interventions, and therefore could not be included in the typology. 118

Description of intervention and mechanisms

All six interventions included in the typology had awareness as a mechanism, and four also had education as a mechanism.

Impact in service use

In Table 16, (+) indicates a statistically significant reduction in service use, (–) indicates a null result and (±) indicates that results were mixed. Six of the studies measured the impact on either emergency or primary care. Two of the five articles showed a reduction in ED use and the single study that measured GP use, expecting to see a reduction, showed a reduction. Hou et al. 122 is an interesting study because the intervention was a community campaign to reduce the use of ambulances for minor problems. The indicator measured was ED use, specifically arriving at an ED by ambulance, so it is described in the ‘impact on ED’ column in Table 16. In the UK, minor problems are dealt with by ambulance personnel at a person’s home/location of the call and are not taken to an ED, so Hou et al. ’s 122 indicator is not a good indicator of ambulance use in the UK.

TABLE 16 - Other types of interventions and their impact on service use

ATS, Australasian Triage Scale; BA, before and after; CBA, controlled before and after; CI, confidence interval; IRR, incidence rate ratio; OR, odds ratio; PCP, primary care provider; PED, paediatric emergency department; SD, standard deviation.

Study	Design	Quality	Mechanisms	Time period of measurement	Data source	Impact on ED	Impact on GP consultations
Beal et al.,116 USA	CBA study	Moderate	Awareness Education	12 months	Records, self-report	All: (–) no significant changes in emergency care visits after accounting for covariates	All: non-urgent health-care use significantly decreased for both enrolled and comparison youth in the year after study launch (t[150] = 5.65; p < .01; and t[150] = 5.13; p < .01, respectively), but the decrease for the intervention group was smaller for than hoped for
DeCamp et al.,111 USA	RCT	Some concerns	Awareness Empowerment Persuasion Tailoring	15 months	Records	All: (+) reduced use of EDs. Salud al Día (n = 79) vs. usual care (n = 78): ED visits, mean 1.23 (SD 1.66) vs. 1.82 (SD 1.64); p = 0.03; IRR 1.48 (95% CI 1.04 to 2.12) for control vs. intervention. Frequency of ED use differed (p = 0.03): 0 visits – intervention, n = 33 (42%); control, n = 17 (22%) 1 visit – intervention, n = 19 (24%); control, n = 23 (29%) ≥ 2 visits – intervention, n = 27 (34%); control, n = 38 (49%)
Hou et al.,122 Australia	BA study	Weak	Awareness Persuasion	3 months	Records	All: (+) fewer patients arrived at the ED by road ambulance or other means, as opposed to own transport (OR 0.90, 95% CI 0.80 to 1.00; p = 0.055) Minor: other results were counterintuitive. Patients arriving at the ED had significantly lower clinical urgency after the campaign than before the campaign. The changes were mainly driven by the decrease of ATS 3 from 46.5% to 44.0% and the increase of ATS 4 from 26.7% to 29.5%. Patients attending the ED were significantly less likely to have higher clinical urgency after the campaign than before the campaign (Wald χ2 test; p = 0.0007)
Rector et al.,125 USA	RCT	Some concerns	Awareness Education Persuasion Tailoring	6 months	Records	All: (–) analysed separately for two Medicaid schemes: plan A and plan B. In plan A the percentage of household members in the intervention group who visited an ED differed from the control group by –1.1% (95% CI –3.1% to 0.8%), that is a reduction that was not statistically significant. The difference in plan B was –1.2% (95% CI –4.1% to 1.4%) Minor: (–) ED use by diagnosis – although 50–60% of the visits had a diagnosis discussed in First Look (booklet), the percentages of ED visits for these specific conditions were similar in the intervention and control groups in both health plans	All: the hypothesis was that GP-type visits would increase. Neither health plan differed significantly in the percentage of intervention households or household members who visited an office-based physician. There were also no differences when subgrouping by age (< 19 years and ≥ 19 years)
Robbins et al.,108 UK	RCT	High risk of bias	Awareness Education Empowerment	7 months	Records		All: (±) parents receiving the intervention visited the child health clinic on significantly fewer occasions than parents in the control group: mean 6 vs. mean 9, respectively (p = 0.039). There were no other significant differences in service use between the two groups: GP visits, prescriptions (e.g. antibiotics), minor illness, nurse telephone and home visits, health visitor telephone and home visits Minor: (–) there were no significant differences in service use between the two groups for minor illness
Sturm et al.,94 USA	RCT	High risk of bias	Awareness Education Tailoring	6 and 12 months	Records	All: (–) PED follow-up at 6 and 12 months. Control (N = 168) vs. intervention (N = 164): high acuity follow-up – 6 months: 18 (10.7%) vs. 22 (13.4%); p = 0.56 12 months: 69 (41.1%) vs. 74 (45.1%); p = 0.53	All: the intervention group had 203 well-child visits (1.23 per patient), compared with 195 (1.16 per patient) in the control group (rate difference 0.02 per person-year (95% CI –0.2 to 0.3; p = 0.87). The intervention group had 139 sick visits during the follow-up period (0.85 visits per patient), compared with 109 in the control group (0.65 per patient) (rate difference 0.19 per person-year (95% CI 0.013 to 0.39; p = 0.036). That is, the treatment group used the PCP more, as planned
						Minor: (±) PED follow-up at 6 and 12 months. Control (N = 168) vs. intervention (N = 164): low acuity follow-up – 6 months: 31 (18.4%) vs. 21 (12.8%); p = 0.14 12 months: 91 (54.2%) vs. 70 (42.7%); p = 0.047

Other outcomes

Only two of the seven articles of this miscellaneous type reported relevant outcomes (Table 17). One was a controlled before-and-after study of moderate quality. 116 This study of adolescents leaving the care system reported that the majority found the intervention, the I Care Guide, helpful. 116 The second was a qualitative study using focus groups to rank caregivers’ preferences regarding different types of educational interventions for childhood illness: the most preferred was the mobile app, and the least preferred were printouts for specific illnesses; this was the same regardless of a participant’s baseline level of health literacy. 112 There was little qualitative information available about why people preferred an intervention, although there were some comments about the accessibility of the mobile app and the comprehensiveness of the information.

TABLE 17 - Other outcomes for ‘other’ types of intervention

CBA, controlled before and after.

Study	Design	Quality	Mechanisms	Data collection time points	Data source	Impact on other outcomes
Beal et al.,116 USA	CBA study	Moderate	Awareness Education	12 months	Records, self-report	(±) Satisfaction: the majority of adolescents reported that the I Care Guide was somewhat (46%) or very helpful (42%), and 78% of participants who responded to surveys at the 12-month follow-up reported that they still had the guide (42% of all participants who received a guide at enrolment). The majority of youth (80%) reported using the I Care Guide at least once during the study, with sections discussing emergency care, reproductive health and symptoms most commonly endorsed
Ohns,118 USA	Qualitative	Strong	Education	Cross-sectional	Focus groups	(+) The preferred method of education identified by the focus groups was the mobile app, Kids Doc, created by the American Academy of Paediatrics. Second, was the 24-hour nurse call line; third was the book, My Child is Sick! Fourth was the booklet, Caring for Your Sick Child: Managing Common Infections at Home; and fifth was the patient education printout specific to diagnosis. Of note, 25 of the 30 caregivers ranked the Patient education printout as their fourth (n = 8) or fifth (n = 17) choice (±) There was no statistical significance found when comparing literacy scores and preferred method of education (ranking the educational option as their first or second choice). A one-way ANOVA with Tukey post hoc comparisons indicated that those who preferred the mobile app were younger (27.6 ± 5.8 years) than those who preferred the 24-hour nurse line (34 ± 4.9 years) and those who preferred the My Child is Sick! book (30 ± 3 years) (p = 0.03)

Current examples in the NHS

Not relevant.

Conclusion and relevance to the NHS

It is difficult to draw conclusions about this miscellaneous type.

Impact on service use: multivariable regression

Using a logistic regression of reduction in ED use versus null, we tested whether or not characteristics of the study, context or intervention were more likely to result in reductions. We repeated this for reduction in GP consultations versus null. We tested the decade in which the study was published, the country, the place of delivery, whether or not the intervention was targeted at people in poverty/with low health literacy, whether it was targeted at a parent of a young child or all people, whether it was given at a specific or a general time, the type of intervention, the study design, the quality of research, the time period measured, whether or not records were used for measurement, the presence of mechanisms of impact, the Flesch score and usability. The Flesch score and usability were available for some papers only. We could not measure approach to intervention development because this was so poorly described in most papers. Numbers were small for this analysis (low statistical power) and a large number of variables were tested, so it is a problematic analysis. There were no statistically significant findings, with the exception of the decade in which articles were published: the 2000s had a greater proportion of studies showing a reduction in ED attendance and GP consultations. There is no explanation for this.

Background

Towards the end of the study, we held a virtual PPI event (in September 2021) followed by a virtual stakeholder event that included PPI members (in October 2021). The aim was to elicit stakeholders’ perceptions of the findings and actions that could be taken. We held the PPI event first so we could hear PPI views alone before hearing the views of all stakeholders together. Both meetings were virtual because of COVID-19. In the stakeholder event, we aimed to involve representatives from the following groups: patients, carers and members of the public, CCGs, NHS England urgent and emergency care, general practice, EDs and emergency ambulance services; we also aimed to involve the NHS England health literacy lead. We report the two meetings together because they discussed similar issues.

Stakeholders attending the events

Ten PPI members attended the PPI event. Nineteen stakeholders attended the stakeholder event along with research team members:

• eight PPI representatives (two provided input outside the meeting)

• five NHS England and CCG commissioners

• one representative from NHS England Health Literacy

• one ED consultant

• one GP

• one NHS trust manager

• two ambulance service representatives (a medical director and a paramedic).

We invited the same stakeholders from the first stakeholder event and some additional stakeholders who had learnt about our review. As the meeting was held virtually, people were able to attend from different parts of England, including the north-east, south-west, London and Yorkshire.

The events

At the PPI event, the team presented study findings in a series of short presentations, leaving most of the time for discussion. We heard useful feedback on how the findings may be relevant to the NHS, the role of family and friends when accessing services, and factors that influence people’s use of urgent care. The second stakeholder event was held on 13 October 2021 and lasted 2 hours. The team presented the findings of the review, including the following:

• the types of interventions identified

• the effectiveness of different types of interventions

• the readability and user-friendliness of interventions

• implications of the review for the NHS, considering how feasible, appropriate, meaningful and effective the interventions were.

We also discussed potential ways of disseminating the findings to influence practice. Margaret Ogden (PPI co-applicant) helped to organise the events, gave a presentation to stakeholders about the user-friendliness of the interventions, and supported the facilitation of the events.

The issues raised and actions taken

After the events, we recorded the points made by attendees at the PPI event and the second stakeholder event and how we would address each point in the report (Table 18). We circulated our write-up to the stakeholders so they could add to the document; none responded.

Summary of findings

A total of 67 articles (64 studies) were included: 37 from the USA, 16 from the UK, 12 from the rest of Europe and two from the rest of the world. There were seven intervention types: navigation tools directing people to the range of services available (n = 7); written education about managing minor health problems in booklet or website format (n = 17); person-delivered education (n = 5); written education with person-delivered education (n = 17); multicomponent of written education, person-delivered education and mass media campaign (n = 5); self-triage (n = 9); and other (n = 7). Most articles reporting satisfaction with the intervention, enablement and perceived changes to behaviour showed positive results. Nineteen out of 30 (63%) articles measuring impact on ED attendances showed a reduction. Sixteen out of 27 (59%) articles measuring impact on GP consultations showed a reduction. Returning to the logic model specified earlier in the report (see Figure 1), the interventions appeared to improve short-term and intermediate-term outcomes (when measured), but there was variation in whether or not they affected the longer-term outcomes of reduced use of EDs or GPs. There was no indication that variation in the evidence base was explained by any of a range of population, research design or intervention characteristics, including complexity of the intervention. Interventions ranged in reading age and user-friendliness. The few articles reporting safety (n = 8) showed that interventions were safe.

Strengths and limitations of the evidence base

The evidence base had six strengths. First, the evidence base was generally very recent (almost half of the included articles had been published since 2010: 32/67), which increases its applicability to current contexts. Second, it derives from the multiple settings of primary care, secondary care and the community, rather than interventions delivered in a single setting only. Third, around one in four articles were conducted in the UK (16/67), and so a sizeable part of the evidence base is highly applicable to the context of interest. Fourth, it was extensive, covering a range of intervention types. Fifth, multiple study designs were used, including a high number of RCTs and quasi-experimental studies. Sixth, there were a considerable number of studies aimed at parents of young children, a group that has been identified as high users of emergency and primary care for minor health problems. 4

The evidence base had seven limitations. First, members of the PPI panel reflected that the evidence base was much smaller than they had expected. Second, the quality of the evidence base was often assessed as being poor, with many RCTs assessed as being at high risk of bias, and all but one non-randomised study being assessed as being of only moderate or weak quality. Third, the evidence base was highly heterogeneous in terms of the study designs, intended populations, the details of the interventions themselves, outcome measurements, and lengths of follow-up. This was especially the case for the measurement of service use outcomes, both in terms of the indicator used and the time period of measurement. This prevented the statistical pooling of interventions and outcomes, so the size of any positive effects for any particular intervention type could not be precisely quantified. Fourth, a large proportion of the studies were conducted in the USA (37/67), which has a financially distinct health and social care system, compared with the UK, which is likely to moderate patient decisions around accessing care. Fifth, cost data or analyses were rarely reported; if they were reported, it was sometimes as a discussion point, rather than details of methods used to calculate costs. So it could not be determined whether or not even a small statistically significant effect might be cost-effective, especially for relatively basic, low-resource intervention types. Sixth, there was almost no explicit assessment of safety. Seventh, there were surprisingly few interventions delivered in emergency care, and none delivered by ambulance services. The emergency care-delivered interventions were dominated by recently published North American studies of interventions aimed at parents of young children, rather than aimed at adults seeking help for their own health.

Relevance of the evidence base

Approximately one-quarter of articles were conducted in the UK (16/67), and so a sizeable part of the evidence base was highly applicable to the UK context. The interventions in the evidence base were similar to interventions in current use in the NHS, for example booklets about minor health problems, NHS website symptom checker, NHS 111 Online.

Fit with other evidence

Strengths and limitations of the review

The systematic review had eight strengths. First, all processes (study selection, data extraction and quality assessment) were carried out by experienced reviewers. Second, the literature search was undertaken by a qualified senior information specialist and was comprehensive, involving the interrogation of multiple databases, supplemented by further techniques, such as reference-checking of included studies and reviews, searching of many grey literature sources (e.g. websites of relevant organisations) and focused searching for the publications of known relevant authors. This was an important issue because of the nebulous nature of the intervention we were interested in. Third, the data were not appropriate for statistical analysis, but the synthesis involved the application of a novel typology, going beyond standard narrative synthesis techniques to deliver practical findings. Fourth, focusing specifically on health literacy interventions, rather than considering them as a subgroup in a wider set of interventions, allowed for a more nuanced understanding of this type of intervention. Fifth, the focus on both emergency and primary care allowed the bringing together of similar interventions used for the same purpose of reducing the use of services. Sixth, assessing the readability and user-friendliness of interventions, as well as effectiveness, drew attention to a strength or weakness of an intervention that might affect its ability to have an impact on service use. Seventh, the extensive involvement of PPI, including having a PPI co-applicant, four events for PPI, and inclusion of members of the Deep End Patient Panel from one of the most socially deprived areas in the UK, grounded the review and findings in the reality of people’s lives. Eighth, the engagement of a wide range of stakeholders helped to ensure that the findings were useful to them. For example, it led to us identifying an intervention produced and used within the NHS currently that we included in our user-friendliness assessment and could recommend for wider use in terms of being a high-quality resource.

The systematic review had seven limitations. First, we found studies mainly from the USA, the UK and the rest of Europe. We may have missed research from the rest of the world, but it is also plausible that interest in improving health literacy to reduce the use of emergency and primary care is limited to those countries. There is some support for this: there has been an exponential growth in publications about health literacy more generally, with a dominance of the USA, Australia and the UK,154 which makes it unlikely that we missed global health literature. Second, we are likely to have missed grey literature and literature that did not measure primary and emergency care use. Towards the end of the review, one of our stakeholders identified a small-scale evaluation of The Little Orange Book undertaken by a patient group. 141 The findings were very similar to those in our review in that users of the intervention found the intervention very helpful. An additional finding was that staff regarded it highly. From discussions with our stakeholders, there are likely to be more of these types of studies that are not published in peer-reviewed journals or easily found using grey literature searches. It is also possible that we missed relevant studies published in databases because we required a focus on reduction of service use. When finalising the report, we found an article reporting the development and evaluation of a component of a mass media campaign to reduce the use of emergency ambulance services for non-emergencies. 155 The campaign was successful in terms of reaching people and changing their attitudes towards ambulance use. That is, it was similar to the findings of our review in that it was successful for short-term outcomes. The next step for the research group was to measure the effect on ambulance use. Third, the majority of each systematic review process (study selection, data extraction and quality assessment) was performed by a single reviewer (CC). However, as an initial step, and to ensure complete clarity and consistency in the application of inclusion criteria and data extraction, four project team members (CC, LP, AOC and AF) all screened 100 titles and abstracts, and extracted three studies, to identify, discuss and resolve any ambiguities. There was also independent checking by a second reviewer (LP) of a minimum of 10% of the article data extractions and critical appraisals (n = 7), and disagreements were found to be rare and easily resolved. All decisions were also checked with reference to a third member of the project team (AOC or AF). Fourth, it was sometimes difficult to stay clear about the boundaries of the research when health professionals such as pharmacists and health visitors offer advice to patients about managing future minor health problems. We have ensured that we are clear that the review was about interventions to help patients make decisions themselves without input from health professionals. Fifth, we could not access all the interventions to do readability testing and user-accessibility assessments. Sixth, we had intended to use context–mechanism–outcome chains in our synthesis, but this did not feel like a helpful process, so we did not do it. Seventh, the ‘feasible, appropriate, meaningful and effective’ framework was not helpful during our second stakeholder event. It may have been more useful if we had identified one intervention or intervention type that stood out as effective and warranted further detailed discussion about its use in the NHS.

Strengths and limitations of the patient and public involvement activity

The GRIPP2 checklist46 and Pollock et al. ’s48 framework were used to help us reflect on our PPI. The PPI activity had seven strengths. First, we involved 14 different people, each with different characteristics and life experiences. Second, we had a PPI member (MO) as an active member of the project team. This ensured that, when conducting the review, there was always a PPI voice who provided valuable advice, feedback and support throughout the project. Third, one of our co-applicants was the stakeholder lead (AF), ensuring that there was researcher and administrator time dedicated to supporting people. Fourth, we involved PPI members throughout the review, including at protocol development stage, developing the UFAT and when considering the implications of the results. Fifth, we kept people informed throughout the study by having a project newsletter, and our stakeholder lead (AF) encouraged telephone and e-mail exchanges if people wanted to contribute outside meetings. Sixth, we remunerated people for their involvement. Seventh, we provided additional support to facilitate involvement, including training on engaging in virtual meetings and on using Twitter.

The PPI activity had three limitations. First, because of COVID-19, we held all our events virtually. We feel that we mitigated against problems with technology, but could have made more use of small breakout rooms during meetings to allow quieter members to contribute. Virtual meetings also reduced any opportunity for informal rapport-building, such as chatting over coffee. Second, there was a long gap of no PPI meetings between the November 2020 and June 2021 meetings. Although we kept the PPI group updated about the review through newsletters, an interim meeting in the spring may have been valuable. We did not undertake any meetings in that period because we were at the search/extraction stage of the review and we wanted to work with PPI members when we had emerging findings to discuss. Third, the university’s financial systems made remuneration complex for some members. This caused stress and resulted in delays in payments for some people. Our department is working with the university to simplify PPI payments so that it does not become a barrier to engagement.

Reflections on doing this type of research

We have five recommendations for researchers doing this work in the future. First, measure actual use of services using records, rather than relying on self-report. Second, be consistent as a research community about the measurement time period. The most common measurement time periods were 6 months and 1 year. Third, we noted that some studies used records from a single hospital only; intervention users may have actively sought care outside this hospital if they felt that their local ED was trying to persuade them not to attend. If possible, we recommend also measuring impact on surrounding EDs. Fourth, not many studies measured safety, and this is a really important issue to patients and the public. It is heartening that those that measured safety found these interventions to be safe. Fifth, we felt that the low scoring in the formal quality assessment of these studies was harsh given the difficulty of undertaking research in such complex environments. This was particularly the case for the RCTs in our review. Researchers could review these assessment tools for application to pragmatic RCTs.

Implications for decision-makers

A number of health literacy interventions are used in the NHS currently. Commissioners and service providers have developed booklets and websites to help parents of young children and the general population to manage minor health problems and decide where best to go for care. Policy-makers have developed NHS 111 Online for self-triage. Rather than develop further new interventions, commissioners and service providers could contact their counterparts in different regions and make use of existing good-quality interventions. It might also be helpful to have a space where local evaluations can be shared. Once commissioners and service providers locate a relevant intervention, they could assess the reading age and user-friendliness, and improve these if necessary. This is important for websites as well as information in booklet format. Our PPI members were keen that commissioners and service providers look for ways to make existing interventions available to more people (e.g. through school) and available in different languages or formats, such as online videos. We know that NHS 111 Online is advertised on bus stops and some navigation tools are posted on social media, yet the PPI members often had not seen them. The PPI members were concerned about a lack of digital literacy in some parts of the population and a lack of access to technology; they described initiatives they had come across such a health champion in general practice who could help people to look things up on a tablet. They also described a system that, when added to a website, reads the website content aloud and translates it into different languages. Finding ways of making existing good-quality interventions more accessible and used by the general population might benefit people’s ability to care for minor health problems and seek health care from the most appropriate place. It is also important to continue to evaluate existing interventions in terms of measuring impact on service use and cost-effectiveness, because this cannot be taken for granted.

Recommendations for further research

• Focus on understanding how to improve access to existing good-quality interventions (particularly those currently used in the NHS), rather than develop new interventions.

• Continue to evaluate the effectiveness and cost-effectiveness of these types of interventions, given that these types of interventions have potential to reduce the use of emergency and urgent care, and that self-care for minor health problems may be necessary during pandemics and to facilitate the sustainability of emergency and urgent care provision.

• Measure the safety of these interventions.

• Focus on understanding why an intervention did or did not work, rather than only measuring impact on service use, that is use mixed-methods evaluations of RCTs/quasi-experimental designs and process evaluation/qualitative research.

• Undertake more rigorous economic evaluation of these interventions.

• Evaluate interventions aimed at adults using EDs for minor health problems because few studies did this.

• Pay more attention to ‘displacement’, whereby people use different services from the ones being measured.

• As a research community, measure changes in service use in a consistent way and at a consistent time post intervention. This will facilitate a future meta-analysis. Consider measuring change in attendances at a service for minor health problems/non-urgent reasons at 1 year post intervention.

• Consider ways of making it easier to monitor change in service use over time using routine data.

Conclusions

Health literacy interventions have potential to reduce emergency and primary care use. They need further rigorous evaluation to determine which work best and for whom.

Acknowledgement for individuals

Thanks to Vera Fibisan for her administrative skills, support for the PPI endeavour and organisation of the stakeholder events.

Thanks to PPI members and stakeholders for their time and insights (some people did not want their names included):

• Aidan Barry

• Kristin Bash

• Jude Beng

• Jonathan Berry

• Mary Bramhall

• Jo Chalmers

• Colleen Ewart

• Debbie Floyd

• Kate Grudzinska

• Carrie Hedderwick

• Carole Hobson

• Sarah Illingworth

• Linda Jones

• Kenye Karemo

• Kerry Knox

• Rashmi Kumar

• Liz Lingard

• Margaret Mellor

• Mike Oliver

• Sam Pilkington

• Vickie Riley

• Niro Siriwardena

• Alan Smith

• Debs Smith

• Jill Smith

• Irene Soulsby

• Josie Soutar

• Helen Twohig

• Frances Verey.

Contributions of authors

Alicia O’Cathain (https://orcid.org/0000-0003-4033-506X) (Professor of Health Services Research) led the study.

Alexis Foster (https://orcid.org/0000-0002-7978-2791) (Research Fellow) led the stakeholder engagement and the user-friendliness of interventions component, and co-led the PPI.

Christopher Carroll (https://orcid.org/0000-0002-6361-6182) (Reader) conducted the systematic review.

Louise Preston (https://orcid.org/0000-0001-7477-4517) (Senior Research Fellow) contributed to design and conduct of the systematic review.

Margaret Ogden (https://orcid.org/0000-0002-4611-5095) (PPI member) co-led PPI input.

Mark Clowes (https://orcid.org/0000-0002-5582-9946) (Research Associate) conducted the searches.

Joanne Protheroe (https://orcid.org/0000-0002-9608-1487) (Director of General Practice Education, Director of Clinical Academic Training, Past Chair of Health Literacy UK, and NHS Clinical Adviser for Health Literacy) advised on health literacy aspects of the review.

Ethics statement

This study is a systematic review with stakeholder involvement and did not need ethics approval.

Data-sharing statement

This report generated no new data; therefore, there are no further data to be shared other than those included in the report. Any further queries should be submitted to the corresponding author.

Disclaimers

This report presents independent research funded by the National Institute for Health and Care 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, the HSDR 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, the HSDR programme or the Department of Health and Social Care.

Wrong population (not patients accessing services for minor illness, non-urgent care, e.g. chronic illness)

1. Agarwal G, Pirrie M, McLeod B, Angeles R, Tavares W, Marzanek F, et al. Rationale and methods of an evaluation of the effectiveness of the community paramedicine at home (CP@home) program for frequent users of emergency medical services in multiple Ontario regions: a study protocol for a randomized controlled trial. Trials 2019;20:75.

2. AHC M. Better patient education can lead to lower medical costs: focus on health literacy. Case Manage Advis 2020;31:1–3.

3. Blancafort Alias S, Monteserin Nadal R, Moral I, Roque Figols M, Rojano I Luque X, Coll-Planas L. Promoting social capital, self-management and health literacy in older adults through a group-based intervention delivered in low-income urban areas: results of the randomized trial AEQUALIS. BMC Public Health 2021;21:84.

4. Botelho A, Dias IC, Fernandes T, Pinto LMC, Teixeira J, Valente M, et al. Overestimation of health urgency as a cause for emergency services inappropriate use: insights from an exploratory economics experiment in Portugal. Health Soc Care Community 2019;27:1031–41.

5. Bronsky ES, McGraw C, Johnson R, Giordano K, Orlando A, Bar-Or D. CARES: a community-wide collaboration identifies super-utilizers and reduces their 9-1-1 call, emergency department, and hospital visit rates. Prehosp Emerg Care 2017;21:693–9.

6. Cambon J, Cordier T, Munnich EL, Renda A, Kapur B, Hoxhaj S, et al. Effects of educational messaging on urgent and emergent care-seeking behaviors among publicly insured populations. Am Health Drug Benefits 2018;11:86–93.

7. Greene JC, Haun JN, French DD, Chambers SL, Roswell RH. Reduced hospitalizations, emergency room visits, and costs associated with a web-based health literacy, aligned-incentive intervention: mixed methods study. J Med Internet Res 2019;21:e14772.

8. Hibbard JH, Greene J. The impact of an incentive on the use of an online self-directed wellness and self-management program. J Med Internet Res 2014;16:e217.

9. Judson TJ, Odisho AY, Neinstein AB, Chao J, Williams A, Miller C, et al. Rapid design and implementation of an integrated patient self-triage and self-scheduling tool for COVID-19. J Am Med Inform Assoc 2020;27:860–6.

10. Kelley L, Capp R, Carmona JF, D’Onofrio G, Mei H, Cobbs-Lomax D, et al. Patient navigation to reduce emergency department (ED) utilization among Medicaid insured, frequent ED users: a randomized controlled trial. J Emerg Med 2020;58:967–77.

11. Michelen W, Martinez J, Lee A, Wheeler DP. Reducing frequent flyer emergency department visits. J Health Care Poor Underserved 2006;17(1 Suppl.):59–69.

12. Nejtek VA, Aryal S, Talari D, Wang H, O’Neill L. A pilot mobile integrated healthcare program for frequent utilizers of emergency department services. Am J Emerg Med 2017;35:1702–5.

13. Pascual FT, Hoang K, Hollen C, Swearingen R, Hakimi AS, King JA, et al. Outpatient education reduces emergency room use by patients with epilepsy. Epilepsy Behav 2015;42:3–6.

14. Pillow MT, Doctor S, Brown S, Carter K, Mulliken R. An Emergency department-initiated, web-based, multidisciplinary approach to decreasing emergency department visits by the top frequent visitors using patient care plans. J Emerg Med 2013;44:853–60.

15. Yu J, Zhang HW, Shao YK, Lei Y, Chen H, Pu ZH, et al. A smartphone-based online tool for prehospital self-triage of COVID-19. Chin J Acad Radiol 2020;3:175–80.

Wrong intervention/not a primary research evaluation

1. Anonymous. Distribution of health handbook leads to reduction in ER visits. Healthc Demand Dis Manag 1999;5:31–2.

2. Anonymous. CMs educate members on appropriate ED use: program targets frequent flyers. Case Manag Advis 2007;18:125–6.

3. Anonymous. Program helps beneficiaries avoid unnecessary ED visits: care coordination steers members toward primary care. Case Manag Advis 2008;19:65–7.

4. Anonymous. Reducing ‘frequent fliers’ in the emergency room: Medicaid members connect with primary care. Case Manag Advis 2008;19:51–2.

5. Anonymous. Patient education program slashes ED readmissions. ED Manag 2009;21:42–3.

6. Anonymous. In review of ED utilization reduction strategies, data regarding impact on safety, outcomes in short supply. ED Manag 2014;26:8–10.

7. Botelho A, Dias IC, Fernandes T, Pinto LMC, Teixeira J, Valente M, et al. Overestimation of health urgency as a cause for emergency services inappropriate use: insights from an exploratory economics experiment in Portugal. Health Soc Care Community 2019;27:1031–41.

8. Chastonay OJ, Lemoine M, Grazioli VS, Canepa Allen M, Kasztura M, Moullin JC, et al. Health care providers’ perception of the frequent emergency department user issue and of targeted case management interventions: a cross-sectional national survey in Switzerland. BMC Emerg Med 2021;21:1–10.

9. DeVries A, Li C-H, Oza M. Strategies to reduce non-urgent emergency department use: experience of a Northern Virginia Employer Group. Medical Care 2013;51:224–30.

10. Eminovic N, Wyatt JC, Tarpey AM, Murray G, Ingrams GJ. First evaluation of the NHS Direct Online clinical enquiry service: a nurse-led web chat triage service for the public. J Med Internet Res 2004;6(2).

11. Grossman LK, Rich LN, Johnson C. Decreasing non-urgent emergency department utilization by Medicaid children. Paediatrics 1998;102(1 Pt 1):20–4.

12. Gustafsson S, Martinsson J, Wälivaara BM, Vikman I, Sävenstedt S. Influence of self-care advice on patient satisfaction and healthcare utilization. J Adv Nurs 2016;72:1789–99.

13. Gustafsson S, Savenstedt S, Martinsson J, Walivaara B-M. Need for reassurance in self-care of minor illnesses. J Clin Nurs 2018;27:1183–91.

14. Houghton J. Minor illness management: empowering parents through shared knowledge. Paediatr Nurs 2005;17:24–5.

15. Kantonen J, Lloyd R, Mattila J, Kauppila T, Menezes R. Impact of an ABCDE team triage process combined with public guidance on the division of work in an emergency department. Scand J Prim Health Care 2015;33:74–81.

16. Nguyen CA, Shih JA, Lin KV, Aladesanmi OA. Targeting national emergency department overuse: a case for primary care, financial incentives, and community awareness. Harvard Health Policy Rev 2014;14:23–6.

17. Peetoom KKB, Smits JJM, Ploum LJL, Verbakel JY, Dinant G-J, Cals JWL. Does well-child care education improve consultations and medication management for childhood fever and common infections? A systematic review. Arch Dis Child 2017;102:261–7.

18. Trevett AJ, Martin JR, Ross WA, Macfarlane E. The introduction of a daily telephone advice service: how is it used and is it worth the effort? Scott Med J 1998;43:57–8.

19. Williams K, White L, Plorde M, Eisenberg M. Empowering the patient: after-care instructions reduce calls and improve quality of life. J Emerg Med Serv 2008;33:42–9.

Wrong outcome

1. Baker MD, Monroe KW, King WD, Sorrentino A, Glaeser PW. Effectiveness of fever education in a paediatric emergency department. Paediatr Emerg Care 2009;25:565–8.

2. Lass M, Tatari CR, Camilla Hoffmann M, Huibers L, Maindal HT. Contact to the out-of-hours service among Danish parents of small children – a qualitative interview study. Scand J Prim Health Care 2018;36(2).

3. Olsen JR, Gallacher J, Piguet V, Francis NA. Development and validation of the Molluscum Contagiosum Diagnostic Tool for parents: diagnostic accuracy study in primary care. Br J Gen Pract 2014;64:e471–6.

4. South J, Darby F, Bagnall A-M, White A. Implementing a community-based self-care training initiative: a process evaluation. Health Soc Care Community 2010;18:662–70.