CHOICE: Choosing Health Options In Chronic Care Emergencies

Article history

The research reported in this issue of the journal was funded by PGfAR as project number RP-PG-0707-10162. The contractual start date was in December 2013. The final report began editorial review in May 2015 and was accepted for publication in September 2016. As the funder, the PGfAR programme agreed the research questions and study designs in advance with the investigators. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The PGfAR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Simon de Lusignan has received funding from Eli Lilly and GlaxoSmithKline (GSK) Biologicals. Funding has also been received from the Department of Health for the National Evaluation of Improving Access to Psychological Therapies (IAPT). Elspeth Guthrie reports grants from the National Institute for Health Research (NIHR) during the conduct of the study. Karina Lovell reports grants from NIHR, the National Institute of Mental Health (NIMH) and Arthritis Research UK during the conduct of the study. Chris Dickens reports grants from NIHR during the conduct of the study. Linda Davies reports grants from the NIHR, Medical Research Council, Economic and Social Research Council, Macmillan, Cancer Research UK, Arthritis Research UK and Central Manchester University Hospitals NHS Foundation Trust (CMFT) during the conduct of this study. Peter Salmon reports grants from Marie Curie Cancer Care, the Liverpool Institute of Health Inequalities Research, Merseycare NHS Trust, Royal Liverpool and Broadgreen Hospitals NHS Trust, the Economic and Social Research Council, and the Medical Research Council during the conduct of the study. Navneet Kapur reports other research funding from NIHR, the Department of Health and the Healthcare Quality Improvement Partnership during the conduct of the study. He was not in receipt of any industry funding. He was involved in National Institute for Health and Care Excellence guidelines and Department of Health (England) advisory groups on suicidal behaviour unrelated to the conduct of the current study.

Disclaimer

The views and opinions expressed herein are those of the authors and do not necessarily reflect those of the National Institute for Health Research Public Health Research programme or the Department of Health.

Copyright statement

© Queen’s Printer and Controller of HMSO 2017. This work was produced by Guthrie et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Background

Unscheduled care is defined as any unplanned contact with the health service by a person requiring or seeking help, care or advice. 1 It includes a wide range of service contacts from specialised hospital support, emergency hospital admissions (EHAs), attendances at emergency departments (EDs), attendances at minor injury units, unplanned primary care call-outs and, finally, self-care. 2 Unscheduled care also includes both urgent care, which refers to conditions that require assessment and treatment within 7 days, and emergency care, which needs assessment and intervention within 24 hours. 2

Use of urgent or unscheduled services is high. Sixteen per cent of the general population access unscheduled care over any 4-week period,3 and attendances at EDs in England have been increasing year on year (Figure 1). The total number of ED attendances in England has risen from 13.9 million in the year 1987–8 to 14.4 million in 1997–8, 19.1 million in 2007–8 and 22.4 million in 2013–14. 4

FIGURE 1.

Total number of annual attendances at EDs in England 2008–15. 4

According to a report published by the Department of Health (DH) in 2013, in 2012–13, there were 5.3 million emergency admissions to hospital, costing approximately £12.5B. 5 In 2014–15, the total number of EHAs had risen to 5.5 million. 4 Reducing unscheduled care has become a DH priority. 6

The increase in EHAs has largely resulted from an increase in short-stay admissions of patients presenting to EDs. These ‘short-stay’ admissions (< 2 days) have increased by 124% over the last 15 years; in comparison, longer admissions have increased by 14%. 5

Avoiding unnecessary EHAs is a major concern for the NHS not only because of the costs associated with these admissions, but also because of the pressure and disruption caused to elective health care and to the people admitted. In a recent report from The King’s Fund, it was suggested that emergency admissions of people with long-term conditions (LTCs) that could have been managed in primary care cost the NHS £1.42B annually and that this could be reduced by 8–18% through investment in primary care- and community-based services. 7

It is estimated that 17.5 million adults in the UK have at least one LTC, by which is meant a chronic physical or mental health condition. 8,9 Over 70% of the health-care budget in England is spent on the care of people with LTCs,10 and many people live with multiple conditions. LTCs account for 50% of general practice consultations and 70% of inpatient bed-days per year in the UK. 11

During this programme of research, there was a growing awareness at a national level of the importance of improving the mental health of people with LTCs. 12 There was also a recognition of the need to better integrate physical and social health care, particularly for the elderly. 13–15 The national mental health strategy for England calls for patient-centred management together with joined-up personalised pathways and systems. The strategy states that interventions must be as efficient as possible at delivering outcomes that are effective, cost-effective and safe. 16 There is the expectation that better joined-up care will reduce costs, particularly the burden on emergency services. A recent report from The King’s Fund, entitled Bringing Together Physical and Mental Health: A New Frontier for Integrated Care, makes a strong case for change to the current ways in which services are organised. 17

There are major challenges to ensure that patients’ needs are met holistically, effectively and efficiently. Importantly, the evidence base underpinning many current recommendations is limited or patchy.

This programme of research addressed several of the key areas described above and its findings are both informative and timely for those involved in policy, practice and research. Instead of focusing on one LTC, the programme included four common exemplar LTCs. This allowed us to compare people with different LTCs but also to study the effects of multimorbidity on use of unscheduled care.

The use of unscheduled care is common in people with LTCs, such as asthma and chronic obstructive pulmonary disease. 18–20 It is argued that, in the case of certain long-term physical conditions, more effective management and treatment in primary care could reduce ED attendance and EHAs, and there has been, and remains, a drive from the DH to reduce urgent care and emergency readmissions to hospitals of those people with long-term physical conditions who could be managed in the community. 11

It is difficult to determine whether or not the use of unscheduled care by an individual is appropriate, as it depends on a value judgement made by a clinician at the point of contact, and the criteria for determining inappropriate contact vary considerably. 21 Although, undoubtedly, some use of unscheduled care is unjustified, it is recognised that, for many people, the use of unscheduled care is entirely appropriate and necessary at the time they access services. An important question, however, is whether or not such contacts could be prevented, in certain groups of patients, by improved management and treatment of their physical health problems, so that crises are avoided.

The provision of unscheduled care is complex, and the relationship between different components of unscheduled care and scheduled care is difficult to disentangle. For example, increased use of general practitioner (GP) out-of-hours (OOH) services, or seeing the same GP on a regular basis, may result in a reduced need to attend the ED. So an increase in one aspect of unscheduled care or scheduled care may result in a decrease in another aspect of unscheduled care. 22 With this in mind, many of the analyses in this programme focus on secondary care aspects of unscheduled care, namely ED attendance and EHAs. These are the two most costly aspects of unscheduled care and the two that government has identified as a priority area to reduce. 11 Where relevant, however, and where possible in the programme, we also comment on other kinds of unscheduled care use.

Several factors have been identified as being associated with increased unscheduled care use in people with LTCs. Higher levels of multimorbidity and greater illness severity are associated with higher rates of EHAs and readmission following discharge. 20,23–25

Organisational factors include proximity to EDs or other kinds of emergency provision, and the availability of scheduled services. 26 Like all health-care decisions, unscheduled care use occurs in a clinical, social and cultural context, which may differ in different countries and health-care settings. There has been relatively little work which has addressed the influence of family and providers of routine care, and the cultural norms which underlie people’s beliefs and their behaviours in relation to unscheduled care use.

Psychosocial factors have also been recognised as increasing the risk of unscheduled care use in people with physical LTCs. Depression and anxiety are associated with higher rates of unscheduled care use in people with a variety of different LTCs. 27–30 Poorer quality of life (QoL) and perceived control of illness are also associated with greater health-care utilisation and increased ED attendances for asthma and chronic obstructive pulmonary disease (COPD). 31–35 The reasons for these associations are unclear. Psychological factors may influence people’s decisions about when to use unscheduled care, or reduce their ability to cope in health emergencies, or may just be markers of greater morbidity.

There are major initiatives to improve the quality of care for people with LTCs11 and primary care is seen as the optimal context to deliver care. 36,37 One of the main drivers for the focus on primary care is to attempt to reduce the use of unscheduled secondary care use, by improved provision of scheduled or routine care.

In this context, the primary purpose of this programme of research was to ‘develop effective psychosocial strategies to reduce the need for unscheduled care in patients with LTCs’. The programme of work began in 2009 and its findings are therefore highly relevant to current health-care initiatives. The programme has become known as the ‘CHOICE’ programme, and this term will be used throughout this document in preference to its longer official title (Choosing Health Options in Chronic Care Emergencies). For the purposes of the programme, we focused on people with physical chronic health conditions, and the term ‘LTC’ will be used throughout this report to refer to physical long-term health conditions.

We chose to study four exemplar LTCs: asthma, COPD, coronary heart disease (CHD) and diabetes. We chose these conditions because (a) they are common; (b) they are all in the leading 15 discharge diagnoses of EDs;38 (c) they are associated with EHAs;39 (d) they are all recognised as ambulatory care-sensitive conditions for which effective care can prevent flare-ups and the need to use secondary care emergency services;40 and (e) GPs in England have a requirement under the Quality and Outcomes Framework (QOF)41 to review people with these conditions at least once per year. Therefore, all primary care practices in England keep electronic databases that enable people with these conditions to be identified.

In a recent European study, unscheduled care accounted for 56% of total health-care costs in adults with asthma,18 regardless of the severity of patients’ symptoms. In the USA, over 12 months, 8.3% of patients with asthma made at least one visit to the ED19 and 13% of COPD patients made six or more visits. 20 Over 300,000 people attend an ED in England and Wales each year because of chest pain. 42

The rate of EHAs for people with LTCs varies considerably across England, and is much higher in socially deprived areas than in the least deprived areas of the country, varying from 38 to 207 admissions per 1000 registered patients. 7 Differences are also apparent in the way people are assessed and treated in an emergency facility, with large variability in services that facilitate discharge. 43

Many of the GP practices participating in the programme are located in areas of high deprivation. Although it could be argued that the findings of the programme may not be representative of the most affluent areas of the UK, it is generally the more deprived and highly populated areas where there is high use of unscheduled care. 26

We chose to focus on psychosocial factors (e.g. depression, anxiety, life stress) that may contribute to unscheduled care use, as these factors (particularly mental health problems) are often hidden in clinical practice, under-reported, and under-researched. A recent systematic review, which examined features of primary care that impact on use of unscheduled care, identified several organisational factors and people-related factors of relevance. 26 Although social deprivation, social isolation, older age and having multiple conditions were identified as potential drivers, mental health did not feature in the studies included in the review.

The inclusion in this programme of four exemplar conditions enabled us to study not only each individual condition, but the effects of one or more of our exemplar conditions (i.e. comorbidity) on use of unscheduled care.

The focus for the programme was the potential relationship between psychosocial factors, LTCs and unscheduled care (primarily ED attendance and EHAs) and the possibility that a tailored psychosocial intervention may reduce the requirement to use unscheduled care in people with LTCs.

Six key factors underpinned the research programme.

Aims and objectives

The programme was given the acronym CHOICE and was divided into three phases of work (Figure 2).

FIGURE 2.

The three phases of the CHOICE programme.

The overall aim was to better understand psychosocial drivers of unscheduled care in people with physical LTCs and develop an intervention to reduce unscheduled care.

Phase 1 involved understanding the problem in greater depth using systematic review methods to identify psychosocial factors that act as drivers for unscheduled care in LTCs, and to identify interventions or strategies to reduce the frequency of unscheduled care.

Phase 2 involved mapping the frequency and pattern of unscheduled care in patients with the four exemplar LTCs (asthma, CHD, COPD and diabetes) over a 12-month period, and the development of a red flag marker to identify patients at risk of becoming high users of unscheduled care.

Phase 3 involved developing and testing the validity and utility of the red flag marker using current NHS databases. We also developed and evaluated a low-intensity psychosocial intervention for use in primary care, with the intention of reducing use of unscheduled care in people with LTCs.

Qualitative methods and analyses were embedded within the programme to provide rich personalised accounts, which were used to:

• triangulate findings with the quantitative methods

• identify mechanisms accounting for the quantitative findings

• evaluate the acceptability and inform the implementation of the intervention.

Our objectives were to:

1. Systematically synthesise the current evidence about psychosocial drivers of unscheduled care, and about interventions or strategies to reduce the frequency of unscheduled care use in patients with LTCs (phase 1).

2. Derive estimates of the frequency and pattern of unscheduled care use in patients with asthma, CHD, COPD and diabetes, as examples of common LTCs (phase 2).

3. Develop and validate a ‘red flag’ system that will identify patients with LTCs who are at risk of becoming frequent users of unscheduled care (phases 2 and 3).

4. Identify personal reasons for unscheduled care use, including barriers to access for routine care, patients’ motivations, expectations and decision-making processes, influences from families and relevant health-care workers, and factors in consultations with active case managers (ACMs) (phase 2). This objective was modified during the life of the programme as ACMs were phased out, so the focus of study was switched to routine health-care reviews in primary care for people with LTCs.

5. Develop and evaluate an intervention that will reduce/prevent unscheduled care use, while maintaining or improving patient benefit (phase 3).

6. Use statistical and health economic modelling to evaluate the costs and benefits associated with a treatment intervention (phase 3). This objective was modified over the life of the programme (there were unavoidable delays in phase 2). These and the findings from phases 1 and 2 meant that the planned decision analyses were unlikely to further inform the development of a red flag system and development of the intervention to reduce/prevent the use of unscheduled care. Accordingly, this objective was adapted to focus on statistical analysis of the costs of scheduled and unscheduled hospital care in phase 2 and exploratory sensitivity analyses of the trial data generated in phase 3.

Patient and public engagement and involvement

During the lifetime of the programme, we involved service users and user-led organisations in workshops/groups/stakeholder meetings to help us evaluate our findings, and to develop appropriate and practical interventions. Our service users were involved in all aspects of the programme and this work was led by our co-applicant, Mrs Jackie Macklin.

The work of the programme is described in this report in a sequential fashion, except for our patient and public engagement and involvement (PPE&I), which we have presented in a separate chapter (see Chapter 10), in order to capture the depth and quality of this important contribution that ran across all areas of our programme.

Terminology

We have used the term ‘unscheduled care’ to refer to all aspects of unplanned health-seeking, in both primary and secondary care. We use the term ‘ED attendance’ to refer to attendances at EDs and the term ‘EHA’ to refer to EHAs that do not included planned admissions to hospital. In this programme, the term ‘LTCs’ is used to refer to long-term physical conditions and does not include long-term mental health conditions. The term ‘scheduled care’ refers to all planned contacts with health care including GP appointments, reviews, outpatient appointments and planned hospital admissions.

Abstract

Overview

In this chapter we report the results from phase 1 of the CHOICE programme of research, which was principally involved with evidence synthesis. The aim was to systematically synthesise current evidence about psychosocial drivers of unscheduled care use and about interventions or strategies, which may reduce the frequency of unscheduled care use in patients with LTCs. As throughout the whole programme, we focused on four exemplar conditions: asthma, CHD, COPD and diabetes.

We carried out five major reviews: two focused on potential psychosocial drivers of use of unscheduled care and two focused on evidence concerning treatment interventions to reduce unscheduled care use. The final review synthesised evidence from studies that had used qualitative research methods to understand why people with LTCs use unscheduled care.

Psychosocial predictors of unscheduled health-care use

For the purpose of the first two reviews, we focused on the two most commonly identified psychosocial predictors of unscheduled care use: depression and anxiety. 55,68 Both are common in people with LTCs69,70 and are associated with poor health outcomes. 71,72 Evidence as to their role in the use of unscheduled care, however, is unclear.

Several studies have shown that depression increases ED attendances or EHAs in patients with LTCs. 63,73–75 However, other studies have found no significant impact of depression on unscheduled care use. 55 The relationship between anxiety and use of unscheduled care in patients with LTCs is also unclear. 68,76,77

Review 1 focused on the relationship between depression and use of unscheduled care, whereas review 2 focused on the role of anxiety and its effect on the use of unscheduled care. Both systematic reviews have been published and will be summarised in this chapter. 78,79 The findings from the review concerning the role of depression as a predictor of unscheduled care (i.e. Dickens et al. 78) are reproduced with the permission from Elsevier. 78

Study 1: does depression predict the use of unscheduled care in patients with long-term conditions?

Results

Figure 3 shows the flow chart for the search strategy used in this review. Sixteen independent prospective studies were identified that had investigated whether or not depression predicted unscheduled care use in patients with either asthma, CHD, COPD or diabetes. 23,24,28–30,55,57,68,90–97 From the 16 studies, there were data for 8477 patients with LTCs (see Figure 3).

FIGURE 3.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2009 flow chart for longitudinal cohort studies of patients with LTCs (asthma, CHD, COPD and diabetes), which included a measure of depression at baseline and a measure of unscheduled care at follow-up. Reprinted from Journal of Psychosomatic Research, Vol. 73, Dickens C, Katon W, Blakemore A, Khara A, McGowan L, Tomenson B, et al. Does depression predict the use of urgent and unscheduled care by people with long term conditions? A systematic review with meta-analysis, pp. 334–42. © 2012, with permission from Elsevier. 78

Table 1 summarises the main characteristics of each of the included studies and is adapted from data in the published paper by Dickens et al. 78 There were eight studies of COPD,23,29,30,55,90,94–96 five of CHD,28,57,91–93 two of asthma24,68 and only one of diabetes. 98 There were no studies from the UK and only two were based in primary care. 68,98 Most studies involved following a cohort of hospital inpatients admitted for an acute episode of illness, or exacerbation of illness, after their discharge to determine readmission rates or ED visits. Depression was assessed in all of the studies using a self-rated measure, and in most as a categorical construct (case or non-case).

TABLE 1 - Main characteristics of studies included in the meta-analysis of depression as a predictor of secondary unscheduled care use: authors, sample size, sample, measure of depression at baseline and measure of unscheduled care use at follow-up78

AMI, acute myocardial infarction; BDI, Beck Depression Inventory; CES-D, Centre for Epidemiological Studies Depression Scale; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition; HADS, Hospital Anxiety and Depression Scale; ITU, intensive treatment unit; NETT, national emphysema treatment trial; PHQ, Patient Health Questionnaire; SCL-90-R, Symptom-Checklist-90-Revised.

Reprinted from Journal of Psychosomatic Research, Vol. 73, Dickens C, Katon W, Blakemore A, Khara A, McGowan L, Tomenson B, et al. Does depression predict the use of urgent and unscheduled care by people with long term conditions? A systematic review with meta-analysis, pp. 334–42. © 2012, with permission from Elsevier. 78

First author and date	Condition of study	Sample size	Mean age	Males	Sample	Depression measure	Urgent health-care utilisation measure
Fan et al., 200755	COPD	603	66.5 years	64.1%	603/611 eligible patients with moderate to severe emphysema self-referred or were referred by a clinician at U.S. clinics to control arm of a lung surgery trial (NETT)	BDI (scores ≥ 10 vs. < 10)	Hospital records for COPD-related inpatient admissions and (ED visits for 1 year
Eisner et al., 200524	Asthma	756	59.9 years	29.8%	All adults admitted to ITU with asthma plus sample of all patients hospitalised (without ITU)	CES-D (≥ 16 = depressed)	ED visits and hospitalisations recorded from hospital computerised records for 12 months
Schneider et al., 200868	Asthma	256	56.3 years	38.3%	Consecutive patients with asthma consulting GPs, Germany	Validated German version of PHQ, from which derive DSM-IV depressive disorder	Patients’ self report of urgent hospitalisations and emergency hospital visits over 1 year
Ng et al., 200723	COPD	376	72.2 years	85.1%	Eligible consecutive inpatients hospitalised for COPD exacerbation in Singapore	Chinese HADS (≥ 8 vs. < 8 on Depression Score)	Patient self-reported on urgent hospitalisation at 6 and 12 months after discharge
Almagro et al., 200630	COPD	141	72.0 years	93%	Eligible, consecutive patients on inpatient ward for acute exacerbation of COPD over 7 months	Yesavage Depression Scale (continuous)	Clinical records checked for readmissions of 24 hours or more over 1 year
Gudmundsson et al., 200690	COPD	416	69.2 years	48.8%	Patients hospitalised with COPD exacerbation	HADS depression score (continuous)	Self reported hospitalisations for acute exacerbations of COPD 1 year after discharge, checked via hospital records
Lauzon et al., 200357	CHD	550	60.0 years	78.9%	Consecutive patients approached after admission in 10 Coronary Care Units	BDI (≥ 10 vs. < 10)	Hospital admissions recorded after 30 days, 6 months and 1 year by self report (postal questionnaire) and chart review
Frasure-Smith et al., 200091	CHD	848	59.3 years	69.0%	Subjects were recruited to 2 separate studies – 1 prospective cohort study and 1 control arm of RCT. All patients were post MI	BDI (≥ 10 vs. < 10)	ED visits (all cause) and associated costs for 1 year post discharge for MI
Kurdyak et al., 200893	CHD	1941	64.0 years	30.4%	MI in-patients from 53 hospitals across Ontario Canada	Depression questionnaire, based on Brief Carroll Depression Rating Scale (5 or more vs. < 5)	ED visits (all cause)
Xu et al., 200829	COPD	491	65.6 years	68.8%	Patients with physician diagnosed COPD, attending 10 general hospitals in Beijing China	Mandarin Chinese HADS (≥ 8 vs. < 8 on Depression scale)	Medical interventions were monitored by a telephone-administered questionnaire over 12 months. Hospitalisations confirmed by chart review
Shiotani et al., 200292	CHD	1086	63.6 years	80.4%	Eligible consecutive patients with AMI, directly admitted or transferred to 25 collaborating hospitals eligible	Zung Self-Rating Depression Scale. Non-depressed – scores < 40: depressed – scores ≥ 40	Data on cardiac events 12 months following discharge from hospital records and telephone interviews with patients and family 12 months after baseline
Ciechanowski et al., 200098	Diabetes	350	61.3 years	44%	Patients from a diabetes register of 2 primary care clinics	SCL-90-R – scores divided into tertiles (low/medium/high) – low vs. others used for meta analysis	ED visits for 6 months following questionnaire assessments collected using General Health Cooperative automated data
Ghanei et al., 200796	COPD	157	58.3 years	63%	Patients attending chest clinic	HADS depression subscale used as continuous measure	Acute hospitalisation resulting from COPD exacerbation
Carneiro et al., 201095	COPD	45	68 years	84.4%	Inpatients admitted due to exacerbation of COPD	BDI used as continuous scale	Hospitalisation resulting from exacerbation of COPD
Farkas et al., 201094	COPD	127	66 years	79%	Hospital outpatients with COPD	CES-D scale used as continuous measure	Hospitalisation resulting from exacerbation of COPD
Pishgoo 201128	CHD	334	57.5 years	67.8%	Cardiology outpatients with ≥ 50% stenosis in at least 1 major coronary artery	HADS depression scale – (Persian translated and validated	Al cause ED visits

Table 2 summarises the results of each individual study, and includes univariate and multivariate analyses where they were reported. Eight of the 16 studies reported significant effects of depression on unscheduled care use (either EHAs or ED attendances)29,30,57,68,91–94 and two showed near-significant effects. 24,96

TABLE 2 - Main multivariate findings for individual prospective cohort studies which measured depression at baseline and unscheduled care at follow-up78

BDI, Beck Depression Inventory; FEV₁, forced expiratory volume in 1 second; FVC, forced vital capacity; GRACE, Global Registry of Acute Coronary Events; HR, hazard ratio; O₂, oxygen; SD, standard deviation; SGRQ, St George’s Respiratory Questionnaire.

Reprinted from Journal of Psychosomatic Research, Vol. 73, Dickens C, Katon W, Blakemore A, Khara A, McGowan L, Tomenson B, et al. Does depression predict the use of urgent and unscheduled care by people with long term conditions? A systematic review with meta-analysis, pp. 334–42. © 2012, with permission from Elsevier. 78

Author and date	Univariate findings	Factors controlled	Multivariate findings
Fan et al., 200755	Patients hospitalised or that had ED visit for COPD were slightly more likely to be depressed at baseline (29.9% vs. 24.8% in non-depressed, p = 0.16)	Adjusted for sex, COPD severity, previous admissions, comorbidity score	States BDI score > 10 not associated with hospitalisation in adjusted analyses
Eisner et al., 200524	Depression did not predict ED visits (HR = 1.36, p = 0.12) Trend for depression to predict hospitalisation (HR = 1.34, p = 0.06)	Age, sex, race, education and smoking	Depression did not predict ED visits (HR, 1.20; p = 0.36) Depression did not predict hospitalisation (HR, 1.34; p = 0.06)
Schneider et al., 200868	Depression predicted hospitalisation (OR, 6.1, p = 0.011) but not emergency GP visits (OR = 1.7, p = 0.30)	Medication guideline adherence, smoking, age and sex	Depressive disorder predicted hospitalisation (p = 0.009)
Ng et al., 200723	61% depressed patients had ≥ 1 urgent hospitalisation 57.8% non-depressed had ≥ 1 urgent hospitalisation, p = 0.31	Age, sex, FVC, previous admissions SGRQ	Depression did not predict hospitalisation [HR = 0.93 (95% CI 0.68 to 1.28)]
Almagro et al., 200630	Readmitted patients showed higher baseline depression scores (5 vs. 3.7, p ≤ 0.05) than those who were not readmitted	Age, sex, FEV1, comorbidity, social support	Depression not independent predictor of readmission
Gudmundsson et al., 200690	Hospitalised and non-hospitalised patients had similar baseline depression scores (5.6 vs. 5.4, p = 0.63)	Age, smoking status, FEV1	Depression HR = 1.09 (95% CI 0.8 to 1.51)
Lauzon et al., 200357	Patients depressed at baseline had higher rates of hospitalisation because of any cardiac complication (30.9% vs. 17.5%)	Age, previous MI, anterior MI, diabetes, hypertension, smoking, sex, previous angina	Readmission due to cardiac complications higher in depressed patients [HR = 1.4 (95% CI 1.05 to 1.86)]
Frasure-Smith et al., 200091	Authors unable to confirm all hospitalisations were urgent Depressed patients had greater mean ED visits (1.3 vs. 0.9, p < 0.0001)	–	Findings of multivariate analysis not presented for urgent health-care utilisation
Kurdyak et al., 200893	Author unable to confirm that all hospital admissions were urgent Depressed patients had significantly more ED visits (mean = 1.7 vs. 1.3, p < 0.001)	Age, sex, income, comorbidity, GRACE score, drugs at discharge, cardiac interventions and symptom burden	Adjusted risk for ED visits not presented. Approx adjusted risk for depression predicting ED visits read from figure = 1.1
Xu et al., 200829	More depressed patients were hospitalised for COPD exacerbations (29.5% vs. 19.8%)	Age, sex, marital, educational and employment status, smoking, FEV1, dyspnea, 6-minute walk, social support, self efficacy, comorbidities, hospital type, drug and O2 use, previous hospitalisations	Adjusted Incidence Rate Ratio for probable depression predicting hospitalisation = 1.72 (95% CI 1.04 to 2.85)
Shiotani et al., 200292	Incidence of cardiac event-related readmission was significantly higher in depressed patients (7.8% vs. 4.3%, p = 0.018)	–	Multivariate results not presented for cardiac event-related readmissions
Ciechanowski et al., 200098	ED visit costs-mean and SD Low depression group = 81 (375) Medium depression group = 128 (479) High depression group = 185 (548)	–	Not reported for urgent health-care utilisation
Ghanei et al., 200796	Rehospitalised patients had higher depression scores than those not rehospitalised (12 vs. 11, p = 0.039)	Monthly income and medical comorbidities	Depression is an independent predictor of urgent readmission (risk ratio = 0.31, p = 0.012) after controlling for monthly income and medical comorbidities
Carneiro et al., 201095	Number of readmission correlated with depression score (p = 0.09)	–	Not reported
Farkas et al., 201094	Depression score was higher for hospitalised compared to non-hospitalised [15 (SD = 11) vs. 11 (SD = 9)]	–	Depression not entered into the multivariate analyses
Pishgoo 201128	Baseline depression score 5.7 among those with ED visits vs. 5.1 in those without, p = 0.22	Sex, angina grade, anxiety and somatic comorbidity score contributed entered into final model	Sex, angina grade, anxiety and somatic comorbidity score contributed significantly to the final model. Depression wasn’t entered into the model as univariate findings were non-significant

The results of the meta-analysis are shown in Figure 4. There is a combined OR for depression across all the studies of 1.49 (95% CI 1.35 to 1.64; p < 0.0005). Effects of individual studies were very homogeneous (Q = 13.2, I² = 0.0, 95% CI 0 to 52; p = 0.58).

FIGURE 4.

Forest plot for associations between depression and subsequent use of unscheduled health care in patients with one of four exemplar LTCs (N = 16; OR 1.49, 95% CI 1.35 to 1.64; p < 0.001; Q = 13.2, I² = 0.0%; 95% CI 0 to 52; p = 0.58). BDI, Beck Depression Inventory; CES-D, Centre for Epidemiological Studies Depression Scale; DRS, Depression Rating Scale; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition; GDS, Geriatric Depression Scale; HADS, Hospital Anxiety and Depression Scale; PHQ, Patient Health Questionnaire; Zung, Zung Self-Rating Depression Scale. Reprinted from Journal of Psychosomatic Research, Vol. 73, Dickens C, Katon W, Blakemore A, Khara A, McGowan L, Tomenson B, et al. Does depression predict the use of urgent and unscheduled care by people with long term conditions? A systematic review with meta-analysis, pp. 334–42. © 2012, with permission from Elsevier. 78

Effects varied with different types of unscheduled care. The combined OR for ED attendances was 1.45 (n = 4; 95% CI 1.26 to 1.66; p < 0.0005), for urgent hospitalisations it was 1.56 (n = 10; 95% CI 1.32 to 1.84; p < 0.0005), for combined urgent hospitalisations or ED visits it was 1.29 (n = 1; 95% CI 0.90 to 1.86; p = 0.17) and for combined urgent hospitalisation and urgent GP visits it was 2.69 (n = 1; 95% CI 1.16 to 6.23; p = 0.021). Comparison across groups using the analogue to analysis of variance (ANOVA) revealed that these differences across different types of unscheduled care were not statistically significant [Q = 2.9, degrees of freedom (df) = 3; p = 0.41].

Effects also varied across the different LTCs, but these were not statistically significant. Only eight of the studies included a measure of severity of illness. 23,28–30,55,57,90,93 In six of these,23,29,30,55,90,93 the independent effects of depression were not significant.

The funnel plot for depression studies appeared asymmetrical, with a relative absence of small negative studies (Figure 5). Egger’s regression method confirmed an association between log-OR and standard error of log-OR (Egger’s intercept = 1.98, 95% CI 0.22 to 3.74; p = 0.03). This suggested that there may be some publication bias in the papers included in our review.

FIGURE 5.

Funnel plot for studies of depression and subsequent use of unscheduled health care in patients with LTCs (Egger’s intercept = 1.98; 95% CI 0.22 to 3.74; p = 0.03). Reprinted from Journal of Psychosomatic Research, Vol. 73, Dickens C, Katon W, Blakemore A, Khara A, McGowan L, Tomenson B, et al. Does depression predict the use of urgent and unscheduled care by people with long term conditions? A systematic review with meta-analysis, pp. 334–42. © 2012, with permission from Elsevier. 78

Table 3 shows the results of the quality review for studies that assessed the impact of depression on unscheduled care. Meta-analysis was repeated for the studies subgrouped according to methodological quality (strong, moderate or weak). There was no apparent association between study quality and effect size. There were two studies that were rated as methodologically strong,23,30 which, when combined, produced an OR of 1.46 (95% CI 0.82 to 2.58; p = 0.20). There were six studies that were rated as methodologically moderate, which resulted in a combined OR of 1.53 (95% CI 1.31 to 1.77; p < 0.0005). 24,29,55,57,68,93 Eight studies were rated as methodologically weak with a combined OR of 1.47 (95% CI 1.26 to 1.72; p < 0.0005). 28,90–92,94–96,98

TABLE 3 - Quality assessment for studies of depression and subsequent health-care use in patients with LTCs78

1, strong; 2, moderate; 3, weak.

Reprinted from Journal of Psychosomatic Research, Vol. 73, Dickens C, Katon W, Blakemore A, Khara A, McGowan L, Tomenson B, et al. Does depression predict the use of urgent and unscheduled care by people with long term conditions? A systematic review with meta-analysis, pp. 334–42. © 2012, with permission from Elsevier. 78

Author and date	Selection bias	Design	Confounding	Blinding	Data collection	Dropouts	Global rating	Discrepancy between reviewers	Reasons for discrepancy	Final rating
Almagro et al., 200630	2	2	1	2	1	1	1	No		1
Carneiro et al., 201095	3	2	3	2	1	1	3	No		3
Ciechanowski et al., 200098	2	2	3	2	1	3	2	Yes	Oversight	3
Eisner et al., 200524	2	2	2	2	1	3	2	No		2
Fan et al., 200755	3	2	1	2	1	1	3	Yes	Oversight, differences in interpretation	2
Farkas et al., 201094	3	2	3	2	1	3	3	No		3
Frasure-Smith et al., 200091	3	2	3	2	1	3	3	No		3
Ghanei et al., 200796	3	2	2	2	1	3	3	Yes	Differences in interpretation	3
Gudmundsson et al., 200690	3	2	1	2	1	3	3	No		3
Kurdyak et al., 200893	3	2	1	2	1	2	2	Yes	Differences in interpretation	2
Lauzon et al., 200357	3	2	1	2	1	1	2	Yes	Oversight	2
Ng et al., 200723	2	2	1	2	1	2	1	No		1
Pishgoo 201128	3	2	3	2	1	1	3	No		3
Schneider et al., 200868	3	2	2	2	1	2	2	No		2
Shiotani et al., 200292	3	2	3	2	1	1	3	Yes	Differences in interpretation	3
Xu et al., 200829	3	2	1	2	1	1	2	No		2

Conclusions

Our main conclusion from this systematic review was that depression was associated with a 49% increase in the odds of using subsequent unscheduled care among people with LTCs. The effects for depression were not significantly different across different types of unscheduled care (EHAs or ED attendances) or across different LTCs. However, only eight of studies included attempts to control for the severity of illness, and this appeared to be an important confounder. 23,28–30,55,57,90,93

Limitations

There are several limitations of this review of the relationship between depression and unscheduled care use. First, none of the studies that we identified had been carried out in the UK, so it is unclear how translatable the results are to the UK health-care system. The delivery and organisation of health care varies across countries and the same factors that drive attendance at an emergency room; for example, the factors that drive attendance in the USA may not be the same as those that drive health-care-seeking in the UK. Second, we were able to identify only one study on diabetes98 and two on asthma,24,68 which limits the generalisability of the results to these conditions and other LTCs. Third, only two of the studies were based in primary care,68,98 with most studies focusing on hospital populations, so the results may not be generalisable to patients in primary care with LTCs. Fourth, the methodological quality of the studies identified varied considerably, though this does not appear to have distorted our results. Fifth, we found evidence of a potential publication bias, which may have inflated the observed association between depression and unscheduled care. Finally, a wide variety of different measures of depression were utilised across the studies, as shown in Table 1. Although there is generally good agreement between measures of depression, there will have been some variability between the studies because of the different measures that were employed. Of note, only three of the studies used a depression measure specifically developed for use in populations with physical illness. 28,90,96

Study 2: does anxiety predict the use of unscheduled care in patients with long-term conditions?

Results

The results of the second systematic review we conducted will now be presented. Figure 6 shows the flow chart for this study. Eight independent studies with a prospective design and which included a measure of anxiety at baseline and a measure of unscheduled care at outcome were identified. 29,68,76,77,97,99–101 From the eight studies we identified, there were data for 28,823 participants. The details of the characteristics of each of the included studies can be found in the published paper by Blakeley et al. 79 and are summarised in Table 4.

FIGURE 6.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart for prospective longitudinal studies of patients with LTCs, which included a measure of anxiety at baseline and a measure of unscheduled care at follow-up. Adapted from Blakeley et al. 79

TABLE 4 - Main characteristics of individual prospective cohort studies included in the meta-analysis to determine whether or not anxiety is a predictor of unscheduled care use79

CCU, coronary care unit; HADS, Hospital Anxiety and Depression Scale; ICD-9, International Classification of Diseases, Ninth Edition; PHQ, Patient Health Questionnaire; PRIME-MD, Primary Care Evaluation of Mental Disorders screening questionnaire.

Table is adapted from Blakeley et al. 79

Author, date, LTC and country of origin	Sample and size	Measure of
		Anxiety	Unscheduled care
Abrams et al., 2011;77 COPD, USA	Veterans with COPD exacerbation, acute or chronic bronchitis admitted to hospital (n = 26,591)	ICD-9	30-day readmission records
Schneider et al., 2008;68 asthma, Germany	Patients from 43 primary care practices (n = 256)	Validated German PHQ	Patient self-reported ED attendances and EHAs over 1 year
Greaves et al., 2002;97 asthma, UK	Community subjects with asthma, one group stable, one group who had suffered recent asthma attack (n = 74)	Seven-item panic fear scale of Asthma Symptom Checklist	Practice records ED attendance and hospital attendance over 1 year
Grace et al., 2004;99 CHD, Canada	Patients with unstable angina and MI admitted to CCUs (n = 913)	Phobic anxiety subscale of Middlesex Hospital Questionnaire Anxiety subscale of PRIME-MD	Self-reported cardiac events over 6 months and 1 year
Kaptein, 1982;100 asthma, the Netherlands	Patients with acute, severe asthma hospitalised for condition (n = 40)	State–Trait Anxiety Inventory plus panic–fear personality scale	Rehospitalisation within 6-month period
Xu et al., 2008;29 COPD, China	COPD patients attending hospitals in Beijing (n = 491)	Mandarin HADS	EHAs confirmed by chart review over 1 year
Gudmundsson et al., 2006;90 COPD, Iceland	Patients hospitalised for COPD (n = 416)	HADS (cut-off point of ≥ 8)	Self-reported readmissions over 1 year
Coventry et al., 2011;76 COPD, UK	Patients admitted to hospital with COPD (n = 79)	HADS	Readmissions over 1 year

Four of the studies involved patients with COPD,29,76,77,90 three involved patients with asthma68,97,100 and one involved patients with CHD. 99 There were no studies on diabetes. Two of the studies were based in the UK. 76,97

The main, multivariate results for each individual study are shown in Table 5. None of the eight studies included in this review found that anxiety had a significant effect on the use of unscheduled health care. One study found near significant effects for anxiety on unscheduled health-care use for patients with asthma (see Table 5). 68

TABLE 5 - Main multivariate findings for individual prospective cohort studies that measured anxiety at baseline and unscheduled care at follow-up79

FEV₁, forced expiratory volume in 1 second; HADS, Hospital Anxiety and Depression Scale; HR, hazard ratio; IRR, incidence rate ratio; NS, not significant; PRIME-MD, Primary Care Evaluation of Mental Disorders screening questionnaire; SD, standard deviation; SGRQ, St George’s Respiratory Questionnaire.

Adapted from Blakeley et al. 79

Author and date	Univariate findings	Factors controlled for	Multivariate findings
Abrams et al., 201177	Patients with anxiety were not more significantly likely to be readmitted than those without anxiety [11.3% vs. 11.5% (NS)]	Smoking status	No significant difference in risk of admission regardless of smoking status. Smoking present (HR 1.22, 95% CI 1.04 to 1.44) and smoking absent (HR 1.22, 95% CI 1.03 to 1.43)
Schneider et al., 200868	Panic disorder did not predict hospitalisation (OR 3.5, 95% CI 0.7 to 18.3; p = 0.145), but did predict emergency visits (OR 4.8, 95% CI 1.3 to 17.7; p = 0.019)
Greaves et al., 200297	There was no main effect of panic (p > 0.05)
Grace et al., 200499	Anxious patients (mean 1.11, SD 1.57) reported more visits to the ED than non-anxious (mean 0.83, SD 1.18) patients (t = –1.37; p = 0.17). However, this was NS	Age, family history of CVD, depression, Killip class, sex, family income, smoking status, diabetes and phobic anxiety	Age (OR 1.02, 95% CI 1.00 to 1.05; p = 0.05), family history of CVD (OR 1.63, 95% CI 1.04 to 2.54; p = 0.03), depression (OR 1.07, 95% CI 1.03 to 1.12; p ≤ 0.01) and PRIME-MD anxiety score at 6 months (OR 0.35, 95% CI 0.19 to 0.65; p < 0.01), were all significant predictors of self-reported recurrent cardiac events. All other factors NS
Kaptein, 1982100	State and trait anxiety not associated with increased length of hospitalisations State anxiety not significantly associated with readmission; however, trait anxiety had slight effect (one-tailed t-test: t = 1.72; p = 0.048)
Xu et al., 200829	Anxiety not associated with increased risk of EHA (p = 0.11); however, length of exacerbation in days was longer for patients with anxiety than for those without (p = 0.03)	Age, sex, smoking, marital status, education, employment, living situation, FEV1, dyspnoea score, 6-minute walk distance, social support, COPD-specific self-efficacy, significant comorbidities, hospital type, use of long-acting bronchodilator and inhaled corticosteroid, long-term oxygen therapy and past hospitalisation	Anxiety was not associated with hospitalisation: IRR 1.63 (95% CI 0.88 to 3.03) for HADS anxiety score of ≥ 11; or with length of hospitalisation for those readmitted: IRR 1.99 (95% CI 0.59 to 6.72)
Gudmundsson et al., 200690	Anxiety had no significant effect on rehospitalisation (p = 0.61). No significant difference between HADS anxiety scores for those who were readmitted [mean 7.1 (SD 4.3)] and those who were not [mean 6.7 (SD 4.0); p = 0.28]	Age smoking status, FEV1, SGRQ	Significant association between the HADS anxiety score and the risk of readmission in patients with a low health status (HR 0.81, 95% CI 0.63 to 1.04). In the same group, anxiety (HADS score of ≥ 8) was related to increased risk of rehospitalisation (HR 0.43, 95% CI 0.25 to 0.74)
Coventry et al., 201176	No significant difference between HADS anxiety scores for those who were readmitted (8.53 ± 4.2) and those who were not (9.47 ± 4.6; p = 0.407)	Age, race, sex, individual medical comorbidities and laboratory values	Depression (OR 1.300, 95% CI 1.06 to 1.60; p = 0.013), FEV1 score (OR 0.962, 95% CI 0.93 to 0.99; p = 0.021), and age (OR 1.092, 95% CI 1.01 to 1.18; p = 0.026) were the only significant predictors of readmission. Anxiety was insignificant

Figure 7 shows a forest plot of the eight studies. 29,68,76,77,90,97,99,100 There was a combined OR for anxiety across all studies of 1.078 (95% CI 0.877 to 1.325; p = 0.476). Effects of individual studies showed a low level of heterogeneity (Q = 9.5, df = 7, I² = 26.07%; p = 0.221).

FIGURE 7.

Forest plot for associations between anxiety and subsequent use of unscheduled health care. ASCL, Asthma Symptom Checklist; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition; HADS, Hospital Anxiety and Depression Scale; PHQ, Patient Health Questionnaire; PRIME-MD, Primary Care Evaluation of Mental Disorders screening questionnaire. From Blakeley et al. 79

A sensitivity analysis was performed excluding the largest study, as this had the largest sample size (n = 26,591) of whom 97% were male. 77 The combined effect in random-effects meta-analysis for the studies excluding Abrams et al. 77 was 1.238 (95% CI 0.969 to 1.551; p = 0.087). Again, the effects of these studies showed low heterogeneity (Q = 5.116, df = 5, I² = 2.27%; p = 0.402).

We investigated whether or not the effects of anxiety varied across the different types of unscheduled care used between studies and the different types of LTCs. The effect of anxiety varied across the different types of unscheduled health-care use and across the four LTCs, but none of the differences was significant.

Four studies conducted multivariate analysis that controlled for the severity of the LTC studied. 76,77,99,101 One out of four studies found that, when severity of the LTC was controlled for, anxiety was significantly related to hospital readmission rates for a subgroup of patients with COPD and poor health status. 101

Meta-analysis was repeated for the studies subgrouped according to methodological quality (strong, moderate or weak). Study quality did not change the results: anxiety had no significant impact on unscheduled care use across studies that were rated as methodologically strong (n = 2; OR 0.927, 95% CI 0.872 to 1.143; p = 0.978), moderate (n = 4; OR 1.243, 95% CI 0.650 to 2.375; p = 0.511) or weak (n = 2; OR 1.258, 95% CI 0.955 to 1.656; p = 0.102).

Publication bias was investigated using a funnel plot. The plot for anxiety studies did not appear to be asymmetrical, except for one small negative study (Figure 8). Egger’s regression method confirmed the lack of association between the log_e-OR and standard error of log_e-OR (Egger’s bias = 1.24, 95% CI –1.01 to 3.48; p = 0.23). The Duval and Tweedie trim and fill procedure created just one imputed study, giving a revised random-effects-combined OR for anxiety across all studies of 1.05 (95% CI 0.82 to 1.33; p = 0.69). The heterogeneity between studies was increased slightly, and was still significant (Q = 12.9, df = 8; p = 0.040).

FIGURE 8.

Funnel plot with pseudo-95% confidence limits for studies of anxiety and subsequent use of unscheduled health care (Egger’s test: bias = 1.24, 95% CI –1.01 to 3.48; p = 0.23). Adapted from Blakeley et al. 79

Conclusions

Our main conclusion from this systematic review was that anxiety does not appear to be associated with an increase in using unscheduled care in patients with LTCs. This is in contrast to our findings from the first systematic review, which showed a relationship between depression and use of unscheduled care.

The number of studies in the anxiety review was small and we did not focus specifically on panic or other forms of anxiety-related symptoms and disorders, which have been shown to be involved in care-seeking. 102,103 However, the findings for our review suggest anxiety does not appear to be an important psychosocial driver of unscheduled care in patients with asthma, CHD, COPD and diabetes.

Discussion for systematic reviews to determine whether depression or anxiety predicts unscheduled care use

The findings from our two first systematic reviews were puzzling. Why should depression, but not anxiety, be associated with unscheduled care use? Both are common in patients with LTCs and both are associated with adverse outcomes in LTCs. 10,69,104,105

One of the reasons for the difference we found between depression and anxiety may be methodological problems with the included studies, which are detailed below, and the relative paucity of studies that have examined the effects of anxiety at all. In addition, depression and anxiety often coexist, particularly in primary care,105 and it may be somewhat of an artificial distinction to separate them in the context of comorbid physical illness. However, as they have been analysed and reported separately in the included studies, it would not have been possible to combine them in a meaningful way for the purposes of the systematic reviews.

Our findings give support to a model of depression in relation to unscheduled care use in people with LTCs. Depression results in greater self-neglect106,107 and less adherence to routine treatment. 108 This in turn results in more symptoms, or more severe symptoms or exacerbations, which may increase the risk of unscheduled care.

This model is illustrated in Figure 9. A similar model has been put forward by Pooler and Beech in relation to COPD and increased hospital admissions,109 although that model also includes a role for anxiety, which is not supported by our findings.

FIGURE 9.

A model illustrating the potential mechanism whereby depression may affect unscheduled care use. Adapted from Pooler and Beech. 109

Anxiety is clearly associated with a variety of poor outcomes in people with LTCs, and is reported from the qualitative literature to be associated with acute exacerbations of illness. 110 It is possible that, once an acute exacerbation of illness is triggered, this is then associated with an increase in worry and anxiety, which heightens the focus on bodily symptoms, resulting in some cases, in fear of death and panic, which necessitates the need to seek immediate health care. Thus, anxiety may not be a long-term predictor of unscheduled care but may play an important role in the immediacy of treatment-seeking which would not be captured by studies employing a prospective longitudinal design, but may be captured by other methods (e.g. qualitative work). This potential mechanism is illustrated in Figure 10.

FIGURE 10.

A model illustrating the potential mechanism whereby depression and anxiety may affect unscheduled care use. Adapted from Pooler and Beech. 109

Complex interventions that reduce unscheduled health-care use

The second aim of this section of the research programme was to systematically synthesise the evidence for existing complex interventions which have been used to reduce the frequency of unscheduled care use in patients with LTCs. Our initial search in 2008 identified that there were over 30 relevant studies for COPD and asthma, but only six studies that examined the effect of complex interventions on unscheduled care use in diabetes and four studies in CHD. We chose to focus on two of our four exemplar conditions [COPD (study 3) and asthma (study 4)], as there was a large number of studies, 32 and 33, respectively, for each condition, which had looked at the effect of complex interventions to reduce unscheduled care use.

Study 3: complex interventions that reduce unscheduled health-care use in chronic obstructive pulmonary disease

Results

The flow chart for this study is shown in Figure 11. Thirty-two independent studies were eligible for inclusion in this review, which included the comparison of 33 independent interventions. 114,117–148

FIGURE 11.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart for studies of the effectiveness of complex interventions in the reduction of unscheduled care use in COPD. Reprinted from Respiratory Medicine, Vol. 108, Dickens C, Katon W, Blakemore A, Khara A, Tomenson B, Woodcock A, et al. Complex interventions that reduce urgent care use in COPD: a systematic review with meta-regression, pp. 426–37. © 2014, with permission from Elsevier. 111

The details of the characteristics of each included study are summarised in Table 6. In the majority of studies, patients with COPD were recruited from secondary care settings, and either ED attendances and/or EHAs were the measure of unscheduled care use.

TABLE 6 - Characteristics of studies included in the systematic review of complex interventions in the reduction of unscheduled care use in COPD111

ICU, intensive care unit.

Reprinted from Respiratory Medicine, Vol. 108, Dickens C, Katon W, Blakemore A, Khara A, Tomenson B, Woodcock A, et al. Complex interventions that reduce urgent care use in COPD: a systematic review with meta-regression, pp. 426–37. © 2014, with permission from Elsevier. 111

Authors	Year of publication	Sample size	Where recruited	Assessment of urgent health-care use	Method of urgent health-care assessment	Duration of follow-up (months)
Bergner et al.119	1988	244	Secondary	ED visits	Records	12
Bourbeau et al.118	2003	191	Secondary	ED visits/urgent hospitalisation	Self-reported	12
Boxall et al.120	2005	46	Combined	Urgent hospitalisations	Self-reported	6
Casas et al.114	2006	155	Secondary	Urgent hospitalisations	Combined	12
Eaton et al.121	2009	97	Secondary	Unscheduled emergency visits to primary or secondary care	Combined	3
Farrero et al.122	2001	94	Primary	ED visits	Records	12
Gadoury et al.123	2005	191	Secondary	ED visits/urgent hospitalisation	Records	24
Gallefoss and Bakke124	2000	71	Secondary	Urgent hospitalisation	Combined	12
Güell et al.125	2000	60	Secondary	Urgent hospitalisation	Unclear	24
Hermiz et al.126	2002	147	Secondary	ED visits	Records	3
Hernandez et al.127	2003	209	Secondary	ED visits	Combined	2
Jarab et al.128	2012	127	Secondary	ED visits/urgent hospitalisation	Unclear	6
Khdour et al.129	2009	173	Secondary	ED visits/urgent hospitalisation/urgent doctor visits	Combined	12
Ko et al.130	2011	60	Secondary	ED visits/urgent hospitalisation	Combined	12
Koff et al.131	2009	40	Secondary	ED visits/urgent hospitalisation	Combined	3
Lee et al.132	2002	89	Secondary	ED visits	Unclear	6
Díaz Lobato et al.133	2005	40	Secondary	Urgent ICU admission	Unclear	1
Man et al.148	2004	34	Secondary	ED visits	Combined	3
Martin et al.134	2004	89	Primary	Ambulance calls	Unclear	12
McGeoch et al.135	2006	154	Primary	ED visits	Unclear	12
Rea et al.117	2004	135	Combined	ED visits	Records	12
Rice et al.137	2010	554	Primary	ED visits	Records	12
Ries et al.138	2003	137	Secondary	ED visits	Self-reported	12
Seymour et al.139	2010	60	Secondary	ED visits/urgent hospitalisation	Combined	3
Smith et al.140	1999	92	Combined	ED visits	Records	12
Soler et al.141	2006	26	Secondary	ED visits	Records	12
Sridhar et al.142	2008	104	Secondary	Urgent doctor visits	Combined	24
Tougaard et al.143	1992	82	Secondary	General practice emergency care	Records	12
Trappenburg et al.144	2011	216	Combined	ED visits/urgent hospitalisation	Combined	6
Wakabayashi et al.145	2011	85	Secondary	ED visits	Unclear	12
Wong et al.146	2005	60	Secondary	ED visits	Self-reported	3
Wood-Baker et al.147	2006	112	Primary	ED visits/urgent doctor visits	Self-reported	12

Table 7 summarises the type of interventions that were employed in the studies included in the review of complex interventions in COPD, and the duration of the intervention.

TABLE 7 - Characteristics of the different kinds of complex interventions employed in the studies included in the systematic review of complex interventions in COPD

1, Treatment included general education; 2, treatment included general discussion; 3, treatment included skills training; 4, treatment included exercise; 5, treatment included behavioural therapy; 6, treatment included relapse prevention; 7, treatment included problem-solving; 8, treatment included cognitive–behavioural therapy; 9, treatment aimed to increase social support; 10, treatment included relaxation; 11, treatment included biofeedback; 12, treatment included miscellaneous; TAU, treatment as usual.

Reprinted from Respiratory Medicine, Vol. 108, Dickens C, Katon W, Blakemore A, Khara A, Tomenson B, Woodcock A, et al. Complex interventions that reduce urgent care use in COPD: a systematic review with meta-regression, pp. 426–37. © 2014, with permission from Elsevier. 111

Study	Intervention	No. sessions	Who delivered the intervention?	Delivery method	Where delivered	Unidisciplinary or multidisciplinary	Intervention components
Bergner et al.119	Respiratory home care vs. TAU and standard home care vs. TAU	Unclear	Non-mental health practitioner	Face to face	Home	Multidisciplinary	12
Bourbeau et al.118	Education and case manager vs. TAU	18	Non-mental health practitioner	Combination	Home	Multidisciplinary	1, 3, 4, 6, 10
Boxall et al.120	Home-based pulmonary rehabilitation vs. waiting list	11	Non-mental health practitioner	Face to face	Home	Multidisciplinary	1, 3, 4
Casas et al.114	Integrated care vs. TAU	1 comprehensive education session	Non-mental health practitioner	Face to face, telephone and online	Home, hospital, online	Multidisciplinary	1, 12
Eaton et al.121	Early pulmonary rehabilitation	Unclear	Non-mental health practitioner	Face to face	Combination	Multidisciplinary	1, 4, 6, 10, 12
Farrero et al.122	Hospital-based home care vs. TAU	4.8	Non-mental health practitioner	Combination	Combination	Multidisciplinary	12
Gadoury et al.123	COPD self-management vs. TAU	18	Unclear	Combination	Home	Unclear	1, 4, 12
Gallefoss and Bakke124	Self-management education vs. TAU	4	Non-mental health practitioner	Face to face	Unclear	Multidisciplinary	1, 3, 6, 12
Güell et al.125	Rehabilitation vs. TAU	31	Non-mental health practitioner	Face to face	Hospital	Unclear	1, 3, 4, 10
Hermiz et al.126	Home-based self-management education and support vs. TAU	2	Non-mental health practitioner	Face to face	Home	Multidisciplinary	1, 12
Hernandez et al.127	Home hospitalisation vs. TAU	Unclear	Non-mental health practitioner	Combination	Home	Unclear	1, 3, 6
Jarab et al.128	Pharmaceutical care programme with emphasis on self-management vs. TAU	Unclear	Non-mental health practitioner	Face to face	Unclear	Unidisciplinary	1, 3
Khdour et al.129	Clinical pharmacy-led; disease and medicine; management programme vs. TAU	1 initial education session, reinforcement at outpatient visit every 6 months plus telephone calls at 3 and 9 months	Non-mental health practitioner	Face to face and telephone	Clinic	Unidisciplinary	1, 2, 3, 4, 6, 12
Ko et al.130	Outpatient pulmonary rehabilitation programme vs. TAU	24	Non-mental health practitioner	Face to face	Hospital	Unidisciplinary	1, 3, 4
Koff et al.131	Proactive integrated care vs. TAU	Unclear	Non-mental health practitioner	Face to face and telephone	Clinic and home	Unidisciplinary	1, 3, 12
Lee et al.132	Treatment guideline implementation vs. TAU	9	Non-mental health practitioner	Combination	Home	Unidisciplinary	1, 2, 3, 6, 12
Díaz Lobato et al.133	Home hospital vs. conventional hospital	Unclear	Non-mental health practitioner	Face to face	Home	Multidisciplinary	1, 12
Man et al.148	Community rehabilitation vs. TAU	16	Non-mental health practitioner	Face to face	Hospital	Multidisciplinary	1, 4
Martin et al.134	Care plan vs. TAU	Unclear	Non-mental health practitioner	Face to face	Combination	Multidisciplinary	1, 3, 12
McGeoch et al.135	Self-management action plan vs. TAU	1	Non-mental health practitioner	Combination	Home	Multidisciplinary	1, 6, 12
Rea et al.117	Written disease management programme vs. TAU	Unclear	Non-mental health practitioner	Face to face	Combination	Multidisciplinary	1, 6, 12
Rice et al.137	Disease management program vs. TAU	1	Mental health practitioner	Combination	Combination	Unidisciplinary	1, 12
Ries et al.138	Self-management education and support vs. TAU	60	Non-mental health practitioner	Combination	Hospital	Unclear	4, 12
Seymour et al.139	Post exacerbation pulmonary rehabilitation vs. TAU	16 (2 × weekly for 8 weeks)	Non-mental health practitioner	Face to face	Hospital	Unidisciplinary	1, 4
Smith et al.140	Home-based nursing intervention vs. TAU	24	Non-mental health practitioner	Face to face	Home	Multidisciplinary	1, 3, 6, 12
Soler et al.141	Self-management education and scheduled clinic review vs. TAU	12	Non-mental health practitioner	Face to face	Hospital	Unidisciplinary	1, 12
Sridhar et al.142	Rehabilitation and ongoing self-management support vs. TAU	40	Non-mental health practitioner	Combination	Combination	Unidisciplinary	1, 4, 12
Tougaard et al.143	Self-management education vs. TAU	Unclear	Non-mental health practitioner	Face to face	Hospital	Multidisciplinary	1, 3
Trappenburg et al.144	Self-management action plan vs. TAU	Unclear	Non-mental health practitioner	Face to face and telephone	Clinic and home	Unidisciplinary	1, 2, 6, 12
Wakabayashi et al.145	Integrated care vs. TAU	Minimum of 6	Non-mental health practitioner	Face to face	Unclear	Multidisciplinary	1, 12
Wong et al.146	Nurse-led telephone follow-up vs. TAU	4	Non-mental health practitioner	Telephone	Home	Unidisciplinary	1, 10
Wood-Baker et al.147	Written action plan vs. TAU	1	Non-mental health practitioner	Face to face	Hospital	Multidisciplinary	6, 12

Table 7 shows the different types of complex interventions that were evaluated in the studies and the different components of each intervention. They were as follows: education (28 studies114,117,118,120,121,123–135,137,139–146,148), general discussion (three studies129,132,144), skills (13 studies118,120,124,125,127–132,134,140,143), exercise (11 studies118,120,121,123,125,129,130,138,139,142,148), behaviour therapy (zero studies), relapse prevention (11 studies117,118,121,124,127,129,132,135,140,144,147), problem-solving (zero studies), cognitive–behavioural therapy (CBT) (zero studies), increased social support (zero studies) and relaxation therapy (four studies118,121,125,146). The forest plot for the meta-analysis is shown in Figure 12. The plot shows that the overall combined effect across the 33 independent interventions was associated with a 32% reduction in the unscheduled care use (OR 0.68, 95% CI 0.57 to 0.80). There was a moderate degree of heterogeneity across the studies (Q = 51.1, df = 32, I² = 37.4%; p < 0.0005).

FIGURE 12.

Forest plot for effect of complex interventions on unscheduled care use in COPD. Reprinted from Respiratory Medicine, Vol. 108, Dickens C, Katon W, Blakemore A, Khara A, Tomenson B, Woodcock A, et al. Complex interventions that reduce urgent care use in COPD: a systematic review with meta-regression, pp. 426–37. © 2014, with permission from Elsevier. 111

When study effects were grouped according to the components of their interventions, significant effects were seen for interventions that included general education (n = 28; OR 0.66, 95% CI 0.55 to 0.81; p < 0.0005), exercise (n = 11; OR 0.60, 95% CI 0.48 to 0.76; p < 0.0005) and relaxation therapy (n = 4; OR 0.48, 95% CI 0.33 to 0.70; p < 0.0005).

When studies were grouped according to their methodological characteristics there was no significant difference in the effect of interventions according to the age of the patient, sex, recruitment site, where treatment was delivered, who delivered treatment, how treatment was organised or the quality of the study.

The funnel plot, which is shown in the published paper,111 appeared asymmetrical with a relative absence of small studies in which interventions were associated with increased unscheduled care use, though Egger’s regression did not confirm a statistically significant association between log-OR and standard error of log-OR (Egger’s intercept = −0.79, 95% CI –1.87 to 0.29; p = 0.15).

Conclusions

The conclusion of this systematic review was that there is evidence that complex interventions have an independent effect on reducing unscheduled care for patients with COPD. Most of the interventions involved low-intensity-type treatments; however, relaxation therapy appeared to show promise.

Study 4: complex interventions that reduce unscheduled health-care use in asthma

The flow chart for this systematic review is shown in Figure 13. Thirty-three independent studies of asthma patients were eligible for inclusion in this review, which included a total of 39 comparisons of independent interventions. 149–181

FIGURE 13.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart for studies of complex interventions in asthma. Reproduced from Blakemore et al. 112

The details of the characteristics of each included study are summarised in Table 8 and further details are available in Blakemore et al. 112 Table 9 shows the characteristics of the different interventions employed in each study in the review.

TABLE 8 - Characteristics of studies included in the systematic review of complex interventions in the reduction of unscheduled care use in asthma

Adapted from Blakemore et al. 2015. 112

First author and year of publication	Sample size	Where recruited	Assessment of urgent health-care use	Duration of follow-up (months)
Aiolfi et al., 1995149	44	Secondary	Hospitalisation; ED visits	12
Bolton et al., 1991150	241	Secondary	Hospitalisation; ED visits	12
Brown et al., 2006151	111	Secondary	Hospitalisation; ED visits; urgent doctor visits	6
Castro et al., 2003152	96	Secondary	Hospitalisation; ED visits; direct and indirect health-care costs	12
Clark et al., 2007153	808	Secondary	Hospitalisation; ED visits; unscheduled clinic visits	12
Coté et al., 1997179	188	Secondary	Hospitalisations; ED visits	12
Donald et al., 2008155	71	Secondary	Hospitalisation; ED visits; urgent doctor visits	12
Galbreath et al., 2008156	429	Primary and secondary	Hospitalisations; ED visits; urgent doctor visits	12
Gallefoss et al., 1999157	78	Secondary	Hospitalisations (n of days)	12
George et al., 1999158	77	Secondary	Hospitalisations; ED visits	6
Huang et al., 2009159	173	Secondary	ED visits; urgent health-care visit	6
Kauppinen et al., 1998160	162	Secondary	ED visits (cost)	12
Kotses et al., 1995161	76	Unclear	Hospitalisations; ED visits	12
Lahdensuo et al., 1996162	115	Secondary	Hospitalisations; urgent ambulatory care visits	12
Levy et al., 2000163	211	Secondary	ED visits	6
Milenkovic et al., 2007164	74	Secondary	ED visits	12 months plus follow-up survey at 60 months
Morice and Wrench, 2001180	80	Secondary	ED visits	18
Osman et al., 2002165	280	Secondary	Hospitalisations; ED visits; urgent calls to doctor	12
Patel et al., 2009166	52	Unclear	Hospitalisations; ED visits	12
Perneger et al., 2002181	115	Secondary	ED visits	6
Pilotto et al., 2004167	170	Primary	Hospitalisations; ED visits	9
Pinnock et al., 2003168	278	Primary	Hospitalisations; ED visits	3
Rasmussen et al., 2005169	253	Unclear	Unscheduled visits to physician; ED visits; hospitalisation	6
Schermer et al., 2002170	193	Primary	Hospitalisation; ED visits (cost)	24
Schott-Baer and Christensen 1999171	36	Secondary	ED visits	1.5
Shelledy et al., 2009172	159	Secondary	ED visits; ED costs	6
Sundberg et al., 2005173	97	Secondary	Hospitalisations; urgent health-care visits	12
van der Meer et al., 2011174	200	Primary	ED visits (cost)	12
Vojita et al., 1999175	121	Unclear	ED visits	6
Wilson et al., 1993176	310	Primary	Urgent health-care visits	24
Yilmaz and Akkaya, 2002177	52	Secondary	Hospitalisations; ED visits	36
Yoon et al., 1993178	76	Secondary	Hospitalisations; ED visits	10

TABLE 9 - Characteristics of the interventions from studies included in the systematic review of complex interventions in asthma

1, Treatment included general education; 2, treatment included general discussion; 3, treatment included skills training; 4, treatment included exercise; 5, treatment included behavioural therapy; 6, treatment included relapse prevention; 7, treatment included problem-solving; 8, treatment included CBT; 9, treatment aimed to increase social support; 10, treatment included relaxation; 11, treatment included biofeedback; 12, treatment included miscellaneous; TAU, treatment as usual.

Adapted from Blakemore et al. 112

First author and year of publication	Intervention	Control	Number of sessions	Where was the intervention delivered	Intervention component
Aiolfi et al., 1995149	Education plus guidelines	TAU	4	Outpatient clinic	1, 3, 6, 12
Bolton et al., 1991150	Education	TAU	3	Hospital	1, 3, 6, 7, 10, 12
Brown et al., 2006151	Education	TAU	2	Clinic and home	1, 10
Castro et al., 2003152	Collaborative care	TAU	2	Hospital and telephone/home	1
Clark et al., 2007153	Education plus counselling	TAU	5	Telephone/home	1, 7
Coté et al., 1997179	Education plus action plan	TAU	Unclear	Unclear	1, 6, 12
Coté et al., 1997179	Education plus action plan based on asthma symptoms	TAU			1, 3
de Oliveira et al., 1999154	Asthma education	TAU	6	Outpatient clinic	1, 3, 6, 12
Donald et al., 2008155	Telephone support plus self-management education	Self-management education	6	Home	12
Galbreath et al., 2008156	Disease management	TAU	6/7	Home	1, 12
Galbreath et al., 2008156	Augmented disease management	TAU	11		1, 3, 12
Gallefoss et al., 1999157	Self-management education	TAU	4	Unclear	1, 3, 6, 10, 12
George et al., 1999158	Education	TAU	Unclear	Hospital	3, 6
Huang et al., 2009159	Individualised education plus peak flow monitoring	TAU	24	Unclear	1, 6, 12
Huang et al., 2009159	Individualised education	TAU	As above	As above	As above
Kauppinen et al., 1998160	Intensive self-management education	TAU	5	Clinic	1
Kotses et al., 1995161	Self-management programme	Wait list	7	Unsure	1, 2, 12
Lahdensuo et al., 1996162	Self-management education	TAU	3	Outpatient clinic	1, 10, 12
Levy et al., 2000163	Self-management education	TAU	3	Clinic	1, 3, 6, 12
Milenkovic et al., 2007164	Peak flow-based self-management	TAU	3	Outpatient clinic	1, 3, 6, 12
Morice and Wrench, 2001180	Education	TAU	2 or more	Hospital	1, 2, 6, 11
Osman et al., 2002165	Self-management education	TAU	2	Hospital	1, 2, 6
Patel et al., 2009166	Telephone questionnaire intervention	TAU	2	Telephone	12
Perneger et al., 2002181	Self-management education	TAU	3	Hospital	3, 6, 7, 10, 11, 12
Pilotto et al., 2004167	Self-management education plus scheduled follow-up	TAU	3	Outpatient clinic	1, 3, 6, 12
Pinnock et al., 2003168	Telephone consultation	TAU	1	Telephone	12
Rasmussen et al., 2005169	Internet-based asthma management	TAU	Unclear	Home	6, 12
Rasmussen et al., 2005169	Specialist monitoring and management	TAU	Unclear	Outpatient clinic	6, 12
Schermer et al., 2002170	Guided asthma self-management	TAU	4	Primary care clinic	1, 3
Schott-Baer and Christensen, 1999171	Self-management training	TAU	4	Unclear	1
Shelledy et al., 2009172	Nurse-led home disease management programme	TAU	5	Home	1, 3
Shelledy et al., 2009172	Respiratory therapist led home disease management programme	TAU	5	Home	1, 3, 6
Sundberg et al., 2005173	Supported computerised education	TAU	4	Outpatient clinic/telephone	1
van der Meer et al., 2011174	Internet-based self management	TAU	Unclear	Home/internet	1, 2, 6, 12
Vojita et al., 1999175	Nurse-led education programme	TAU	4	Home	1, 6, 3
Wilson et al., 1993176	Group self-management education	TAU	4	Unclear	1, 2, 3, 5, 6, 7, 9, 10, 12
Wilson et al., 1993176	Individual self-management education	TAU	3–5	Unclear
Yilmaz and Akkaya, 2002177	Asthma education	TAU	6	Outpatient clinic	1, 3, 6, 12
Yoon et al., 1993178	Self-management education	TAU	1	Clinic	1, 2, 3, 6, 12

Figure 14 shows the combined effect across the 39 independent studies. The results show that complex interventions were associated with a 21% reduction in the use of unscheduled care (OR 0.79, 95% CI 0.67 to 0.94). A moderate degree of heterogeneity was seen across the studies (Q = 58.1, df = 38, I² = 34.6%; p = 0.020).

FIGURE 14.

Forest plot for effect of complex interventions on unscheduled health-care use in asthma. Reproduced from Blakemore et al. 112 TAU, treatment as usual; UC, unscheduled care.

When study effects for asthma were grouped according to the components of their interventions, significant effects were seen for general education (n = 32; OR 0.77, 95% CI 0.64 to 0.91; p = 0.003), skills training (n = 20; OR 0.64, 95% CI 0.48 to 0.86; p = 0.003) and relapse prevention (n = 23; OR 0.75, 95% CI 0.57 to 0.98; p = 0.04). No studies used relaxation or CBT.

When studies were grouped according to their methodological characteristics, there was no significant difference in the effect of interventions according to the age of patients, sex, duration of illness, recruitment site, how or where treatment was delivered, who delivered treatment, number of treatment sessions or whether or not there was a structured management plan or follow-up. Nor were the effect sizes significantly associated with the type of unscheduled care or quality of the study.

Publication bias was assessed using a funnel plot (Figure 15) and Egger’s regression method. The funnel plot appeared to be symmetrical. Egger’s regression method did not confirm statistically significant associations between log-OR and standard error of log-OR (Egger’s intercept = –0.37, 95% CI –1.21 to 0.48; p = 0.38).

FIGURE 15.

Funnel plot with pseudo-95% confidence limits for studies of complex interventions to reduce unscheduled health-care use in asthma. Reproduced from Blakemore et al. 112

Conclusion

The conclusion of study 4 was that there is evidence that complex interventions have a small, independent, effect on reducing unscheduled care for patients with asthma. The results were similar to the previous review on COPD, with most of the interventions falling into the category of ‘very low-intensity treatments’ such as skills training.

Why do people use unscheduled care? A qualitative meta-synthesis

In the final part of our evidence synthesis, we carried out a review of qualitative studies which have examined reasons for unscheduled care use in patients with LTCs. Qualitative research is better able to identify personal and contextual reasons for using health care than quantitative research. Factors such as poor access to primary care and poor preventative care have been identified as being associated with higher risk of hospital admissions. 20,182–184 However, the decision-making process that patients with LTCs use to determine whether or not to use unscheduled care is unclear.

Overall discussion and conclusions of our evidence synthesis

The five studies that we carried out in this phase of the programme helped to clarify the existing evidence base. There was some evidence that depression, but not anxiety, is prospectively associated with an increased probability of using unscheduled care in patients with LTCs, but it is unclear from the work carried out to date whether or not this is mediated via the severity of physical illness.

As there is a bidirectional relationship between severity of physical illness and depression, our findings suggest that further work to fully elucidate the relationship between depression, severity of physical disease and unscheduled care use is required.

There were also very few studies that had been conducted in a primary care setting, yet this is the major health-care arena in which care of LTCs is provided in the UK. Further work based in primary care is essential to determine if depression or other psychosocial factors are important predictors of unscheduled care use in people with LTCs.

The results from the qualitative review suggested that patients make clear choices about when and what type of health care to use. Their main considerations are shaped by previous experience of health-care use, and the values they place on different kinds of health care, for example seeing hospitals as places of safety and expertise. Low mood or depression, or even anxiety, were cited in very few of the qualitative papers as being factors that people are aware may influence health-seeking at a time of crisis. This suggests that whatever role psychosocial factors play in increasing the likelihood of health-care-seeking behaviour; their influence is hidden or not uppermost in people’s minds when accounting for health-care use.

Abstract

Overview

In this chapter, we report the findings from the main quantitative study of the second phase of the programme. This was a prospective longitudinal cohort study in primary care, which we carried out to identify factors that predict use of unscheduled care in patients with at least one of our four exemplar LTCs: asthma, CHD, COPD and diabetes. We were particularly interested in what role psychosocial factors play in the use of unscheduled care, as these have been relatively underinvestigated in comparison with physical factors. We had two major objectives:

1. to derive estimates of the frequency and pattern of unscheduled care use in patients with at least one of the four exemplar conditions we chose to study in this programme: asthma, CHD, COPD and diabetes

2. to develop predictors (i.e. a red flag system) in primary care to identify patients with LTCs who are at risk of becoming frequent users of unscheduled care, with a specific focus on potential psychosocial predictors.

Background

The focus for the CHOICE programme was to investigate the potential relationship between psychosocial factors, LTCs and unscheduled care, and to explore the possibility that a tailored psychosocial intervention may reduce the requirement to use unscheduled care in people with LTCs. We were therefore interested in potential psychosocial factors that may be risk factors for use of unscheduled care.

The evidence synthesis that we carried out in the first phase of the programme identified depression, but not anxiety, as being a potential predictor of use of unscheduled care in patients with LTCs. Depression is two to three times more common in people with a LTC than in the general population and is seven times more common in people with two or more LTCs. 69

Depression is particularly associated with poor QoL,69 and the presence of comorbid depression can have a greater effect on the functional status and QoL of people with LTCs than the level of severity of their physical illness. 230,231 Health-care utilisation and costs are substantially increased for patients with LTCs who are depressed in comparison with those who are not,111 and this effect persists after adjusting for medical severity. 232,233 It has been estimated that between 12% and 18% of all expenditure on LTCs in England is linked to poor mental health (mainly depression). 39

To identify potential risk factors for use of unscheduled care in patients with LTCs, we carried out a large longitudinal prospective study in primary care. We included key factors, identified from our own evidence synthesis, to measure at baseline, and we focused on patients with at least one of our four exemplar LTCs. The key factors included physical comorbidity, severity of physical illness, demographic characteristics, prior use of unscheduled care, distance to nearest ED, depression and life stress.

Specifically, we asked what factors independently predict use of unscheduled care in people with LTCs in primary care?

Method

The study was conducted in Manchester, which is a large city in the UK with a population of approximately 500,000 and a wider conurbation area population of 2.5 million. 234 According to the 2010 Index of Multiple Deprivation (IMD), Manchester is the fourth most deprived local place in England. 235 Unemployment throughout 2012–13 averaged 11.9%, which was above the national average, but lower than some of the country’s other comparably sized cities. The population density is approximately 4000 people per km², which is below that of London, but similar to other urban conurbations in the UK including the West Midlands and Merseyside. Statistics from the 2011 census showed that 66.7% of the population of the city of Manchester were white, 17.1% Asian, 8.6% black, 4.7% were mixed race, 1.9% Arab and 1.2% of other ethnic heritage. Manchester is therefore fairly representative of a large UK urban population.

We conducted a primary care-based prospective cohort study of patients aged ≥ 18 years with LTCs. All patients with at least one of the four exemplar conditions (asthma, CHD, COPD and diabetes) were identified from QOF registers in 10 general practices in the city of Manchester. A variety of strategies were used to recruit practices, including direct invitation and presentations at local meetings. There were potentially 100 eligible practices within the Manchester area, and we targeted those with the highest combined prevalence rates for the four exemplar conditions; of these, 31 showed some interest in participation. We invited these practices to participate in a sequential fashion until we considered we had a sufficient baseline sample. In all, 21 practices were approached, of which 10 participated in the study and 11 declined. No incentives were given to practices to participate in the study, although they were provided with support costs to cover work they carried out in relation to the study.

All patients with at least one of the four exemplar LTCs were identified, and the lists were checked by practice GPs, who excluded patients who were receiving palliative care, were terminally ill or would not have capacity to participate in the study.

Eligible patients were sent a postal questionnaire between June 2010 and December 2010, with a further reminder questionnaire pack 3 weeks later. We used the following strategies to maximise response: a personalised letter addressed to the individual from their GP; an explanation which focused on the importance of both physical and mental health; coloured ink; stamped addressed envelopes; and an explanation that the research was funded by the NHS and conducted by the university, as opposed to a commercial organisation. We also offered translation services for participants who wished to complete the questionnaire in a language other than English.

Participants were asked whether or not they would give permission for review of their medical records for 1 year prior to the date of completion of the questionnaire and for the year of the prospective study period. This enabled the severity of patients’ physical health to be assessed, and the number of EHAs to be verified, for the year before the baseline questionnaire, and for the 12-month prospective period. The notes were reviewed by research staff and junior doctors who were gaining research experience on the project. Any queries by the research staff were checked by a medically qualified practitioner. Participants were also sent a follow-up questionnaire 12 months after the first one, when they were asked again about their use of unscheduled care in the previous 12 months, in order to establish their self-reported use of unscheduled care over the prospective year. The main results from this study have been published and full details are available. 236

Ethics approval and guidelines for reporting

The study received ethics approval from the Northwest 8 Research Ethics Committee (REC) – GM East (reference number 09/H1013/80). All participants provided written informed consent. We followed the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines for reporting observational studies. 243,244

Statistical analysis

For all demographic variables, other LTCs reported by the participant and hospital admissions data, we present numbers and percentages for categorical variables for the whole group and for each of the four exemplar conditions. As there was considerable overlap in the four LTC groups with many patients having multiple diagnoses, patients with and without each of the four conditions were compared using Fisher’s exact tests followed by Bonferroni adjustments for multiple testing (four tests per variable).

Results

Recruitment

The flow of participants into the study is shown in Figure 16. Baseline questionnaires were sent to 6682 participants, with 2553 responding (38.2%). Of those returned, there were 1860 usable questionnaires (27.8%). More women responded than men (28.6% vs. 25.6%; p = 0.007), and older patients responded more than younger patients (response rate was 36.0% for patients aged 70–79 years, decreasing to 9.9% for patients aged 18–29 years, but was 30.1% for patients aged ≥ 80 years; p < 0.001). The response rates ranged from 16.7% to 35.2% at the 10 different practices (p < 0.001). There were no significant differences on any variables between the patients who returned the questionnaire without prompting and those who returned it after receiving a reminder (n = 467; 25.1%). Eight out of the 10 practices were in the top 10% of the most deprived areas in England, with five in the top 5% and two in the top 1%. 235

FIGURE 16.

Prospective longitudinal study of patients with LTCs in primary care: flow of study participants for analyses using EHAs as main outcome. Adapted from Guthrie et al. 236

A total of 1488 patients (80%) provided consent for their medical records to be reviewed, and these were retrieved and examined for 1415 patients altogether, 1411 for the year before, and 1398 patients for the prospective year (four sets of notes were obtained for the prospective year but had been missed for the baseline year and were unavailable when researchers went back to them, which explains the discrepancy between 1415 and 1411 for the year before baseline). The disparity between the number of patients consented and notes being reviewed is largely accounted for by the death of one of the GPs working in a single-handed practice and we were unable to retrieve the notes as patients were subsequently registered in a number of practices across the city. This also accounts for the lower number of notes reviewed for the prospective year (1398 out of 1415).

Patients whose records were examined were significantly more likely to be male (p = 0.026) and to have reached a moderate educational standard [i.e. some ‘O’ Levels, General Certificates of Secondary Education (GCSEs), or higher] (p = 0.002) than patients whose records were not examined.

We achieved our predetermined sample size (n = 500) for all of the four LTCs except for COPD. Out of the 1860 patients who returned questionnaires, the QOF diagnoses from the GP databases were as follows: 590 had ischaemic heart disease, 708 had asthma, 617 had diabetes and 449 had COPD. There was considerable multimorbidity, with over 20% of patients having at least two of the four exemplar diagnoses, which explains why the number of patients with the four LTCs adds up to more than 1860. There were 963 females in the cohort (51.8%), the mean age was 62.3 years [standard deviation (SD) 15.4 years] and 81.6% identified themselves as white British. Approximately half of the participants were single, widowed, divorced or separated. Table 11 shows the basic demographic details for the cohort as a whole and for the four exemplar LTCs.

TABLE 11 - Demographic variables for participants in the study cohort as a whole and for each of the four exemplar LTCs

p < 0.05.

Poor education is defined by not achieving any ‘O’ Levels, GCSEs or any higher education.

Demographic variable	All patients (N = 1860), n (%)	Exemplar LTC
		Asthma (N = 708), n (%)	CHD (N = 590), n (%)	COPD (N = 449), n (%)	Diabetes (N = 617), n (%)
Female	963 (51.8)	471 (66.5)a	228 (38.6)a	226 (50.3)	268(43.4)a
Aged ≥ 65 years	919 (49.4)	233 (32.9)a	396 (67.1)a	276 (61.5)a	323 (52.4)
Single	349 (18.8)	178 (25.1)a	72 (12.2)a	50 (11.1)a	115 (18.6)
Widowed, separated or divorced	598 (32.2)	196 (27.7)a	229 (38.8)a	180 (40.1)a	194 (31.4)
Poor educationb	1110 (59.7)	346 (48.9)a	418 (70.8)a	333 (74.2)a	379 (61.4)
Not working due to ill health	275 (14.8)	136 (19.2)a	70 (11.9)	68 (15.1)	86 (13.9)
Unemployed and seeking work	50 (2.7)	30 (4.2)a	9 (1.5)	6 (1.3)	9 (1.5)

Participant-reported use of unscheduled care during 12-month follow-up period

Shortly before the 12-month follow-up questionnaires were mailed, the practices excluded a further 141 patients: 35 had died, 48 had left the practice and a further 58 were excluded for other reasons. Follow-up questionnaires were mailed to 1719 of the baseline cohort, of whom 1203 (70.4%) returned the completed questionnaire. Figure 17 shows the modified flow chart, which includes the responses to the 12-month follow-up questionnaire.

FIGURE 17.

Prospective longitudinal study of patients with LTCs in primary care. Flow of study participants, including eligibility and responses to the 12-month follow-up questionnaire.

Out of 1170 patients who completed the section on use of unscheduled care in the follow-up questionnaire, 455 (38.9%) reported having used unscheduled care in the prospective 12 months; 302 reported attending an ED; 176 reported attending a different emergency facility to an ED; 51 reported calling an ambulance; and 188 reported receiving an emergency GP home visit. For individual exemplar LTCs, these numbers of participants were as follows: asthma, 185 out of 430 (43.0%); CHD, 153 out of 385 (39.7%); COPD, 136 out of 270 (50.4%); and, diabetes, 149 out of 398 (37.4%). Only 310 patients (26.5%) reported having used unscheduled care more than once in the prospective 12 months.

Participant self-reported unscheduled care in the follow-up year was significantly associated with older age (p = 0.035); being widowed, separated or divorced (p < 0.001); a lower level of education (p = 0.010); a HADS depression score of ≥ 8 (p < 0.001); COPD (p < 0.001); asthma (p = 0.030); self-reported arthritis and/or joint problems (p < 0.001); self-reported stomach or bowel problems (p < 0.001); more severe physical illness (p = 0.021); experiencing a threatening life event (p < 0.001); and use of unscheduled care in the previous year (p < 0.001).

In logistic regression, the baseline variables which were significantly independently associated with self-reported prospective unscheduled care use were having no partner; asthma; COPD; reporting a threatening life experience; used unscheduled care in the year before baseline; and a HADS depression score of ≥ 8 (Table 15). Maximum severity of QOF diagnosis was not significant. The jackknife procedure found lack of a partner (p = 0.001), asthma (p = 0.015), COPD (p = 0.045), unscheduled care in the previous year (p < 0.001) and a HADS depression score of ≥ 8 (p = 0.008) to be significant.

TABLE 15 - Results of multiple logistic regression analysis with dependent variable participant self-reported unscheduled care use in the prospective year (total n = 1127)

Sig = significance probability or p-value.

Notes

Analysis is adjusted for non-completion of unscheduled care variables using relevant sampling weights. Age, sex, poor education, not working because of ill health, CHD, diabetes, self-reported cancer, stomach problems, high blood pressure, arthritis, and distance to the nearest ED were included, but were not significant, and not shown in the table. Stata imputation for missing data on distance to hospital (n = 3) and maximum severity (n = 356) was used, based on other independent variables in the analysis.

Possible risk factor	Model
	OR	95% CI	Siga
No partner	1.90	1.34 to 2.69	< 0.001
Asthma	1.81	1.16 to 2.81	0.009
COPD	1.94	1.14 to 3.29	0.014
Threatening life experiences (per each experience)	1.16	1.00 to 1.34	0.043
Used any unscheduled care in year before baseline (participant report)	3.64	2.47 to 5.37	< 0.001
Maximum severity of QOF diagnoses	1.26	0.97 to 1.62	0.080
HADS depression score of ≥ 8	1.79	1.22 to 2.64	0.003

There was a significant variation between the 10 GP practices in the percentage of patients who self-reported using unscheduled care in the prospective year ranging from 26% to 62% (χ² = 19.8, df = 9; p = 0.019). There was, however, no significant correlation between the IMD for each practice and percentage of patients who reported using unscheduled care (r = 0.16; p = 0.67).

Discussion

In phase 2 of the programme, we conducted a large longitudinal prospective cohort study of patients with at least one of our four exemplar LTCs in primary care in the UK. In relation to our first main aim, which we stated at the beginning of this chapter, we were able to derive estimates of the frequency, and the pattern, of unscheduled care use in patients with our four exemplar LTCs.

The prospective study enabled us to collect data on patients over a 2-year period. During that time, we found that very few patients were actually ‘frequent users’ of unscheduled care, although many used it once or twice. Approximately 16% of patients in the cohort had an EHA in the baseline year, and the proportion was similar in the prospective year.

In relation to our second aim, we were not able to identify predictors of frequent use of unscheduled care, as so few patients were frequent users. Our findings suggest that the ED attendances and admissions to hospital of the participants in the study were most likely necessary and entirely justified, and prompted by an exacerbation in their physical condition.

In our analyses, we focused attention on those patients who used any unscheduled care during the relevant study time periods, in comparison with those who had not.

We used two measures of unscheduled care: EHAs recorded by contemporaneous data from GP records; and participant self-reported use of unscheduled care, via questionnaire data.

We found a similar pattern of results, for both objective data about EHAs from the GP records and participant self-reported data. The following factors were independent predictors of unscheduled care in both sets of analyses: use of unscheduled care in the previous year, depression, not having a partner and stressful life experiences. Other factors that were independent predictors in at least one of the analyses were having more than one of the four exemplar LTCs and severity of physical illness.

A prior history of use of unscheduled care in the year before completion of the questionnaire was by far the most powerful predictor of use of unscheduled care in the prospective study period (whether measured by EHAs or self-reported use of unscheduled care). This factor alone increased the risk of unscheduled care use by at least three and a half times.

We also found that baseline depression was a significant predictor of both prospective EHAs and participant self-reported use of unscheduled care. The ORs for depression that we found in the cohort study were of a very similar order to the results of the systematic review we carried out in the first phase of the programme. Depression was independently associated with both EHAs (OR 1.58, 95% CI 1.04 to 2.40; p = 0.031) and attendances at EDs (OR 1.79, 95% CI 1.22 to 2.64; p = 0.003). 236 More severe depression was associated with a greater than twofold increased risk of using unscheduled care (EHAs, OR 2.42, 95% CI 1.12 to 5.23; p = 0.025; and participant-reported use of unscheduled care, OR 3.58, 95% CI 1.88 to 6.82; p < 0.001). 236 To our knowledge, this is the only UK study, to date, which has shown that baseline depression is an independent predictor of future use of unscheduled care in people with LTCs, even when controlling for prior use of unscheduled care and severity of physical illness.

Over 70% of the participants in the prospective study had multimorbidity with at least two LTCs. We examined the effect of multimorbidity on use of unscheduled care by entering the number of exemplar conditions that each participant had (recorded by the QOF) into a separate analysis. Each additional LTC increased the risk of using unscheduled care by approximately 40%. This is consistent with other work in this area,26 but of note, depression remained an independent predictor in these separate analyses.

There is currently a debate whether or not there is benefit attached to routine case-finding for depression in primary care,245 and during the lifetime of this programme, case-finding for depression in certain LTCs has been withdrawn as a QOF target for primary care in England. There is insufficient evidence at present to support its reintroduction, but our findings suggest the value of case-finding for depression in particular high-risk groups has not been fully explored.

As we used the HADS in the cohort study, we also had the opportunity to investigate the role of anxiety, but not panic, as a predictor of unscheduled care in patients with LTCs. The analyses are not presented in this chapter, as they were not the main focus of our investigation. However, we found no strong evidence for anxiety acting as an independent predictor of unscheduled care. These findings, again, support the results of the systematic review we carried out in the first phase of the programme, which found no evidence of an association between anxiety and use of unscheduled care.

We found that lack of a partner and experiencing threatening life events were also risk factors for using unscheduled care. Although the ORs for life stress were relatively small, the ORs in each regression are for each additional item experienced. Hence, several threatening experiences in the same person would increase their risk of using unscheduled care quite substantially. Severity of illness was also included in one of the predictor models and is a well recognised risk factor for unscheduled care use. 20,23,30,57

The duration of our prospective follow-up period was 12 months, and beyond this period of time, the predictive power of factors appears to wane. 246 Even so, over 30% of our cohort reported using unscheduled care at least once during this 12-month period, and 17% had at least one EHA. This suggests that there may be a potential to reduce unscheduled care in this population if some of the contributing factors can be addressed.

The strengths of our study included the following: a large cohort of primary care patients; recruitment target achieved in three out of the four exemplar LTCs; an independent measurement of health care which employed scrutiny of GP records as well as a measure of participant self-report; use of a standardised instrument to measure depressive symptoms; and the independent assessment of severity of illness using GP records. A further strength was the inclusion of patients with several LTCs as opposed to focusing on a single disease, with most patients in the baseline cohort having multimorbidity. The follow-up data of the patients who were entered into the prospective study were good, with nearly 80% giving consent for their medical records to be checked. In addition, there was a good response to the follow-up questionnaire (over 70%).

The response rate to the baseline questionnaire was disappointingly low, despite efforts to maximise recruitment. Achieving good response rates to postal-based questionnaires is becoming a challenge and two other large-scale studies, which have been recently carried out, have reported similar low response rates to our own: the GP Patient Survey247 and a recent large primary care cohort study from Oxford that targeted patients with chronic physical disease. 248 In the area where we carried out our study, response rates to postal questionnaires have declined dramatically over the last 15 years, dropping by over 20%. 249

Our main concern with the low response rate to the baseline questionnaire was the potential for bias in our study sample. Participants in our study were more likely to be female and older than non-participants. They were also over-represented by people with a white British background and the study was carried out in a relatively deprived area of Manchester.

We attempted to adjust for differences in age and sex between responders and non-responders using inverse probability sampling weights based on age, sex and GP practice, which were the only variables available to us for the non-responders. We also compared the responses of patients who returned and completed the questionnaire spontaneously with the responses of those who returned the questionnaire after a reminder. The latter group could be argued to be more representative of non-responders, as they would not have responded without prompting. There was no difference between these two groups with any of the variables.

We also considered whether patients who participated in our study were either over- or under-represented in terms of depression. Approximately 25% of people in the study had a HADS depression score of ≥ 11 and approximately 40% had a HADS depression score of ≥ 8. If a score on the HADS depression subscale of ≥ 8 is broadly equivalent to subthreshold depression and a score of ≥ 11 is broadly equivalent to clinical depression, the prevalence of depression in our sample was commensurate with that expected for a population of patients with LTCs. 69

Our results cannot be generalised to all populations of patients with LTCs, but they suggest that, in a substantial proportion, depression is a potential important psychosocial determinant of use of unscheduled care, and could be a potential red flag in addition to other more recognised predictors of unscheduled care (e.g. previous use of unscheduled care). The results of our systematic review, conducted in phase 1, also suggested that depression should be considered as a strong candidate for inclusion as a red flag variable, and the similarity between the magnitude of the ORs for depression in the systematic review and the cohort study provides some support for the validity of the findings from the cohort study, despite the low response rate.

Our findings suggest a potential model that links physical and psychosocial factors in people with LTCs, increasing the risk of unscheduled care use through exacerbations of illness (Figure 18).

FIGURE 18.

A model that links together physical and psychosocial factors in people with LTCs as risk factors for the use of unscheduled care.

The model shows that the biggest risk factor for unscheduled care use is prior use of unscheduled care in the previous 12 months. Admission to hospital with an acute exacerbation of illness is linked both to severity of illness20,23,24,30,57,250 and to depression. 27,29,30,49,92,251 There is also a well-recognised bidirectional association between the severity of physical illness and depression. 252 Living alone is linked both to an increased risk of depression in people with LTCs253 and to worse outcome,254 and life stressors are also associated with an increased risk of depression in elderly people. 255 All the factors in the model have an independent influence on unscheduled care use, but there are also multiple interactions between different components of the model, which may worsen overall outcome. We also found that the number of LTCs increases the risk of unscheduled care use and this is represented by the boxes and arrows to the left of Figure 18. 256,257

We recognise that the relationship between physical illness and depression is more complex than represented by the model in Figure 18. The way in which depression and severity of illness interact with each other has been elegantly described by Katon in a detailed and thorough review of the relevant literature. 252 Katon’s conceptual model brings together genetic and childhood factors that predispose individuals to both physical and mental health problems later in life. Comorbid depression in physical illness affects the patient–physician relationship, the ability to seek comfort and support from others, and the ability to self-care. Depression is a major risk factor for non-compliance with medical treatment108 and, compared with non-depressed patients, depressed patients are three times more likely to be non-compliant with medical recommendations. Depression often involves a degree of hopelessness, impaired concentration and poor sleep, all of which may interfere with medication adherence as well as attendance at appointments. Problems with adherence to treatment may lead to exacerbations in physical illness, which then necessitate urgent treatment. In addition to effects on adherence, depression has also been linked to other forms of poor self-care in physical disease, which may include such behaviours as overeating, increased alcohol consumption, smoking, etc. 252 All these behaviours can have a deleterious effect on physical illness and increase the risk of exacerbations of illness in people with LTCs. Katon’s model will be discussed further in Chapter 11, when the overall findings of the programme are brought together.

Limitations

There were several limitations of the prospective study. The issue of low response rate has already been discussed above. There may have been recall bias of self-reported use of unscheduled care by the study participants. It was not feasible to send out follow-up questionnaires at either 3 months or 6 months, in the context of the overall programme. Limiting the follow-up to 12 months may have meant that some participants inaccurately recorded their contacts with unscheduled care.

Although great efforts were made to record severity using data recorded in the GP notes, there remain several methodological problems with the severity ratings we used. Each rating according to the specific LTC will be discussed in turn. Asthma severity was classified according to the Global Strategy for Asthma Management and Prevention. 237 This is based on the type of medication that is prescribed to control the disease. The rating does not take into account whether or not the patient adheres to the medication that is prescribed, or even collects the medication from the chemist. COPD was rated using the FEV₁, as recommended by the GOLD criteria, as defined at the time of the study. 239 The GOLD criteria, however, have subsequently been updated and suggest that COPD severity should now be rated using a composite of COPD symptoms, plus FEV₁ divided by forced vital capacity and risk of exacerbations. Although we recognise that the updated guidelines are an improvement on the ones that we used in the study, it would not have been possible to use the updated criteria had they been available at the time, as the kind of detailed information they require is not recorded routinely in GP records. CHD severity was based on how much patients were limited by their disease during physical activity. Such data, however, were not always recorded in the GP notes. HbA_1c was used as a proxy for severity for diabetes because of difficulties in identifying reliable ways of recording severity from the data recorded in GP records. We recognise that HbA_1c is more a marker of adherence to treatment than severity of disease, and this may have led to possible confounding in our results as, discussed earlier, adherence is influenced by depression. We were also unable to distinguish between type 1 or type 2 diabetes.

In the next chapter, we will report on the costs associated with use of unscheduled care.

Abstract

Overview

In this chapter, we describe the economic analysis of the longitudinal cohort study that we described in the previous chapter (see Chapter 3).

The overall aim of the cost analyses was to explore the costs of unscheduled care in the exemplar conditions. Specific research questions to address this aim were:

• What are the costs of scheduled and unscheduled health services used by people with the exemplar LTCs?

• What is the relative contribution of unscheduled and other care to the total cost of health services used by people with the exemplar LTCs?

Background

People with LTCs account for 50% of all GP appointments, 64% of outpatient appointments and 70% of all inpatient bed-days. In total, around 70% of the total health and care spend in England (£7 out of every £10) is attributed to caring for people with LTCs. This means that 30% of the population account for 70% of the spend on health care. 258

All four of the exemplar conditions included in the CHOICE programme carry enormous economic health burdens. The total annual cost of all patients with CHD was £7.06B in 1999,259 the highest of all diseases in the UK for which comparable analyses have been done. 260,261 The UK has some of the highest prevalence rates for asthma in the world,262 with England, Wales and Scotland ranked among the top nine countries. The total annual cost of COPD to the NHS is estimated to be over £800M for direct health-care costs. 263,264 In the late 1990s, 24 million working days per year were lost because of COPD, with the cost of lost productivity being estimated at around £2.7B. 9 In a recent European study, unscheduled care accounted for 56% of total costs in adults with asthma,68 regardless of the severity of patients’ symptoms. In the USA, over 12 months, 8.3% of patients with asthma made at least one visit to the ED and 13% of COPD patients made six or more visits. 19 The costs of health care for patients with type 2 diabetes in the UK will rise by 25% over the next 30 years and the economic burden of the disease will rise by 40%. 265

Comorbid mental health problems in people with LTCs raise total health-care costs by at least 45% for each person, and it has been estimated that between 12% and 18% of all NHS expenditure on LTCs is linked to poor mental health and well-being. 38

Methods

The design and main methods of the longitudinal cohort study were described in Chapter 3. The following section focuses specifically on the cost analysis. Inverse probability survey weights, as described in Chapter 3, were also applied here.

In this analysis, we included data on both scheduled and unscheduled care. The items of service use included in the scheduled care costs were use of routine GP practice services at the surgery or at home (GP, nurse and other health-care workers, and telephone calls); hospital outpatient and day case visits; and routine or elective hospital admissions. The service use categories included in the unscheduled care costs included GP OOH service and OOH home visits, accident and emergency (A&E) services, including walk-in centres and minor injury facilities, non-routine hospital day case visits and inpatient admissions. The detailed service use data about whether or not a participant had used a service and the intensity or frequency of use were collected by participant questionnaire and GP case note review as part of the longitudinal prospective cohort study (see Chapter 3). Data were collected from GP records using a standardised data extraction form. The data collection form was pilot tested in two GP practices, for 72 patients’ records. The data collection form was revised to take account of any issues that arose. Data on scheduled and unscheduled care were also collected in the participant questionnaires at baseline and at the 12-month follow-up. Participants were asked to recall service use over the previous 12 months. The data from GP records were used as the primary source of data, supplemented by information from the participant questionnaire that reported service use not available from the GP notes. The data were recorded and costed per patient, to give a total cost per person in the study. This cost was used in all the analyses, with no attempt to attribute parts of the cost to any of the four exemplar conditions. These service use data were costed using published national health- and social-care unit costs. This was to facilitate the transferability of the results to different NHS settings. The two main sources of unit costs were the National Schedules of Reference Costs 2012–13 (NSRC) (hospital costs), which is published annually by the DH266 and The Unit Costs of Health and Social Care 2012, which is published annually by the Personal Social Services Research Unit. 250 The price year used was 2011–12, which was the latest available when the unit cost data were collected.

Results

Discussion

The total health-care costs for patients with LTCs in the longitudinal cohort study we conducted were similar to the national average cost per person. Health-care spending per person in the UK was £2268 in 2012. 234 The average costs for each condition in this study were also comparable to the most recent published estimates of health-care costs in the UK.

A focused search of the literature in October 2014 found a limited number of robust cost studies that provided robust and up-to-date data that are relevant to the UK health-care system. The results of these are compared with the findings from our longitudinal prospective cohort study below.

The total direct health-care costs of CHD for the UK were reported as £3859M in 2004 prices, in a study that used a top-down analysis of total health-care service use attributed to CHD. 260 Applying an annual prevalence of CHD of 3.5%261 to the population of the UK in 2004 (59,950,364),267 the average cost per person is approximately £1839 in 2004 prices. This equates to approximately £2821 per person if the costs are inflated to 2011 prices,250 which is similar to that found in this study.

The total direct health-care costs of diabetes for the UK were reported as £9777M in 2010–11, in a study that used a synthesis of literature review and total recorded health service use attributed to diabetes. 265 This total cost was based on an annual prevalence of diabetes of 3,818,545 people (adults and children with type 1 or type 2 diabetes) in 2010–11. 265 This indicates an average cost of £2560 per person with diabetes in 2010–11 prices. Inflating these to 2011–12 prices250 gives a cost of approximately £2790, which, again, is similar to that of year 2 costs found in this study.

A study using registry data estimated the average cost per person of severe refractory asthma to be in the range of £2912 to £4217, in 2011 prices. 268 The costs of non-severe asthma in a smaller sample of 80 people were estimated at £1670 to £2788. 268

The costs of COPD in this study are comparable to those found in a recent study using a similar methodology, in a sample of 58,589 participants with COPD. 269 This estimated the cost of COPD (excluding medications) at £1806 per person in the year prior to baseline and £2108 in the year following baseline (2011 unit costs). The costs ranged from £1523 for patients with no exacerbations to £3396 for patients with two or more moderate to severe exacerbations.

Overall, the descriptive and multivariate analyses indicated that the costs of unscheduled care are an important component of total care, and account for around 50% of annual costs for people with asthma, CHD, COPD or diabetes. Although the costs are similar between the conditions included in this study, there was a high level of variation within conditions. This is highlighted by the substantially lower costs of those participants who did not use any unscheduled care over the 2 years of the study (year 1: mean £838, 95% CI £716 to £961; year 2: mean £841, 95% CI £715 to £967), compared with those who used unscheduled care in both years (year 1: mean £5927, 95% CI £3283 to £8572; year 2: mean £6212, 95% CI £3540 to £8884).

The multivariate analyses indicated that the presence of depression (i.e. a HADS depression score of ≥ 8) and the LTC being rated as severe or very severe were associated with higher costs of scheduled and unscheduled care. The analyses included a broad range of participants’ demographic and clinical characteristics, most of which were not statistically associated with the use or costs of unscheduled care or the costs of scheduled care. This may indicate the need for further work to identify whether or not there are other factors that it would be useful to measure/include in future research.

The results reported in Chapter 3 and the current chapter emphasise the key role that depression plays in the use of unscheduled care in people with LTCs. The results from this cost analysis also indicate increased costs of scheduled care for patients with LTCs who suffer from depression. The severity of patients’ LTC was important in the cost analyses, as were total previous costs. These findings are similar to the activity-based analyses in Chapter 3, where previous use of health care was a very important predictor. Severity of physical condition just failed to reach significance in the analyses in Chapter 3, but is clearly an important contributor to overall cost of health care in people with LTCs.

There was great variability in terms of cost of health care for patients with LTCs between the practices, with costs, on average, £1000 more per annum per patient for some practices than others. This was also true for individual patients, as those who used unscheduled care in either year of the study had significantly higher costs than those who did not. This suggests that a ‘one size fits all’-type intervention may not be applicable and that interventions to reduce the use of unscheduled care may need to be tailored to the practice and the patient.

Limitations

The response rate to the baseline questionnaire in the cohort study was 27%, and all the cost analyses and the analyses reported in Chapter 3 need to be viewed in this context. It may be that the patients who consumed most health care in the practices in the cohort study were those who would be excluded from participation by their GPs, because they were too ill. The costs may therefore underestimate the actual costs of people with LTCs in the practices. However, all the cost analyses used inverse probability sampling weights to reduce the impact of this selection bias.

The costs of ambulance services to transport participants to and from unscheduled and scheduled care were not included in this study, which will underestimate the total costs of care. These data are not typically recorded on GP notes and it was not felt that asking for detailed data about the use of ambulance services in the patient health survey would provide consistent data. The costs of medications were also excluded from the study, which will again underestimate total costs. Information about the prescription and use of medications is complex to collect and analyse – it was beyond the resources available for this study to collect robust and consistent data on these items for costing and analysis.

The relative costs of non-elective hospital admissions were high and appeared to account for the main component of the costs of unscheduled care, when compared with the use of other unscheduled care services. However, this may in part be due to an artefact of the data. There was concern that GPs were only consistently notified by the admitting hospital about emergency inpatient admissions. Information about use of ED services and OOH services may be less consistently reported to GPs. Combined with issues of patient recall, this could mean that the costs of other unscheduled care are underestimated to a greater degree than the costs of emergency inpatient admissions.

Conclusions

The costs of CHD, diabetes, asthma and COPD estimated in this cohort study are within the range of other published studies in the UK.

Overall, the descriptive and multivariate analyses indicated that the costs of unscheduled care are an important component of total care, and account for around 50% of annual costs for people with asthma, CHD, COPD or diabetes.

The costs of participants who did not use any unscheduled care over the 2 years of the study were substantially lower than for participants who used unscheduled care in both years.

The results from Chapter 3 and this analysis of costs indicated that the presence of depression (i.e. HADS depression score of ≥ 8) and whether the LTC was rated as severe or very severe were associated with higher use and costs of scheduled and unscheduled care.

Abstract

Overview

In this chapter we describe a qualitative study that we undertook in parallel with the longitudinal cohort study described in Chapter 3. The main aim of this study was to improve our understanding of why patients with LTCs use unscheduled care. To do this, we carried out detailed interviews with HCPs who were involved in the care of patients with LTCs, and we also interviewed a subset of patients who were participants in the longitudinal cohort study. The work has been written up and published elsewhere. 270

Our main objective from the original programme application was to identify personal reasons for use of unscheduled care including barriers to access for routine care, patients’ motivations, expectations and decision-making processes, and influences from families and relevant health-care providers.

Background

Methods

Recruitment

Health-care practitioners

Health-care practitioners from different professions and varying responsibilities in unscheduled care were given the opportunity to participate in the study. All HCPs were identified through either purposive sampling or recruitment through local networks (Table 27 and Figure 19). Information sheets were given to prospective participants, and informed written consent was obtained prior to commencing interviews. Although some participants were recruited through local networks, interviewers only interviewed participants with whom they had no prior relationship.

TABLE 27 - Recruitment methods of HCPs

Profession	Recruitment method
GP	Purposive: identified from practices already involved in the CHOICE programme
PN	Purposive: identified from practices already involved in the CHOICE programme
OOH GP	Purposive: identified through local OOH provider
ED doctor	Purposive: identified from three local hospitals
ACM	Purposive: identified from contacting local networks
SN	Purposive: identified from local networks
DN	Purposive: identified from contacting local networks

FIGURE 19.

The Consolidated Standards of Reporting Trials (CONSORT) flow diagram demonstrating sampling and recruitment of patients and practitioners.

Patients

The sample of patients for this study was drawn from patients who had agreed to participate in the large, longitudinal cohort study to identify predictors of unscheduled care use, which was described in Chapter 3. The target patient population were patients aged ≥ 18 years with one or more of the four exemplar LTCs:

1. COPD

2. CHD

3. asthma

4. diabetes.

Six out of the 10 general practices involved in the cohort study were used for the qualitative study. Patients were identified using the QOF registers of general practices and invited to take part in the cohort study (see Chapter 3). The Consolidated Standards of Reporting Trials (CONSORT) flow diagram (see Figure 19) demonstrates that, of 939 patients across six different general practices who returned questionnaires for the cohort study, 474 (50%) consented to be contacted for interviews. From these consenting patients, the team strategically sampled 212 individuals on age, sex, type and number of LTCs, and different levels of self-reported use of routine primary care and unscheduled care use. The purpose was to achieve a variation of characteristics and to minimise sampling bias. Invitation packs including a patient information sheet (PIS), a reply slip, and a stamped addressed envelope were sent out to this sample of 212 individuals. A total of 67 participants agreed to be interviewed by returning a reply slip or calling the researchers directly; recruitment ceased when category saturation had been achieved, with the final sample comprising 50 patients. Informed written consent was obtained from all participants prior to commencing interviews. Patients were not known to the interviewers prior to the interview, and it was made clear on information sheets and in person what the study was about and that the interviewers were non-clinical and not involved in their health care in any way.

Results

Discussion

Owing to the high financial costs of unscheduled care services and the increasing burden of LTCs, reducing unscheduled care use by patients with LTCs is a major policy priority across all levels of the UK health system and internationally. 270,277,278 This chapter reports the first study to explore the perceptions of a range of HCPs and patients about unscheduled care use in people with LTCs.

Across our sample of different HCPs involved in the provision, and reduction of unscheduled care use for patients with LTCs, unscheduled care use was viewed as a necessary component of care, with exacerbations recognised as inevitable in patients with LTCs. Thus, ED doctors did not see it as their role to reduce patients’ unscheduled care use. However, some HCPs saw unscheduled care use as a failure to meet policy targets, on which they were performance managed. Thus, GPs discussed the need to optimise the system to direct patients to more appropriate services, and PNs, ACMs and SNs, driven by policy targets to reduce unscheduled care use, described mitigating their relative lack of power and influence by negotiating the system. Most HCPs also described maximising their resources by using consultations with patients to attempt to enact behaviour change and reduce unscheduled care use. However, they only described information giving, and there is little evidence that this is effective in changing behaviour. Therefore, approaches to reducing unscheduled care use depended on the HCP’s role.

Consistent with HCPs’ views, patients viewed unscheduled care services as one of several health-care options available to them, complementing routine care, and using unscheduled care was not necessarily seen as a failure of self-management or routine care. 185 Patients drew on previous experiences of services and practitioners when choosing how to respond to illness exacerbations. The choice of unscheduled care compared with routine primary care was shaped by patients’ perceptions of urgency, which were in turn influenced by previous responses from health-care practitioners,26 and by involvement of friends or family. Choosing between different unscheduled care providers was also shaped by perceptions of those services, formed by previous experiences of their accessibility and technological expertise. These findings, from patients’ detailed accounts of specific instances when they decided to use or avoid unscheduled care, therefore confirm and expand on previous qualitative evidence from studies in which patients were asked about their use of unscheduled care in more general terms. 185

Two theoretical concepts from the health-care access literature, hitherto unapplied to the problem of unscheduled care use in people with LTCs, provide an interpretive framework for understanding how patients make choices between health-care services. 279,280 The first, ‘candidacy’, describes how access to health care is framed as often requiring work for patients to achieve, and eligibility to access care is continuously negotiated in patient–practitioner interactions. 279 Developed from interpretive synthesis of literature on access to health care in socioeconomically disadvantaged groups,279 the concept has been applied to health-care use in other vulnerable populations. 281,282 The second concept, ‘recursivity’, describes how future demand for services, and the process of help-seeking, is determined by a patient’s previous experiences. 280

The establishment of candidacy was evident in patients’ accounts of interactions with practitioners in both primary and secondary care services. Figure 17 describes a pivotal instance of health care in response to palpitations (perceived fast or irregular heart beat) wherein the specialist and hospital staff ratified the patient’s decision to use unscheduled care. Negotiations of candidacy were sometimes bypassed by family and friends who acted on behalf of patients. Patients were sensitive both to practitioners’ responses to a request for help, and to the responses of family and friends; both recursively shaped patients’ candidacy when making future health-care decisions, demonstrating that help-seeking is a social process involving more than just patients’ decisions.

Recursivity was seen in patient accounts of how they chose between health-care services, particularly in the choice to use unscheduled care. They framed these choices by drawing on previous experiences of help-seeking. Although patients described using unscheduled care as inevitable, their judgements of urgency and their understanding of why unscheduled care use was ‘inevitable’ were socially conditioned, arising out of previous encounters with health-care practitioners, family and friends, and particular services. Box 1 illustrates recursivity in how the judgement of urgency, and ultimately candidacy for accessing care, is established through previous encounters. Similarly, Box 2 illustrates how previous experience of particular qualities in a health-care service (in this case, easy accessibility and technological capability) ensures future reliance on that service for similar problems. That is, previous experiences of a service can build a foundation of trust, which strengthens patients’ confidence in choosing that service in future.

When considered together, the concepts of candidacy and recursivity highlight that the key determinants of patient choice of health care evolve over time and through social interactions, with future health-care use contingent on prior service responses to patients’ requests for care, and on previous experiences of the social process of care. 279,283 Patients rely on the knowledge they develop through previous experiences of services and practitioners to choose between services and to establish their candidacy for accessing services. The HCPs talked about changing individual patient behaviour as part of an ongoing relationship. This approach would fit well with the way patients make decisions, if the patient–practitioner relationship also addressed the wider health-care experiences of the patient. The HCPs highlighted difficulties with the lack of consistency across services in terms of responses to patients. In the case of the ACMs this lack of consistency undermined attempts to reduce unscheduled care. Patients make choices dependent on their knowledge of and confidence in the services they have used before, and judge their candidacy to use services based on previous responses. If services are inconsistent, this can make patient decisions more difficult and patients are likely to prioritise based on the most salient and positive service experiences.

Patients experienced barriers to using primary care in times of urgent need, and this recursively shaped future choices between services. ‘Permeability’ offers a way to conceptualise the impact of these barriers. 279 Highly permeable services require less work and fewer resources from patients who access them, for example EDs in the UK, which are open at all times. The permeability of emergency services is reinforced by ED doctors, who believe that their role is to see all patients regardless of need. A service that seems accessible may in fact be impermeable to particular patient groups. 281 For example, despite general practices being locally available, with designated systems for urgent access, patients in our study described that they were, in fact, impermeable because of factors such as receptionists’ gate-keeping, and travel cost or mobility problems. In our study, the combination of high permeability and technological expertise led most patients to choose the hospital ED in times of perceived urgent need.

In seeking to reduce unscheduled care use, health-care policy defines patients as in need of education to use services effectively, or suggests the need for reorganisation of health-care systems to reduce use of costly emergency care services, especially the ED. 284 This ‘deficit’ model also dominates previous research investigating unscheduled care use, with research focusing on characteristics of the patient20,278 or the health-care system20,285 that increase unscheduled care use. In this study, the GPs hinted at a deficit model of patients, suggesting changes to the health-care system such as triage and fee-for-service to force patients to use the system more appropriately. In contrast, this qualitative work demonstrates that patients understood the array of unscheduled care services available and were discriminating in their use of them, influenced primarily by previous experiences of services that recursively shaped their future health-care choices. It contributes to a growing body of research emphasising the social processes of help-seeking, and the expertise patients bring to decision-making around health-care use. 286,287

When patients with LTCs feel vulnerable in health crises, it is their previous experience of services that shapes their perception of candidacy and thus their choice of service to access, with patterns of under- or over-use of services becoming established recursively based on these responses. This is consistent with the ACMs and SNs who described making their systems permeable so that patients used them instead of unscheduled care. However, this does have implications for the short-term nature of their services and what happens once patients have learned about these services and are then excluded from using them following discharge.

Both HCPs and patients recognised the limitations of availability and accessibility of primary care and the need for unscheduled care as a last resort. As previous experiences of health-care practitioners and services have a recursive influence on patients’ health-care decisions, it may be that better management in primary care, utilising pre-existing practitioner–patient relationships to address health-care use, and improved access to primary care for patients with LTCs could reduce the frequency with which patients feel they need to draw on the last resort of unscheduled care.

There were several important strengths of the above study. The patient sample was large and heterogeneous, and patient interviews covered a broad range of health-care service use, asking about decisions to use or not use unscheduled care. The HCP interviews likewise included a broad range of practitioners from across the health-care system, and data collection and analysis were conducted by researchers of different professional backgrounds, which reduced the possibility of some form of professional bias.

Limitations

The study also had several limitations. It was undertaken in north-west England and patient participants had to opt into the study when they completed the baseline questionnaire from the cohort study. The participants were therefore a self-selected group. The participants’ recall of unscheduled care use may have been limited and there was a limited ethnic diversity in our sample. These limitations mean that caution should be applied when applying the findings to other geographical and cultural settings. The practitioner group was somewhat smaller than the patient group, and some professional groups within our sample were under-represented because of difficulties in recruitment (e.g. PNs and ED doctors). For those professional groups which are under-represented, it may be that there are other perspectives on unscheduled care use that have not been captured in this study and, therefore, further research may be required to test the findings of this research.

Conclusions

Health-care practitioners do not generally see unscheduled care use as a problem, describing it as inevitable for patients with LTCs. Patients similarly described feeling that they had no alternative but to seek unscheduled care in the face of perceived urgent need. Patients’ judgements that they needed unscheduled care, and their choice of unscheduled care provider, were influenced by previous experiences of health-care services and practitioners, in particular of the ED, valuing the ease of access and technical expertise available there. Although most HCPs saw it within their role to reduce unscheduled care use, the ED doctors we interviewed did not. They reported that attendance by patients with LTCs at EDs was legitimate. This is likely to reinforce help-seeking at EDs by this group of patients. Apart from ED doctors, HCPs described attempting to change the behaviour of the patient, but described limited strategies to do so. In addition, GPs described the need to optimise the system in which patients have to seek help, whereas PNs, ACMs and SNs described negotiating (and often bypassing) systems that constrain them and the care they perceive they can offer patients.

In all the interviews we conducted, there was very little consideration, by either HCPs or patients, of any psychosocial factors that may influence care-seeking. Although the quantitative data from our prospective cohort study identified depression, life stress and living alone as important independent predictors of unscheduled care use, neither patients themselves nor HCPs appear to be aware of such influences. Contacts with unscheduled care are framed almost entirely in terms of a physical context, both at the time the person decides to seek care and on reflection, after care has been sought.

Although, undoubtedly, physical health reasons are crucially important in decision-making about seeking unscheduled care, other, perhaps more subtle, contextual factors (e.g. the fact someone lives alone or is feeling very low) fail to be recognised. Even if HCPs have the opportunity to influence patients’ behaviour, the absence of any kind of dialogue or discourse about psychosocial factors, suggests problems of this nature will remain unchanged.

In the next chapter (Chapter 6), we explore this question further, by investigating what happens in consultations between HCPs and patients with LTCs to see how, if at all, psychosocial factors and planning for action during exacerbations of illness are addressed in routine primary care.

Abstract

Overview

Patients with common LTCs are now routinely assessed on an annual basis in primary care as part of the QOF. In this chapter, we describe further work we carried out to explore what happens in these QOF reviews and in other consultations in primary care involving patients with LTCs. We were particularly interested in how the QOF reviews shaped the focus of consultations; how aspects of routine consultations may have a potential to influence use of unscheduled care; what opportunities there may be in routine reviews for behaviour change; and how any relevant psychosocial factors are addressed. More details of this work have been published. 288,289

Our main objective from the original programme application was to identify barriers to, and facilitators of, those aspects of HCP behaviour with the potential to reduce unscheduled care.

Background

The Chronic Care Model (URL: www.improvingchroniccare.org) and the NHS and Social Care Long Term Conditions Model (URL: http://webarchive.nationalarchives.gov.uk/+/www.dh.gov.uk/en/Healthcare/Longtermconditions/DH_4130652) highlight key system components that need to be addressed in order to deliver effective care for people with LTCs,9,290 including delivery system design, decision support, clinical information systems and self-management support. The QOF incentivises primary care professionals’ activity, financially rewarding the practice when evidence-based care in line with clinical quality indicators has been demonstrated,41 as well as providing a record to ensure payment to the practice against the quality indicators.

Two core activities in managing LTCs within primary care are the promotion of self-management behaviours, and enabling and supporting behavioural change. The chronic care model suggests that 70–80% of people with LTCs should be able to self-manage their conditions with support from health-care services. 291 Behavioural change with regard to risk factors such as smoking, alcohol use, diet and exercise are key targets for improving outcomes for those with LTCs, such as reducing disease-related complications, exacerbations and premature morbidity and mortality. Primary care consultations are crucial for enabling and supporting patients to become proactive managers of their own conditions.

In the qualitative literature review we carried out (see Chapter 2)185 and in our interview study of patient and practitioner perspectives (see Chapter 5), the importance of previous experiences of health care was highlighted in people’s accounts of using unscheduled care. This indicated that primary care practitioners may have a role in influencing patients’ choice of service in times of crisis. However, these data explored patients’ perspectives of decision-making retrospectively. Our intention in the study reported in this chapter was to investigate the role of primary care consultations in influencing patients’ health-care choices prospectively.

Methods

Data collection and analysis

Data collection

Patients and practitioners were interviewed separately after the recorded consultations, using a semistructured interview guide and stimulated recall; short snippets of the consultation were played back during the interview to prompt thoughts and reflections. Prior to the commencement of each initial interview, informed written consent was taken. For follow-up contacts and interviews, consent was checked and retaken to ensure that patients still felt informed and willing to participate in the study.

Topic guides for this study were developed within the research team, taking into account what had been learnt from the earlier portions of the programme. The interviews were framed around understanding the context of routine appointments and the structure of care for the patients, and the content of the specific consultations. The interviewers familiarised themselves with the topic guides prior to commencing interviews. Topic guides were refined over time through discussion within the research team of emerging themes and experiences of interviewing (see Appendix 3, Box 7).

Practitioners were interviewed once about all consultations they recorded, patients were interviewed once following their consultation and then invited to be followed up for a period of 3 months in order to gain insight into patients’ choices around health-care use over time. Follow-up with patients involved regular telephone contact and the option of completing health-care use diaries in between telephone contacts. Patients were then asked to participate in a final interview at the end of 3 months (Figure 20).

FIGURE 20.

Longitudinal, qualitative evaluation of primary care consultations between health-care practitioners and patients with a LTC.

Stimulated recall involved the researchers reviewing consultations prior to interviews, marking points in the consultations to play back to practitioners or patients. These points were identified using a topic guide developed by, and regularly reviewed and discussed within, the research team. These extracts were then played back during interviews and participants were asked to reflect on what was occurring at that point in the consultation, what they were thinking at that moment, their reflections on the topic being discussed or raised, and what had happened following on from that event. The focus was on gathering reflections on topics pertinent to health-care use (retrospective or prospective) and management of LTCs (e.g. medication advice, behavioural change issues, self-management).

During the final interviews with patients, at the end of 3 months, researchers used any patient-completed health-care use diaries and notes from telephone contacts to frame the interviews and focus in on specific decision-making activities around health-care use and management of health concerns during the follow-up period.

Analysis

All audio-recordings (consultations and interviews) were transcribed verbatim and anonymised. Data were analysed using an integrative framework approach, adapting the Ritchie and Spencer’s approach276 for the incorporation of multiple perspectives on the same event (consultation). Specifically, the research team analysed transcripts inductively, comparing data within, and between, the case. 291 For each case (identified as a single patient), short narrative summaries were constructed from the analysis, tracking themes for that patient from consultation to final interview, including discussion of that patient in the practitioner interview. Owing to the complexity of the analysis, with up to five different sources of data per patient (consultation, first patient interview, follow-up contacts, final patient interview and practitioner interview), a combination of Microsoft Word and Excel (Microsoft Corporation, Redmond, WA, USA) was used to handle the data. The research team was multidisciplinary, incorporating expertise in primary care, psychology and social anthropology. Analysis took place concurrently with data collection, with developing themes discussed within the team and incorporated into further recruitment and topic guides. Data collection ended when the team was satisfied that analytical saturation was reached, when the aim of characterising and understanding the experience of primary care consultations from multiple perspectives was achieved.

Results

Participant characteristics

Thirty-nine primary care practices were approached and six practices agreed to take part. Using publicly available QOF data, practices with high prevalence rates in at least one of the four LTCs compared with the median for England were identified and approached, with the aim of achieving a sample with a diverse socioeconomic and geographic spread within the region of recruitment. Practice recruitment ceased when a sufficiently large and diverse sample of practices had agreed to participate. From these six practices, five GPs (three senior partners and two GP registrars) and five PNs took part in audio-recorded consultations and interviews about these consultations (Table 31).

TABLE 31 - Characteristics of HCPs who participated in the qualitative study

Practice	Health-care practitioner
	ID	Role
A	HCP1	PN
B	HCP2	PN/advanced nurse practitioner
C	HCP3	GP (senior partner)
C	HCP4	GP registrar
D	HCP5	PN/specialist practitioner
D	HCP6	GP (senior partner)
E	HCP7	PN
E	HCP8	GP (senior partner)
E	HCP9	GP registrar
F	HCP10	PN

Twenty-nine consultations were audio-recorded by practitioners, with the number recorded per practitioner varying between 1 and 8 (mean 3). On five occasions, consultations were not recorded because of technical errors; in these instances, patients were still invited to take part in an interview about the consultation.

Sixty-five patients were approached within primary care practice waiting rooms, and 34 patients agreed to audio-recording of their consultations (Table 32). Patients were mainly white British (82.4%), male (65%) and aged from 34 to 87 years. Most patients had more than one LTC (73.5%; 29.4% had at least two of COPD, CHD, asthma and diabetes).

TABLE 32 - Characteristics of patients who participated in the longitudinal, qualitative study

The table is adapted from Hunter et al. 289 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Practice	Patient ID	Age (years)	Sex	Condition(s)
A	1	70	Male	COPD and cancer
A	2	62	Male	COPD and depression
A	3	51	Female	Asthma
B	4	46	Male	COPD
B	5	Not known	Female	Diabetes and COPD
B	6	85	Male	COPD, atrial fibrillation, dropped foot and balance problems
C	7	51	Male	Hypertension
C	8	Not known	Male	Diabetes
C	9	82	Male	Diabetes and asthma
C	10	54	Male	Diabetes and bowel problems
C	11	47	Female	Diabetes and cancer
C	12	87	Male	CHD, diabetes and depression
D	13	65	Female	Diabetes, COPD and angina
D	14	60	Male	Diabetes
D	15	Not known	Female	CHD and asthma
D	16	50	Female	Diabetes and nerve spasms
D	17	76	Female	CHD, cancer and high cholesterol
D	18	69	Female	COPD and arthritis
D	19	74	Male	CHD, asthma, COPD and meningioma
D	20	50	Male	CHD, depression and blindness
D	21	43	Male	Diabetes
D	22	62	Male	CHD, diabetes, hypertension and chronic kidney disease
D	23	58	Male	CHD, diabetes, cancer and piles
D	24	57	Female	Asthma, sarcoidosis and bronchiectasis
D	25	Not known	Female	Diabetes
E	26	41	Female	Asthma
E	27	51	Male	Asthma and hypertension
E	28	73	Male	CHD, diabetes and gout
E	29	30s	Female	Asthma, depression and irritable bowel syndrome
E	30	76	Male	CHD, hypertension and arthritis, asbestosis
E	31	76	Female	Diabetes and arthritis
E	32	67	Male	Diabetes, hypertension and glaucoma
F	33	Not known	Male	Asthma and CHD
F	34	67	Male	CHD and hypertension

There were 18 patients from whom consultation data, patient first interview, patient second interview and practitioner interview were all available – these constituted full cases (Figure 21). Of the remaining 16 patients, seven declined interviews but gave consent to retain their consultation; four had a recorded consultation and first interview, but declined participation in the second follow-up interview; and five took part in either one or two interviews, but lacked a recorded consultation. Retention during follow-up was high (81.5%), with 22 out of 27 patients who completed a first interview going on to complete the follow-up interview as well. Figure 21 details patient recruitment and retention throughout the study.

FIGURE 21.

Patient recruitment and retention. The figure is adapted from Hunter et al. 289 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

All of the participants who took part in the initial interview also took part in regular telephone calls, in which the researcher asked about their health status, health-care use and decision-making. Researchers kept a written record of each telephone call. If a participant described a problem with their health during these telephone calls, they were asked by the researcher to fill in a health-care log of the incident. These health-care logs asked participants to record their symptoms, when the incident occurred, where they were at the time and who they were with, and what they decided to do at the time. All but four of these participants then took part in a follow-up interview. Health-care logs and telephone calls were then analysed for content relating to health-care use and health-care decisions, and were used to inform the discussion in the follow-up interviews and the analysis of health-care use (Table 33).

TABLE 33 - Sources of data for the longitudinal qualitative study

Source of data	Number of data
Number of patients	34
Consultations recorded	29
Patient baseline interviews	27
Health-care logs	14 completed logs, from four patients
Telephone calls	Between 6 and 15 calls, with 22 patients
Patient follow-up interviews	22
Number of health-care practitioners	10
Health-care practitioner interviews	10

Discussion

In this study, primary care consultations for LTCs were characterised by a retrospective focus on meeting clinical targets, to the point at which issues such as prospective health-care use, behaviour change and talk of unmet patient needs was neglected. Except in two cases, psychological and social issues were sidelined in favour of completing the technical and biomedical aspects of review work. The patients were socialised into an expectation that primary care was for ‘checking’ their health and fulfilling the practitioner’s agenda, and that they themselves were responsible for instigating any lifestyle or dietary changes indicated by their LTCs. Patients were not clear about what practitioners could do for them, beyond provide medication, and, although patients appreciated receiving regular reviews, they tended not to see them as the place to bring up further needs.

Perhaps as a result of the focus on the QOF, practitioners often missed or did not follow-up cues around unmet needs, including informational, emotional and biomedical needs unrelated to the LTC under review. Having experienced consultations wherein their concerns were not picked up, or where the focus was primarily on checking biomedical test results and medications, patients tended to adapt their expectations accordingly, and became socialised into a particular way of participating in reviews. This socialisation is potentially damaging, as patients may delay bringing up serious issues, and it could also impact on how patients use primary care and unscheduled care, as patients may choose to use other services because of a mistaken belief that primary care cannot help them.

With regard to health-care use, it was unclear whether or not primary care reviews influenced patients’ decisions over time. This was in part due to the infrequency of unscheduled care use within the sample, but also in part due to the infrequency of health-care use or crisis management as a topic in the consultations. Practitioners rarely addressed prospective health-care use or crisis management with patients in review consultations, and when they did address it, it was usually in a prescriptive manner, giving advice on what to do but not discussing options with the patient. Despite this, there was evidence of patients using unscheduled care as a result of past practitioner advice, and also evidence of patients using self-care tactics that aligned with those which practitioners had recommended in consultations. The key features of unscheduled care decision-making, as seen in this study, mapped on to those seen in the previous qualitative study described in Chapter 5, with both prior experience of health care and the person’s perception of their own legitimacy to use unscheduled care in that instance being essential. From this perspective, this chapter offers further support for the candidacy and recursivity theories of the preceding chapter, and reinforces the notion of primary care practitioners as potential influences on this decision-making process by virtue of how they have responded to previous incidents of need experienced by the patients.

Behaviour change is a current UK priority for LTC management, and in discussing behaviour change in interviews, practitioners emphasised their belief in its importance and its centrality in LTC management. They described a role for themselves in providing information to patients about the importance of behaviour change. However, they also emphasised constraints on their ability (and, therefore, their responsibility) to effect behaviour change. Behaviour change work was constrained by a perceived lack of effective techniques to influence patient motivation, by the need to ensure a continued patient–practitioner relationship and, ultimately, by the patients themselves. 275,292–295 This included attempts to give directive information about when to seek help. The relegation of behaviour change work to the responsibility of the patients or other practitioners offered a way for practitioners to disengage with this work in their consultations.

The key finding in this study was that reviews were often focused on offering one form of care (evidence-based care as represented by the QOF) to the detriment of other, equally valuable aspects of care (a patient-centred approach which addressed patient-defined needs; behaviour change work essential to successful management of LTCs). The unintended, and concerning, consequence of this focus is that patients adapt their behaviour and their expectations accordingly, becoming more passive in reviews and less likely to raise problematic issues or engage in proactive work with the practitioner.

It is acknowledged that the practitioners in this study and in primary care in general are embedded in an increasingly busy and scrutinised context in which certain tasks are valued and counted (such as meeting biomedical targets),296 and others (such as less measurable behaviour change work) are not. 288 It may be that practitioners feel both ill-equipped to achieve behaviour change in patients and pressurised to achieve other targets, and the rhetoric of patient control and responsibility enables them to legitimately distance themselves from this work. 297 This may be why there was little discussion of health-care use and why patients felt disengaged from their GPs, as they offered only medication. Likewise, practitioners may be missing or not responding to patient cues about physical or emotional needs that fall outside the review consultation’s standard focus because they are under pressure to keep to time and achieve their targets. Depression and stressful life experiences are independent drivers of unscheduled care use, yet such factors are unlikely to be brought up by health-care staff, and patients are also reluctant to disclose or talk about such problems. Practitioners struggle to bring up or discuss depression within consultations where the agenda is driven by QOF and medication review, concerned that such discussions might derail this agenda or identify needs they do not have the skills or resources to manage. 298,299 Practitioners are likely to engage with psychosocial factors and behavioural change work only if the structures in primary care promote and value this work. 300,301

Addressing the tensions between patient-centred care and delivering evidence-based care might be achieved by longer consultation times,302 and greater continuity of care. 302 Careful consideration is also required of how practitioners and patients communicate, and how this influences the work that can be achieved by both parties around health outcomes; too much of a focus on an individual patient’s responsibility when it comes to behaviour change and managing LTCs can harm the patient–practitioner relationship and impede open discussion. 303 Health-care practitioners need to be aware that patients have to do more than biomedical work to manage their LTC, and require social resources in order to achieve LTC work, and consideration is needed of how practitioners can work with patients in these broader aspects of their lives, cognisant of multiple problems,37 and can support and encourage patients to bring their concerns to the consultation.

There are several strengths of this study. Its innovative design, involving repeated patient interviews over time, generated a rich understanding of patients’ attitudes, needs and experiences. The combination of multiple perspectives allowed for a richer and deeper analysis of the data than is usually possible in qualitative work. Stimulated recall helped situate discussions in specific consultations, thereby reducing recall bias, and the regular patient contact over time enabled participant–researcher relationships to form that encouraged greater openness about sensitive topics.

Conclusions

Routine reviews of patients with LTCs have become well established in primary care and represent a golden opportunity to provide high-quality, patient-centred care. Our findings from the current study suggest, however, that reviews have a narrow, ‘tick-box’ focus and fail to address people’s needs in a holistic fashion. Use of unscheduled care or crisis management is rarely discussed, so there is little opportunity to implement strategies that make overt use of unscheduled care. There is little space or time in consultations for behavioural change, which may improve the management of patients’ LTCs, and this kind of work is currently not valued as highly as ‘physical health’ procedures. Finally, emotional and social problems are regarded as peripheral to people’s physical health problems and, as such, are sidelined during routine reviews.

We also found examples of good practice, involving practitioners who engaged with their patients in a collaborative style, and were comfortable moving beyond the confines of ‘QOF’ to provide sensitive and personalised health care.

Limitations

There were several limitations regarding the study. The time frame for follow-up was kept short in order to reduce impact on patients; however, it may not have been long enough to fully explore episodes of unscheduled care use as these episodes were rare in the patients interviewed. Recordings of single consultations can only offer a snapshot of the care being given, and patients with multiple LTCs are likely to have input from multiple practitioners, which was not captured in this approach. As a result, it is possible that some of the cues or needs missed within the study were being addressed elsewhere. Only a small number of practitioners and practices were included in the study and, as the study was located in north-west England, the results may not be generalisable to other parts of the country.

Abstract

Introduction

In this chapter, we describe a replication study which we carried out to test the validity of our findings from the longitudinal cohort study conducted in phase 2 of the CHOICE programme. The findings from our longitudinal cohort study (see Chapter 3) showed that several potential red flag variables were associated with an increased risk of use of unscheduled care in primary care patients, with at least one of the four following LTCs: asthma, CHD, COPD or diabetes. These variables were prior use of unscheduled care; a diagnosis of any of the four LTCs; depression; lack of a partner; and threatening life experiences. The aim of this replication study was to determine whether or not our findings from the longitudinal study could be replicated using routinely collected NHS data.

The main objective, as described in the original grant application, was to ‘evaluate the feasibility and validity of establishing a red flag system to identify patients with LTCs who are at risk of becoming frequent users of unscheduled care’. As we had already established that ‘frequent use’ of unscheduled care was relatively rare in our primary care population of people with LTCs, we focused instead on prospective use of unscheduled care.

The collection and analysis of data were carried out by Professor Simon de Lusignan and his team at the University of Surrey. In the original grant application, we had specified that we would use a large data set in Salford to carry out the replication study. However, there were logistical problems in being able to access the data set during the time frame that was required for the CHOICE programme.

We looked therefore for viable alternatives and we chose to collaborate with the Surrey team for three reasons. First, they were able to access data in a different part of the country to Manchester, so if the findings of the cohort study were replicated, this would strengthen the generalisability of the results. Second, they had considerable expertise in the analysis of large primary care data sets. 304 Third, they had prior experience of extracting mental health variables from routine primary care data. 305

The main plan was to use routine primary care data of patients with at least one of our four exemplar conditions and collect data over a 3-year time period. This would then be divided into a ‘baseline’ period of 2 years and a ‘prospective’ period of 1 further year. Data collected over the baseline period would be used to predict use of unscheduled care over the prospective 12-month period. It was decided to increase the duration of the baseline period to better capture health-care activity and morbidity, but the follow-up period remained 12 months, as there is evidence that the predictive power of baseline factors appears to wane over a longer period of time. 246

Patients with asthma, CHD, COPD or diabetes were identified using the patient registers used for the QOF306 in GP practice information technology (IT) systems.

We worked with the Surrey team to identify variables that they would be able to extract from routine electronic GP data, which would map on to the predictor and outcome variables from the cohort study that we had undertaken in phase 2. Two variables, attendances at EDs and EHAs, were chosen for the main outcome (i.e. use of unscheduled care). Both these variables were also used to represent measures of previous use of unscheduled care during the baseline period. We chose these two variables because they were easily identifiable from NHS-linked data sets and were both a priority area for government.

Unfortunately, it was not possible from the data sets to identify people who did not have a partner. Although the Surrey team were able to identify some routine demographic data, such as age and sex, marital status was not available and there was no way to identify a proxy measure for either ‘living alone’ or ‘lack of partner’ from the electronic data. There was also no equivalent measure in the routine data sets for threatening life experiences, which we had used in the prospective longitudinal study.

The Surrey team, however, had previously developed a composite mental health variable which they termed CMHP,305 which, in addition to depression and anxiety codes, also included entries relating to stress in the GP notes (e.g. feeling stressed, stress-related problem, problems at work, home problems). Although this was not a direct indicator of ‘threatening experiences’ and CMHP also included all entries relating to depression and anxiety, we decided to carry out analyses using CMHP interchangeably with the depression variable to see if it performed better than depression, because of its inclusion of depression plus several indicators of life stress.

We also used the IMD,235 which could be mapped from patient postcodes, as another potential proxy measure for threatening experiences. The IMD measure is a composite of several factors, which overlap with some items on the threatening experiences scale. The IMD includes the following: income, employment, health deprivation and disability, education skills and training, barriers to housing and services, crime and living environment.

Read codes for depression in GP IT systems were used to construct a depression variable. The details are provided in Methods.

Methods

The data were collected from a large geographical area in north London. We conducted a medical records review study of adult patients with LTCs in primary care; the sources of information consisted of patient records from GP IT systems in primary care practices in this geographical area.

Our sample involved the total GP-registered population available at the time of this study (150,000+) from which we identified those records corresponding to people suffering from one or more of four specified QOF-defined LTCs, namely asthma, CHD, COPD and diabetes. There were 23,422 patients. Owing to regional differences in the way data are collected, we further subdivided the population into the geographical regions north-east and north-west London – there were 8722 patients in the north-east and 14,700 patients in the north-west data sets.

Index dates

Each anonymised patient record consists of an index date, prior to which there are two full years of GP records, as well as a full year of follow-up records (post-index date). For the north-west London data from the ICP, the index date varies per practice according to the practice’s last successful data extract. The date is calculated from the last successful extract date to allow 2 years of observations before and 1 year of follow-up afterwards. For the north-east London data, the index date used was 28 January 2013, as the Hospital Episode Statistics data were complete up to 28 January 2014. The GP records provided information on emergency admissions to hospital and ED attendances.

We constructed indicator variables for the presence of each of the LTCs at index date, as well as emergency admissions and ED attendances in the pre-index and follow-up periods. We also constructed binary indicator variables for events in the pre-index date period: depression incident (none vs. at least one) and CMHP incidents (none vs. at least one). The codes that were used to construct the depression variable are shown in Table 34.

The variable common mental health problem was defined as having either depression or anxiety, stress or panic. The codes used to construct this variable included all the depression codes plus all the codes in Table 35.

We banded age into the groups: 18–65, 66–80 and > 80 years.

The IMD variables were mapped from full postcodes included in patient GP records using geographical information system methods. The IMD is a composite score based on the following deprivation measures (communities and local government 2010):235

• income

• employment

• health deprivation and disability

• education skills and training

• barriers to housing and services

• crime

• living environment.

We treated IMD as a continuous variable, and in multivariate analyses, we analysed these data with and without IMD included as an independent variable.

Results

Data were obtained for 23,422 patients: 8722 patients in the north-east and 14,700 patients in the north-west.