Article Text

PDF

A systematic review of the predictors of health service utilisation by adults with mental disorders in the UK
  1. Conal D Twomey1,
  2. David S Baldwin2,
  3. Maren Hopfe3,4,
  4. Alarcos Cieza5,6,7
  1. 1Faculty of Social and Human Sciences, University of Southampton, Southampton, UK
  2. 2Faculty of Medicine, University of Southampton, Southampton, UK
  3. 3Swiss Paraplegic Research, Nottwil, Switzerland
  4. 4Department of Health Sciences & Health Policy, University of Lucerne, Lucerne, Switzerland
  5. 5Faculty of Social and Human Sciences, School of Psychology, University of Southampton, UK
  6. 6Department of Medical Informatics, Biometry and Epidemiology—IBE, Research Unit for Biopsychosocial Health, Ludwig-Maximilians-University (LMU), Munich, Germany
  7. 7Swiss Paraplegic Research, Nottwil, Switzerland
  1. Correspondence to Conal D Twomey; c.twomey{at}soton.ac.uk

Abstract

Objectives To identify variables that predict health service utilisation (HSU) by adults with mental disorders in the UK, and to determine the evidence level for these predictors.

Design A narrative synthesis of peer-reviewed studies published after the year 2000. The search was conducted using four databases (ie, PsycINFO, CINAHL Plus with full text, MEDLINE and EMBASE) and completed on 25 March 2014.

Setting The majority of included studies were set in health services across primary, secondary, specialist and inpatient care. Some studies used data from household and postal surveys.

Participants Included were UK-based studies that predicted HSU by adults with mental disorders. Participants had a range of mental disorders including psychotic disorders, personality disorders, depression, anxiety disorders, eating disorders and dementia.

Primary outcome A wide range of HSU outcomes were examined, including general practitioner (GP) contacts, medication usage, psychiatrist contacts, psychotherapy attendances, inpatient days, accident and emergency admissions and ‘total HSU’.

Results Taking into account study quality, 28 studies identified a range of variables with good preliminary evidence supporting their ability to predict HSU. Of these variables, comorbidity, personality disorder, age (heterogeneous age ranges), neurotic symptoms, female gender, a marital status of divorced, separated or widowed, non-white ethnicity, high previous HSU and activities of daily living, were associated with increased HSU. Moreover, good preliminary evidence was found for associations of accessing a primary care psychological treatment service and medication use with decreased HSU.

Conclusions The findings can inform decisions about which variables might be used to derive mental health clusters in ‘payment by results’ systems in the UK. The findings also support the need to investigate whether combining broad diagnoses with care pathways is an effective method for mental health clustering, and the need for research to further examine the association between mental health clusters and HSU.

Statistics from Altmetric.com

Strengths and limitations of this study

  • The review was limited to UK studies, meaning the list of identified variables is not exhaustive and the findings may not be applicable to services in other countries.

  • There was wide heterogeneity in the operationalisation of health service utilisation (HSU) by included studies, which limits the validity of comparisons across studies. Addressing this issue, the operationalisation of HSU in included studies was documented in considerable detail (table 1).

  • The study benefits from its use of structured checklists for assessments of study quality.

  • The majority of literature searching was undertaken by one study author. However, in order to minimise bias and error, 20% of abstracts were independently screened by another author.

Introduction

Many stakeholders with differing needs are involved in the delivery of public health services. Patients seek the best obtainable care, providers aim to deliver optimal care but also strive for self-regulation and autonomy, and policymakers need to balance meeting high public demand with controlling health service expenditure.1 To meet these differing needs, fair and efficient health service payment systems are required. Contemporary, ‘activity-based’ payment systems aim to achieve this fairness and efficiency by financially incentivising competing providers to treat more patients, cut costs and reduce waiting list times.2

Table 1

Observational studies of HSU (n=17)

In typical activity-based payment systems, resources are allocated towards distinct patient ‘clusters’ (or groups). These clusters are comprised of patients with similar clinical characteristics and expected health service utilisation (HSU) patterns.3 Each patient treated by a health service provider is assigned to a specific cluster based on collected information about a range of ‘case-mix’ variables (eg, diagnosis, comorbidity, age) which are associated with HSU.4 Thereafter, health service providers receive a fixed payment based on the cluster each patient is allocated to, with clusters with higher expected HSU generating higher payments than those with lower expected HSU.5 Paying providers fixed payments based on ‘clustering’ of treated patients allows policymakers to distribute resources in a systematic and equitable manner.3

In recent years, there have been ongoing efforts by the National Health Service (NHS) in England to develop (a potential UK-wide) activity-based payment system for its mental health services, in what is referred to as Mental Health Payment by Results. Initially, the system will cover secondary care services with various service types excluded (eg, those relating to primary care psychotherapy, acquired brain injury and autism).6 A subject of much debate in this development surrounds how to define ‘mental health clusters’ for use in this system. In contrast to typical activity-based payment systems, diagnostic information has so far not been used to define these clusters. Instead, clusters have been defined using the newly-developed Mental Health Clustering Tool (MHCT). The MHCT assesses the domains of behaviour, symptoms, impairment, social functioning and risk factors, and is used to assign patients to one of 21 clusters, falling under one of three broad ‘super-classes’ (non-psychotic, psychotic and organic).7

One of the main reasons for not using diagnostic information for clustering in Mental Health Payment by Results was that mental disorder diagnosis was shown to be a poor predictor of HSU in studies involving national and multisite trial data sets.8–11 On the other hand, it has been argued that although mental disorder diagnosis alone is not sufficient for clustering purposes, information about broad diagnoses and care pathways can be combined, in a simple and practical manner, to form reliable clusters with homogenous resource patterns.12 Moreover, the MHCT has also been criticised because its development did not take HSU and costs into account,13 and, currently, very little evidence exists for the ability of the MHCT to predict HSU in patient populations.

In the context of the ongoing development of Mental Health Payment by Results, and the debate surrounding the use of diagnostic information and the MHCT, it is important to provide evidence that can inform decisions about which variables might be used to derive mental health clusters. To date, no UK-based systematic reviews informing this process have been undertaken. A review of relevant studies set in the UK would address UK-specific HSU patterns, increasing the applicability of findings to the Mental Health Payment by Results system. Therefore, the general objective of this systematic review is to identify variables with sufficient evidence supporting their ability to predict HSU. The review has two specific aims. First, to identify the variables examined in relation to the prediction of HSU by adults with mental disorders in the UK. Second, to determine the level of evidence that exists for identified predictors of this HSU.

Method

Inclusion and exclusion criteria

Only the following types of studies were included in the review: (1) observational and intervention studies that predicted HSU by adults with mental disorders. (For the purposes of this review, mental disorders included adults experiencing elevated symptoms of mental disorders or adults formally diagnosed with a mental disorder. Studies with participants with intellectual disability were excluded due to the specific additional needs of this population that have to be met beyond the healthcare system, eg, in the education or labour systems); (2) studies based in the UK, with UK participants; (3) peer-reviewed studies published in scientific journals, in the year 2000 or after. (This cut-off point was chosen so that included studies were approximately in line with the overall Payment by Results scheme introduced in 2003. Intervention costing studies that did not predict HSU were excluded).

Literature search

Based on these criteria, the first author searched four databases: PsycINFO, CINAHL Plus with full text, MEDLINE and EMBASE. The final search was conducted on 25 March 2014. Additional records were identified from manually searching reference lists of included studies. Search terms and database subject headings related to HSU (ie, health care utilisation [subject heading] OR health care utili* OR health service utili* OR health care use OR health service use) were combined with those terms for mental disorders (ie, Mental disorders [subject heading] OR psychiatric) and the UK location (ie, UK [subject heading) OR NHS). Owing to the differing search procedures deployed by the four databases, slightly altered versions of this search strategy were used in each database. Independent screening of 20% of abstracts was undertaken by the third author. When the first author and third author disagreed regarding the screening outcome of an abstract, the abstract was included in screening at ‘full-text’ level (by the first author).

Data extraction

Data from included studies were extracted using an Excel spreadsheet. Extracted data pertained to basic study description, study design, records source, data collection times, participants, mental disorder investigated, operationalisation of HSU outcomes, the prediction of HSU and statistics. In addition, each study was assessed for quality using the STROBE statement14 (for observational studies) and the National Institute for Health and Clinical Excellence (NICE) checklist for Randomised Controlled Trials (RCTs).15 The former is a checklist of 22 items related to the reporting of title (one item), introduction (two items), methods (nine items), results (five items), discussion (four items) and funding information (one item).14 The latter assesses bias in RCTs in four sections: selection bias, performance bias, attrition bias and detection bias.15

Data analysis

Owing to the heterogeneity in study designs, samples and mental disorders investigated, a meta-analysis was not possible. Narrative synthesis was deemed the most appropriate method of data analysis.

Results

Literature search flow

The literature search flow is displayed in figure 1. In total, 1364 records were identified. Database-searching yielded 1347 records and manually searching yielded 17 additional records. After duplicates were removed, 928 studies were screened at ‘abstract’ level. For screening of abstracts, there was a 94.1% agreement rate between the first author and the third author. After abstract screening, 133 studies were assessed for eligibility at ‘full-text’ level. Twenty-eight studies were included in the final review.

Figure 1

Literature search flow.

Overview of included studies

To provide an overview of included studies, extracted data were summarised in two tables (tables 2 and 3). Table 2 summarises observational studies of HSU, and table 3 summarises studies of interventions (of both observational and experimental design) aiming to reduce HSU. As can be seen in both tables, the data source of included studies varied. Most frequently, it included routine NHS service data or databases (n=14), different versions of the Adult National Psychiatric Morbidity Survey (n=6) and other household and postal surveys (n=3). The sample composition also varied and included adults with a psychotic disorder (n=7), personality disorder (n=5), depression (n=3), an anxiety disorder (n=2), an eating disorder (n=1), ‘common mental health problems’ (n=2) and dementia (n=1). It also included health service users (n=6) and former adolescent psychiatric patients (n=1). The quality of included studies was mixed. STROBE statement14 scores for observational studies (n=25) ranged from 9 to 20 (mean (M)=15.5; SD=3.05), out of a possible maximum score of 22. Of the three RCTs assessed using the NICE checklist,15 two indicated the absence of bias, and one indicated the possible presence of bias. As can be seen in tables 2 and 3, both the operationalisation of HSU outcomes and the identified predictors of HSU in individual studies varied widely.

Table 2

Intervention studies of HSU (n=11)

Table 3

Frequency of HSU outcomes used across included studies (n=28)

Operationalisation of HSU outcomes

To determine the level of evidence for identified predictors of HSU, it was beneficial to first summarise the operationalisation of HSU outcomes across included studies. This summary is provided in table 1. Across the 28 studies, 60 different HSU outcome variables were assessed 155 times in total: 24 of these related to primary care HSU, 79 to specialist HSU, 40 to inpatient HSU and 12 to ‘total and other’ HSU. Across all categories apart from the ‘total and other’ HSU category, 65 outcomes related to mental health HSU and 78 related to general health HSU.

HSU outcomes used in three or more studies were: medication usage (n=12); inpatient days (n=9); accident and emergency (A & E) admissions (n=8); inpatient admissions (n=8); total HSU (n=8); GP contacts (n=7); GP contacts for psychological problems (n=6); psychotherapy attendances (n=6); community psychiatric nurse contacts (n=5); psychiatrist contacts (n=5); psychiatric inpatient admissions (n=5); psychologist contacts (n=5); nurse contacts (n=4); outpatient attendances (n=4); counsellor contacts (n=3); and home carer visits (n=3). Remaining HSU outcomes are shown in table 1.

Summary of evidence for identified predictors of HSU

Table 4 provides a summary of the evidence for identified predictors of HSU. The table is structured as follows. First, identified predictors are categorised by ‘demographics’, ‘diagnosis’, ‘interventions’, ‘symptoms’, ‘functioning’ and ‘behaviour’. Second, the table displays the number of times each identified predictor variable was assessed in relation to HSU, and the number of times each identified variable significantly predicted HSU (and vice versa). Third, using the broad categories of ‘primary care HSU’, ‘specialist HSU’, ‘inpatient HSU’ and ‘total HSU’, the table documents the operationalisation of HSU outcomes in relation to the prediction of HSU. Fourth, study quality information is provided to aid evaluation of the evidence. For simplicity, a study was arbitrarily deemed to be of ‘satisfactory’ quality if it scored >16 on the STROBE statement,14 or if bias was not present on three out of four domains on the NICE checklist for RCTs.15

Table 4

Summary of the evidence for examined predictors of HSU

As an overview, the review identified 31 predictor variables that were examined in relation to the prediction of HSU. By category, these were: 12 demographic variables, six intervention variables, five diagnostic variables, four symptom variables, three functioning variables and one behavioural variable.

The 12 demographic variables significantly predicted increased HSU 41 of 65 times assessed (63.1%). Six demographic variables predicted increased HSU in two or more assessments and in over 50% of assessments made. These variables, in order of frequency of increased HSU prediction, were: comorbidity (both mental and physical), age (heterogeneous age ranges), female gender, a marital status of divorced, separated or widowed, non-white ethnicity and high previous HSU. Regarding the age variable, several heterogeneous age ranges (eg, 35–54, 31–49, 35+, 50–64) were associated with increased HSU, thus it was not possible to draw conclusions relating to specific age ranges. Specific age ranges associated with increased HSU in individual studies are viewable in table 2. As study quality was satisfactory in the vast majority of these assessments, it can be concluded that good preliminary evidence exists for these six demographic variables in relation to the prediction of increased HSU.

The six intervention variables significantly predicted decreased HSU 10 of 17 times assessed (58.8%). Two intervention variables predicted decreased HSU in two or more assessments and in over 50% of assessments made. These variables, in order of frequency of decreased HSU prediction, were: accessing an Improving Access to Psychological Therapies (IAPT) service and medication. As study quality was satisfactory in all but one of these assessments (an assessment of IAPT), it can be concluded that good preliminary evidence exists for both IAPT and medication in relation to the prediction of decreased HSU.

The five diagnostic variables significantly predicted increased HSU 13 of 15 times assessed (86.6%). Two diagnostic variables predicted increased HSU in two or more assessments and in over 50% of assessments made. These variables, in order of frequency of increased HSU prediction, were: personality disorder and obsessive compulsive disorder. Whereas all (eight) assessments of personality disorder came from studies of satisfactory quality, none of the (four) assessments of obsessive compulsive disorder came from studies of satisfactory quality. Therefore, it can only be concluded that good preliminary evidence exists for personality disorder in relation to the prediction of increased HSU.

The four symptom variables significantly predicted increased HSU 7 of 15 times assessed (46.6%). One symptom variable—neurotic symptoms—predicted increased HSU in six of six assessments made. Although two assessments came from studies of unsatisfactory quality, it can be concluded that good preliminary evidence exists for neurotic symptoms in relation to the prediction of increased HSU.

The three functioning variables significantly predicted increased HSU five of nine times assessed (55.6%). Two functioning variables predicted increased HSU in two or more assessments and in over 50% of assessments made. These variables, in order of frequency of increased HSU prediction, are: cognitive deficits and activities of daily living (ADLs). Whereas all (two) assessments of ADLs came from studies of satisfactory quality, none of the (three) assessments of cognitive deficits came from studies of satisfactory quality. Therefore, it can only be concluded that good preliminary evidence exists for ADLs in relation to the prediction of increased HSU.

In the final variable category, a behavioural variable—self-harm—significantly predicted increased HSU once, of the one time assessed. This assessment came from a study of satisfactory quality. However, as just one assessment was undertaken, it cannot be concluded that good preliminary evidence exists for self-harm in relation to the prediction of increased HSU.

In summary, taking into account frequency of prediction and study quality, several predictor variables have good preliminary evidence supporting their ability to predict HSU by adults with mental disorders in the UK. Of these variables (in order of frequency of prediction), comorbidity, personality disorder, age (heterogeneous age ranges), neurotic symptoms, female gender, a marital status of divorced, separated or widowed, non-white ethnicity, high previous HSU and activities of daily living, were associated with increased HSU. Moreover, good preliminary evidence was found for associations of accessing a primary care psychological treatment service and medication use with decreased HSU. Figure 2 illustrates the relative frequencies of predictors of HSU, by category.

Figure 2

Frequency of HSU prediction by variable category. HSU, health service utilisation; frequencies were obtained by counting some studies various times for one variable category; for interventions, the count concerned the prediction of decreased HSU.

Discussion

Summary of main findings

Taking into account study quality, 28 studies identified a range of variables with good preliminary evidence supporting their ability to predict HSU. Of these variables, comorbidity, personality disorder, age (heterogeneous age ranges), neurotic symptoms, female gender, a marital status of divorced, separated or widowed, non-white ethnicity, high previous HSU and activities of daily living, were associated with increased HSU. Moreover, good preliminary evidence was found for associations of accessing a primary care psychological treatment service and medication use with decreased HSU.

Comparison of main findings with other reviews

Few existing reviews of the predictors of HSU in mental health populations were available for comparison of results. Nevertheless, comorbidity—the most evidenced predictor of increased HSU in the present review—was also shown in a review of 72 studies to predict increased psychiatric service utilisation by ‘heavy users’ of psychiatric services.16 This previous review found that several variables not examined by studies in our review (ie, substance abuse, psychotic illness, isolation, homelessness and social support) were predictive of increased psychiatric service utilisation. In line with the present review, another review of eight studies found that high previous utilisation predicted increased psychiatric service utilisation.17 On the other hand, this review found that the variables of living alone and psychosis diagnosis—not examined by studies in the present review—were predictive of increased psychiatric service utilisation.

Overall, the findings from previous reviews add robustness to our finding of good preliminary evidence for the variables of comorbidity and high previous HSU in relation to the prediction of increased HSU by adults with mental disorders in the UK. In addition, despite the sole focus of the previous reviews on psychiatric services, which limits their comparability, it is possible that several additional variables—in particular, a psychosis diagnosis—may also predict increased HSU by adults with mental disorders in the UK.

Comparison of main findings with international studies of HSU

As the review was limited to UK studies, it is informative to compare the findings with those from international studies of HSU by adults with mental disorders. Three recent international studies were chosen for comparative purposes because of their large samples comprising adults with a range of mental health problems.18–20

The first was set in Canada, and had a sample of 243 adults diagnosed with various mental disorders.18 In line with our review, it found that increased social withdrawal, female gender and (mental disorder) comorbidity were associated with increased HSU. Additional predictors of increased HSU not identified by studies in our review were emotional problems, income, major depression diagnosis and alcohol dependence.

The second study was set in Australia and had a sample of 822 adults who had previously participated in a school-based epidemiological study in their youth.19 In line with our review, it found that age (treated as continuous variable), comorbidity and a marital status of divorced, were associated with increased HSU. Additional predictors of increased HSU not identified by studies in our review were psychological distress, affective disorder diagnosis and exposure to childhood trauma, while rural living predicted reduced HSU.

The third study20 used data from a cross-national health survey and involved 8688 adults from the USA and Canada. It found that comorbidity (various health comorbidities), female gender and non-white ethnicity, were associated with increased HSU. Additional predictors of increased HSU not identified by studies in our review were emotional problems, income, having a regular doctor and having insurance.

The findings from these international studies add robustness to our finding of good preliminary evidence for the variables of comorbidity, female gender and a marital status of divorced, in relation to the prediction of HSU by adults with mental disorders in the UK. In addition, it is possible that several additional variables identified in international studies—in particular, emotional problems—may also predict HSU by adults with mental disorders in the UK.

Implications of findings for mental health payment by results

Our findings can inform the debate surrounding the use of diagnostic information and the MHCT for clustering purposes. The findings also highlight several additional variables that are worthy of consideration in the clustering process.

Regarding the use of diagnostic information, in contrast to previous large-scale studies, which showed mental disorder diagnosis to be a poor predictor of increased HSU,9–11 the review yielded good preliminary evidence for personality disorder diagnosis in relation to the prediction of increased HSU. In addition, it is noted that diagnoses of psychosis, major depression and affective disorder, were identified as predictors in previous reviews and international studies.16–19 Although methodological differences (eg, in the operationalisation of HSU) in these reviews and studies mean that firm conclusions cannot be drawn, a possible explanation for the discrepancy in findings is that some but not other mental disorder diagnoses may be significantly associated with increased HSU. The uncertainty regarding the ability of mental disorder diagnoses to predict increased HSU means that this review neither refutes nor supports the argument that reliable mental health clusters can be formed by combining broad diagnoses with care pathways, in a simple and practical manner.12

Findings relating to the domains of the MHCT (ie, behaviour, symptoms, impairment, social functioning and risk factors) can aid assessments of its suitability for clustering purposes. Although some variables relating to these domains were examined, good preliminary evidence for the prediction of increased HSU was found for just two relevant variables—neurotic symptoms and ADLs. Therefore, this review does not provide sufficient evidence to settle the debate regarding the use of the MHCT. However, it highlights the need for further investigation of the link between the MHCT and increased HSU, especially since this link was not taken into account in the initial development of the MHCT.13

Regarding additional variables worth considering in the clustering process, various demographic (ie, comorbidity, age, female gender, marital status, non-white ethnicity, high previous HSU) and intervention (ie, IAPT, medication) variables with good preliminary evidence relating to their ability to predict HSU were identified. Future research could investigate if adding these variables into the ‘case mix’ of the MHCT adds to the economic validity and reliability of mental health clusters. However, it is worth noting that variables that are predictive of HSU are not always suitable for clustering and resource allocation purposes. For example, concerning demographic variables, it could be argued that it would be unfair to distribute resources on the basis of increased HSU by females (relative to males). Similar arguments could be made regarding other population groupings with contrasting HSU levels (eg, certain ethnic groups). Moreover, the benefit of using intervention variables for clustering purposes may be somewhat limited because it is relatively easy for providers to use these variables to ‘game’ the system (ie, when patients are inappropriately and deliberately allocated to clusters that attract higher fixed payments) in order to generate additional revenue.7

Methodological considerations

There is relevant research relating to HSU by people with mental disorders not included in this review. This was for various methodological reasons, for example, differing conceptualisations of HSU in investigations by Killapsy and Zi,21 and Trieman and Leff.22 These studies focused on the stability of HSU over time, and were excluded because they do not address our study question, which concerns identifying predictive variables contributing to an increase or decrease in HSU. In addition, various methodological factors should be taken into account when interpreting our findings. First, the quality of included studies was mixed. Specifically, using arbitrarily cut-off points on the STROBE statement14 and the NICE checklist for RCTs,15 18 of the 28 studies (64.2%) were deemed to be of ‘satisfactory’ quality. This mixed quality limits the strength of conclusions that can be drawn. Second, there was wide heterogeneity in the operationalisation of HSU by included studies, which limits the validity of comparisons across studies. A possible reason for this heterogeneity is that 23 out of 28 (82%) studies collected secondary data from NHS service databases or household surveys, and thus their operationalisation of HSU was constrained. Addressing this issue, the operationalisation of HSU in included studies was documented in considerable detail (table 1). Third, the review was limited to UK studies, meaning the list of identified variables is not exhaustive, and the findings may not be applicable to services in other countries. Indeed, this applicability is particularly limited given that only a few other countries (eg, Australia, New Zealand, Canada, the Netherlands, Norway, USA) have made progress implementing mental health payment systems, using heterogeneous clustering and resource distribution methodologies.23 Fourth, the majority of literature searching was undertaken by one study author. However, in order to minimise bias and error, 20% of abstracts were independently screened by another author. Fifth, the age variable was reported with heterogeneous age ranges across studies. Thus, conclusions in relation to specific age ranges could not be made. Finally, the study benefits from its thorough reporting process and use of structured checklists for assessments of study quality.

Additional future research directions

Five future research directions not already discussed in relation to Mental Health Payment by Results are provided. First, as the operationalisation of HSU in included studies was largely constrained by the use of secondary data from service databases, future HSU studies may benefit from the administration of measures such as the Client Services Receipt Inventory,24 alongside secondary data. Second, an international systematic review of the predictors of HSU by mental health populations could provide a more comprehensive list of predictor variables. Third, the HSU of people with intellectual disabilities was not examined in this review due to the specific additional needs of this population that have to be met beyond the healthcare system. However, it is an important area of research since UK-based studies have highlighted the widespread failure of health services to make required additional accommodations (eg, extended appointment hours) for this patient group, with no additional funding currently allocated for these purposes to NHS acute trusts.25 Determining how the inadequate provision of additional accommodations impacts on the HSU of people with intellectual disabilities could inform future decisions surrounding allocation of resources. Fourth, the review identified a number of variables (ie, attending a community outreach service, attending a psychiatric liaison service, unspecified International Classification of Diseases Tenth Edition (ICD-10) diagnosis, insomnia symptoms, self-harming behaviour) examined in relation to HSU in just one study yet predictive of HSU. Therefore, the associations of these variables with HSU could be explored in future research. Finally, further large-scale case register studies (including participants from shared service catchment areas) would address the study heterogeneity found in this review and provide more robust evidence on the predictors of HSU by people with mental disorders in the UK.

Conclusions

This review provides evidence that can inform decisions about which variables might be used to derive mental health clusters in the Mental Health Payment by Results system. Several variables—in particular, comorbidity, female gender, age (heterogeneous age ranges) high previous HSU and a marital status of divorced—have good preliminary evidence supporting their ability to predict HSU by adults with mental disorders in the UK, and thus are relevant for clustering purposes. The findings support the need to determine the association of the MHCT (and its domains of behaviour, symptoms, impairment, social functioning and risk factors) with HSU, the need to investigate whether combining broad diagnoses with care pathways is an effective alternative method for mental health clustering, and the need for research to further examine the association between existing mental health clusters and HSU. Overall, this review has highlighted important unresolved issues related to the Mental Health Payment by Results system. Addressing these issues could improve how health service resources are distributed, helping to ensure that people experiencing mental health problems can access the most appropriate services at their time of need.

References

View Abstract

Footnotes

  • Contributors CDT, AC and DSB conceived the paper. CDT and AC designed the paper, and undertook analysis and interpretation of data. CDT drafted the paper. MH and DSB inputted into the literature search process. All authors critically reviewed the paper and suggested revisions. All authors gave final approval for the paper for submission.

  • Funding The research is funded by the People Programme (Marie Curie Actions) of the European Union’s Seventh Framework Programme FP7/2007–2013/ (REA grant agreement n° 316795).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement No additional data are available.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.