Systematic review of the effect of on-site mental health professionals on the clinical behaviour of general practitionersBMJ 2000; 320 doi: https://doi.org/10.1136/bmj.320.7235.614 (Published 04 March 2000) Cite this as: BMJ 2000;320:614
- Peter Bower (), research fellow,
- Bonnie Sibbald, professor of health services research
- National Primary Care Research and Development Centre, Williamson Building, University of Manchester M13 9PL
- Correspondence to: P Bower
- Accepted 13 December 1999
Objectives: To review the published literature concerning the effects of on-site mental health professionals on general practitioners' management of mental health.
Design: Systematic review of controlled trials.
Setting: Primary care.
Participants: General practitioners and mental health professionals.
Main outcome measures: Consultation rates, prescribing of psychotropics, and referrals to secondary care mental health services by general practitioners.
Results: The effect of on-site mental health professionals on consultation rates was inconsistent. Referral to a mental health professional reduced the likelihood of a patient receiving a prescription for psychotropics or being referred to secondary care, although the effects were not consistent. An on-site mental health professional did not alter prescribing and referral behaviour towards patients in the wider practice population.
Conclusions: The secondary effects of mental health professionals on the clinical behaviour of general practitioners are comparatively modest and inconsistent and seem to be restricted to patients directly under the care of the mental health professional.
A variety of mental health professionals such as counsellors work in primary care.1 As yet the cost effectiveness of such provision is unclear when compared with routine general practitioner care.2 The presence of an on-site mental health professional, however, represents an organisational change that may have benefits beyond immediate patient outcome, such as changes in established clinical routines among general practitioners. This could have profound effects on the cost effectiveness of mental health provision in primary care.
On-site mental health professionals may have two distinct effects on the behaviour of general practitioners. The direct effect concerns the influence of on-site mental health professionals on the behaviour of general practitioners towards patients referred to the mental health professionals—for example, are general practitioners less likely to prescribe antidepressants for depressed patients whom they refer to the practice counsellor compared with those who remain under the doctor's care alone? Direct effects are examined in studies comparing the behaviour of general practitioners towards patients who are allocated to mental health professionals or to routine general practitioner care. The indirect or spillover3 effect concerns the influence of on-site mental health professionals on the behaviour of general practitioners towards the wider practice population who have not been referred to the mental health professional—for example, does the prescription rate for psychotropics throughout the practice decline when a practice employs a psychologist? Indirect effects are examined in studies that compare clinical behaviour in practices with and without on-site mental health professionals.
Several different models exist of the relationship between on-site mental health professionals and general practitioners, 4 5 but for the purposes of our review two main models were distinguished. In the “replacement” model the mental health professional assumes primary responsibility for the management of the patient's mental health problem. In the “consultation-liaison” model the mental health professional aims to support the general practitioner's management of the patient's mental health problem through education and support.4 Our review was concerned with the direct and indirect effects of the replacement model, which is the more common model found in primary care in the United Kingdom. Consultation-liaison studies are the subject of a separate review.
We aimed to test the hypothesis that on-site mental health professionals reduce the frequency of consultations, prescriptions, and referrals to off-site services by general practitioners.
Our review was conducted as part of the effective practice and organisation of care module of the Cochrane Library.6 We electronically searched Medline (1966-98), PsycInfo (1984-98), Embase (1980-98), the Cochrane Clinical Trials' Register, the specialised register of the Effective Practice and Organisation of Care group, and Counselling in Primary Care Trust Counsel.Lit database (see website). We also searched the reference lists of all relevant studies.
Direct effects are examined by studies that compare the behaviour of general practitioners towards patients who are allocated to the care of either a mental health professional or general practitioner. We included randomised controlled trials reporting objective measures (for example, searches of medical records) of consultations, rates and costs of prescribing psychotropics, or referrals to secondary care. Although consultation rates are sometimes considered a patient behaviour, an important proportion of consultations in the United Kingdom are initiated by doctors, and we therefore regarded them as a relevant indicator of management by general practitioners. Although we examined non-randomised controlled “before and after” trials, the presence of a major number of randomised trials examining direct effects led to the post hoc decision to exclude non-randomised trials owing to the difficulties associated with their interpretation.7
Indirect effects are examined by studies that compare the behaviour of general practitioners in practices with and without on-site mental health professionals. Because of difficulties in the random allocation of practices to intervention and control groups, we broadened the inclusion criteria to consider randomised controlled trials, controlled before and after studies, and interrupted time series reporting objective measures of rates and costs of psychotropic prescribing or referrals to secondary care at practice level. Several studies reported the effect of community psychiatric services on admissions to mental hospital. 8 9 Admissions were considered an indicator of specialist, not general practitioner, management and were therefore excluded.
Where possible we tabulated results in terms of means and standard deviations for consultations and proportions for prescribing and referrals. Other data are presented as reported in the original source. In controlled before and after studies, baseline differences were recorded wherever possible to provide some indication of the comparability of study groups before the intervention.
Although there is significant heterogeneity in the professional background and therapeutic approach of mental health professionals in primary care, such differences may be comparatively unimportant in terms of changes in the behaviour of the general practitioner, and thus pooling of outcomes from randomised controlled trials may be justified. Such pooling was not, however, undertaken because data, such as variance statistics, were frequently not reported. Our analysis was therefore qualitative.
We identified 40 relevant outcomes: 13 randomised controlled trials of direct effects on consultation rates (table A on website)10–23; 12 randomised controlled trials of direct effects on prescribing (table B on website) 10–14 18–25; six randomised controlled trials of direct effects on referrals (table C on website) 11 15 18 22–25; three controlled before and after studies of indirect effects on prescribing (table D on website)26–29; and six controlled before and after studies of indirect effects on referrals (table E on website). 26 27 29–32
Studies of direct effects
Concealment of allocation was considered adequate in two studies, 15 23 open to bias in three, 10 16 24 and unclear from information provided by the remainder. Follow up rates of greater than 80% were reported in five studies, 10–12 16 19 less than 80% in two, 13 17 and eight did not provide information. Sample sizes ranged from 44 to 429 (mean 152).
Behaviours were assessed through searches of medical records, reviews of charts, or automated data. The reliability of data from searches of medical records was not assessed. Although several studies reported power analyses, these were always related to mental health outcomes of patients (the primary focus of all the studies) rather than the outcomes of general practitioner behaviour, so their utility for our review is unclear.
Studies of indirect effects
The unit of analysis in studies of indirect effects should be the practice but only two studies clearly analysed at that level. 26 28 Information concerning the comparability of the control sites was provided in only one study.26 One study used random selection of controls from a sample but did not provide any descriptive statistics.28 Other studies described qualitative similarities between practices30–32 or used practices in the same geographical area.29 The comparability of control and intervention practices in terms of outcome variables at baseline was examined statistically in only two studies 28 32 and confirmed in one.28
In these studies, most general practitioner behaviours were assessed objectively and reliability enhanced through the use of automated recording systems. Control practices were chosen so as not to have access to the intervention under test, although it was not clear in most cases whether these practices had access to other mental health professionals not included in the intervention. Samples sizes ranged from three to 87 practices. None of the indirect studies reported a power analysis.
Of the 13 studies of consultation rates only three reported statistically significant effects, with lower rates in the mental health professional groups. 16 18 20 Of the 12 studies of prescribing behaviour, five found significant reductions in the mental health professional group. 10 11 19–21 The effects were not always consistent within studies in terms of the different drugs examined and the duration of the effect. Three of the six randomised controlled trials of referral behaviour reported significant reductions in the mental health professional group. 11 22 24 Two others reported lower rates and costs in the mental health professional group, and one reported no difference.
None of the studies reported a significant association between on-site mental health professionals and practice prescribing of psychotropics. One controlled before and after study reported a significant association between an on-site mental health professional and higher rates of referral to secondary care.26 The authors, however, also conducted a second analysis using a smaller subset of practices matched for deprivation, population size, fundholding status, and location, and they found no differences in referral rates.
Our review was restricted to controlled trials thereby excluding qualitative research that may add insights into why the behaviour of general practitioners may or may not change with on-site mental health professionals. We found no discernible publication bias favouring positive effects on behaviour of general practitioners, possibly because most included studies were primarily concerned with the clinical effectiveness of mental health professionals and not their impact on the general practitioner. For the same reasons the quality of information on behaviour of general practitioners was variable.
The quality of the included studies was variable, but statistically significant results were not restricted to studies with methodological weaknesses. Concerns have also been raised about the external validity of such primary care studies. Prescribing behaviour in trials may differ from routine care, 33 because of Hawthorne effects or more specific mechanisms, such as the confirmation of a diagnosis in trial patients. Future studies would also benefit from longer term follow up.
On-site mental health professionals did not cause substantial changes in the clinical behaviour of general practitioners. In terms of direct effects, referral to an on-site mental health professional did not consistently reduce consultations with general practitioners. The effects on prescribing behaviour were also inconsistent, but suggest that fewer patients referred to a mental health professional are given a prescription for psychotropics, especially in the short term. The evidence for a direct effect was strongest in relation to general practitioner referrals to secondary care providers, which were reduced with an on-site mental health professional, although even here the effect was not totally consistent. Although of comparatively modest size, such effects may have an effect on the overall cost effectiveness of treatment from mental health professionals compared with routine general practitioner care: a recent trial found no obvious cost advantage associated with either treatment, due to these direct effects.15
These findings are interpretable in that on-site mental health professionals may provide an accessible alternative to drugs and off-site referral. The apparent inconsistency of these effects, however, requires explanation. The relapsing nature of some psychiatric disorders or lack of patient response to treatments may encourage general practitioners to revert to traditional management options over time. Adding mental health professionals to primary care teams may, however, alter interprofessional as well as interpersonal working relationships in ways that are poorly understood. Future studies would benefit from the addition of qualitative research to increase knowledge about the conditions—relating to person, profession, and practice—that facilitate or prevent behaviour change in general practitioners.
What is already known on this topic
Mental health professionals are increasingly working in the primary care setting yet randomised controlled trials have not provided evidence that their treatments are superior to routine general practitioner care in the long term
The addition of such professionals to primary care represents an organisational change that may have an impact beyond the immediate health outcomes of patients managed by such professionals—for example, additional costs of such specialists could be recouped from reductions in prescribing and referral to secondary care. These reductions can occur in patients referred to the mental health specialist or may generalise to the ways that general practitioners manage the wider practice population
What this study adds
Although conclusions from this review are restricted by shortcomings in the methodology and reporting of studies, referring a patient to a mental health professional reduces the likelihood of a general practitioner prescribing psychotropics or referring patients to specialist psychiatric services, at least in the short term
The effects on consultation rates are less consistent. On-site mental health professionals do not seem to affect the behaviour of general practitioners towards the wider practice population who are not referred directly to the mental health professional
Given the modest direct effects, it is not surprising that indirect effects were uncommon. Indeed, the available evidence suggests that on-site mental health professionals may increase referral rates to certain mental health services, possibly through sensitising the general practitioner to psychosocial problems that cannot be managed within the practice. The number of studies reporting increases was, however, small. Little information was available on the number of hours mental health professionals were employed to work in the practice, and indirect effects may only occur when they undertake sufficient work.34
The use of randomisation to examine indirect effects is problematic because of difficulties in randomising service provision at practice level. Differences between control and intervention groups in the characteristics of both providers and practices complicated the interpretation of outcomes. More detailed reporting of these characteristics would allow better assessment of comparability. It is inevitable that interpretative difficulties caused by lack of control over allocation may only be offset by a weight of evidence from several studies showing consistent results. Studies in our review have shown that service evaluations using automated databases, such as prescribing data, can provide both large samples and long term evaluation.
Referral to an on-site mental health professional may reduce referrals and prescribing by general practitioners, but there is no evidence that such changes are enduring or particularly broad in scope.
We thank Jeremy Grimshaw and Graham Mowatt (effective practice and organisation of care group) for their advice and support, Steve Rose (National Primary Care and Research Development Centre, University of Manchester) for assistance with the searches, and the staff at John Rylands Library (University of Manchester) for assistance with interlibrary loans.
Contributors: BS had the idea for the review and PB and BS wrote the protocol. PB carried out the searches and administration of the review. Both PB and BS extracted data from the studies, interpreted the results, and wrote the paper. PB will act as guarantor for the paper.
Funding The National Primary Care Research and Development Centre is funded by the Department of Health.
Competing interests None declared.
website extra Details of the search strategy and included and excluded trials appear on the BMJ's website www.bmj.com