Systematic review of the effect of on-site mental health professionals on the clinical behaviour of general practitioners

Direct effects

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

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.

Statistical analysis

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.

Results

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

Study quality

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.

Direct effects

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.

Indirect effects

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.

Discussion

Methological issues

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.

Direct effects

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.

Indirect effects

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.

Conclusion

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.