Community Treatment Orders for Psychiatric Patients: The Emperor with No Clothes

Abstract

How does society protect some of its most vulnerable and disadvantaged patients while respecting individual autonomy? This is the dilemma facing many Canadian jurisdictions as they consider the extension of compulsory treatment to community settings. Canada is not unique but part of a trend to enforce outpatient treatment in many jurisdictions, including Australia, New Zealand, the United States, and the European Union (1,2). What is the basis for this enthusiasm? Certainly, research published in this journal has endorsed the measure to varying degrees (3), as does the position paper of the Canadian Psychiatric Association (4). We wonder whether this confidence is misplaced, and we critically examine several troubling aspects of the literature on CTOs. Opinion Is Not Evidence One approach to the evaluation of CTOs has been to ask practising psychiatrists for their opinion on the utility of the orders (3). These suggest general support for the measure. However, surveys of psychiatrists' views on CTOs provide only Level 3 evidence, in contrast to the Level 1 evidence from appropriately conducted RCTs and the Level 2 evidence from well-designed non-RCT studies, such as case-control studies or interrupted time series. Such surveys have little place in an era of evidence-based practice and, in the absence of other data, would not be accepted as a reason to introduce any other psychiatric intervention. No Controls, Little Evidence Early uncontrolled studies, largely from the United States, suggested that patients on CTOs had reduced rates of rehospitalization and shorter stays in hospital, associated with increased compliance. However, there were no significant differences between the 2 groups when control subjects not subject to a compulsory treatment order were included (5-7). The weakness of uncontrolled designs is the difficulty in determining the reason for any change in outcome. Aside from the effect of the intervention, other possible explanations include regression to the mean, other treatment, life events, or changes in social circumstances. These often overestimate the effect of the intervention of interest. Controlled studies without randomization also have shortcomings. There may always be another reason why patients were placed on treatment and the control subjects were not. They may be less insightful about their illness or more likely to have a history of aggressive behaviour. Only one matched control study controlled for forensic history (7). What About RCTs? RCTs address many of these problems but are very difficult to conduct where mental health legislation policy is implemented at a state, provincial, or national level. There have therefore only been 2 RCTs, and both were in the United States (New York and North Carolina) (8-11). These trials have been the subject of much debate in terms of subjects, design, analysis, and generalizability. Selection and Follow-Up Bias: Were the Patients Typical? Both RCTs excluded patients with a history of violence from randomization. Although understandable from an ethical and legal standpoint, this limits their applicability because recent dangerousness, particularly violence against others, is often the reason for compulsory treatment in hospital or in the community. One study did include a nonrandomized violent subgroup in the analysis, but this negates the whole point of an RCT (see below) (9-11). Selection bias was further compounded by high dropout rates. Of the 577 patients identified as eligible for participation in the New York and North Carolina studies, including the violent subgroup, only 292 (51%) were followed up 1 year later. Although the North Carolina study had a higher completion rate among randomized patients than the New York study (82% and 45%, respectively), data were not available for all patients for all of the outcomes. The 1 -year follow-up is therefore of a highly selected and potentially unrepresentative population that was not dangerous and was sufficiently compliant to participate in baseline and follow-up assessments. …