To the Editor: Having observed patient progress during conditional release (a form of involuntary outpatient commitment) and found the policy useful in protecting some individuals with serious mental illness, I anticipated positive results from the randomized clinical trial by Dr. Swartz et al. The investigators randomly assigned patients released from a mental hospital to one of two conditions: involuntary outpatient commitment (analogous to parole) or unconditional release. All patients were followed for 12 months. The finding of no difference between the involuntary outpatient commitment and control groups in the number and duration of rehospitalizations at follow-up was instructive regarding the limitations of clinical judgment.

It was, therefore, with some chagrin that I read the authors’ attempts to move beyond comparisons of their randomized groups. They dismissed their findings of no group differences and extended their data analysis to demonstrate results supporting involuntary outpatient commitment. They split the involuntary outpatient commitment group into two subgroups: one with 180 or fewer days of involuntary outpatient commitment and the other with more than 180 days of involuntary outpatient commitment. Since a person in outpatient commitment is, by definition, out of the hospital, analyzing the data by comparing the experience of the group with more than 180 days of involuntary outpatient commitment to the experience of the control group creates a statistical artifact for a result. By definition, the group with more than 180 days of involuntary outpatient commitment could have been hospitalized only during a total of 185 days in the follow-up year. The control group had a full year in which to be hospitalized. The authors arrived at unwarranted conclusions on the basis of this artifact.

The post hoc selection of the group with more than 180 days of outpatient commitment might have been acceptable had the authors compared this group to the control group that had spent more than 180 days in the community with mental health services but without involuntary outpatient commitment. Had the authors compared the two, they would have compared groups with equal risks of hospitalization during the follow-up period. Instead, they compared the involuntary outpatient commitment group with 185 days of risk for hospitalization to the control group with 365 days of hospitalization risk. They found that the involuntary outpatient commitment group with 185 days of risk had approximately one-half the hospitalization experience of the control subjects with a 365-day risk. They compounded their error by attributing this group difference to the intervention of involuntary outpatient commitment.

Another way of understanding the problem with their analysis is to imagine that the authors compared control group members with more than 180 days in the community without involuntary outpatient commitment with the total experimental group’s 1-year follow-up experience. The authors might then have concluded that long-term community residence without involuntary outpatient commitment leads to less frequency and a shorter duration of rehospitalization than living with involuntary outpatient commitment. Perhaps because of having half the risk, this "extended community residence" subset of the control group would have been found to have half the hospitalization of the total involuntary outpatient commitment group.

Involuntary outpatient civil commitment is one of the most controversial issues in psychiatry today. The authors owe the Journal readership appropriate comparisons with the subgroups of the control group. Should direct comparisons between the two groups with more than 180 days of treatment sustain their findings, they will have vastly enhanced the credibility of their conclusions.