Debate ensues as to whether randomised controlled trials are realistically the best method of evaluating the impact of out-patient commitment (OPC) on hospitalisation. ^{Reference Swanson and Swartz1} Even a large and well-funded prospective randomised controlled study such as the OCTET ^{Reference Burns, Rugkåsa, Molodynski, Dawson, Yeeles and Vazquez-Montes2} has been demonstrated to be potentially fundamentally flawed. ^{Reference Mustafa3}

In this context, naturalistic studies evaluating ‘real world’ patients, such as that reported by Castells-Aulet et al, ^{Reference Castells-Aulet, Hernández-Viadel, Jiménez-Martos, Cañete-Nicolás, Bellido-Rodriguez and Calabuig-Crespo4} are welcome and potentially useful. However, I would like to point out three issues that the authors may wish to respond to.

First, given that both controls and OPC patients had their index admissions within the same month, one could reasonably assume that the treating physicians must have had clinical grounds for choosing to place only patients forming the latter group on OPC. Those physicians may have drawn on their knowledge of individual patients (which is not necessarily reflected by the general characteristics described in the study) in reaching their decisions. For instance, the treatment adherence status before the index admission (which, remarkably, differs between the two groups) may have been used, understandably, as an indication of the suitability of patients for the OPC. Hence, one could justifiably doubt the similarity of the two groups, undermining any conclusions that could be drawn from the results.

Second, there potentially could have been another detrimental selection bias in the control group. Patients who were initially discharged informally, but were subsequently readmitted within the following 2 years and then discharged again on an OPC, would have been automatically excluded from the control group, which eventually comprised only patients who, even when re-hospitalised, were not considered by their physicians as requiring OPC, and thus introducing a type II error.

Thirdly, the authors fail to elaborate on the apparent general trend of reduced hospitalisation over the 4-year period, which may have been driven by factors that could potentially confound the results of this study.