Good science proceeds by taking approaches that differ from those that have previously “failed to show effects” (1). It dictates that outcomes address the question. The CTO is not a vaccine for preventing future illness. No statute prescribes the CTO outcome of symptom reduction nor requires the prevention of hospitalization; that would constitute the denial of needed treatment. Studies cited by Kisely et al. as “usual” use the outcomes of post-CTO hospital days and rehospitalization, which conflate intervention (needed rehospitalization) with outcome (rehospitalization). Both studies randomized at the stage of hospital release, ignoring hospital days saved from the current inpatient episode, time in which—without the CTO—the patient would remain hospitalized. We chose average inpatient episode duration because that is where the savings in hospital days are most likely to occur, because the CTO uses early conditional release in 95% of cases. Kisely et al., quoting our 2006 paper, omitted the words indicating CTO patients “experienced briefer inpatient episodes,” words preceding the report that they also had “more inpatient days.” Our current article clearly states that the CTO population is more seriously ill and had more days in care because of more episodes of illness. The real question—because “least restrictive alternative” is defined in court decisions as limiting hospital days—is, Are days saved in each hospital episode, days that would have been spent in hospital without the availability of the CTO early conditional release?

Kisely et al.’s emphasis on crude statistics is misleading because all comparisons need to account for between-groups differences. Contrary to Kisely et al.’s assertions, the study models accounted for time of follow-up. Patients with dementia constituted 10% of the CTO sample, not 20%. The mortality models adjusted for dementia by giving it causal priority over the CTO with the use of multiple partial slopes on which the odds ratios are based. The CTO effect was reported only after dementia and all other controls explained all they could in all-cause mortality outcome.

Kisely et al. note that the association of CTOs with reduced all-cause mortality disappeared after adjustment for post-CTO outpatient contacts in his study, suggesting that increased contact, rather than CTO supervision, accounted for the reduced mortality. Almost half (46%) of CTO patients experience more than one CTO episode and refuse treatment. They participate in outpatient treatment to the extent of other patients only when on a CTO, failing to do so outside of such supervision (2). Thus, treatment post-CTO episode is likely to result from further CTO requirements and is likely to be conflated with those requirements in a model, making such a conclusion inappropriate.

The “usual” methods in CTO evaluations lack validity (3,4). Failure to survive longer periods without rehospitalization may be positive (in that patients receive needed treatment) or negative (indicating failure to provide needed community treatment) or both; survival analysis conflates the two, making the result uninterpretable. Surviving death for two years might be a useful alternative approach, although the fixed follow-up time is problematic because of the use of multiple CTOs. We focused on replicating previous reduced mortality results and finding a mechanism by which the CTO might contribute to them. We demonstrated that CTO patients “saved more years of life” during the study period and linked the CTO to the receipt of life-sustaining, acute medical care. Most important in the mortality paper is the interaction of the CTO with obtaining an acute medical care diagnosis in influencing noninjury-related deaths—a direct outcome associated with protecting health as opposed to an inappropriate proxy.

Kisely et al. imply that Victoria’s patient population “is atypical of anywhere else” and indicate that “clinicians and decision makers should therefore be wary of changing practice on the basis of these findings.” Victoria rapidly deinstitutionalized and appears, to its credit, to be protecting the health and safety of its most vulnerable population, as opposed to abandoning them as other jurisdictions have. Victoria’s law and practices are actually typical of most law and are described in our articles and online appendices to facilitate comparisons with other jurisdictions. Generalizability in all science depends on replication. In 2009, our 2006 findings on hospital-day savings were replicated in Western Australia (5). We replicated them again in 2017. Kisely et al. caution that Victoria’s inclusion of dementia cases in its CTO law may make the findings in our mortality article less generalizable. The mortality study replicated similar findings in Victoria in 2006. Ironically, Kisely et al., fully concerned about the dementia issue, replicated the all-cause mortality findings in 2013 in another jurisdiction exclusive of dementia cases (6). Perhaps their interpretation of the results has been forgotten: “Community treatment orders were associated with a reduction in all-cause mortality after adjusting for confounders using matching, multivariate or propensity score techniques” (6).

We hope, contrary to Kisely et al.’s assertions, that clinicians and decision makers will pay close attention to these articles and evaluate them within the context of their own jurisdiction.