Study protocol for a stepped-wedge, randomized controlled trial to evaluate implementation of a suicide risk...

Age-adjusted suicide rates have increased in the U.S. over the past 25 years. Algorithm-based methods for identifying individuals at risk for suicide based on electronic health record and claims data have been validated but few studies have evaluated implementation or effects on population-level suicide attempt rates.

This hybrid type I effectiveness-implementation pragmatic clinical trial will test a suicide risk identification model in behavioral health clinics at three large health systems. Local decision-makers will determine implementation specifics at each site. Clinics within each health system will be randomized to determine order of implementation. A stepped-wedge design using repeated measures pre/post-implementation maximizes statistical efficiency and power with fewer participants compared to a parallel design while allowing all clinics to participate. A pre-implementation period will serve as the baseline. The primary outcome will be the rate of suicide attempt per 1000 visits at 90- and 180-days following a behavioral health visit in which an individual was identified by the suicide risk model compared with the baseline period (no use of suicide risk model). Secondary outcomes include identification of suicide risk and recognition of individuals at risk for suicide (e.g., completed risk assessment), both compared to the baseline period. Generalized linear mixed models will be used to account for clustering within clinics and repeated measures over time, adjusting for relevant covariates to estimate the effect of the suicide risk model on outcomes. Implementation outcomes, including system-level determinants and clinician acceptance and use of the suicide risk model, will also be measured.

Few suicide risk models derived from administrative and clinical data have been tested in real world care settings. This trial will determine whether the use of such a risk model reduces suicide attempts compared to usual care. By describing important implementation factors, use of such risk models, if effective, may be accelerated for other health care systems.

ClinicalTrials.gov NCT06060535.