Defining Suicidal Thought and Behavior Phenotypes for Genetic Studies

BackgroundSuicidality, including suicidal ideation (SI), attempt (SA), and death (SD), represents complex and partially overlapping phenotypes. This complexity contributes to study population heterogeneity in suicidality research, impeding replication efforts and data consolidation by research consortia. The standardization of suicidality definitions would help but has been insufficiently addressed in existing literature. Here, the Suicide Workgroup of the Psychiatric Genomics Consortium (PGC) provides International Classification of Disease (ICD) definitions, a critical real-world data source, for SA and SI. MethodsThe PGC Suicide Workgroup used published definitions coupled with expert consensus to develop ICD lists to serve as suicidality phenotype definitions. One SI and two SA lists were produced and evaluated for performance against patient screening responses in two independent cohorts (N = 9,151 and 12,621) with differing ascertainment strategies. OutcomesICD list suicidality definitions were produced. Evaluation of generated ICD lists versus patient responses across two cohorts demonstrated varied sensitivity (15{middle dot}4% to 71{middle dot}1%), specificity (67{middle dot}6% to 96{middle dot}3%), and positive predictive values (0{middle dot}57-0{middle dot}92). SI ICD code performance also varied in sensitivity (29{middle dot}4%-86{middle dot}1%), specificity (64{middle dot}2% to 90{middle dot}6%), and positive predictive values (0{middle dot}67 to 0{middle dot}98). InterpretationGuidelines were developed to provide more consistent and comparable suicidality definitions. However, real-world application of ICD codes leads to a wide range of performance, dependent on cohort characteristics, that will need to be carefully considered in implementation. Future efforts would benefit from consistent training in use of ICD codes between sites to improve generalizability, and should include validation in diverse populations. FundingThis work was funded by NIMH R01MH132733 (Mullins), R01MH132733 (Ruderfer), R01MH123619 (Docherty), R01MH123489 (Coon), R01MH124839 (PGC4), R01MH118233 and MH117599 (Smoller), Brain and Behavior Research Foundation No. 31248 (Monson), the Huntsman Mental Health Institute, National Science Foundation Graduate Research Fellowship Program Grant #1842169, and by grant # I01BX005881 and #IK6BX006523 (Kimbrel) from the Department of Veterans Affairs.