ONE CONCEPT, MANY COHORTS: IMPACT OF COHORT PHENOTYPE DEFINITIONAL VARIANCE IN DEPRESSION, MAJOR DEPRESSIVE DISORDER AND TREATMENT-RESISTANT DEPRESSION ON REAL-WORLD RESEARCH
Author(s)
Aaron Kamauu, MPH, MS, MD1, Jared Harrison Wailana Kamauu, BA2, Michael Buck, PhD3, Craig G Parker, MD, MS4, Allise Kamauu, MS5, Scott L. DuVall, PhD6.
1Navidence, Inc., Bountiful, UT, USA, 2Navidence, Inc., Lehi, UT, USA, 3Navidence, Inc., Aurora, CO, USA, 4Navidence, Sandy, UT, USA, 5Navidence, Bountiful, UT, USA, 6PurpleLab, Wayne, PA, USA.
1Navidence, Inc., Bountiful, UT, USA, 2Navidence, Inc., Lehi, UT, USA, 3Navidence, Inc., Aurora, CO, USA, 4Navidence, Sandy, UT, USA, 5Navidence, Bountiful, UT, USA, 6PurpleLab, Wayne, PA, USA.
OBJECTIVES: Complex psychiatric concepts: depression, major depressive disorder, and treatment-resistant depression, are operationalized through conceptual and computable operational definitions (CODefs) used to build cohort phenotypes in real-world research (RWR). Regulators caution that even minor CODef differences “may have a large impact on study results” [FDA 2023]. Building on our review of published depression definitions, we assessed the downstream impact of definitional variance on cohort phenotypes in real-world data (RWD).
METHODS: We replicated 10 published definitions of depression from peer-reviewed literature and clinical references (Martens 2026; Pilon 2019 [depression; MDD]; Engels 2020; Xu 2025; Solberg 2006; Hwang 2015; Veterans Affairs Frailty Index (VA-FI); and CMS CCW Chronic Conditions [broader; narrower]). Using PurpleLab® CLEAR Claims, we operationalized each definition by applying its exact ICD-10-CM code list, then cross-compared the resulting cohorts to quantify patient overlap and the divergences most likely to alter study results.
RESULTS: The 10 depression CODefs ranged from 16 to 28 distinct ICD-10-CM codes. Core single- and recurrent-episode codes (F32.0-F32.3, F32.9; F33.0-F33.4, F33.9) appeared in all definitions, but cohorts diverged on inclusion of dysthymia (F34.1), unspecified/“other” depressive episodes, mood disorders due to physiological conditions (F06.3x), and adjustment disorder with depressed mood (F43.21). The broadest definition (VA-FI) uniquely incorporated bipolar disorder and manic-episode codes (F30.x, F31.x). Cohort patient counts ranged from 87m to 110m; 86m patients qualified for all 10 definitions, reflecting the strong influence of the shared core codes across all definitions. The code F32.9 alone included 50m+ patients where F30.3 F30.4, and F33.4 each included less than 10k patients.
CONCLUSIONS: Despite sharing core codesets, the definitions produced materially different cohorts, showing that cohort size tracks which specific codes are chosen more than how many. These results argue for empirically grounded definition selection, since the operational definition itself can substantially shape who enters a cohort and the endpoints derived from it.
METHODS: We replicated 10 published definitions of depression from peer-reviewed literature and clinical references (Martens 2026; Pilon 2019 [depression; MDD]; Engels 2020; Xu 2025; Solberg 2006; Hwang 2015; Veterans Affairs Frailty Index (VA-FI); and CMS CCW Chronic Conditions [broader; narrower]). Using PurpleLab® CLEAR Claims, we operationalized each definition by applying its exact ICD-10-CM code list, then cross-compared the resulting cohorts to quantify patient overlap and the divergences most likely to alter study results.
RESULTS: The 10 depression CODefs ranged from 16 to 28 distinct ICD-10-CM codes. Core single- and recurrent-episode codes (F32.0-F32.3, F32.9; F33.0-F33.4, F33.9) appeared in all definitions, but cohorts diverged on inclusion of dysthymia (F34.1), unspecified/“other” depressive episodes, mood disorders due to physiological conditions (F06.3x), and adjustment disorder with depressed mood (F43.21). The broadest definition (VA-FI) uniquely incorporated bipolar disorder and manic-episode codes (F30.x, F31.x). Cohort patient counts ranged from 87m to 110m; 86m patients qualified for all 10 definitions, reflecting the strong influence of the shared core codes across all definitions. The code F32.9 alone included 50m+ patients where F30.3 F30.4, and F33.4 each included less than 10k patients.
CONCLUSIONS: Despite sharing core codesets, the definitions produced materially different cohorts, showing that cohort size tracks which specific codes are chosen more than how many. These results argue for empirically grounded definition selection, since the operational definition itself can substantially shape who enters a cohort and the endpoints derived from it.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
RWD123
Topic
Epidemiology & Public Health, Methodological & Statistical Research, Real World Data & Information Systems
Topic Subcategory
Data Protection, Integrity, & Quality Assurance, Reproducibility & Replicability
Disease
Mental Health (including addiction)