SAME CONCEPT, DIFFERENT COHORT? COMPARING DIAGNOSIS CODE-BASED (CLAIMS) AND PHQ-9-BASED (SPECIALTY EMR) OPERATIONAL DEFINITIONS OF DEPRESSION IN REAL-WORLD DATA
Author(s)
Aaron Kamauu, MPH, MS, MD1, Reyna Klesh, MS2, Jared Harrison Wailana Kamauu, BA3, Michael Buck, PhD4, Craig G Parker, MD MS5, Scott L. DuVall, PhD6.
1Navidence, Inc., Bountiful, UT, USA, 2Behavioral Health Insights, Nashville, TN, USA, 3Navidence, Inc., Lehi, UT, USA, 4Navidence, Inc., Aurora, CO, USA, 5Navidence, LLC, Sandy, UT, USA, 6PurpleLab, Wayne, PA, USA.
1Navidence, Inc., Bountiful, UT, USA, 2Behavioral Health Insights, Nashville, TN, USA, 3Navidence, Inc., Lehi, UT, USA, 4Navidence, Inc., Aurora, CO, USA, 5Navidence, LLC, Sandy, UT, USA, 6PurpleLab, Wayne, PA, USA.
OBJECTIVES: Depression cohorts in real-world research (RWR) are often built from administrative diagnosis codes, whereas the clinical reference standard for identifying and grading depression is based on validated instruments, such as the Patient Health Questionnaire-9 (PHQ-9). Diagnosis codes and symptom scores capture different information and reside in different data sources, so the operational definition may determine who is studied. We compared a code-based (claims) depression phenotype with a PHQ-9-based phenotype derived from specialty electronic medical record (EMR) data.
METHODS: Using PurpleLab® CLEAR Claims, we built depression cohorts from published ICD-10-CM code-list definitions. In parallel, Behavioral Health InsightsTM characterized a clinical cohort in a U.S. specialty behavioral-health EMR: patients with >e;1 PHQ-9 score %gte;5, indicating at least mild depressive symptoms (Kroenke 2001). We compared cohort size, composition, and data availability rather than direct patient-level overlap.
RESULTS: The PHQ-9 >e;5 cohort comprised 464,274 patients with 1,302,698 assessments; 57.7% had only one assessment. Patients assessed annually rose from 75,595 (2020) to 147,926 (2025). The cohort concentrated in the Midwest (44.1%) and South (32.0%); 56.4% had ≥1 e-prescription, most commonly trazodone (15.9%), sertraline (11.3%), and quetiapine (11.1%). This clinically defined population would be largely unidentifiable from diagnosis codes alone. Claims-based ICD-10-CM cohorts yielded 87m-110m patients across 10 definitions. The number of patients who had only one type of depression diagnosis code ranged from 52% to 57% across the cohorts with patients averaging 17 to 22 instances. All 10 cohorts had the same most common prescriptions in this order gabapentin, sertraline, escitalopram, bupropion, and trazadone.
CONCLUSIONS: Code-based and PHQ-9-based definitions identify depression through fundamentally different mechanisms and data sources, producing cohorts likely to diverge in size and composition. The choice of definition and data source can therefore materially shape depression cohorts and outcomes in RWR; quantifying patient-level overlap in linked data is an important next step.
METHODS: Using PurpleLab® CLEAR Claims, we built depression cohorts from published ICD-10-CM code-list definitions. In parallel, Behavioral Health InsightsTM characterized a clinical cohort in a U.S. specialty behavioral-health EMR: patients with >e;1 PHQ-9 score %gte;5, indicating at least mild depressive symptoms (Kroenke 2001). We compared cohort size, composition, and data availability rather than direct patient-level overlap.
RESULTS: The PHQ-9 >e;5 cohort comprised 464,274 patients with 1,302,698 assessments; 57.7% had only one assessment. Patients assessed annually rose from 75,595 (2020) to 147,926 (2025). The cohort concentrated in the Midwest (44.1%) and South (32.0%); 56.4% had ≥1 e-prescription, most commonly trazodone (15.9%), sertraline (11.3%), and quetiapine (11.1%). This clinically defined population would be largely unidentifiable from diagnosis codes alone. Claims-based ICD-10-CM cohorts yielded 87m-110m patients across 10 definitions. The number of patients who had only one type of depression diagnosis code ranged from 52% to 57% across the cohorts with patients averaging 17 to 22 instances. All 10 cohorts had the same most common prescriptions in this order gabapentin, sertraline, escitalopram, bupropion, and trazadone.
CONCLUSIONS: Code-based and PHQ-9-based definitions identify depression through fundamentally different mechanisms and data sources, producing cohorts likely to diverge in size and composition. The choice of definition and data source can therefore materially shape depression cohorts and outcomes in RWR; quantifying patient-level overlap in linked data is an important next step.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
RWD164
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)