MEASUREMENT OF PREVALENCE AND INCIDENCE IN OPEN CLAIMS DATA: ASSESSMENT OF DIFFERENT METHODOLOGIES
Author(s)
Shirley Xiaolei Li, BSc1, James B. Young, MPH1, Adam Idica, PhD2, aishwarya Uday, MTech1, Ekta Bhatnagar, MSc3, Hemanth Nair, PhD, MPH4;
1Clarivate, Toronto, ON, Canada, 2Clarivate, Chandler, AZ, USA, 3Clarivate, Bangalore, India, 4Clarivate, Arlington, VA, USA
1Clarivate, Toronto, ON, Canada, 2Clarivate, Chandler, AZ, USA, 3Clarivate, Bangalore, India, 4Clarivate, Arlington, VA, USA
OBJECTIVES: Real‑World Data (RWD) are widely used to estimate disease prevalence and incidence, yet differences in data sources, analytic choices, and assumptions yield substantial variation and limit comparability. Open claims data present added challenges because of absence of fixed denominator population. This study evaluated how methodological specifications affect prevalence and incidence estimates derived from an open claims database.
METHODS: 2018-2023 annual prevalence and incidence for chronic obstructive pulmonary disease (COPD), cystic fibrosis, and schizophrenia were estimated using Clarivate RWD claims and electronic health record (EHR). We varied three specifications: (1) disease identification, defined as either any claim or ≥2 claims separated by ≥30 days within a 365‑day period; (2) lookback period, defined as prior observation time of 1 year, 2 years, or all history beginning in 2016; and (3) continuous enrollment (CE) proxy, defined as no CE, ≥1 claim every 12‑month interval, or ≥1 claim every 6‑month interval.
RESULTS: Stricter disease identification reduced estimates across conditions: in 2023, requiring ≥2 claims (vs any claim) decreased prevalence by 1.8-2.5× and incidence by 3.8-9.4×, with cystic fibrosis most affected. Extending the lookback from 1 to 2 years modestly changed 2023 estimates (incidence decreased 1.2-1.3×; prevalence increased 1.15-1.36×). Using all available history produced the largest shifts relative to 1 year (incidence decreased 1.46-2.46×; prevalence increased 1.68-1.97×). CE proxies also influenced results: 6-month CE yielded the highest estimates, no proxy the lowest. Across 2018-2023, all history lookback showed increasing prevalence and decreasing incidence for COPD and stable prevalence, decreasing incidence for cystic fibrosis. However, temporal variability existed amongst methods.
CONCLUSIONS: Methodological choices - especially disease identification rules, look‑back duration, and CE proxies - affect prevalence and incidence estimated from open claims. Rigorous, consistent cohort definitions and transparent parameter selection are essential to produce interpretable, comparable RWD outputs for healthcare planning and drug development.
METHODS: 2018-2023 annual prevalence and incidence for chronic obstructive pulmonary disease (COPD), cystic fibrosis, and schizophrenia were estimated using Clarivate RWD claims and electronic health record (EHR). We varied three specifications: (1) disease identification, defined as either any claim or ≥2 claims separated by ≥30 days within a 365‑day period; (2) lookback period, defined as prior observation time of 1 year, 2 years, or all history beginning in 2016; and (3) continuous enrollment (CE) proxy, defined as no CE, ≥1 claim every 12‑month interval, or ≥1 claim every 6‑month interval.
RESULTS: Stricter disease identification reduced estimates across conditions: in 2023, requiring ≥2 claims (vs any claim) decreased prevalence by 1.8-2.5× and incidence by 3.8-9.4×, with cystic fibrosis most affected. Extending the lookback from 1 to 2 years modestly changed 2023 estimates (incidence decreased 1.2-1.3×; prevalence increased 1.15-1.36×). Using all available history produced the largest shifts relative to 1 year (incidence decreased 1.46-2.46×; prevalence increased 1.68-1.97×). CE proxies also influenced results: 6-month CE yielded the highest estimates, no proxy the lowest. Across 2018-2023, all history lookback showed increasing prevalence and decreasing incidence for COPD and stable prevalence, decreasing incidence for cystic fibrosis. However, temporal variability existed amongst methods.
CONCLUSIONS: Methodological choices - especially disease identification rules, look‑back duration, and CE proxies - affect prevalence and incidence estimated from open claims. Rigorous, consistent cohort definitions and transparent parameter selection are essential to produce interpretable, comparable RWD outputs for healthcare planning and drug development.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
RWD63
Topic
Real World Data & Information Systems
Topic Subcategory
Health & Insurance Records Systems, Reproducibility & Replicability
Disease
SDC: Respiratory-Related Disorders (Allergy, Asthma, Smoking, Other Respiratory)