A SIMULATION FRAMEWORK FOR QUANTIFYING STRUCTURAL UNCERTAINTY IN CLAIMS-BASED RARE DISEASE PREVALENCE ESTIMATION
Author(s)
Astrid Genet1, Susanne Schwenke, .2.
1Berlin, Germany, 2IQVIA Commercial GmbH & Co. OHG, München, Germany.
1Berlin, Germany, 2IQVIA Commercial GmbH & Co. OHG, München, Germany.
OBJECTIVES: Rare disease prevalence estimates derived from claims data are subject to substantial structural uncertainty which cannot be fully quantified in standard analyses. This research aims to develop a simulation‑based framework to help quantify the impact of such structural uncertainty on prevalence estimates using a case study of ATTR‑CM in the German SHI population.
METHODS: A simulation‑based framework for point prevalence estimation was developed using the SHI‑derived number of prevalent ATTR‑CM cases (942 [95% CI: 689-1,256]), as reported in the tafamidis HTA dossier, as the baseline estimate. Stochastic variations of this baseline prevalence was modelled using a uniform distribution. Structural uncertainty in case identification was operationalized as multiplicative limitation factors reflecting key sources of bias, including ICD‑10 coding uncertainty, exclusion of AL amyloidosis, capture of diagnostic procedures, identification of prevalent cases, and underdiagnosis. A MC simulation combined stochastic variation of baseline prevalence with scenario‑based variation of limitation factors within predefined ranges (0.80-1.50).
RESULTS: In the reference scenario accounting only for statistical input uncertainty, the median estimated ATTR‑CM prevalence was 970 [709; 1,235] patients. When all uncertainty sources varied simultaneously, the prevalence distribution showed a median of 1,083 [650; 1,745] patients. Compared with the reference scenario, this corresponds to a moderate median increase (+12%) but a substantial widening of the uncertainty interval (+108%). Those results indicate that while central prevalence estimates remain relatively stable across assumptions, structural uncertainties substantially increase the variability of prevalence estimates.
CONCLUSIONS: Claims‑based approaches provide a valid basis for estimating rare disease prevalence. However, our analyses highlight that structural uncertainties substantially affect result precision. Decisions should therefore not rely on point estimates alone, and triangulation with complementary data sources (e.g., disease registries, clinical cohorts, published epidemiological studies, or expert input) is recommended to account for these uncertainties.
METHODS: A simulation‑based framework for point prevalence estimation was developed using the SHI‑derived number of prevalent ATTR‑CM cases (942 [95% CI: 689-1,256]), as reported in the tafamidis HTA dossier, as the baseline estimate. Stochastic variations of this baseline prevalence was modelled using a uniform distribution. Structural uncertainty in case identification was operationalized as multiplicative limitation factors reflecting key sources of bias, including ICD‑10 coding uncertainty, exclusion of AL amyloidosis, capture of diagnostic procedures, identification of prevalent cases, and underdiagnosis. A MC simulation combined stochastic variation of baseline prevalence with scenario‑based variation of limitation factors within predefined ranges (0.80-1.50).
RESULTS: In the reference scenario accounting only for statistical input uncertainty, the median estimated ATTR‑CM prevalence was 970 [709; 1,235] patients. When all uncertainty sources varied simultaneously, the prevalence distribution showed a median of 1,083 [650; 1,745] patients. Compared with the reference scenario, this corresponds to a moderate median increase (+12%) but a substantial widening of the uncertainty interval (+108%). Those results indicate that while central prevalence estimates remain relatively stable across assumptions, structural uncertainties substantially increase the variability of prevalence estimates.
CONCLUSIONS: Claims‑based approaches provide a valid basis for estimating rare disease prevalence. However, our analyses highlight that structural uncertainties substantially affect result precision. Decisions should therefore not rely on point estimates alone, and triangulation with complementary data sources (e.g., disease registries, clinical cohorts, published epidemiological studies, or expert input) is recommended to account for these uncertainties.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MSR127
Topic
Epidemiology & Public Health, Methodological & Statistical Research, Real World Data & Information Systems
Disease
Cardiovascular Disorders (including MI, Stroke, Circulatory), Rare & Orphan Diseases