Comparative predictive performance of the PopGrouper: internal validation of a population-based classification system for morbidity adjustment using German administrative data

Abstract Background To compare the predictive performance of a newly developed population classification system, the PopGrouper, with established morbidity measures used in administrative data research for 1-year mortality, hospitalisation, and healthcare costs. Methods This study presents an internal validation using an independent hold-out sample derived from 2022 to 2023 claims data from BARMER, a large German statutory health insurer covering about 10% of the national population (> 8 million). All adults (≥ 18 years) continuously insured in 2022–2023 or deceased in 2023 were included of which a 20%-random sample ( n = 1,310,273) was used as validation cohort. Regression models were estimated for different outcomes, using age-sex groups, the Charlson Comorbidity Index, Elixhauser comorbidities, PopGroup severity quintiles, Macro PopGroups, or PopGroups as morbidity adjustment methods. For mortality and hospitalisation, sparsely populated PopGroups were hierarchically aggregated to ensure a minimum number of events per group. Predictive performance was assessed using the coefficient of determination (R²) and root mean square error (RMSE) for log-transformed cost outcomes, and c-statistics for mortality and hospitalisation, based on 1,000 bootstrap samples. Results The PopGroup model explained 35.8% of the variation in log-transformed total healthcare costs, compared with 11.5% for age-sex groups, 13.2% for Charlson, and 18.6% for Elixhauser. Discrimination for mortality was excellent (c-statistic = 0.884 [95% CI: 0.882–0.886]) and acceptable for hospitalisation (0.714 [0.712–0.715]). Differences compared with conventional morbidity measures were most pronounced for healthcare cost prediction, whereas gains for mortality and hospitalisation discrimination were smaller. Sensitivity analyses yielded consistent findings, confirming robustness across model specifications. Conclusions The PopGrouper is an open and reproducible classification system which captures multimorbidity by building clinically and economically meaningful groups derived from administrative data. In this internal validation, it demonstrated strong performance across both clinical and economic outcomes, with the largest improvements observed for healthcare cost prediction. The PopGrouper may therefore provide a useful alternative tool for morbidity adjustment, particularly in applications involving healthcare planning, resource allocation, and population health management.

Authors

Institutions

Publication Details

Journal
BMC Health Services Research
Published
2026-11-12
DOI
https://doi.org/10.1186/s12913-026-16108-3
Primary Topic
Medical Coding and Health Information
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Comparative predictive performance of the PopGrouper: internal validation of a population-based classification system for morbidity adjustment using German administrative data

Anika Kreutzberg, Reinhard Busse, Wilm Quentin, Chrissa Tsatsaronis
BMC Health Services Research
Medical Coding and Health Information
article

Comparative predictive performance of the PopGrouper: internal validation of a population-based classification system for morbidity adjustment using German administrative data

Anika Kreutzberg, Reinhard Busse, Wilm Quentin, Chrissa Tsatsaronis
article en

Abstract

Abstract Background To compare the predictive performance of a newly developed population classification system, the PopGrouper, with established morbidity measures used in administrative data research for 1-year mortality, hospitalisation, and healthcare costs. Methods This study presents an internal validation using an independent hold-out sample derived from 2022 to 2023 claims data from BARMER, a large German statutory health insurer covering about 10% of the national population (> 8 million). All adults (≥ 18 years) continuously insured in 2022–2023 or deceased in 2023 were included of which a 20%-random sample ( n = 1,310,273) was used as validation cohort. Regression models were estimated for different outcomes, using age-sex groups, the Charlson Comorbidity Index, Elixhauser comorbidities, PopGroup severity quintiles, Macro PopGroups, or PopGroups as morbidity adjustment methods. For mortality and hospitalisation, sparsely populated PopGroups were hierarchically aggregated to ensure a minimum number of events per group. Predictive performance was assessed using the coefficient of determination (R²) and root mean square error (RMSE) for log-transformed cost outcomes, and c-statistics for mortality and hospitalisation, based on 1,000 bootstrap samples. Results The PopGroup model explained 35.8% of the variation in log-transformed total healthcare costs, compared with 11.5% for age-sex groups, 13.2% for Charlson, and 18.6% for Elixhauser. Discrimination for mortality was excellent (c-statistic = 0.884 [95% CI: 0.882–0.886]) and acceptable for hospitalisation (0.714 [0.712–0.715]). Differences compared with conventional morbidity measures were most pronounced for healthcare cost prediction, whereas gains for mortality and hospitalisation discrimination were smaller. Sensitivity analyses yielded consistent findings, confirming robustness across model specifications. Conclusions The PopGrouper is an open and reproducible classification system which captures multimorbidity by building clinically and economically meaningful groups derived from administrative data. In this internal validation, it demonstrated strong performance across both clinical and economic outcomes, with the largest improvements observed for healthcare cost prediction. The PopGrouper may therefore provide a useful alternative tool for morbidity adjustment, particularly in applications involving healthcare planning, resource allocation, and population health management.

BMC Health Services Research
Technische Universität Berlin (DE), University of Bayreuth (DE)
Technische Universität Berlin
Good health and well-being
Medical Coding and Health Information
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon