Generalizing Multimorbidity Models Across Countries: A Comparative Study of Austria and Denmark

TL;DR

This study tests the cross-national generalizability of data-driven multimorbidity trajectory models by applying hierarchical clustering to population-wide health records from Austria and Denmark. It derives 132 health-state multimorbidity clusters per country and cross-validates by assigning Danish records to Austrian clusters, using Normalized Mutual Information and Adjusted Rand Index to quantify alignment. The results reveal near-identical cluster structures across the two health systems (ARI ≈ 0.998, NMI ≈ 0.88) and show similar disease trajectories for key conditions, supporting the idea that multimorbidity progression reflects shared biological and epidemiological mechanisms rather than country-specific data artifacts. These findings bolster the use of trajectory-based approaches for cross-national health surveillance and planning, with implications for harmonized prevention strategies and policy development.

Abstract

Chronic diseases frequently co-occur in patterns that are unlikely to arise by chance, a phenomenon known as multimorbidity. This growing challenge for patients and healthcare systems is amplified by demographic aging and the rising burden of chronic conditions. However, our understanding of how individuals transition from a disease-free-state to accumulating diseases as they age is limited. Recently, data-driven methods have been developed to characterize morbidity trajectories using electronic health records; however, their generalizability across healthcare settings remains largely unexplored. In this paper, we conduct a cross-country validation of a data-driven multimorbidity trajectory model using population-wide health data from Denmark and Austria. Despite considerable differences in healthcare organization, we observe a high degree of similarity in disease cluster structures. The Adjusted Rand Index (0.998) and the Normalized Mutual Information (0.88) both indicate strong alignment between the two clusterings. These findings suggest that multimorbidity trajectories are shaped by robust, shared biological and epidemiological mechanisms that transcend national healthcare contexts.

Figures (12)

• Figure 1: Workflow of the analysis presented in this article. Boxes represent data entities, arrows represent analysis steps. (a) Identification of national multimorbidity patterns. Patient-year health states were derived from national hospital registers, including all individuals with at least one in-patient hospital stay between 2003 and 2014 and no stay between 1997 and 2002. For each patient and year, binary vectors were constructed to indicate whether diagnoses had been recorded in each diagnose block, covering ICD-10 diagnoses A00–N99. Applying a hierarchical clustering algorithm to these health states identified 132 multimorbidity clusters per country. (b) Cross-country comparison. Danish patients were assigned to both Austrian and Danish clusters. Information recovery metrics were calculated, clusters were matched based on overlap, and pairwise cluster comparisons were performed to assess cross-country differences in multimorbidity patterns.
• Figure 2: Share of cohort that received a diagnosis from a specific diagnosis block within the observation period for the Austrian cohort, the Danish cohort and the age & gender adjusted Danish cohort. Only diagnosis blocks are shown where the difference between the Austrian and the Danish sample exceeded 1pp.
• Figure 3: Comparison between Austrian and Danish cluster characteristics and overlap patterns.
• Figure 4: Birthyear & gender distribution of all individuals with (1) at least one hospital diagnosis from ICD-10 blocks A00 to Z99 between 1997 and 2014 and (2) no diagnosis from A00 to N99 between 1997 and 2002
• Figure 5: 20 most frequent diagnoses in the Austrian, Danish and matched Danish cohorts. Diagnoses are only counted once per individual.