An ensemble clustering typology of long-term care policies: evidence from 30 OECD countries

Population ageing is pushing long-term care (LTC) to the centre of welfare-state reform debates, yet countries differ markedly in how they finance, regulate, and supply LTC. Existing typologies are largely qualitative, suffer from limited reproducibility and inconsistent country coverage, and constrain cross-national policy learning.

To construct an empirically grounded, reproducible typology of LTC systems across 30 OECD countries and to evaluate its stability and limits through systematic validation.

We use a clustering ensemble combining fuzzy c-means (FCM), self-organising maps (SOM), and partitioning around medoids (PAM). The classification draws on ten indicators spanning four dimensions: supply, demand, financing, and access, regulation and outcomes. We validate the typology through internal cluster indices, descriptive cluster profiling, and sensitivity analyses, including random subsampling.

We identify seven LTC system types. Three broadly recover established welfare-state groupings: a high-supply, publicly financed Nordic type; a low-supply, family-reliance Eastern European type; and a Southern European familialism type. Four reveal finer distinctions or substantive departures from institutional classifications: Sweden and the United States cluster together along outcome-based dimensions, the United Kingdom aligns with continental Europe rather than the Anglo-Saxon liberal regime, and two further types capture mixed or transitional systems. The seven-cluster solution is stable across algorithmic parameters and random subsampling.

The typology offers a transparent and reproducible benchmark for cross-country comparison and a tool for evidence-based policy learning. By identifying countries that converge on outcomes despite divergent institutional characteristics, it highlights peer comparators relevant to LTC reform under demographic and fiscal pressure.