From adverse outcome prediction to e-health indicators of frailty in older populations

Prof.ssa Giovanna BOCCUZZO
Professor of Social Statistics - University of Padua

Abstract
This contribution presents a framework for constructing a health-related frailty index based on the best prediction of multiple adverse events, such as death, hospitalisation, dementia, disability, and others. The index is intended to measure the level of frailty across the entire older population of a Local Health Authority or a region, using administrative healthcare data. This objective is motivated by events that have made it particularly important to assess population frailty (COVID-19, heatwaves) and to implement stratification approaches that identify individuals at higher risk of adverse health outcomes.
Frailty is a latent and multidimensional concept and, as such, is not directly measurable. A widely adopted criterion for its assessment is to consider the main health outcomes more frequently experienced by frail individuals and, from these, to identify a core set of explanatory variables for such outcomes, which can then be used to build the frailty index.
From a methodological perspective, this process entails several challenges:
1. identifying the most representative set of health outcomes, as redundant or overly limited sets may lead to the selection of an inadequate core of variables for the index;
2.  identifying a statistical approach that yields the best subset of predictors for multiple outcomes simultaneously;
3.  identifying an appropriate index construction procedure, given that it relies on dichotomous and ordinal variables, for which conventional aggregation methods are not suitable. In this work, the use of POSET theory (Partial Order Theory) is proposed;
4. proposing an approach that results in an index based on a limited number of variables, so that it can be easily computed within Health Authorities;
5.  proposing a method that is as data-independent as possible, avoiding, for example, fixed weights to be assigned to the variables included in the index;
6.  proposing methods for comparing populations.
The contribution presents the methodological solutions adopted, the results obtained, and the research questions that remain open.

References:
 Silan M., Nicolaio M., Boccuzzo G. (2025) Profiling Frailty: A parsimonious Frailty Index from health administrative data based on POSET theory. arXiv:2506.23158 [stat.AP]. https://doi.org/10.48550/arXiv.2506.23158M.
 Silan, M. Nicolaio, E. Banzato, G. Boccuzzo. (2025). “Identifying core adverse health outcomes for frailty assessment in older adults using administrative data". Frontiers in medicine, 12, doi: 10.3389/fmed.2025.1678317.
 Silan, M., Signorin, G., Ferracin, E., Listorti E., Spadea T., Costa G., Boccuzzo G. (2022) Construction of a Frailty Indicator with Partially Ordered Sets: A Multiple-Outcome Proposal Based on Administrative Healthcare Data. Social Indicators Research, 160, 989–1017. https://doi.org/10.1007/s11205-020-02512-7.
 Silan, M., Caperna, G. & Boccuzzo, G. Quantifying Frailty in Older People at an Italian Local Health Unit: A Proposal Based on Partially Ordered Sets. Social Indicators Research, 146, 757–782 (2019). https://doi.org/10.1007/s11205-019-02142-8.

With the collaboration of Center for Applied Statistics in Business and Economics