New R Package Untangles How Diseases Shape Healthy Life Expectancy

Researchers have introduced an open-source R package, shapleyHALE, that decomposes changes in health-adjusted life expectancy (HALE) without the ordering biases that afflict conventional decomposition methods. The software, described in BMC Medical Research Methodology, treats each disease as a player in a cooperative game and applies the Shapley value from game theory to attribute shares of HALE change to individual causes of death and disability.

Conventional decompositions of HALE change assume that diseases act independently and produce results that depend on the order in which causes are evaluated. They also cannot quantify whether the combined effect of two causes is larger or smaller than the sum of their separate effects. The Shapley value solves both problems: it provides an ordering-free, additive attribution with no residual, and it can quantify population-level complementarities among causes. According to the authors, its adoption in population health had been hindered by computational complexity and the absence of reusable software.

The package works by averaging each disease’s marginal contribution over 5,000 randomly sampled cause orderings by default. It separates mortality-driven from disability-driven components, so users can see whether a cause reduced healthy life expectancy mainly through deaths or through years lived with illness. Second-order Shapley interaction indices are computed to quantify pairwise synergies and antagonisms between causes. A parametric bootstrap with 1,000 iterations propagates uncertainty in the input data to all outputs by default. The package accepts standard age-cause matrices of mortality and morbidity rates and is compatible with any population-based surveillance data.

To demonstrate the tool, the team applied it to Global Burden of Disease Study 2023 data covering 22 causes across five Socio-demographic Index (SDI) quintiles over the period 1990 to 2023. The decomposition ranked cause contributions to HALE change, separated mortality from disability effects, and quantified pairwise interactions among the causes.

One striking result concerns mental disorders. They appeared among the 10 leading causes of HALE change in the Low, Middle, and High SDI groups, and were a negative contributor in each. The estimated contributions were −0.336 years (95% uncertainty interval −0.482 to −0.252) in the Low SDI group, −0.295 years (−0.440 to −0.213) in the Middle SDI group, and −0.342 years (−0.518 to −0.234) in the High SDI group.

The separation of mortality and disability effects showed that mental disorders, along with diabetes and kidney diseases, contributed negatively primarily through disability rather than through premature death.

The interaction analysis revealed predominantly positive complementarities among the major causes, meaning that for most pairs the combined contribution exceeded what would be expected from their separate effects. Mental disorders, however, exhibited systematic antagonism with other major causes in every group where they ranked among the top 10, indicating that their combined effects with other diseases were smaller than the sum of the individual contributions.

The authors conclude that shapleyHALE provides an open-source, ordering-free decomposition tool for HALE with population-level interaction indices, offering a standardised and reproducible platform to generate evidence for integrated, multi-disease health strategies in an era of rising multimorbidity. Because the software accepts any population-based surveillance data, its use is not limited to global burden of disease estimates. The article is open access, and the version shared carries a permanent DOI; it is citable while remaining subject to further editorial edits before the final version of record replaces it.

Subject of Research: An R package for ordering-free decomposition of health-adjusted life expectancy change with disease interaction indices

Article Title: ShapleyHALE: an R package for ordering-free decomposition of health-adjusted life expectancy change with population-level interaction indices, demonstrated using Global Burden of Disease Study data

Article References: Li, X.-Q., Hu, W., Yan, B., Sun, J., & Xi, J.-Y. (2026). ShapleyHALE: an R package for ordering-free decomposition of health-adjusted life expectancy change with population-level interaction indices, demonstrated using Global Burden of Disease Study data. BMC Medical Research Methodology. https://doi.org/10.1186/s12874-026-03061-2

Image Credits: AI Generated

DOI: 10.1186/s12874-026-03061-2

Keywords: Shapley value, health-adjusted life expectancy, R package, Global Burden of Disease Study, decomposition methods, mental disorders, disability, multimorbidity, population health metrics, cooperative game theory, uncertainty quantification, open-source software