Adapting a probabilistic redistribution approach for Ill-defined deaths in Slovenia: implications for national burden of disease estimates

Abstract Introduction Ill-defined deaths (IDDs) reduce the usefulness of mortality statistics for public health and burden of disease analyses. In Slovenia, they account for a substantial share of registered deaths, but no redistribution of IDDs has been performed so far and its potential impact on cause-specific mortality patterns remains unknown. This study aimed to apply a probabilistic redistribution approach to Slovenian mortality data and assess how redistribution affects the distribution and ranking of causes of death (CoD). Methods The primary analysis included all registered deaths in Slovenia in 2021 ( n = 23,261). A sensitivity analysis was additionally performed for 2019, the last full pre-pandemic year. Underlying CoD coded in ICD-10 were mapped to the Global Burden of Disease (GBD) cause hierarchy list. IDDs were redistributed using a modified probabilistic approach based on the Belgium Burden of Disease (BoD) study, including ICD-based redistribution, package redistribution, and all-cause redistribution. Redistribution was performed at GBD Level 4 and presented at Levels 1–3. Probabilities were estimated using Slovenian mortality data and Belgian multiple cause-of-death data from the corresponding five-year reference periods. Uncertainty was assessed across 100 probabilistic iterations. Results In Slovenia, 15.9% (3,701) of all deaths were classified as IDDs and were redistributed to specific causes. The largest share was reassigned through the package (50.5%), followed by ICD-based (31.0%) and all-cause redistribution (18.5%). Redistribution primarily affected the magnitude of specific causes, while the ranking of leading CoD remained largely unchanged across GBD levels. The greatest increase was observed for non-communicable diseases at GBD Level 1. At Level 3, stroke, ischaemic heart disease, and several chronic conditions showed notable increases after redistribution. The contribution of redistribution varied across causes, with ICD-based reassignment dominating for some causes and package-based redistribution contributing substantially to others. A sensitivity analysis for 2019 showed a very similar distribution across the three redistribution steps and stable ranking of the leading causes. Conclusions A probabilistic redistribution approach was successfully adapted to Slovenian mortality data and enabled more specific attribution of cause-specific mortality. While redistribution had little effect on the ranking of leading causes, it substantially affected the magnitude of several causes and supported the inclusion of IDDs in national burden of disease assessments. These findings are particularly relevant for years of life lost estimation, where redistribution directly influences the quantified burden of specific causes.