Nordic Multimorbidity Index, Charlson, Elixhauser and count-based comorbidity indices for mortality risk: a comparative nationwide cohort study using national health registers in Sweden
Björn Zethelius, Mats Talbäck, Rickard LjungObjectives
The aim of the present study was to compare Area Under the Receiver Operating Characteristic curves (AUROCs) of age-and-sex alone, and in addition respectively, the Nordic Multimorbidity Index (NMI), Charlson Comorbidity Index (CCI) and Elixhauser Comorbidity Index (ECI), and explore simple count-based measures on in-patient care and filled drug prescriptions, with the outcome mortality using Swedish health registries.
Study design and setting
Study population: All persons born in 1975 or earlier who were continuously resident in Sweden between 1 January 2010 and 31 December 2014 (n=5 010 261) with a follow-up for all-cause mortality from 1 January 2015 to 31 December 2019.
Methods
AUROCs were calculated with 1- to 5-year-look-back for age-and-sex, each index and explored measures for 1- to 5-year mortality.
Results
AUROC was for age-and-sex alone for 5-year mortality 0.8731. Age-and-sex alone outperformed, respectively: NMI, difference in AUROC, 0.0695 (95% CI 0.0687 to 0.0704); CCI, 0.1509 (95% CI 0.1500 to 0.1518) and ECI 0.1631 (95% CI 0.1622 to 0.1640). AUROC for NMI when added to age-and-sex, 0.9072 was marginally higher than age-and-sex alone. AUROC for explored measures as numbers of hospitalisations, days hospitalised, or numbers of filled drug prescriptions were around 0.75–0.80 and when added to age-and-sex around 0.90.
Conclusion
Age-and-sex alone showed higher AUROC than any other measure alone and NMI provides modest additional discrimination, with CCI and ECI both less helpful, that is, informative. Explored measures; days hospitalised or numbers of filled prescriptions rendered AUROC of somewhat lower magnitude when added to age-and-sex. Thus, the latter could be considered for use in addition to age-and-sex when full background data are lacking.