MLCI Outperforms Charlson and Elixhauser Scores Across Multiple EHR Outcomes

A Machine-Learned Comorbidity Index (MLCI) uses normalized Hilbert-Schmidt Independence Criterion to map ICD diagnosis codes to a single scalar score, capturing nonlinear risk relationships that linear rule-based indices like Charlson miss. Benchmarks on multiple EHR datasets show MLCI outperforms existing baselines across several clinical outcome metrics.