Explainable Machine Learning for Public Health Informatics in HEDIS Childhood Immunization Status Combo 10.

BACKGROUND: HEDIS Childhood Immunization Status (CIS) Combination 10 is a pediatric quality measure within a widely used health care performance framework; HEDIS is used by more than 90% of U.S. health plans, covering more than 190 million people in plans that report HEDIS quality results. Because HEDIS immunization measures support population-level monitoring and quality improvement, they are directly relevant to public health practice. However, aggregate reporting limits public health use by obscuring component-level drivers of noncompletion. OBJECTIVE: To apply explainable machine learning as a public health informatics approach to identify vaccine components associated with CIS Combo 10 completion among U.S. children aged 24-35 months. METHODS: We analyzed 2021-2023 National Immunization Survey-Child public-use files. The age-eligible cohort included 32,997 children; weighted modeling included 16,021 children. Survey-weighted logistic regression estimated national trends with 95% CIs, and Random Forest modeling with SHAP and cross-validation identified component-level predictors. RESULTS: CIS Combo 10 completion declined from 53.7% in 2021 to 44.6% in 2023; the survey-weighted regression model was statistically significant (annual OR 0.83, 95% CI 0.831-0.834; P<.001). Influenza and rotavirus had the largest mean absolute SHAP values. CONCLUSIONS: CIS Combo 10 completion declined substantially from 2021 to 2023. An explainable public health informatics approach identified influenza and rotavirus as actionable targets for immunization quality improvement.