Analyzing 11 years of US Medicaid and Medicare claims across 137,293 adults with Down syndrome, gradient-boosted machine learning models identified dementia claims, pneumonia, recurring cardiovascular disease (within three years of death), heart failure, and epilepsy as the strongest mortality predictors. Among the 30,894 deaths (22.5% of cohort), mean age at death was 55 years overall — just 59 for those with Alzheimer's disease and only 52 for those without, underscoring that Alzheimer's is not the sole driver of premature mortality in this population.
Down syndrome is characterized by trisomy 21, which confers near-universal Alzheimer's pathology risk by midlife due to triplication of the APP gene on chromosome 21 — a well-established mechanistic link. Yet this dataset, the largest of its kind, reveals a substantial fraction dying younger without Alzheimer's, pointing to undertreated cardiovascular and respiratory vulnerabilities that are potentially modifiable. The identification of pneumonia as a top predictor is particularly actionable: vaccination adherence, aspiration risk management, and respiratory monitoring could meaningfully shift outcomes. Cardiovascular surveillance beginning in the fourth decade of life appears warranted.
Critical limitations apply: this is an administrative claims-based study, meaning diagnoses reflect billing rather than clinical confirmation. Causal inference is not established. As an unreviewed preprint posted on medRxiv, these findings await peer scrutiny and should be interpreted cautiously. Still, the cohort scale and machine learning approach represent a methodological step forward — this is confirmatory yet clinically valuable work with clear preventive medicine implications.
