Abstract:
Background: Background: Observational comparisons of dementia-related outcomes between neurodivergent and non-neurodivergent populations are prone to strong confounding by age. We compared risks before and after age-based propensity score matching using a global federated electronic medical record (EMR) dataset.
Methods: Methods: We conducted a Compare Outcomes analysis on TriNetX, including adult females with (cohort B: ICD-10 F90/F90-F98/F84.0) versus without neurodivergence (cohort A). The composite outcome included dementia diagnoses (F01-F03, G30-G31.0) and speech assessment (UMLS F00) ≥180 days after index. Analyses were run before and after age-based propensity score matching. Kaplan-Meier survival and risk measures were extracted from both reports.
Results: Results: Unadjusted analysis showed a higher absolute risk among non-neurodivergent women (RR 1.55, 95% CI 1.39-1.73, HR 1.85, 95% CI 1.66-2.06, p<0.001). After age matching (n=1,211,916 per cohort), the association reversed, revealing a higher risk in neurodivergent women (RR 0.266 for A/B, 95% CI 0.21-0.34; HR 0.353, 95% CI 0.28-0.45, p<0.001). This suggests a 3.8-fold increased relative risk in neurodivergent women after adjusting just for age. Mean follow-up was longer for neurodivergent women (1,620 vs. 1,258 days).
Conclusion: Conclusions: The apparent protective association of neurodivergence was fully explained by age confounding. After age control, neurodivergent women demonstrated a significantly higher rate of dementia-related outcomes. Further studies incorporating psychiatric and vascular covariates are warranted to clarify this association.