Health-Related Digital Engagement and Incident Stroke Among Older Adults: Prospective Cohort Study

Most studies examining internet use and health outcomes in older adults rely on cross-sectional designs and binary exposure measures, which are insufficient to capture the multidimensional nature of health-related digital engagement over time. The high collinearity between digital engagement and socioeconomic factors makes it challenging to disentangle independent effects from marker effects. Longitudinal evidence linking health-related digital engagement to incident stroke remains limited.

This study aimed to examine the longitudinal association between a composite Health-Related Digital Engagement Index (HDEI) and incident stroke among community-dwelling older adults, and to quantify the extent to which socioeconomic factors account for this association.

This prospective cohort study used data from the National Health and Aging Trends Study, Waves 1-10 (2011-2020). The HDEI (range 0-4) was constructed from 4 health-related internet behaviors assessed at baseline. The primary outcome was incident stroke ascertained by self- or proxy-reported physician diagnosis. Discrete-time hazard models with a complementary log-log link were fitted across 4 nested models, progressively adjusting for demographics, socioeconomic factors, chronic disease burden, functional disability, and social isolation.

Among 5384 participants followed for a median of 5 (IQR 2-9) years, 472 incident stroke events were observed; 81.6% (4395/5384) had an HDEI score of 0, 10.5% (566/5384) had a score of 1, and 7.9% (423/5384) had a score of 2 or higher. In the unadjusted model, each 1-point increase in HDEI was associated with a lower risk of stroke (hazard ratio [HR] 0.76, 95% CI 0.66-0.88; P<.001). After adjustment for age and sex, the association was attenuated but remained statistically significant (HR 0.84, 95% CI 0.72-0.96; P=.01). After further adjustment for race or ethnicity, education, and household income, the association was no longer statistically significant (HR 0.92, 95% CI 0.79-1.06; P=.23); fully adjusted analyses yielded similar results (HR 0.91, 95% CI 0.79-1.05; P=.21). Subgroup patterns observed in demographically adjusted analyses were attenuated after socioeconomic adjustment, and no statistically significant interaction remained in the primary model 3 framework. Sensitivity analyses showed similar patterns of attenuation.

Greater health-related digital engagement was associated with a lower risk of incident stroke in unadjusted and demographically adjusted models; however, this association was substantially attenuated and was no longer statistically significant after adjustment for socioeconomic factors. These findings are consistent with socioeconomic confounding and suggest that health-related digital engagement may, at least in part, reflect broader socioeconomic advantage among older adults. Future studies should further examine whether digital engagement has an independent role in stroke prevention beyond the socioeconomic and structural determinants of health.