Internet use and incident myocardial infarction in community-dwelling older adults: a 13-year longitudinal cohort study

Internet use is increasingly common among older adults, but its relationship with incident myocardial infarction (MI) is uncertain, and internet use is strongly patterned by socioeconomic position. We examined the association between internet use and incident MI over 13 years, with particular attention to socioeconomic confounding. We used the National Health and Aging Trends Study (NHATS, Rounds 1–14, 2011–2024). Internet use at baseline was the exposure and incident self-reported MI during follow-up the outcome. Covariates were selected from a directed acyclic graph: the primary model adjusted for sex, age group, education, race or ethnicity, household income, Medicaid coverage, marital status, five comorbidities and smoking history, while variables considered to lie on the causal pathway were examined only in an exploratory model. Sensitivity analyses included inverse probability of treatment weighting, latency analyses, E-values, and replication in a later-baseline (2015) cohort that is not independent of the 2011 sample. Among 4,120 participants free of MI at baseline (429 incident events; 26,121 person-years; median follow-up 5 years), incidence was 12.8 per 1,000 person-years in internet users and 20.4 in non-users. Users and non-users differed markedly in socioeconomic position (median household income $50,000 vs. $22,000; Medicaid coverage 4.1% vs. 18.8%). The unadjusted hazard ratio (HR) was 0.64 (95% CI 0.53–0.78); demographic adjustment attenuated this to 0.76 (95% CI 0.61–0.95) and socioeconomic adjustment attenuated it further, to 0.80 (95% CI 0.63–1.00). In the primary model the HR was 0.78 (95% CI 0.62–0.99; P = .04), and was similar under inverse probability of treatment weighting (0.76, 95% CI 0.60–0.96) and in the 2015 cohort (0.80, 95% CI 0.62–1.02), despite internet use prevalence rising from 45% to 52%. No subgroup interaction was significant. E-values were 1.87 for the point estimate and 1.14 for its upper confidence limit. Baseline internet use was associated with a modestly lower rate of incident MI, an association substantially attenuated by adjustment for socioeconomic position. Given the low E-value, residual confounding by unmeasured dimensions of socioeconomic position remains a plausible explanation. These findings are hypothesis-generating and do not establish that internet use lowers cardiovascular risk.