Aging
Intimate partner violence and aging outcomes: an application of transportability using an observational, occupational cohort study Audrey R. Murchland* Audrey R. Murchland Sebastien Haneuse Karestan C. Koenen M. Maria Glymour Eleanor Hayes-Larson
Background: Few population-based aging cohorts measure intimate partner violence (IPV). Transportability methods may facilitate extrapolation of results from selected observational cohorts to US nationally representative target populations. However, there has been little use of these methods in observational settings to date.
Methods: To obtain US population average effects of IPV victimization on incident stroke or myocardial infarction events (CVD), we used data from the Nurses’ Health Study II (NHS2) and a US population survey-weighted sample, the Health and Retirement Study (HRS). The target population-average estimands included RRs and RDs of incident CVD (2001-2019), comparing individuals exposed and unexposed to IPV by 2001. We used inverse odds of selection weight (IOSW) estimators to transport sample average exposure effects, while accounting for confounding and censoring in NHS2.
Results: Compared to the weighted HRS (unweighted n=2,475; weighted n=16,998,380), NHS2 participants (n=64,615) were born earlier, more likely to be White, and higher socioeconomic status. IPV victimization was reported among 46.5% of NHS2 (42% emotional, 23% physical, and 11% sexual). If the entire NHS2 sample had been exposed versus unexposed to IPV victimization, risk of incident CVD would have been significantly higher (RR=1.23 (95% CI 1.08, 1.40) and RD=38/100,000 (95% CI: 14, 63)) over follow-up. After applying IOSW, if the entire US female target population was exposed versus unexposed to IPV victimization, magnitude of incident CVD risk would be slightly higher relative to NHS2 (RR=1.33 (95% CI: 1.05, 1.70) and RD=53/100,000 (95% CI: 10, 99)). Results were generally consistent across IPV subtypes (Figure 1) though CIs are wide.
Conclusions: Exposure to IPV victimization was associated with increased CVD risk in NHS2. The magnitude of associations generally increased when transporting estimates, but increased precision is needed for improved evaluation of differences.
