We described differences in baseline demographics between our analytic sample and excluded individuals, and by hearing group using χ2 (categorical variables) and Kruskal–Wallis analysis of variance (continuous variables). Given the discrete nature of our data we estimated a discrete-time proportional hazard model using a cloglog model to estimate hazard ratios (HRs) (22 ) and 95% confidence intervals (CIs) for losing USOC. We verified the proportional hazard assumption by looking at the correlation between Schoenfeld residuals and survival times.
We performed unweighted and weighted (using 2011 enrollment weights) estimations to account for NHATS’ complex survey design. All main analyses were adjusted for participant’s baseline characteristics including: age, sex, race/ethnicity, marital status, education, household income, self-reported health, number of chronic conditions, dementia status, additional health coverage, and depression. All covariates were treated as time-invariant in our analyses. Survival functions by hearing group were estimated for the unweighted unadjusted model.
In secondary analyses, we explored the potential moderation effect of (i) depression and (ii) transportation barriers on the association between hearing groups and self-reported loss of USOC. These secondary analyses were driven by previous findings showing the importance of unmet transportation needs and depression as risk factors for loss of USOC. For the case of analyses pertaining to transportation barriers, a sample of N = 320 was used due to missing data.
Finally, as a sensitivity analysis we estimated our main model including all study participants who satisfied our inclusion criteria (community-dwelling, having a USOC, full set of covariates), but who were lost to follow-up during the study period. As participants in residential care facilities were assumed to have a USOC, we excluded at-risk participants who transitioned into residential care from this analysis. All our estimations were performed using Stata/SE 17.0.
We performed unweighted and weighted (using 2011 enrollment weights) estimations to account for NHATS’ complex survey design. All main analyses were adjusted for participant’s baseline characteristics including: age, sex, race/ethnicity, marital status, education, household income, self-reported health, number of chronic conditions, dementia status, additional health coverage, and depression. All covariates were treated as time-invariant in our analyses. Survival functions by hearing group were estimated for the unweighted unadjusted model.
In secondary analyses, we explored the potential moderation effect of (i) depression and (ii) transportation barriers on the association between hearing groups and self-reported loss of USOC. These secondary analyses were driven by previous findings showing the importance of unmet transportation needs and depression as risk factors for loss of USOC. For the case of analyses pertaining to transportation barriers, a sample of N = 320 was used due to missing data.
Finally, as a sensitivity analysis we estimated our main model including all study participants who satisfied our inclusion criteria (community-dwelling, having a USOC, full set of covariates), but who were lost to follow-up during the study period. As participants in residential care facilities were assumed to have a USOC, we excluded at-risk participants who transitioned into residential care from this analysis. All our estimations were performed using Stata/SE 17.0.
Free full text: Click here
