Use of multivariate joint modeling in an epidemiologic study of mortality in the presence of measurement error, correlation, and skewness
64th ISI World Statistics Congress - Ottawa, Canada
Format: CPS Abstract
Session: CPS 77 - Statistics and mortality
Wednesday 19 July 8:30 a.m. - 9:40 a.m. (Canada/Eastern)
A substantial proportion of global deaths from all causes and diseases is attributed to lifestyle risk factors such as unhealthy diets. The risk factors can either be assessed at baseline or longitudinally, and are often correlated, skewed distributed and measured with error in epidemiologic studies such as the US National Health and Nutrition Examination Survey (NHANES) linked to the National Death Index (NDI) mortality.
We applied multivariate joint modeling (MJM) approach to assess the association of four longitudinal dietary intakes (cholesterol, total fat, dietary fiber, and energy) and a set of baseline factors (sex, age, and Body Mass Index) with time-to-death due to all causes among NHANES participants accounting for measurement error, skewness, and correlation in the longitudinal dietary intakes. We subsequently applied MJM with a random intercept association structure to estimate the baseline effects of longitudinal dietary intakes and compared the results with those from the standard method that uses observed baseline values.
The MJM resulted in stronger associations than the mean method for the longitudinal dietary intake variables. With MJM, the logarithm of hazard ratio (log HR) for dietary fiber intake increased by ~ 14 times (from logHR = -0.04 to -0.60), translating into relative hazard of death of 0.55 (95% Credible Interval, CI: 0.45; 0.65) with the MJM and 0.96 (95% CI: 0.95; 0.97) with the mean method. The logHR for baseline effect estimates from the standard method were severely attenuated, even masking the significance of some variables. For instance, we estimated the relative hazard of death at baseline for dietary fiber intake as 0.79 (95%CI: 0.73; 0.85) with MJM and 0.97 (95%CI: 0.96; 0.98) with the standard method. The MJM provides an efficient modeling approach to adjust for random measurement error in correlated and skewed distributed multiple longitudinal outcomes, when estimating their associations with an event.