Dynamic risk prediction of infertility based on joint modeling of multivariate longitudinal processes and discrete survival time

Infertility affects approximately 16% of couples in the United States. It is estimated that roughly one-third of infertility is caused by female disorders, and one-third by combined male and female disorders. Motivated by concerns of identifying couples' at-risk for infertility, our goal has been to build a dynamic risk predictor of infertility based on biological processes, example menstrual cycle length and behavior, intercourse pattern of couples and other well-known risk factors to develop personalized risk prediction for couples. Here, we will focus on joint analysis and prediction of multivariate longitudinal processes, like menstrual cycle lengths (a skewed longitudinal process), intercourse pattern (a binary longitudinal process), cycle-specific peaks of reproductive hormonal profiles within woman with time-to-pregnancy (a discrete survival time). We will present a rigorous formulation for the joint modeling of these processes with time to event under a shared parameter framework. Our proposed approach will be investigated through simulations and application to multiple real data studies including Oxford Conception Study and the LIFE Study.

Auxiliary data

Longitudinal modeling

Multivariate longitudinal data

Survival modeling