A nonlinear model for highly unbalanced repeated time-to-event data: Application to labor progression.

The collection of repeated measurements over time on an experimental unit to study the changes over time of a certain characteristic is common in biological and clinical studies. Data of this type are also often referred to as growth curve data or repeated measures data. There arise situations when one is interested in an estimate of the time to an event, based on a characteristic that indicates progression towards the event. The assessment of the progression of labor during childbirth based on cervical dilation is one such example. Here increase in the dilation of the cervix indicates progression towards delivery. Based on how long one has been in labor and an estimate of the time to complete dilation one might make crucial decisions like the decision to administer a drug or to perform a C-section. Here a repeated measures approach is developed to model the time to the event. The parameters of the model are estimated by a maximum likelihood approach. A general model is developed for a class of data structures and a nonlinear model is developed specific to the labor progression data. Simulations are performed to assess the methodology and conditions are suggested for predicting the time to an event.