Efficiency of Longitudinal First-Hitting Time Models Versus Interval-Censored Survival Analysis with Application to Labor Duration Curves

• Activity Number: 318
• Type: Topic Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Alexander McLain*+ and Caroline Mulatya
• Companies: University of South Carolina and University of South Carolina
• Keywords: First-hitting time ; longitudinal data ; interval censoring ; Weiner process

Abstract

Characterizing the change in cervical dilation measurements during labor is an important obstetrics problem. For example, an estimate of the distribution of the time it takes women in labor to go from 4 to 5cm dilation can help obstetricians decide whether a stalled labor should be allowed to proceed or should be stopped in favor of other options. A common method of analyzing such data is to define the event as the time from 4 to 5cm dilation, and implement interval censored survival models. However, survival analysis ignores the underlying longitudinal process (e.g., cervical dilation), and does not use all the available data. Alternatively, implementing a longitudinal first hitting time model uses all the available data, and can still estimate all the desired quantities (i.e., the survival distribution of going from 4 to 5cm dilation). In this talk, I will discuss the trade offs of the longitudinal and interval censored approaches, and propose that sections of the longitudinal cervical measurements can be represented as a Weiner process with random drift. The proposed methodology will be illustrated with longitudinal cervical dilation data from the Consortium of Safe Labor Study.