Innovative Statistical Approaches to Mitigate Data Challenges in Women's and Maternal Health

The Childhood Cancer Survivor Study (CCSS) is an enriching resource for survivorship research. It provides valuable information on the incidence and prevalence of various health conditions and the ages at their first occurrence. However, the data may exhibit partial incompleteness due to survivors' challenges in recalling the exact onset age, leading to interval censoring. A complicated cancer treatment received, which potentially affects their memory, together with the evident fact of memory fading with time, makes the censoring informative. We propose a model where the probability of not recalling the onset age is influenced by the time elapsed since the onset of GHD and radiotherapy to the brain. This talk proposes a Cox regression model. The model introduces a novel semi-parametric inference for estimating regression parameters and such data's baseline hazard (or distribution). To demonstrate the method's applicability, we apply it to the analysis of the time to deficiency of growth hormones using data from the CCSS. 

Accurate assessment of gestational age at birth is necessary for optimal pediatric care. In high resource countries, several methods using ultrasound have been proposed to assess gestational age at birth; however, these methods are not easily accessible for low-resource populations. We develop a shared random parameter model for estimating gestational age at birth from longitudinal maternal anthropometry that incorporates additional maternal information from the last menstrual period, a measure of gestational age collected with sizable measurement error. The proposed methodology is evaluated using simulation studies under a training-test set paradigm. In addition, we propose methodology to validate prediction when some measurements of the gold standard are collected with measurement error. We illustrate the proposed methodologies with data from the NICHD Fetal Growth Studies. 

The risk of preterm birth increases with maternal age, and it is possible that hypertensive disorders, like gestational hypertension, mediate this maternal age-preterm birth relationship. Previous studies, however, have found low diagnostic accuracy of gestational hypertension. Thus, any mediation analysis conducted with this potentially misclassified binary mediator variable may be severely biased. This bias is especially challenging to address when the misclassification is covariate-dependent and when no gold standard measures are available. In this study, we develop methods to handle misclassification in the gestational hypertension mediator variable by modelling misdiagnosis based on patient-level factors. We present an expectation-maximization algorithm to estimate the model and provide an R package to implement the proposed methods. Using these methods, we assess the misclassification-corrected effect of maternal age on preterm birth risk, while simultaneously estimating misclassification rates of gestational hypertension. 

Maternal mortality is of national priority. This is resulting in increased attention to labor progression, resulting in sometimes unnecessary medical intervention during spontaneous labor.
Thus, guidelines for obstetrical care are getting much closer attention. Motivated by these issues, we focus on data-driven approach for women who will deliver naturally outside of the clinical guidelines. Spontaneous labor in women is a muti-state process, with the duration of first stage of labor denoting the time to full cervical dilation of 10 centimeters and the second stage of labor denoting the time to pushing the fetus out. The first stage of labor is a length-biased data due to start of dilation not known to the woman and the second stage of labor is a competing risks event type as the woman either delivers spontaneously or through various medical interventions, like cesarean section. Due to considerable increase in the C-section rates world wide, it is of considerable interest in identifying women at risk for C-section delivery. We propose a shared parameter model, where we predict the duration of second stage of labor based on her first stage of labor. This will help identify women who are at risk for prolonged second stage of labor and at higher risk for medical intervention. The proposed method will be assessed through simulation studies and a full analysis of NICHD Consortium of Safe Labor will be presented.