Use of Large Healthcare Databases for Understudied Patient Populations

Abstract

Pregnant women with multiple sclerosis (MS) are an understudied patient population due to the difficulty of recruiting sufficient numbers of participants to evaluate relatively rare outcomes. This dissertation used administrative datasets to develop a large cohort of which the pregnancy experience in women with MS (from frequency of relapses and patterns of medication dispensing, to adverse outcomes) could be described. In Chapter One, we identified a cohort of 2,738,787 pregnancies in the Truven Health MarketScan Database (2011-2015). Algorithms to identify pregnancies were based on prior literature and iterations of patient claim reviews. In Chapter Two, we compared recording of neurologic conditions (MS and epilepsy) in the claims during delivery hospitalization and a 270-day period before delivery. Most women with the conditions had at least one code recorded at delivery (60% MS; 73% epilepsy). Having an obstetric outcome was generally not associated with recording of neurologic conditions at delivery. However epilepsy was less likely confirmed in pregnancy when it was recorded during deliveries complicated with preeclampsia, presumably due to coding of eclamptic convulsions. In Chapter Three we ascertained MS in the year before delivery and compared the risks of obstetric outcomes between those with and without an MS diagnosis. Women with MS (n= 3,875) were at an increased risk for infections in pregnancy (Risk ratio: 1.22 [95% CI: 1.16-1.27]) and preterm delivery (1.19 [1.04-1.35]). Their risks for other measured pregnancy outcomes were not increased. Finally, in Chapter Four, we assessed disease modifying therapy (DMT) in pregnancy. While 35% of the cohort was prescribed a DMT before pregnancy, dispensing dropped in the second trimester (5%) and increased again post-pregnancy (30%). We did not observe increased risks of outcomes in women exposed versus unexposed to DMTs in the pre-pregnancy interval. Altogether, the relatively low risks for adverse outcomes found in women with MS, even for those with active disease should be reassuring for this special patient population. The methodology developed here to create linked pregnancy cohorts in large administrative healthcare databases will facilitate future research in the natural history and safety of treatment for pregnant women living with chronic health conditions.