Hospital variation in cesarean delivery rates: contribution of individual and hospital factors in Florida

Abstract

Primary cesarean deliveries are a major contributor to the large increase in cesarean delivery rates in the United States over the past 2 decades and are an essential focus for the reduction of related morbidity and costs. Studies have shown that primary cesarean delivery rates among low-risk women in the United States vary 3-fold across hospitals and are not explained by differences in patient case-mix. However, the extent to which maternal vs hospital characteristics contribute to this variation remains poorly understood because previous studies were limited in scope and did not assess the influence of factors such as maternal ethnicity subgroups or prepregnancy obesity. We assessed the contribution of individual- and hospital-level risk factors to the hospital variation in primary cesarean delivery rates among low-risk women in Florida. Our population-based retrospective cohort study used Florida's linked birth certificate and hospital discharge records for the period of 2004-2011. The study population was comprised of 412,192 nulliparous, singleton, vertex, live births with labor at 37-40 weeks gestation in 122 nonmilitary delivery hospitals. Data were analyzed with logistic mixed-effects regression with cesarean delivery as the outcome. This approach provided adjusted risk estimates at an individual and hospital level and the estimated percent of hospital variation statewide that was explained by these factors. The primary cesarean delivery rate in the study population was 23.9%, with hospital-specific estimates that ranged from 12.8-47.3%. Leading risk factors for cesarean delivery were maternal age ≥35 years (adjusted relative risk, 2.22), prepregnancy obesity (body mass index, ≥30 kg/m(2); adjusted relative risk, 1.73), medical risk conditions (adjusted relative risk, 1.72), labor induction (adjusted relative risk, 1.52), and delivery in hospitals located in Miami-Dade County (adjusted relative risk, 1.73). Hospital geographic location was a significant effect modifier for prepregnancy obesity, medical conditions, and labor induction (P < .05), with a tendency towards lower adjusted relative risks for these factors in Miami-Dade County relative to other Florida regions. Conversely, Miami-Dade County had an increased prevalence of higher-risk ethnic subgroups, such as Cuban or Puerto Rican mothers, and also substantially higher adjusted relative risks that were associated with practice-related factors, such as delivery during weekday hours. Whereas hospital geographic location contributed to 39.6% of the observed variation statewide, the estimated contribution of maternal ethnicity ranged from 1.6-15.7% among Florida regions. Hospital geographic location contributes to hospital variation in primary cesarean delivery rates among low-risk women in Florida. In contrast to previous studies, our findings suggest that individual level risk factors such as maternal ethnicity also contribute to some of this variation, with differing extent by region. These individual factors likely interact with practice factors and add to the variation. This study was limited by not including maternal Bishop score before induction or obstetrics provider in the analysis. These were not available on the dataset but likely contribute to the variation. Our findings suggest potential issues to consider in quality improvement efforts, such as the need for future qualitative research that focuses on mothers in higher-risk ethnic subgroups and providers in high-rate hospitals, particularly those in Miami-Dade County. These studies may help to identify potential cultural differencesin maternal beliefs and expectations for delivery and maternal reasons for differences in obstetrics practices.