The relationship between variations in cesarean delivery and regional health care use in the United States

Abstract

Cesarean delivery rates vary widely across the United States. Health care usage in many other areas of medicine also varies widely across the United States; it is unknown whether the variation in cesarean delivery rates across US communities is correlated with this broader underlying variation in health care usage patterns. The purpose of this study was to determine whether the variation in cesarean delivery rates across US communities is correlated with other measures of health care usage in that community. We performed a population-based observational study that combined multiple national data sources, which included 2010 birth certificate data and Medicare claims data. Cesarean delivery rates in each US community, as defined by the Hospital Service Area, Medicare total spending per beneficiary, and hospital days in the last 6 months were calculated. Cesarean delivery and Medicare spending were on different patient populations; the Medicare variables were used to characterize the broader health care usage and spending pattern of that community. We examined the relationship between a community's cesarean delivery rates and these measures of health care usage using Pearson correlation coefficients. We also stratified by quartile of Medicare spending and hospital use in the last 6 months of life and calculated the cesarean delivery rates per quartile, adjusting for underlying differences in patient characteristics, demographics, hospital structure, and the malpractice environment using a least-squared means method. We compared the amount of variation in cesarean delivery rates across communities that could be explained by differences in health care usage patterns to the amount of variation that was explained by other factors using the R-squared from multivariable models. Cesarean delivery rates varied from 4-65% across communities in the United States. Cesarean delivery rates were correlated positively with total Medicare spending (r= 0.48; P < .001) and hospital use in the last 6 months of life (r= 0.45; P < .001). Similar variation was seen in nulliparous women with a term fetus in vertex presentation (nulliparous, term, singleton, vertex cesarean deliveries), which is a common subset used for analysis of cesarean delivery rates. Communities in the lowest quartile of Medicare spending had the lowest rates of cesarean delivery (29.1% vs 35.7% in the highest quartile; P < .001 for differences across quartiles), which is a difference that persisted after adjustment (29.5% vs 31.8%; P < .001). Similar results were seen for nulliparous, term, singleton, vertex cesarean deliveries and when data were stratified by hospital days in the last 6 months of life. Overall, 28.6% of the total variation in cesarean delivery rates was explained by differences in health care usage patterns, as compared with 16.6% by differences in obstetric procedures, 7.9% by hospital structure, and 2.3% by variations in the malpractice environment. Of the 56.3% of variation that was unexplained by differences in patient characteristics and area demographics, 8.2% could be accounted for by differences in health careusage patterns, as compared with 4.6% by differences in obstetric procedures, 2.1% by hospital structure, and 1.2% by variation in the malpractice environment. Cesarean delivery rates vary widely across US communities; this variation is correlated broadly with the variation that is seen in other measures of health care usage across US communities.