Effect of severity of illness on cesarean delivery rates in Washington State

Abstract

Hospitals and providers are increasingly held accountable for their cesarean delivery rates. In the perinatal quality improvement arena, there is vigorous debate about whether all hospitals can be held to the same benchmark for an acceptable cesarean rate regardless of patient acuity. However, the causes of variation in hospital cesarean delivery rates are not well understood. We sought to evaluate the association and temporal trends between severity of illness at admission and the primary term singleton vertex cesarean delivery rate among hospitals in Washington State. We hypothesized that hospitals with higher patient acuity would have higher cesarean delivery rates and that this pattern would persist over time. In this cross-sectional analysis, we analyzed aggregate hospital-level data for all nonmilitary hospitals in Washington State with ≥100 deliveries/y during federal fiscal years 2010 through 2014 (287,031 deliveries). Data were obtained from the Washington State Comprehensive Hospital Abstract Reporting System, which includes inpatient demographic, diagnosis, procedure, and discharge information derived from hospital billing systems. Age, admission diagnoses and procedure codes were converted to patient-level admission severity-of-illness scores using the All Patient Refined Diagnosis Related Groups classification system. This system is widely used throughout the United States to adjust hospital data for severity of illness. Mean admission hospital-level severity-of-illness scores were calculated for each fiscal year among the term singleton vertex population with no history of cesarean delivery. We used linear regression to evaluate the association between hospital admission severity of illness and the primary term singleton vertex cesarean delivery rate, calculated Pearson correlation coefficients, and compared regression line slopes and 95% confidence intervals for each fiscal year. Hospitals were diverse with respect to delivery volume, level of care, and geographic location within Washington. Hospital aggregate admission severity-of-illness score correlated with primary term singleton vertex cesarean delivery rate in all fiscal years (R2 0.38-0.58, P < .001). For every year in the study interval, as admission severity of illness increased so did the primary term singleton vertex cesarean rate. The slope of the regression line decreased during the study interval, suggesting that statewide decrease in primary term singleton vertex cesarean rate occurred across the range of severity of illness. Admission severity-of-illness score is strongly associated with the primary term singleton vertex cesarean delivery rate among hospitals in Washington State. Approximately 50% of variation in hospital primary term singleton vertex cesarean delivery rates appeared to be related to admission severity of illness. This relationship persisted over time despite a statewide decrease in cesarean delivery, suggesting that patient acuity will likely continue to contribute to hospital variation in cesarean delivery rates despite perinatal quality improvement efforts. The major implication of this study is that patient acuity should be considered when determining optimal cesarean delivery rates. High-acuity hospitals are likely to have high cesarean rates because they provide a specific role in serving regional needs. To hold these centers to an arbitrary benchmark may jeopardize the funding necessary to support regional safety net institutions.