Data from the 2009 Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID) were used to examine variation in discharge rates for primary cesarean delivery by payer [19]. HCUP is a family of health care databases and related software tools and products developed through a federal-state-industry partnership and sponsored by the Agency for Healthcare Research and Quality (AHRQ). HCUP databases bring together the data collection efforts of state data organizations, hospital associations, private data organizations, and the federal government to create a national information resource of patient-level health care data. HCUP includes the largest collection of longitudinal hospital care data in the United States, with all-payer, encounter-level information beginning in 1988. The HCUP SID contain the universe of inpatient discharge abstracts from participating states, translated into a uniform format to facilitate multi-state and local market comparisons and analyses. All investigators signed a Data Use Agreement; because HCUP does not involve human subjects, IRB approval was not required for this study.
The following 44 states were included in the analysis: Arizona, Arkansas, California, Colorado, Connecticut, Florida, Georgia, Hawaii, Illinois, Indiana, Iowa, Kansas, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Minnesota, Missouri, Montana, Nebraska, Nevada, New Hampshire, New Jersey, New Mexico, New York, North Carolina, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Utah, Vermont, Virginia, Washington, West Virginia, Wisconsin, and Wyoming. HCUP 2009 data was not available for the following 6 states: Alabama, Alaska, Delaware, Idaho, Mississippi and North Dakota.
We aggregated HCUP inpatient data from community, acute care hospitals to the Core-Based Statistical Area (CBSA) level using patient ZIP Code. CBSAs are the universe of U.S. metropolitan statistical areas and micropolitan areas [20] and are a readily available, transparent unit of analysis with established use in variation studies [21],[22].
Our analytic file included 804 CBSAs (representing 86.5 percent of CBSAs in the United States). We obtained characteristics on population size, education, income, and race/ethnicity from the U.S. Census Bureau at the CBSA level. Data on physician and hospital resources, including the total number of primary care physicians, obstetric and gynecologic physicians, physician assistants, and midwives per capita were obtained from the Area Health Resource Files. Information on hospital type and beds per capita was obtained from the American Hospital Association. Average malpractice payment data were from the National Practitioner Data Bank.
We used the AHRQ Inpatient Quality Indicator (IQI) definition of primary cesarean rate to define the population studied. IQIs reflect procedures whose use varies significantly across hospitals or geographic areas and include measures of utilization of procedures for which there are questions of overuse, underuse, or misuse [23]. High rates of these indicators may suggest inappropriate or inefficient delivery of care.
Primary cesarean delivery rate was defined as the number of cesarean deliveries, identified by diagnosis-related group (DRG) (370–371), Medicare severity diagnosis-related group (MS-DRG) (765–766), and International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedure codes (740, 741, 742, 744, 7499) without a hysterectomy procedure code (7491) per 1,000 deliveries. Deliveries were defined by delivery DRG (370–375) and MS-DRG (765–768; 774–775), and excluded deliveries with the following: any diagnosis of abnormal presentation, preterm birth, fetal death, or multiple gestation diagnosis codes; any breech procedure codes; and previous cesarean delivery diagnosis in any diagnosis field.
Payer was based on the expected payer as indicated in the discharge record. Medicaid includes fee-for-service and managed care Medicaid patients. Patients covered by the State Children's Health Insurance Program (SCHIP) may be included. Because most state data do not identify SCHIP patients specifically, it is not possible to present this information separately. Private insurance includes Blue Cross, commercial carriers, and private HMOs and preferred provider organizations (PPOs). One CBSA with a higher than expected percentage of births with Medicare as the primary expected payer was excluded.
First, we calculated unadjusted and risk-adjusted CBSA cesarean delivery rates by payer. Risk-adjusted rates were calculated as observed cesarean delivery rate divided by expected cesarean delivery rate, multiplied by the overall CBSA average cesarean delivery rate. The expected delivery rate was estimated using a hierarchical logistic model where the outcome was type of delivery (1 = cesarean; 0 = vaginal) and CBSA was included as the second level. Specifically, we used the SAS (SAS Institute, Inc; Cary, NC) GLIMMIX procedure that fits statistical models to data with correlations or nonconstant variability, where the outcome may not be normally distributed [24]. Because type of delivery was specified as a dichotomous outcome, we specified a logit link and binomial distribution.
In the model, we adjusted for maternal and neonatal characteristics associated with an increased risk of cesarean delivery. These factors included maternal age and race, primary expected payer (in the all-payer model), infant birth weight, and maternal conditions including maternal distress, placenta previa, hypertension, pre-eclampsia, pre-existing or gestational diabetes, herpes, HIV, and prior myomectomy. We included race in the risk adjustment model despite the absence of clinical evidence that race should affect cesarean rates because previous research has found wide variability in the rate of indications for primary cesarean section by race/ethnicity [25],[26]. We included primary expected payer in the all-payer model because we expected rates to differ by payer and did not want these differences to confound the overall rate calculation. Finally, we included the set of maternal conditions in the risk adjustment model as these clinical factors may be considered medical indications for cesareans [16],[27]-[32]. We look specifically at the influence of all of these factors (race, payer, maternal conditions) in the second part of our study. We calculated the correlation in cesarean delivery rate among Medicaid and private insurance using Pearson's correlation weighted for population size.
Second, we measured the predictors of having a cesarean delivery for each hospitalization. We estimated models with all hospitalizations together (regardless of payer) and separately by the primary payer for the hospitalization. We included patient-level and CBSA-level predictors to evaluate factors associated with cesarean delivery. Variables included patient-level measures (detailed above), as well as population measures (e.g., race, income, education) and market measures (e.g., hospital market share, bed size, teaching status, provider density, average malpractice payments, and HMO enrollment). Continuous variables were centered at population means. We excluded 46 CBSAs for which complete patient, population, and market data were unavailable.