Racial Differences in the Use of Epidural Analgesia for Labor

This study is based on data from the New York State Perinatal Database and includes all inpatient admissions for birth in the Finger Lakes region of New York State between 1998 and 2003 (83,376 patients). “The Finger Lakes region of New York State includes Rochester and extends south to the Pennsylvania border, as well as east and west to approximately halfway between Rochester and Syracuse and Buffalo, respectively.”52This data set has been previously described.52Briefly, trained data collectors recorded and submitted data (using a common software platform) on demographic information, medical and obstetric conditions, and maternal and perinatal outcomes to a central repository overseen and funded by the New York State Department of Health. Data are subjected to data edit checks by the New York State Department of Health.53“There are regular meetings with hospital coders and representatives to minimize variability in the data gathering process and to standardize field definitions, as directed by the New York State Department of Health.”52There is however, no formal data auditing mechanism in place. A data reabstraction study in which birth certificate data from New York State was compared with medical records data showed that “most of the information provided on birth certificates is accurate.”53This data set included encrypted patient, provider, and hospital identifiers. We excluded 1.8% of the patients from the analysis because of missing data. This study was exempted from institutional review board approval because it involved the secondary use of preexisting data. In the study sample, 90.5% of the patients received obstetric care from physicians and 5.5% received care from certified nurse midwives (4.0% were missing).

Because the unit of analysis was the mother, a single pregnancy with multiple gestations was counted as one observation. The association between baseline characteristics and race/ethnicity were examined using analysis of variance and chi-square tests, as appropriate. In the exploratory analysis, the association between use of epidural analgesia and age, insurance, race, and patient characteristics were explored using bivariate logistic regression analysis.

To examine whether race and insurance status were associated with the likelihood of receiving epidural analgesia after adjusting for patient characteristics, we constructed a fixed effects logistic regression model that included provider identifiers as fixed effects. Fixed effects modeling54was used to adjust for the impact of differences in the use of epidural analgesia between obstetricians. We chose not to use a random effects model because of the possibility that patient risk factors would be correlated with provider effect, which could lead to inconsistent estimates of the model coefficients in a random effects model.54To account for the fact that outcomes for patients treated by the same obstetrician may be correlated, we used robust variance estimators.55The method of fractional polynomials56was used to test the linearity of the log odds of the outcome as a function of continuous variables and to help select optimal cutoffs when continuous variables were transformed to categorical variables.55,57We estimated a separate logistic model to test the interaction between patient race/ethnicity and insurance status. This analysis was repeated after excluding patients treated by midwives and patients whose provider status was unknown.

Model discrimination was evaluated using the C statistic. All statistical tests were two-tailed, and P  values less than 0.05 were considered significant. Data management, logistic regression models, and regression diagnostics were performed using STATA SE/9.1 (STATA Corp., College Station, TX).

The study sample consisted of 81,883 patients treated by 405 obstetricians at 17 hospitals. Patient characteristics are described in tables 1 and 2; 78.6% of the patients were self-identified as white/non-Hispanics, 4.4% as white/Hispanics, 13.9% as black, 0.5% as Asian, 0.27% as American Indian, 1.45% as other/non-Hispanic, and 0.8% as other/Hispanic. Overall, 38.3% of the patients received epidural analgesia for labor.

There were significant differences in clinical and demographic characteristics across racial and ethnic groups. White/non-Hispanic patients were somewhat older and had lower parity (P < 0.001). There were significant differences in insurance coverage, with white/non-Hispanic patients being more likely either to have private insurance or to belong to a health maintenance organization (P < 0.001). Nonwhite patients, on average, completed fewer years of schooling than white/non-Hispanic patients (P < 0.001).

Bivariate and multivariate odds ratios for clinical risk factors and socioeconomic variables are shown in table 2. In the unadjusted analyses, white/Hispanic, black, and American Indians were less likely to receive epidural analgesia than white/non-Hispanic patients. After adjusting for clinical risk factors, socioeconomic status, and provider effects, white/Hispanic and black patients remained less likely to receive epidural analgesia: The adjusted odds ratios (adj ORs) were 0.85 (95% confidence interval [CI], 0.78–0.93) for white/Hispanics and 0.78 (0.74–0.83) for blacks relative to non-Hispanic whites. Other significant predictors of epidural use in the multivariate model included insurance status, education level, multiple gestation, and parity. Compared with patients with private insurance, patients with health maintenance organization coverage (adj OR, 0.92; 95% CI, 0.87–0.97), Medicaid (adj OR, 0.87, 95% CI, 0.82–0.93), and without insurance (adj OR, 0.76; 95% CI, 0.64–0.89) were less likely to receive epidural analgesia. In general, patients who had completed fewer years of schooling were also less likely to have an epidural for labor: The adj OR for patients who had not completed the eighth grade was 0.72 (95% CI, 0.63–0.83). Multiple gestation was associated with greater use of epidural analgesia (adj OR, 1.7; 95% CI, 1.47–1.99), whereas higher parity was associated with lower use of epidural analgesia (adj OR, 0.86; 95% CI, 0.85–0.87). Patients undergoing induction of labor were more likely to have epidurals placed (adj OR, 2.12; 95% CI, 2.01–2.24).

We tested the interaction between patient race/ethnicity and insurance status in a separate multivariate model (table 3). The rate of epidural analgesia did not vary for black patients regardless of the type of insurance coverage (P = 0.56). Black patients with private insurance had rates of epidural use similar to those of white/non-Hispanic patients without insurance coverage: The adj OR was 0.66 (95% CI, 0.53–0.82) for white/non-Hispanic patients without insurance versus  0.69 (95% CI, 0.57–0.85) for black patients with private insurance. In comparison, white/Hispanic patients had similar rates of epidural use compared with white/non-Hispanic patients who had the same level of insurance coverage. The interaction between other racial/ethnic groups and insurance status could not be tested because of sample size considerations. Excluding patients treated by midwives and patients whose provider status was unknown did not substantially change the results of the analyses and are not shown (results are available on request).

In this study, we find that black and Hispanic women are less likely than white women to receive epidural analgesia for labor. This finding persists after adjusting for clinical risk factors, socioeconomic status, and provider effects. We also find evidence that patients with private insurance have the highest rate of epidural analgesia, whereas women without insurance have the lowest rate of epidural analgesia. Most striking, in light of the association between epidural use and insurance coverage, we find that black patients with private insurance have the same rate of epidural analgesia as white patients without any insurance. Further, there is no difference in the rate of epidural use among blacks with private insurance compared with blacks with Medicaid or without insurance.

Although differences in epidural analgesia across racial and ethnic groups have been documented in previous studies,45,49–51our study adds to previous research in important ways. It is now recognized that racial disparities in health care are due, in part, to the fact that racial and ethnic minorities are treated by different hospitals and physicians than whites.24,25,27Whether racial disparities persist when racial and ethnic minorities are treated by the same provider is only now being investigated.26,28In contrast to previous studies, we used fixed effects modeling to examine the variation in the use of epidural analgesia in racial and ethnic minorities within the practice of individual obstetricians and providers.

We chose to use fixed effects modeling rather than hierarchical modeling because of the evidence that nonwhites are treated by different physicians and hospitals than are whites. When patient characteristics (e.g. , race) are correlated with provider effects, hierarchical models and conventional models will yield inconsistent estimates of the effect of race, whereas fixed effects models will be unbiased.54Further, our use of robust variance estimators accounts for the fact that the outcomes for patients treated by the same provider are correlated, thus avoiding the potential problem of overestimating the precision of the impact of race and ethnicity on the rate of epidural use in our analysis.

Unlike most of the previous studies examining the impact of race on the use of epidural analgesia for labor, our study was based on a comprehensive clinical data set that included detailed data on patient demographics and clinical characteristics. Last, although previous studies analyzed the impact of both race and insurance on the use of epidural analgesia, we extended our analysis to evaluate the interaction between race and insurance status to gain a better understanding of whether differences in insurance coverage accounted for racial differences in the use of epidural analgesia.

Are there reasons why racial and ethnic minorities are less likely to receive epidural analgesia for labor than white patients? A number of different factors may account for the observed differences in epidural use across different racial and ethnic groups. These factors included differences in pain perception and patient preferences, and provider bias. However, despite experimental evidence suggesting that blacks display increased sensitivity to painful stimuli compared with whites,58–62there is widespread evidence that acute36,37,39,43,44,63and chronic pain33,38,40are undertreated in black and Hispanic patients. There is no evidence that minority patients have different pain management expectations36,64or that physicians' ability to assess pain severity differs for minority patients compared with white patients.35Nonetheless, it is possible, that minority patients might prefer intravenous or opioid analgesics instead of the use of the more invasive modality, epidural analgesia, despite the better pain relief associated with epidural analgesia. Mistrust of the healthcare system by minorities1may contribute to the underuse of epidural analgesia in this population. Alternatively, provider beliefs and behavior may contribute to the observed disparity65in the provision of epidural analgesia to minority patients.

Our results should be interpreted with caution because this study has several important limitations. First, as with all observational studies, our study can not establish a cause-and-effect relation between an outcome (use of epidural analgesia) and a clinical characteristic (race and ethnicity). Second, although we believe that our analysis controlled for all potentially relevant confounders, it is possible that the absence of unmeasured variables may have biased our findings. In particular, we have no information on patient preferences for the type of analgesia, likely the most important factor in determining whether a patient receives epidural analgesia. Ideally, this analysis would have also controlled for patient preference to determine whether there are racial differences in epidural use among women who actually request epidural analgesia. Our finding that Asian women tended to be less likely to receive epidural analgesia compared with non-Hispanic white women, even after adjusting for clinical risk factors, insurance, and education, strongly suggests that patient preferences plays an important role in the decision to place an epidural during labor. Although we cannot rule out the possibility that black patients are less likely to request epidural analgesia compared with other racial groups, we believe that it is highly unlikely that the observed disparity in epidural analgesia is solely due to differences in patient preferences given the extensive evidence of racial disparities in other areas of health care.

Third, this study was based on patients living in New York State and may not be generalizable to other regions or states. Finally, race and ethnicity were self-identified in our data set. If provider bias is an important determinant of racial disparities, basing such analysis on race and ethnicity as classified by the providers may be preferable.

In summary, we find that black and Hispanic women in labor are less likely than white patients to receive epidural analgesia. Further, we find that these differences remain after accounting for differences in insurance coverage and clinical characteristics. In keeping with the goals of Healthy People 2010, efforts to eliminate racial disparities in health care should include efforts to insure that all women in labor have access to adequate pain relief.