International Journal of Obstetric Anesthesia (2014) 23, 161–167 0959-289X/$ - see front matter c 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijoa.2013.09.001
ORIGINAL ARTICLE
www.obstetanesthesia.com
Preferred spoken language mediates differences in neuraxial labor analgesia utilization among racial and ethnic groups J.A. Caballero, A.J. Butwick, B. Carvalho, E.T. Riley Department of Anesthesia, Stanford University School of Medicine, Stanford, CA, USA ABSTRACT Introduction: The aims of this study were to assess racial/ethnic disparities for neuraxial labor analgesia utilization and to determine if preferred spoken language mediates the association between race/ethnicity and neuraxial labor analgesia utilization. Methods: We performed a retrospective cohort study of 3129 obstetric patients who underwent vaginal delivery at a tertiary care obstetric center. Bivariate analyses and multivariate logistic regression models were used to assess the relationships between race/ ethnicity, preferred spoken language and neuraxial labor analgesia. Results: Hispanic ethnicity (adjusted OR 0.77, 95% CI 0.61–0.98) and multiparity (adjusted OR 0.59, 95% CI 0.51–0.69) were independently associated with a reduced likelihood of neuraxial labor analgesia utilization. When preferred spoken language was controlled for, the effect of Hispanic ethnicity was no longer significant (adjusted OR 0.84, 95% CI 0.66–1.08) and only non-English preferred spoken language (adjusted OR 0.82, 95% CI 0.67–0.99) and multiparity (adjusted OR 0.59, 95% CI 0.51–0.69) were associated with a reduced likelihood of neuraxial labor analgesia utilization. Conclusions: This study provides evidence that preferred spoken language mediates the relationship between Hispanic ethnicity and neuraxial labor analgesia utilization. c 2013 Elsevier Ltd. All rights reserved.
Keywords: Disparities; Utilization; Race; Ethnicity; Neuraxial analgesia; Labor analgesia
Introduction The availability and use of neuraxial techniques for the provision of labor analgesia has been increasing among USA hospitals with obstetric care facilities.1,2 However, important disparities are known to exist in the use of neuraxial labor analgesia among patients of different racial/ethnic groups.2–5 Based on birth certificate data from a nationwide sample of USA patients undergoing vaginal delivery, the rate of neuraxial labor analgesia among Hispanic women was found to be lower than that of Caucasian women (48% vs. 69%, respectively).2 This healthcare disparity has important ramifications for the future delivery of high-quality obstetric anesthetic care, as a recent USA Census Bureau report indicated that more than half of the growth in the population between 2000 and 2010 (15.2 million of 27.3 million) was due to an increase in the Hispanic population.6 There is growing interest in leveraging health information technologies as tools for outcomes-based research with the goal of identifying and addressing Accepted September 2013 Correspondence to: Alexander Butwick, Department of Anesthesia, Stanford University School of Medicine, 300 Pasteur Avenue, H3580, Stanford, CA 94305-5640, USA. E-mail address: [email protected]
health disparities that affect racial/ethnic minorities and non-English speaking patients.7,8 Previous studies have suggested that racial/ethnic disparities persist even after adjusting for relevant clinical risk factors, socioeconomic factors and provider effects (based on insurance status).3–5 However, residual confounding is likely to have been an important limitation in these studies, with preferred spoken language in particular being notably absent from the models used to assess predictors for utilization of neuraxial analgesia. There is growing appreciation that language barriers can contribute to important healthcare disparities, with studies conducted in a non-obstetric setting reporting that patients with limited English proficiency had higher rates of adverse outcomes and reduced compliance with medical treatment compared with patients who are English-speaking.9,10 However, the impact of language proficiency on healthcare disparities in the labor and delivery setting is incompletely understood. The aims of this study were to assess racial/ethnic disparities for neuraxial labor analgesia utilization and to determine if preferred spoken language mediates the association between race/ethnicity and neuraxial labor analgesia utilization in a multiethnic cohort of women who underwent vaginal delivery.
162
Methods This retrospective study was conducted at Lucile Packard Children’s Hospital (LPCH) after approval by the Institutional Review Board (IRB) of Stanford University Medical Center; informed consent for this study was waived by the IRB. LPCH is a university hospital and tertiary obstetric center with approximately 5000 deliveries per year and has a level III neonatal intensive care unit. Patients who deliver at LPCH comprise a multiethnic cohort with low- and high-risk pregnancies. The labor and delivery unit provides obstetric care to patients with different payer or insurance coverage (insured, Medicaid, or self-pay). Obstetricians who provide obstetric care at LPCH are employed by private practice groups, Stanford University Medical Center, or the San Mateo County Health System. The obstetric anesthetic coverage is provided for all inpatients by a team of anesthesiologists based at Stanford University Medical Center (comprising of at least one dedicated anesthesia attending and one anesthesia resident) who are available in-hospital 24 h per day. Hospital intake workflow is the same for all patients, regardless of race, ethnicity, religion, sexual orientation or preferred spoken language. Our study population consisted of patients who underwent vaginal delivery at LPCH from January 1, 2008 to December 31, 2008. To identify our cohort, we accessed an institutional electronic data warehouse that stores patient health information sourced from our institution’s electronic medical records (Cerner PowerchartÒ, Cerner Corp., Kansas City, MO, USA). The year 2008 was chosen for the study because it was the first year that complete electronic records were available for research. We performed an electronic search of the data warehouse for patients who underwent vaginal delivery during the study period. Information about neuraxial labor analgesia, maternal age at the time of delivery, racial/ethnic group, parity, insurer/payer status and preferred spoken language were abstracted from electronic medical records. Clinical elements were abstracted from nursing and physician entries whereas insurance and demographic information was abstracted from administrative and clerical entries, respectively. We excluded patients who underwent a trial of labor before intrapartum cesarean delivery, as we were unable to clearly ascertain whether latent or active labor preceded cesarean delivery using search queries within the data warehouse. To verify data integrity and account for missing data, one of the authors (JC) manually reviewed the medical charts of 450 patients (14% of study cohort). Women who received either epidural, combined spinal-epidural or single-shot spinal techniques for labor analgesia were defined as receiving neuraxial labor analgesia. For our primary dependent variable, we dichotomized patients into groups either receiving or not receiving neuraxial labor analgesia. Maternal race/
Language and uptake of neuraxial analgesia ethnicity and preferred spoken language were self-reported by patients at the time of inpatient admission to the labor and delivery unit and then entered manually by labor and delivery nurses into each patient’s electronic medical record. We categorized race/ethnicity as non-Hispanic White, non-Hispanic Black, Hispanic, Asian/Pacific Islander (hereafter called Asian), and non-Hispanic Other in accordance with the Office of Management and Budget’s Directive No. 15 (Race and Ethnic Standards for Federal Statistics and Administrative Reporting).11 We initially examined the proportion of patients within each racial/ethnic group who spoke English, Spanish, and languages other than English or Spanish. We created a dichotomous variable for preferred spoken language (English or non-English) because the proportions of patients speaking languages other than English or Spanish were small (Whites, 5.3%; AfricanAmericans, 0.03%; Hispanics, 0.4%; Asians, 3.8%; non-Hispanic Others, 0.2%). We categorized health insurance status as public insurance (comprising patients on Medi-Cal) or private insurance, aggregating all types of private insurance into a single group. Medi-Cal is California’s Medicaid program that provides health insurance coverage for 19% of the non-elderly population, including 2.7 million adults.12 After manual review of the medical records, 35 patients were excluded from the analysis because we could not clearly ascertain their health insurance status.
Statistical analysis Frequency distribution, mean and standard deviation (SD) of maternal and obstetric characteristics were determined by race/ethnicity. We used the chi-squared test to compare independent categorical variables (preferred spoken language, insurance type, parity and utilization of neuraxial labor analgesia) by race/ethnicity. We performed one-way analysis of variance (ANOVA) to compare mean ages by race/ethnicity. The association between neuraxial labor analgesia and each of the explanatory variables was initially assessed using univariate logistic regression models. Crude odds ratios (OR) and 95% confidence intervals (CI) were calculated. The potential for collinearity (which refers to a linear relationship between two explanatory variables) was determined by calculating variance inflation factors. We did not detect evidence of collinearity as the variance inflation factors for each racial/ethnic group ranged from 1.0 to 2.3. We incrementally constructed two full multivariable logistic regression models to assess the associations between race/ethnicity and neuraxial labor analgesia utilization. The OR for neuraxial labor analgesia utilization among Hispanics, non-Hispanic Blacks, Asians, and non-Hispanic Others were compared with non-Hispanic Whites (as the reference group) within each multivariate logistic regression model. The first model included only age, race/ethnicity, parity and insurance type. The
J.A. Caballero et al.
163
second model investigated whether preferred spoken language mediated the relationship between race/ethnicity and utilization of neuraxial labor analgesia by adding preferred spoken language to all covariates in the first model. To assess the sensitivity of our models, we also performed multivariate logistic regression using backward and forward variable selection approaches. The goodness of fit of the final model was evaluated using the Hosmer-Lemoshow statistic.13 This test provides a measure for how well the model fits the data; if the chi-squared goodness of fit is not significant, then the model has adequate fit, whereas if the test is significant, the model does not adequately fit the data. The ability of the model to classify patients (predictive performance) was assessed using the area under the receiver-operating characteristic curve (AUROC or c-index).14 We tested for mediation using the causal step approach.15,16 The test of whether preferred spoken language mediated the relationship between race/ethnicity and utilization of neuraxial labor analgesia included three phases: (1) a regression model of utilization of neuraxial labor analgesia adjusted for race/ethnicity, (2) a regression model of utilization of neuraxial labor analgesia adjusted for race/ethnicity and preferred spoken language, and (3) a regression model of race/ethnicity adjusted for preferred spoken language. For primary spoken language to mediate the association of race/ethnicity and neuraxial labor analgesia, the following would be required: race/ethnicity would need to be statistically significant in the first regression, race/ethnicity would not be significant and language would be significant in the second regression, and language would be significant in the third regression. We also performed a secondary analysis to assess factors (maternal age, preferred spoken language [English or
Table 1
Spanish], insurance type, and parity) associated with neuraxial labor analgesia utilization among Hispanic patients. Within this study cohort only 12 Hispanic patients preferred speaking a language that was neither English nor Spanish; therefore these patients were excluded from the secondary analyses. Data analysis was conducted with SAS software version 9.2 (SAS Institute, Cary, NC. USA).
Results We identified 3129 women who underwent vaginal delivery during the pre-defined study period, of whom 1692 (54%) utilized neuraxial labor analgesia. Overall, 1516 patients (48.4%) were covered by Medi-Cal, and 1421 patients (45.4%) were nulliparous. English (1192 [62.7%] patients) and Spanish (981 [31.4%] patients) were the predominant preferred spoken languages. Maternal age, preferred spoken language, insurance type, parity and utilization of neuraxial labor analgesia varied among women of differing race/ethnicity (Table 1). Data for the unadjusted and adjusted logistic regression analyses are presented in Tables 2 and 3. In the unadjusted analyses, the rates of neuraxial labor analgesia utilization were significantly lower among the following groups: Hispanic patients, patients who did not have English as the preferred spoken language, patients with Medi-Cal insurance, and multiparous patients. After adjusting for insurance type and parity (Model 1), patients of Hispanic ethnicity had a reduced likelihood of utilizing neuraxial labor analgesia compared with nonHispanic White patients (Table 3). The AUROC for Model 1 was 0.60 (95% CI 0.58–0.62). In the second multivariate model (Model 2), which comprised the covariates in Model 1 and preferred spoken language,
Patient characteristics in the cohort according to ethnicity/race
Age (years) Primary Language English Non-English Insurance Type Medi-Cal Private Parity Nulliparous Multiparous Neuraxial block for labor analgesia Yes No
Non-Hispanic Other (n = 57)
P value*
(n = 1367)
Asian/Pacific Islander (n = 725)
26.7 ± 6.1
32.4 ± 5.0
30.3 ± 5.9
ORIGINAL ARTICLE
www.obstetanesthesia.com
Preferred spoken language mediates differences in neuraxial labor analgesia utilization among racial and ethnic groups J.A. Caballero, A.J. Butwick, B. Carvalho, E.T. Riley Department of Anesthesia, Stanford University School of Medicine, Stanford, CA, USA ABSTRACT Introduction: The aims of this study were to assess racial/ethnic disparities for neuraxial labor analgesia utilization and to determine if preferred spoken language mediates the association between race/ethnicity and neuraxial labor analgesia utilization. Methods: We performed a retrospective cohort study of 3129 obstetric patients who underwent vaginal delivery at a tertiary care obstetric center. Bivariate analyses and multivariate logistic regression models were used to assess the relationships between race/ ethnicity, preferred spoken language and neuraxial labor analgesia. Results: Hispanic ethnicity (adjusted OR 0.77, 95% CI 0.61–0.98) and multiparity (adjusted OR 0.59, 95% CI 0.51–0.69) were independently associated with a reduced likelihood of neuraxial labor analgesia utilization. When preferred spoken language was controlled for, the effect of Hispanic ethnicity was no longer significant (adjusted OR 0.84, 95% CI 0.66–1.08) and only non-English preferred spoken language (adjusted OR 0.82, 95% CI 0.67–0.99) and multiparity (adjusted OR 0.59, 95% CI 0.51–0.69) were associated with a reduced likelihood of neuraxial labor analgesia utilization. Conclusions: This study provides evidence that preferred spoken language mediates the relationship between Hispanic ethnicity and neuraxial labor analgesia utilization. c 2013 Elsevier Ltd. All rights reserved.
Keywords: Disparities; Utilization; Race; Ethnicity; Neuraxial analgesia; Labor analgesia
Introduction The availability and use of neuraxial techniques for the provision of labor analgesia has been increasing among USA hospitals with obstetric care facilities.1,2 However, important disparities are known to exist in the use of neuraxial labor analgesia among patients of different racial/ethnic groups.2–5 Based on birth certificate data from a nationwide sample of USA patients undergoing vaginal delivery, the rate of neuraxial labor analgesia among Hispanic women was found to be lower than that of Caucasian women (48% vs. 69%, respectively).2 This healthcare disparity has important ramifications for the future delivery of high-quality obstetric anesthetic care, as a recent USA Census Bureau report indicated that more than half of the growth in the population between 2000 and 2010 (15.2 million of 27.3 million) was due to an increase in the Hispanic population.6 There is growing interest in leveraging health information technologies as tools for outcomes-based research with the goal of identifying and addressing Accepted September 2013 Correspondence to: Alexander Butwick, Department of Anesthesia, Stanford University School of Medicine, 300 Pasteur Avenue, H3580, Stanford, CA 94305-5640, USA. E-mail address: [email protected]
health disparities that affect racial/ethnic minorities and non-English speaking patients.7,8 Previous studies have suggested that racial/ethnic disparities persist even after adjusting for relevant clinical risk factors, socioeconomic factors and provider effects (based on insurance status).3–5 However, residual confounding is likely to have been an important limitation in these studies, with preferred spoken language in particular being notably absent from the models used to assess predictors for utilization of neuraxial analgesia. There is growing appreciation that language barriers can contribute to important healthcare disparities, with studies conducted in a non-obstetric setting reporting that patients with limited English proficiency had higher rates of adverse outcomes and reduced compliance with medical treatment compared with patients who are English-speaking.9,10 However, the impact of language proficiency on healthcare disparities in the labor and delivery setting is incompletely understood. The aims of this study were to assess racial/ethnic disparities for neuraxial labor analgesia utilization and to determine if preferred spoken language mediates the association between race/ethnicity and neuraxial labor analgesia utilization in a multiethnic cohort of women who underwent vaginal delivery.
162
Methods This retrospective study was conducted at Lucile Packard Children’s Hospital (LPCH) after approval by the Institutional Review Board (IRB) of Stanford University Medical Center; informed consent for this study was waived by the IRB. LPCH is a university hospital and tertiary obstetric center with approximately 5000 deliveries per year and has a level III neonatal intensive care unit. Patients who deliver at LPCH comprise a multiethnic cohort with low- and high-risk pregnancies. The labor and delivery unit provides obstetric care to patients with different payer or insurance coverage (insured, Medicaid, or self-pay). Obstetricians who provide obstetric care at LPCH are employed by private practice groups, Stanford University Medical Center, or the San Mateo County Health System. The obstetric anesthetic coverage is provided for all inpatients by a team of anesthesiologists based at Stanford University Medical Center (comprising of at least one dedicated anesthesia attending and one anesthesia resident) who are available in-hospital 24 h per day. Hospital intake workflow is the same for all patients, regardless of race, ethnicity, religion, sexual orientation or preferred spoken language. Our study population consisted of patients who underwent vaginal delivery at LPCH from January 1, 2008 to December 31, 2008. To identify our cohort, we accessed an institutional electronic data warehouse that stores patient health information sourced from our institution’s electronic medical records (Cerner PowerchartÒ, Cerner Corp., Kansas City, MO, USA). The year 2008 was chosen for the study because it was the first year that complete electronic records were available for research. We performed an electronic search of the data warehouse for patients who underwent vaginal delivery during the study period. Information about neuraxial labor analgesia, maternal age at the time of delivery, racial/ethnic group, parity, insurer/payer status and preferred spoken language were abstracted from electronic medical records. Clinical elements were abstracted from nursing and physician entries whereas insurance and demographic information was abstracted from administrative and clerical entries, respectively. We excluded patients who underwent a trial of labor before intrapartum cesarean delivery, as we were unable to clearly ascertain whether latent or active labor preceded cesarean delivery using search queries within the data warehouse. To verify data integrity and account for missing data, one of the authors (JC) manually reviewed the medical charts of 450 patients (14% of study cohort). Women who received either epidural, combined spinal-epidural or single-shot spinal techniques for labor analgesia were defined as receiving neuraxial labor analgesia. For our primary dependent variable, we dichotomized patients into groups either receiving or not receiving neuraxial labor analgesia. Maternal race/
Language and uptake of neuraxial analgesia ethnicity and preferred spoken language were self-reported by patients at the time of inpatient admission to the labor and delivery unit and then entered manually by labor and delivery nurses into each patient’s electronic medical record. We categorized race/ethnicity as non-Hispanic White, non-Hispanic Black, Hispanic, Asian/Pacific Islander (hereafter called Asian), and non-Hispanic Other in accordance with the Office of Management and Budget’s Directive No. 15 (Race and Ethnic Standards for Federal Statistics and Administrative Reporting).11 We initially examined the proportion of patients within each racial/ethnic group who spoke English, Spanish, and languages other than English or Spanish. We created a dichotomous variable for preferred spoken language (English or non-English) because the proportions of patients speaking languages other than English or Spanish were small (Whites, 5.3%; AfricanAmericans, 0.03%; Hispanics, 0.4%; Asians, 3.8%; non-Hispanic Others, 0.2%). We categorized health insurance status as public insurance (comprising patients on Medi-Cal) or private insurance, aggregating all types of private insurance into a single group. Medi-Cal is California’s Medicaid program that provides health insurance coverage for 19% of the non-elderly population, including 2.7 million adults.12 After manual review of the medical records, 35 patients were excluded from the analysis because we could not clearly ascertain their health insurance status.
Statistical analysis Frequency distribution, mean and standard deviation (SD) of maternal and obstetric characteristics were determined by race/ethnicity. We used the chi-squared test to compare independent categorical variables (preferred spoken language, insurance type, parity and utilization of neuraxial labor analgesia) by race/ethnicity. We performed one-way analysis of variance (ANOVA) to compare mean ages by race/ethnicity. The association between neuraxial labor analgesia and each of the explanatory variables was initially assessed using univariate logistic regression models. Crude odds ratios (OR) and 95% confidence intervals (CI) were calculated. The potential for collinearity (which refers to a linear relationship between two explanatory variables) was determined by calculating variance inflation factors. We did not detect evidence of collinearity as the variance inflation factors for each racial/ethnic group ranged from 1.0 to 2.3. We incrementally constructed two full multivariable logistic regression models to assess the associations between race/ethnicity and neuraxial labor analgesia utilization. The OR for neuraxial labor analgesia utilization among Hispanics, non-Hispanic Blacks, Asians, and non-Hispanic Others were compared with non-Hispanic Whites (as the reference group) within each multivariate logistic regression model. The first model included only age, race/ethnicity, parity and insurance type. The
J.A. Caballero et al.
163
second model investigated whether preferred spoken language mediated the relationship between race/ethnicity and utilization of neuraxial labor analgesia by adding preferred spoken language to all covariates in the first model. To assess the sensitivity of our models, we also performed multivariate logistic regression using backward and forward variable selection approaches. The goodness of fit of the final model was evaluated using the Hosmer-Lemoshow statistic.13 This test provides a measure for how well the model fits the data; if the chi-squared goodness of fit is not significant, then the model has adequate fit, whereas if the test is significant, the model does not adequately fit the data. The ability of the model to classify patients (predictive performance) was assessed using the area under the receiver-operating characteristic curve (AUROC or c-index).14 We tested for mediation using the causal step approach.15,16 The test of whether preferred spoken language mediated the relationship between race/ethnicity and utilization of neuraxial labor analgesia included three phases: (1) a regression model of utilization of neuraxial labor analgesia adjusted for race/ethnicity, (2) a regression model of utilization of neuraxial labor analgesia adjusted for race/ethnicity and preferred spoken language, and (3) a regression model of race/ethnicity adjusted for preferred spoken language. For primary spoken language to mediate the association of race/ethnicity and neuraxial labor analgesia, the following would be required: race/ethnicity would need to be statistically significant in the first regression, race/ethnicity would not be significant and language would be significant in the second regression, and language would be significant in the third regression. We also performed a secondary analysis to assess factors (maternal age, preferred spoken language [English or
Table 1
Spanish], insurance type, and parity) associated with neuraxial labor analgesia utilization among Hispanic patients. Within this study cohort only 12 Hispanic patients preferred speaking a language that was neither English nor Spanish; therefore these patients were excluded from the secondary analyses. Data analysis was conducted with SAS software version 9.2 (SAS Institute, Cary, NC. USA).
Results We identified 3129 women who underwent vaginal delivery during the pre-defined study period, of whom 1692 (54%) utilized neuraxial labor analgesia. Overall, 1516 patients (48.4%) were covered by Medi-Cal, and 1421 patients (45.4%) were nulliparous. English (1192 [62.7%] patients) and Spanish (981 [31.4%] patients) were the predominant preferred spoken languages. Maternal age, preferred spoken language, insurance type, parity and utilization of neuraxial labor analgesia varied among women of differing race/ethnicity (Table 1). Data for the unadjusted and adjusted logistic regression analyses are presented in Tables 2 and 3. In the unadjusted analyses, the rates of neuraxial labor analgesia utilization were significantly lower among the following groups: Hispanic patients, patients who did not have English as the preferred spoken language, patients with Medi-Cal insurance, and multiparous patients. After adjusting for insurance type and parity (Model 1), patients of Hispanic ethnicity had a reduced likelihood of utilizing neuraxial labor analgesia compared with nonHispanic White patients (Table 3). The AUROC for Model 1 was 0.60 (95% CI 0.58–0.62). In the second multivariate model (Model 2), which comprised the covariates in Model 1 and preferred spoken language,
Patient characteristics in the cohort according to ethnicity/race
Age (years) Primary Language English Non-English Insurance Type Medi-Cal Private Parity Nulliparous Multiparous Neuraxial block for labor analgesia Yes No
Non-Hispanic Other (n = 57)
P value*
(n = 1367)
Asian/Pacific Islander (n = 725)
26.7 ± 6.1
32.4 ± 5.0
30.3 ± 5.9
