Association of the United States Affordable Care Act Dependent Coverage Provision with Labor Neuraxial Analgesia Use

The study protocol was deemed exempt by the Institutional Review Board of the Columbia University Irving Medical Center (New York, New York). We report the study according to the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines. No statistical analysis plan was established a priori, nor was a clinically meaningful change in labor neuraxial utilization rate associated with the Dependent Coverage Provision defined a priori. The currently presented analysis was based upon the initial analysis combined with changes requested during the peer review process.

We analyzed U.S. birth certificate data contained in the restricted access Natality File of the National Vital Statistics System (National Center for Health Statistics, Centers for Diseases Control and Prevention, Atlanta, Georgia). The Natality File is based on the 2003 revised U.S. Standard Certificate of Live Birth.¹⁷  The Standard Certificate of Live Birth was implemented gradually across states, from two states in 2003 to 50 states and the District of Columbia in 2015. The Natality File is a census of all live births in the United States and contains comprehensive information on the woman, labor, and delivery. It also provides county identifiers for the woman’s residence and for the delivery hospital. These identifiers allow abstraction of county characteristics from the Area Health Resource File.¹⁸  Area Health Resource File data contain detailed information abstracted from greater than 50 data sources on healthcare professions, health facilities, population characteristics, and economics measured at the county level.

We used the difference-in-differences method to exploit the natural experiment created by the arbitrary age cutoff of the Dependent Coverage Provision (fig. 1). This method has been used in many previous evaluation studies of the Dependent Coverage Provision, as well as other health insurance expansions.^12,19–21  Briefly, the difference-in-differences method compares the difference in the mean labor neuraxial analgesia utilization rate before and after the implementation of the Dependent Coverage Provision in women exposed to the Dependent Coverage Provision (i.e., women aged 25 yr or younger; first difference) to the difference in the mean labor neuraxial analgesia utilization rate before and after the implementation of the Dependent Coverage Provision in women not exposed to the Dependent Coverage Provision (i.e., women aged 27 yr or older; second difference). The difference between these two differences (i.e., difference-in-differences) can be estimated using the regression coefficient of a two-way interaction term between the age group (indicator of the 21- to 25-yr-old group) and the period (indicator of the postintervention period) in a linear regression model. For example, a regression coefficient of 0.015 indicates that the Dependent Coverage Provision is associated with a 1.5% (absolute) increase in the probability of labor neuraxial analgesia in women aged 21 to 25 yr compared to women aged 27 to 31 yr.

For a post hoc analysis requested during the peer review process and analyzing the proportion of women with private health insurance coverage, we used the interrupted time series method instead of the difference-in-differences method.²² 

The study sample included birth certificates for nonoperative vaginal deliveries from January 2009 to December 2013 in the 28 states that had implemented the 2003 revised U.S. Standard Certificate of Live Birth as of January 1, 2009 (California, Colorado, Delaware, Florida, Georgia, Idaho, Indiana, Iowa, Kansas, Kentucky, Michigan, Montana, Nebraska, New Hampshire, New Mexico, New York, North Dakota, Ohio, Oregon, Pennsylvania, South Carolina, South Dakota, Tennessee, Texas, Utah, Vermont, Washington, and Wyoming). These 28 states accounted for 66% of national births during the study period. We further limited the study sample to women aged 21 to 25 yr and to women aged 27 to 31 yr.

We did not include birth certificate data after December 2013 because of the implementation in January 2014 of the Medicaid expansion, authorized by the Patient Protection and Affordable Care Act. The 2014 Medicaid Expansion gave states the option to expand Medicaid coverage to nonelderly adults with incomes up to 138% of the Federal Poverty Level and may have superseded the Dependent Coverage Provision.

We did not include birth certificate data before January 2009 because it would have excluded too many states. Only 19 states accounting for 49% of U.S. births had implemented the 2003 revised U.S. Standard Certificate of Live Birth as of January 1, 2006; 22 states accounting for 53% of U.S. births as of January 1, 2007; and 27 states accounting for 65% of U.S. births as of January 1, 2008.

We identified vaginal deliveries without forceps or vacuum extraction using a specific checkbox on the birth certificate.¹⁷ 

Exclusion criteria were (1) missing information on labor neuraxial analgesia (0.3%); (2) birth not occurring in a hospital or missing information on birth location (2.2%); (3) woman not residing in the United States or missing information on residence country (0.2%); and (4) missing information on woman’s residence county or hospital of delivery county (0.0%; fig. 2).

The intervention was the Dependent Coverage Provision, which requires private health insurers to allow young adults to remain on their parents’ plans until their 26th birthday. Depending on the parents’ health insurance plan, the young dependent adult is covered until the end of the month in which the dependent turns 26 yr or until the end of the year in which the dependent turns 26 yr. The law took effect for insurance plan renewals beginning on September 23, 2010, and applies to all plans in the individual market and to all employer plans.¹⁶  The preintervention period was January 2009 to August 2010, and the postintervention period was September 2010 to December 2013. In a sensitivity analysis, we excluded data from the year 2010 (washout period). In post hoc analyses requested during the peer review process, we limited the postintervention period to January 2012 to December 2013 and to January 2013 to December 2013.

The exposure of interest was women’s age group, categorized into exposed (21 to 25 yr) and unexposed (27 to 31 yr). In a sensitivity analysis, we used a narrower definition of the two women’s age groups and defined exposed as women aged 24 to 25 yr (instead of 21 to 25 yr) and unexposed as women aged 27 to 28 yr (instead of 27 to 31 yr).

The primary outcome was labor neuraxial analgesia utilization rate. In the birth certificate, labor neuraxial analgesia is reported in a specific checkbox (“Epidural or spinal anesthesia during labor”), and defined as the “administration to the mother of a regional anesthetic for control of the pain of labor (i.e., delivery of the agent into a limited space with the distribution of the analgesic effect limited to the lower body).”²³  Using individual medical records as the accepted standard, the reported sensitivity of labor neuraxial analgesia in birth certificate data in a study conducted in two states in 2009 to 2011 was greater than 80%.²⁴  The Natality Files do not contain detailed information on the type of neuraxial analgesia, precluding the analysis of specific neuraxial techniques (epidural, spinal, or combined spinal–epidural).

The secondary outcome assessing the quality of prenatal care was early initiation of prenatal care, defined as initiation during the first 3 months of pregnancy. In a post hoc analysis requested during the peer review process, we analyzed the proportion of women with private health insurance coverage as a secondary outcome.

Women’s characteristics and comorbidities directly recorded from birth certificate data included race and ethnicity; education level (less than high school, high school with no diploma, high school graduate or general educational diploma, and college or higher); marital status (married or unmarried); health insurance (Medicaid, private insurance, self-pay, and other); prepregnancy body mass index (18.4 or less, 18.5 to 24.9, 25.0 to 29.9, 30.0 to 34.9, or 35 or greater kg/m²); and preexisting or gestational diabetes or hypertension. Maternal race and ethnicity included four mutually exclusive groups: (1) non-Hispanic White (hereafter referred to as White); (2) non-Hispanic Black (Black); (3) non-Hispanic Other; and (4) Hispanic. We acknowledge that race is a social construct, and we used race as a proxy for racism and discrimination. Since birth certificate data do not provide information on women’s residence, we estimated the following women’s characteristics at the county of residence level using the Area Health Resource File: residence location (urban, suburban, or rural), proportion of persons in poverty, and proportion of persons unemployed.

Obstetrical characteristics directly recorded from birth certificate data included previous cesarean section; delivery during a weekend; woman transferred in (i.e., transfer from another facility for maternal medical or fetal indications for delivery); parity (nulliparous or parous); gestational age at delivery (33 or less completed weeks, 34 to 38 completed weeks, or 39 or more completed weeks); multiple gestation; noncephalic presentation; induction of labor; augmentation of labor; antibiotics during labor; fever or chorioamnionitis during labor; attendant at birth (doctor of medicine, doctor of osteopathy, midwife, or other); and birth weight (2,499 or less g, 2,500 to 4,000 g, or greater than 4,000 g).

Since birth certificate data do not provide a hospital identifier, we estimated the following characteristics at the hospital county level using the Area Health Resource File: hospital location (urban, suburban, or rural); number of hospital beds; number of hospital births; number of medical doctors; number of obstetricians and gynecologists; number of physician anesthesiologists; and number of certified registered nurse anesthetists.

Statistical analysis was performed with R version 3.6.2 (R Foundation for Statistical Computing, Austria).²⁵ 

In each age group, we compared women’s, obstetrical, and hospital county characteristics between the pre- and postintervention periods using the absolute standardized difference, with a value greater than 0.10 (10%) used to define a clinically important imbalance.²⁶  For continuous variables estimated at the county level (e.g., unemployment rate), the absolute standardized difference was calculated using the median and not the mean.

We estimated labor neuraxial analgesia utilization rates overall and according to health insurance type (Medicaid, private, self-pay, or other), race and ethnicity (White, Black, Other, or Hispanic), and parity (nulliparous or parous). Then, in each age group, we estimated labor neuraxial analgesia utilization rates in the pre- and postintervention periods. For each age group, we calculated the absolute difference in labor neuraxial analgesia utilization rates between the post- and preintervention periods.

The crude difference-in-differences estimator was the coefficient of a two-way interaction term between the variable intervention and the variable exposure in the following linear regression model (Model 1): Y = β₀ + β₁ Intervention + β₂ Exposure + β₃ Intervention × Exposure_. In this model, Y denotes labor neuraxial analgesia, intervention the intervention period (postintervention period coded as 1 and preintervention period as 0), and exposure the women’s age group (21 to 25 yr group coded as 1 and 27 to 31 yr group as 0).

The main assumption of the difference-in-differences approach is that the trends in labor neuraxial analgesia utilization rate before the implementation of the Dependent Coverage Provision would not differ between the two age groups (i.e., the “parallel trends assumption”). To validate this assumption, we estimated the statistical significance of the coefficient of an interaction term between the age group and the month of delivery treated as a continuous variable in the preintervention period in a linear regression model. The result indicated that there was no significant divergence in labor neuraxial analgesia trends before the implementation of the Dependent Coverage Provision between the two women’s ages (Supplemental Table 1, https://links.lww.com/ALN/D165).

To adjust the difference-in-differences estimator (i.e., β₃), we added to Model 1 (1) variables with a plausible association with the exposure and the outcome (Supplemental Figure 1 and Supplemental Table 2, https://links.lww.com/ALN/D165); (2) a year fixed effect; (3) a state fixed effect; and (4) secular trends (month of delivery treated as a continuous variable, with January 2009 coded as 1 and December 2013 coded as 60). Health insurance and early initiation of prenatal care were not used for adjustment because they were treated as the outcomes of the Dependent Coverage Provision. We performed a complete case analysis with 4.8% of birth certificates excluded because of missing values for the variables used for adjustment.

To estimate the association between the Dependent Coverage Provision and labor neuraxial analgesia use for specific subgroups, we conducted stratified analyses by estimating the adjusted difference-in-differences coefficient separately by women’s race and ethnicity and by parity.

Of the 4,515,667 birth certificates analyzed, 3,033,129 (67.2%) indicated use of labor neuraxial analgesia (table 1). Labor neuraxial analgesia utilization rates ranged from a minimum of 36.2% in the state of New Mexico to a maximum of 82.7%% in the state of Utah (Supplemental Figure 2, https://links.lww.com/ALN/D165). The labor neuraxial analgesia utilization rate was higher for women with private insurance (73.6%) than for Medicaid beneficiaries (62.9%). It was also higher for White women (73.8%) than for Black women (67.5%), Other race and ethnicity women (64.0%), or Hispanic women (54.7%), and for nulliparous women (74.5%) than for parous women (63.9%).

Comparison in each women’s age group of demographic, obstetrical, and hospital county characteristics between the pre- and the postintervention periods are presented in table 2. In each women’s age group, no significant difference was observed between the pre- and postintervention periods, except for a decrease in both groups in the unemployment rate and an increase in the proportion of persons in poverty in the county of residence during the postintervention period.

For women aged 21 to 25 yr, the labor neuraxial analgesia utilization rates were 64.9% during the preintervention period and 68.9% during the postintervention period (difference, 4.0%; 95% CI, 3.9 to 4.2; table 3, fig. 3). For women aged 27 to 31 yr, the labor neuraxial analgesia utilization rates were 64.9% during the preintervention period and 67.7% during the postintervention period (difference, 2.8%; 95% CI, 2.7 to 2.9). There was no evidence of difference in prepolicy labor neuraxial analgesia utilization trends between the two women’s age groups (fig. 3 and Supplemental Table 1, https://links.lww.com/ALN/D165).

Before adjustment, the Dependent Coverage Provision was associated with a statistically significant 1.2% increase in the labor neuraxial analgesia utilization rate among women aged 21 to 25 yr compared with women aged 27 to 31 yr (crude difference-in-differences estimator, 0.012; 95% CI, 0.011 to 0.014; table 3). After adjustment (Supplemental Table 3, https://links.lww.com/ALN/D165), the Dependent Coverage Provision was associated with a statistically significant 1.0% increase in the labor neuraxial analgesia utilization rate (adjusted difference-in-differences estimator, 0.010; 95% CI, 0.008 to 0.012). A similar finding was observed for early initiation of prenatal care (adjusted difference-in-differences estimator, 0.012; 95% CI, 0.010 to 0.014; Supplemental Table 4, https://links.lww.com/ALN/D165). In the post hoc analysis, we also observed a significant increase in the proportion of women with private health insurance coverage associated with the Dependent Coverage Provision (Supplemental Figure 3, https://links.lww.com/ALN/D165).

Results were robust in the sensitivity analysis excluding data for the year 2010 and in the sensitivity analysis using a narrower definition of the two women’s age groups (table 4). Results were also robust in the post hoc analyses restricting the postintervention period to 2012-2013 and to 2013, instead of 2010 to 2013 (Supplemental Table 5, https://links.lww.com/ALN/D165).

The increase in the labor neuraxial analgesia utilization rate associated with the Dependent Coverage Provision was significant among White women (adjusted difference-in-differences estimator, 0.008; 95% CI, 0.006 to 0.011), Hispanic women (adjusted difference-in-differences estimator, 0.007; 95% CI, 0.004 to 0.011), nulliparous women (adjusted difference-in-differences estimator, 0.012; 95% CI, 0.009 to 0.015), and parous women (adjusted difference-in-differences estimator, 0.010; 95% CI, 0.008 to 0.012), but not among Black women (adjusted difference-in-differences estimator, 0.004; 95% CI, –0.001 to 0.009) and women of Other race and ethnicity (adjusted difference-in-differences estimator, 0.006; 95% CI, –0.002 to 0.014; table 4).

In this natural experiment study of spontaneous vaginal deliveries from 2009 to 2013, we report that the Dependent Coverage Provision was associated with a 1.0% increase in the labor neuraxial analgesia utilization rate among women aged 21 to 25 yr compared with women aged 27 to 31 yr.

It is noteworthy that albeit statistically significant, this increase in the labor neuraxial analgesia utilization rate associated with the Dependent Coverage Provision is small and unlikely of clinical significance. This small effect size can be explained by the small increases in the proportion of women with private health insurance coverage and with early initiation of prenatal care (as a proxy for better prenatal education) associated with the Dependent Coverage Provision. Of note, the reported effect size for adverse obstetrical or maternal outcomes associated with health policies aiming to increase health insurance coverage is usually modest. For example, Daw and Sommers reported that the 2010 Dependent Coverage Provision was associated with a 0.20% (95% CI, 0.03 to 0.30) adjusted decrease in the odds of preterm birth in the United States;¹²  we reported that the 2014 Medicaid expansion was associated with a 0.42% (95% CI, 0.07 to 0.75) adjusted decrease in the odds of severe maternal morbidity in New York State.²²  The small increase in the labor neuraxial analgesia utilization rate associated with the Dependent Coverage Provision also suggests that health insurance coverage alone may not be sufficient to increase the utilization of labor neuraxial analgesia. Efforts focusing on the patient’s anticipated use of neuraxial analgesia may be more effective. For example, reducing misconceptions and fears (e.g., fear of needles or of chronic back pain) about neuraxial analgesia through hospital-based prenatal educational programs have been associated with reduced misconceptions about labor neuraxial analgesia and increased labor neuraxial analgesia use.¹⁴  Of note, increased labor neuraxial analgesia use associated with the Dependent Coverage Provision was statistically significant for White and Hispanic women but not for Black and Other race or ethnicity women. This may be explained by the lower proportion of parents with private health insurance coverage in these minority groups.²⁷ 

Previous research on the association of provision of health insurance coverage with labor neuraxial analgesia use is limited to one study by Xiao et al. on the 2014 Medicaid expansion authorized by the Patient Protection and Affordable Care Act.²⁸  The 2014 Medicaid expansion led to a significant reduction in insurance discontinuity for pregnant people, and to an increase in preconception, prenatal, and postpartum healthcare access and utilization.^10,11  Analyzing birth certificate data, Xiao et al. reported that the Medicaid expansion was associated with a 1.5% (95% CI, 0.2 to 2.9) increase in the labor neuraxial analgesia utilization rate among nulliparous women in states that expanded Medicaid compared to nulliparous women in states that did not expand Medicaid.²⁸  This effect size is similar to the one observed in our study, and although statistically significant, is also unlikely of clinical significance.

First, we have no information in the Natality file on health insurance coverage for parents of the women. Since only women whose parents have private health insurance coverage may have benefited from the Dependent Coverage Provision, our analysis should be restricted to these women. In the absence of data, we included everyone who was eligible based on age to benefit from the Dependent Coverage Provision. Therefore, the estimated increase in labor neuraxial analgesia utilization rate refers to the overall impact of the Dependent Coverage Provision on all women aged 21 to 25 yr irrespective of their beneficiary status. The increase in labor neuraxial analgesia utilization rate associated with the Dependent Coverage Provision among the beneficiaries might be greater than the estimated 1% as reported in this study. Second, the duration of the preintervention period (from January 2009 to August 2010, or 20 months) is relatively short, which may make the estimate of the preintervention labor neuraxial analgesia trends unreliable. However, the duration of the preintervention period in our study is similar to the one used in other studies on the effect of the Dependent Coverage Provision.^12,15  Third, our results are susceptible to unmeasured confounding. Some important variables known to influence labor neuraxial analgesia utilization, such as actual participation in prenatal educational programs and anticipated use of labor neuraxial analgesia, is not available in the Natality file.^14,29  Fourth, we estimated some hospital characteristics (e.g., number of physician anesthesiologists or certified registered nurse anesthetists) at the hospital–county level as a proxy for the hospital because the Natality file does not provide a hospital identifier. Furthermore, some hospital characteristics (e.g., teaching status) could not be estimated using the hospital–county—level data. Last, our study sample was limited to 28 states. Since labor neuraxial analgesia utilization rates vary markedly by state, the specific states that are included in the study may not necessarily be nationally representative.⁵  Therefore, our findings may not be directly generalizable to the nation as whole.

Implementation of the Dependent Coverage Provision was associated with a statistically significant increase in labor neuraxial analgesia use, but the small effect size is unlikely of clinical significance.

Dr. Guglielminotti is supported by grant R21 MD016414 from the National Institute on Minority Health and Health Disparities (Bethesda, Maryland) and grant R21 MH126096 from the National Institute of Mental Health, National Institutes of Health (Bethesda, Maryland).

Dr. Landau was a consultant for Pacira Pharmaceuticals Inc. (Parsippany-Troy Hills, New Jersey) and serves on the editorial board of the journal Regional Anesthesia and Pain Medicine. The other authors declare no competing interests.

Supplemental Digital Content, https://links.lww.com/ALN/D165

Supplemental Figure 1: Conceptual framework

Supplemental Figure 2: Labor neuraxial analgesia utilization rate according to states

Supplemental Figure 3: Temporal trends in the proportion of women with private health insurance

Supplemental Table 1: Parallel trend assumption

Supplemental Table 2: Variables used for the adjustment of the difference-in-differences estimator

Supplemental Table 3: Regression coefficients from the linear regression model used to estimate the adjusted difference-in-differences estimator

Supplemental Table 4: Early initiation of prenatal care according to women’s age groups and intervention periods

Supplemental Table 5: Post hoc analysis for labor neuraxial analgesia utilization rate using different durations of the postintervention period