Assessment of Racial and Ethnic Disparities in the Use of Medication to Treat Opioid Use Disorder Among Pregnant Women in Massachusetts

Key Points Question Do differences by maternal race and ethnicity exist in the use of methadone or buprenorphine medications for the treatment of opioid use disorder during pregnancy? Findings In this cohort study of 5247 women with opioid use disorder who delivered a live infant, black non-Hispanic and Hispanic women with opioid use disorder were significantly less likely to use any medication for treatment and were less likely to consistently use medication for treatment during pregnancy compared with white non-Hispanic women with opioid use disorder. Meaning This study found racial and ethnic disparities in the use of medications for the treatment of opioid use disorder during pregnancy among a large population-level sample of women with opioid use disorder in Massachusetts; further investigation is warranted to explore the factors associated with inequitable access to and receipt of medication.


Introduction
3][14][15][16][17] In addition, racial and ethnic disparities in the use of medications to treat OUD have been described, including differential access to buprenorphine prescribers by neighborhood, 18 greater buprenorphine prescription rates for white non-Hispanic individuals compared with black non-Hispanic individuals, 19 and less timely receipt of medication to treat OUD among black youths compared with white youths. 20e prenatal period offers an opportunity to assess disparities in treatment use, as most pregnant individuals are eligible for health insurance, federal regulations emphasize priority access to addiction treatment during pregnancy, and the use of medication for the treatment of OUD (in contrast with medically assisted withdrawal from opioids) is recommended by all professional societies and public health agencies. 2,3To our knowledge, 2 studies have examined the use of medications to treat OUD by race among pregnant women with OUD.Among a cohort of Medicaid enrollees in Pennsylvania, pregnant women of color were reported to be less likely to receive any medication to treat OUD, and among women receiving buprenorphine treatment, women of color had higher rates of early discontinuation and decreasing adherence to buprenorphine treatment during pregnancy compared with white non-Hispanic women. 9,21However, it remains uncertain if racial and ethnic health disparities persist after adjusting for health status. 22,235][26] Understanding at what point along the OUD treatment cascade (which includes OUD diagnosis, engagement in care, treatment use, and adherence in treatment) 27 racial and ethnic disparities may be present is an important first step to addressing potential inequities in the use of medication for the treatment of OUD.Therefore, the objective of our study was to explore the extent to which maternal race or ethnicity was associated with (1) any use of medication to treat OUD during pregnancy, (2) the duration of the use of medication to treat OUD during pregnancy, and (3) the type of medication used to treat OUD.Data were obtained from a population-level linked public health data set of women with OUD who delivered a live infant in Massachusetts.We hypothesized that, among women with OUD, white non-Hispanic women would be more likely to receive any medication for the treatment of OUD, more likely to consistently use medication to treat OUD during pregnancy, and more likely to receive buprenorphine treatment compared with their black non-Hispanic and Hispanic counterparts.

Design
We performed a retrospective analysis of a cohort identified through the Public Health Data Warehouse, which is a linked statewide data set.This data set was established as part of a 9][30]  the data structure, and the linkage rates across data sets has been previously described. 31This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.The institutional review board of Partners HealthCare reviewed this study and deemed it non-human subjects research that was exempt from the need for informed consent.

Participants
We identified Massachusetts residents who delivered a live infant with a documented gestational age of 20 weeks or more in Massachusetts using birth certificate data.Each participant's individual pregnancy period was calculated based on gestational age at delivery.We included women who delivered an infant between October

Outcomes
Our main outcomes were the use of any medication for the treatment of OUD, the extent of medication used to treat OUD, and the type of medication used to treat OUD.Because treatment data were reported monthly, any use of medication to treat OUD was defined as the receipt of buprenorphine or methadone treatment, starting with the month of conception and ending with the month of delivery.Data on the use of medication to treat OUD were identified from the following: (1) insurance claims for methadone treatment (Healthcare Common Procedure Coding System code H0020), (2) receipt of methadone treatment from state-funded treatment programs, and (3) filled prescriptions for buprenorphine or buprenorphine/naloxone identified from prescription monitoring program data.Naltrexone was not included in our definition of medication to treat OUD because it is not currently recommended for use during pregnancy.
The extent of the use of medication to treat OUD was defined as consistent (monthly medication to treat OUD that was measured for at least 6 months of treatment before delivery to estimate the number of women receiving treatment throughout their second and third trimesters), inconsistent (any medication to treat OUD in the year before delivery but with gaps in treatment months), and no medication (no indication of the receipt of methadone or buprenorphine treatment).The type of medication used to treat OUD was categorized as buprenorphine, methadone, or neither.Deliveries among women who received both buprenorphine and methadone therapies were classified as methadone, as most individuals transition from buprenorphine to methadone treatment when clinically indicated.

Exposures
Our primary exposure was maternal race and ethnicity, documented from self-reported birth certificate records.If race and ethnicity data were missing on the birth certificate (1.2% of deliveries) but available across the linked data set, that value was included (accounting for 0.7% of deliveries).
Women were categorized as white non-Hispanic, black non-Hispanic, Hispanic, or other race/ ethnicity.Multiracial individuals were assigned to the self-reported racial classification with the smallest total representation in the general population.Other races represented 1% of the sample and were excluded owing to small sample sizes.Additional maternal demographic variables included age at delivery, highest educational level, enrollment in Medicaid (MassHealth) during the month of delivery, marital status, and geographic location of residence (categorized as either rural or nonrural based on total population, density, census tract, or hospital licensure).
Psychosocial and health care use characteristics included a maternal diagnosis of anxiety or depression during pregnancy, an opioid prescription (excluding buprenorphine) that was filled in the 3 months before delivery (excluding delivery month), incarceration (release from a Massachusetts prison or jail during the study period), homelessness (during the study period), high use of unscheduled care (Ն3 emergency department and/or obstetric triage visits during pregnancy), and adequacy of prenatal care use (using the Kotelchuck Index from the birth certificate). 32

Statistical Analysis
We used descriptive statistics to compare the characteristics of our cohort by race/ethnicity and level of treatment engagement.Fisher exact and χ 2 tests were performed to compare across groups.In our multivariable models, we used statistical model-building criteria (ie, P < .05for both race/ ethnicity and any of the 3 medications for OUD outcomes; eTable 4 and eTable 5 in the Supplement) for their inclusion.Based on these criteria, we included age, rural residence, emergency department service use, and opioid prescriptions in the last 3 months of pregnancy.In addition, we included education and health insurance as an adjustment for socioeconomic status and maternal diagnosis of anxiety and/or depression as an adjustment for the association of mental health conditions and treatment receipt.We compared consistent and inconsistent use of medication with no use of medication for the treatment of OUD.Next, using nominal logistic regression analysis, we compared buprenorphine treatment with methadone treatment and buprenorphine treatment with no treatment.To assess the association of race/ethnicity with treatment engagement, we removed race/ethnicity and kept other maternal covariates in our model to calculate a pseudo-R 2 value before and after treatment using Nagelkerke mirrors. 33We assessed the significance of the interaction of race/ethnicity and all included covariates and retained significant interaction terms in the final model.When a significant interaction was identified, separate effect measures were presented for each level of the relevant covariate.
For the sensitivity analyses, we first performed an analysis excluding black non-Hispanic and Hispanic women to address the potential of a differential diagnosis of NAS by maternal and infant race/ethnicity because these women were more likely to be identified as having OUD by NAS diagnosis alone.Second, we performed an analysis including all individuals who had an OUD diagnosis code but had been excluded (no indication of a clinical need for medication to treat OUD).
Third, to determine whether the group of individuals receiving both methadone and buprenorphine therapies differed, we performed an analysis excluding this group.

Demographic Characteristics
Of 274 234 deliveries in Massachusetts resulting in a live birth, we identified 5247 deliveries to women with indicators of having OUD, after excluding deliveries owing to possible iatrogenic NAS, other race/ethnicity, no identified clinical need for medication to treat OUD, multiple deliveries in the study period, and missing variables (Figure 1).The mean (SD) age of all participants was 28.7 (5.0) years.Among deliveries to women with OUD, 4551 women (86.7%) were white non-Hispanic, 462 women (17.3%) were Hispanic, and 24 989 women (9.2%) were black non-Hispanic.Compared with white non-Hispanic women, black non-Hispanic and Hispanic women were older, had lower educational levels, were more likely to live in an urban area, had higher unscheduled emergency department use, were less likely to receive an opioid prescription in the 3 months before delivery, had lower use of publicly funded addiction programs, and were more likely to be identified in the cohort by an infant diagnosis of NAS alone (Table 1).Black non-Hispanic and Hispanic women also had lower rates of exclusive use of buprenorphine treatment.

Medication Use and Type
Overall, 3474 deliveries (66.2%) in our cohort were to women who received any medication for the treatment of OUD in the year before delivery: A total of 1999 women (38.1%) consistently used age for any treatment use and (2) maternal anxiety/depression for the extent of medication use and the type of medication used to treat OUD (Figure 2).Among those 25 years and younger, black non-Hispanic and Hispanic women were 0.23 times (95% CI, 0.14-0.38)and 0.29 times (95% CI, 0.20-0.42)more likely, respectively, to receive any medication for the treatment of OUD compared with white non-Hispanic women.Among women aged 26 to 34 years at delivery, black non-Hispanic and Hispanic women were 0.46 times (95% CI, 0.32-0.67)and 0.46 times (95% CI, 0.35-0.60)more likely, respectively, to receive any medication to treat OUD compared with white non-Hispanic women.Among women 35 years and older, black non-Hispanic and Hispanic women were 0.43 times (95% CI, 0.22-0.83)and 0.64 times (95% CI, 0.39-1.03)more likely, respectively, to receive any medication to treat OUD compared with white non-Hispanic women.than white non-Hispanic women of receiving buprenorphine treatment compared with methadone treatment.The results of stratifying by maternal diagnosis of anxiety or depression for both the extent of medication use and the type of medication used to treat OUD revealed that racial and ethnic differences were less substantial (consistent use of medication vs no use of medication to treat OUD and treatment with buprenorphine vs no treatment with medication) or nonsignificant (treatment with buprenorphine vs methadone) for women who had a psychiatric diagnosis (Figure 2).Among those without a diagnosis of anxiety or depression, black non-Hispanic and Hispanic women times (95% CI, 0.22-0.68)and 0.50 times (95% CI, 0.12-0.30)more likely, respectively, to receive consistent medication vs no medication to treat OUD.In addition, black non-Hispanic and Hispanic women with no anxiety or depression were 0.41 times (95% CI, 0.25-0.69)and 0.59 times (95% CI, 0.41-0.85)more likely, respectively, to receive buprenorphine treatment vs methadone treatment compared with white non-Hispanic women.Race/ethnicity explained only 2.7% to 3.0% of the total variance (using the Nagelkerke pseudo-R 2 ) in our models of the use and type of medication used to treat OUD.

Sensitivity Analysis
In our first sensitivity analysis, we excluded women identified as having OUD based on an infant diagnosis of NAS alone (n = 766) given the racial/ethnic differences in cohort inclusion based on this variable.The adjusted likelihood of the use of any medication to treat OUD in this group was 42% lower (95% CI, 17%-59%) for black non-Hispanic women and 34% lower (95% CI, 15%-49%) for Hispanic women than for white non-Hispanic women compared with the original cohort, in which the adjusted likelihood was 63% lower (95% CI, 51%-72%) for black non-Hispanic women and 58% lower (95% CI, 48%-65%) for Hispanic women than for white non-Hispanic women.In the second sensitivity analysis, we expanded our sample to include individuals without a clinical indication for medication (n = 5776), and our findings were similar.
The adjusted likelihood of the use of any medication to treat OUD in this larger group was 62% lower (95% CI, 50%-80%) for black non-Hispanic women and 57% lower (95% CI, 48%-64%) for Hispanic women than for white non-Hispanic women compared with the original cohort, in which the adjusted likelihood was 63% lower (95% CI, 51%-72%) for black non-Hispanic women and 58% lower (95% CI, 48%-65%) for Hispanic women.All outcomes and stratified models are available in eTable 6 and eTable 7 in the Supplement.We excluded individuals who received both methadone and buprenorphine therapies (n = 285), and no differences were found in the main models (eTable 8 in the Supplement).

Discussion
In this study of 5247 pregnant women with OUD who delivered a live infant in Massachusetts, we found that white non-Hispanic women were more likely to have a diagnosis of OUD than black non-Hispanic or Hispanic women.In our sample, the consistent receipt of medication to treat OUD was low among all groups.However, black non-Hispanic and Hispanic women were significantly less likely to receive any medication for the treatment of OUD or to consistently use medication to treat OUD in the 6 months before delivery, with and without adjusting for other maternal characteristics.
In addition, among those without depression or anxiety, black non-Hispanic and Hispanic women were significantly less likely to receive buprenorphine treatment compared with methadone treatment or no medication treatment.
Our findings support the analysis by Krans et al, 9 which indicated that, in a cohort of Medicaid enrollees in Pennsylvania, black non-Hispanic women (27%) and Hispanic women (36%) were less likely than white non-Hispanic women (59%) to receive any medication for the treatment of OUD and were more likely to receive methadone treatment than buprenorphine treatment.Our analyses extended this research by characterizing the extent to which these differences could be associated with race and ethnicity after controlling for other maternal characteristics, and we observed that these disparities were even greater among younger women.Furthermore, we found that black non-Hispanic and Hispanic women were less likely to consistently use medication for the treatment of OUD before delivery, suggesting that not only do disparities exist in treatment initiation but they may also be observed in treatment continuation during pregnancy.
We identified that white non-Hispanic women with OUD were more likely to use buprenorphine compared with black non-Hispanic or Hispanic women, a finding similar to that of the Lagisetty et al 19 study of buprenorphine prescriptions in a primary care setting.Our analysis, however, was strengthened by accounting for OUD prevalence by racial/ethnic group in our population-based sample.Krawczyk et al 34 found that black non-Hispanic and Hispanic clients were more likely to access medication for the treatment of OUD from publicly funded treatment programs than were Our study benefited from having a robust measure for both buprenorphine and methadone treatments; we found that methadone treatment was similar across all racial/ethnic groups, but we identified a lower use of buprenorphine treatment among black non-Hispanic and Hispanic women without depression or anxiety compared with white non-Hispanic women.6][37] We hypothesize that women with depression/anxiety may have been more engaged in medical or psychiatric care for the treatment of their depression and thus initiated buprenorphine treatment at similar rates, accounting for the similar receipt of officebased buprenorphine treatment among racial/ethnic groups.The reasons underlying inequitable medication use are not well understood, but Hansen and Netherland 38 and Hansen et al 39 have suggested that the marketing campaigns of buprenorphine manufacturers have specifically targeted white individuals and that fewer programs and clinicians providing buprenorphine treatment are located in low-income communities of color in New York City. 18,40 importance, race/ethnicity explained only 2.7% to 3.0% of the total variance (using the Nagelkerke pseudo-R 2 ) in our models of the use and type of medication used to treat OUD, suggesting that many unmeasured factors are associated with treatment engagement and adherence during pregnancy.Pregnancy represents a potentially 9-month opportunity during which frequent engagement with the health care system can support assessment, medication initiation, and continued engagement in services.We identified higher rates of the use of medication to treat OUD during pregnancy than after other high-risk events, such as a single encounter for nonfatal overdose. 41However, additional investigation is needed to better understand why one-third of the women with OUD in this cohort were not treated with any medication; further research that includes an examination of maternal age, marital status, insurance status, and geography, which all were associated with differences in the use of medication to treat OUD in our sample, is warranted.
3][44] In addition, it is necessary to further elucidate the treatment trajectories of pregnant women with OUD.Lo-Ciganic et al 21 characterized distinct trajectories of the use of medication to treat OUD, finding that more than 25% of women who initiate treatment report low adherence or early discontinuation.
We hypothesize that a confluence of current and historical factors may be associated with our findings.[51][52] Persistent racial inequities in maternal morbidity and mortality, even after adjusting for other maternal comorbid conditions, suggest that structural racism may be associated with a lower standard of care and fewer treatment options for women of color. 24,53,54

Limitations
Our study has several limitations.First, our findings may not be generalizable outside of Massachusetts, a state that provides increased services for pregnant and postpartum women with OUD.Second, race and ethnicity are proxies for a complex number of factors, including not only Assessment of Racial/Ethnic Disparities in Use of Medication for Opioid Use Disorder in Pregnancy JAMA Network Open.2020;3(5):e205734.doi:10.1001/jamanetworkopen.2020.5734(Reprinted) May 26, 2020 2/15 Downloaded From: https://jamanetwork.com/ on 08/04/2020

First, we used
logistic regression analysis to estimate the strength of the association between maternal race/ethnicity and any use of medication to treat OUD.Second, we used nominal logistic JAMA Network Open | Obstetrics and Gynecology Assessment of Racial/Ethnic Disparities in Use of Medication for Opioid Use Disorder in Pregnancy JAMA Network Open.2020;3(5):e205734.doi:10.1001/jamanetworkopen.2020.5734(Reprinted) May 26, 2020 4/15 Downloaded From: https://jamanetwork.com/ on 08/04/2020 regression to examine the extent of treatment use and the association with maternal race/ethnicity.

Figure 2 .A
Figure 2. Adjusted Odds Ratios for Extent of Medication Use and Type of Medication Used for Treatment of OUD in Pregnant Women

JAMA Network Open | Obstetrics and Gynecology Assessment
of Racial/Ethnic Disparities in Use of Medication for Opioid Use Disorder in Pregnancy 1, 2011, and December 31, 2015, to allow for 9 months of treatment data before delivery.Data on fetal deaths were not available.Singleton and multiple births were included, with multiple births treated as a single delivery episode.When a person had multiple deliveries during the study period, only the first delivery in the period was included.The birth certificate linkage rate with the main data set (ie, the all-payer claims database) for our study period was 91.7%.

Table 1 .
Characteristics of Pregnant Women With Opioid Use Disorder by Race/Ethnicity a Assessment of Racial/Ethnic Disparities in Use of Medication for Opioid Use Disorder in Pregnancy b Values of fewer than 11 deliveries were not included in accordance with privacy rules.c At any time from October 1, 2011, to December 31, 2015.JAMA Network Open | Obstetrics and Gynecology JAMA Network Open.2020;3(5):e205734.doi:10.1001/jamanetworkopen.2020.5734(Reprinted) May 26, 2020 6/15 Downloaded From: https://jamanetwork.com/ on 08/04/2020

JAMA Network Open | Obstetrics and Gynecology Assessment
of Racial/Ethnic Disparities in Use of Medication for Opioid Use Disorder in Pregnancy Hispanic clients, but most of these programs dispensed only methadone and had no data available on buprenorphine prescriptions.