Neonatal Outcomes in a Medicaid Population With Opioid Dependence

Neonatal Outcomes in a Medicaid Population With Opioid Dependence Abstract Confounding may account for the apparently improved infant outcomes after prenatal exposure to buprenorphine versus methadone. We used Massachusetts Medicaid Analytic eXtract (MAX) data to identify a cohort of opioid-dependent mother-infant pairs (2006–2011), supplemented with confounder data from an external Boston, Massachusetts, cohort (2015–2016). Associations between prenatal buprenorphine exposure versus methadone exposure and infant outcomes in the MAX cohort were adjusted for measured MAX confounders and were additionally adjusted for unmeasured confounders with bias analysis using external cohort data. A total of 477 women in MAX were treated with methadone and 543 with buprenorphine. More buprenorphine users than methadone users were white and used psychotropic medications. After adjustment for MAX confounders, risk ratios among infants exposed to buprenorphine versus those exposed to methadone were 0.45 (95% confidence interval (CI): 0.34, 0.61) for preterm birth (birth at <37 weeks) and 0.75 (95% CI: 0.51, 1.11) for low birth weight for gestational age. The mean difference in infant hospitalization was −7.35 days (95% CI: −9.16, −5.55). After further adjustment with bias analysis, the risk ratios were 0.53 (95% CI: 0.39, 0.71) for preterm birth and 1.14 (95% CI: 0.77, 1.69) for low birth weight for gestational age, and the mean difference in infant hospitalization was −3.66 days (95% CI: −5.46, −1.87). External confounder data can be used to adjust for unmeasured confounding in studies of prenatal outcomes among women on opioid agonist therapy based on administrative databases. buprenorphine, confounding factors (epidemiology), methadone, neonatal abstinence syndrome, opioids, pregnancy Editor’s note: An invited commentary on this article appears on page 1162. High rates of opioid dependence in pregnant women and associated birth outcomes are major public health issues (1). Prenatal opioid agonist therapy (OAT) with buprenorphine, a partial opioid agonist, or methadone, a full opioid agonist, is recommended for pregnant opioid-dependent women (2). Some observational studies have suggested improved fetal growth for prenatal treatment with buprenorphine as compared with methadone, including higher gestational age at birth (3–6), higher birth weight (3, 5, 7), and fewer infants born small for gestational age (8). Pregnant opioid-dependent women with better clinical and psychosocial profiles are more likely to be treated with buprenorphine, raising the possibility that associations are confounded (9, 10). In a clinical trial, Jones et al. (6) found shorter hospital stays and less morphine treatment for gastrointestinal, respiratory, autonomic, and central nervous system disturbances from opioid withdrawal after delivery (neonatal abstinence syndrome) among infants prenatally exposed to buprenorphine as compared with infants prenatally exposed to methadone, but high dropout rates in that study raised concerns about selection bias. Administrative health-care databases can be an efficient data source for epidemiologic studies due to the reduced time and cost otherwise required to assemble large cohorts. Use of psychotropic medication during pregnancy has been studied in Medicaid beneficiaries (11, 12). Massachusetts Medicaid data are efficient for studying OAT, because OAT coverage is virtually unrestricted in the state (13). Most clinic-based studies of prenatal OAT have been small, with fewer than 300 mother-infant pairs (9). To address small sample sizes and the possibility of confounding in prior studies, we designed a study that used extant Medicaid data paired with data from a prospective cohort study and bias analysis methods to examine the safety of buprenorphine as compared with methadone in terms of infant outcomes. Herein, we describe the methods, limitations, and findings of this novel approach. METHODS Medicaid Analytic eXtract cohort Study population The study population was identified using Medicaid Analytic eXtract (MAX) data (14) from Massachusetts. Women were included if they were 14 years of age or older, delivered between 2006 (when prenatal buprenorphine became available in Massachusetts) and 2011 (most recent MAX data) (14), and 1) had a Medicaid claim with a diagnosis of opioid dependence within 2 years before delivery or 2) had a Medicaid claim with an unspecified type of drug dependence within 2 years before delivery and were on prenatal OAT (see Web Table 1, available at https://academic.oup.com/aje, for codes). Inpatient and outpatient deliveries were identified using previously identified codes (15), in addition to others (Web Table 1). Mothers were linked to infants by matching family case identifiers (Figure 1) and an infant birth date that fell within the delivery date window (15). Medicaid Statistical Information System Identifiers (MSIS ID) identify unique individuals within the Medicaid system (Figure 1). Multifetal deliveries may have a different MSIS ID for each infant, and some infants (singleton or multifetal) may have multiple MSIS IDs because of assignment of temporary MSIS IDs at birth followed by a permanent MSIS ID (15). We therefore could not distinguish repeat records of the same infant with different MSIS IDs and multiple births from the same delivery; thus, 1 infant per delivery was retained. These criteria were applied to MAX inpatient, other therapy, patient summary, and prescription databases provided by the Centers for Medicare and Medicaid Services. Per the Confidentiality of Alcohol and Drug Abuse Patient Records Regulations (16), the Centers for Medicare and Medicaid Services does not release postal codes and sets birth dates to the first day of the month for the given month and year of birth. Figure 1. View largeDownload slide Linkage of mother-infant pairs in Medicaid Analytic eXtract (MAX) data provided by the Centers for Medicare and Medicaid Services (CMS) for deliveries to opioid-dependent women in Massachusetts, 2006–2011. Case ID, family case identifier; MSIS ID, Medicaid Statistical Information System Identifiers; OAT, opioid agonist therapy; PS, personal summary. Figure 1. View largeDownload slide Linkage of mother-infant pairs in Medicaid Analytic eXtract (MAX) data provided by the Centers for Medicare and Medicaid Services (CMS) for deliveries to opioid-dependent women in Massachusetts, 2006–2011. Case ID, family case identifier; MSIS ID, Medicaid Statistical Information System Identifiers; OAT, opioid agonist therapy; PS, personal summary. Prenatal OAT exposure The prenatal OAT exposure period was defined as the period from 294 days before the matched infant’s birth date through 30 days after the matched infant’s birth date. A window of 30 days after the infant’s birth date was used because birth dates were set to the first day of the month by the Centers for Medicare and Medicaid Services. Prescriptions for prenatal buprenorphine, selective serotonin reuptake inhibitors (SSRIs), and benzodiazepines were identified in the Medicaid prescription claims database (National Drug Codes are provided in Web Table 1). Methadone was identified with procedure code H0020 in the other therapy database. A small number of women switched treatment from buprenorphine to methadone (n = 26) and vice versa (n = 1); an intent-to-treat approach with the initial OAT for classification was used. Medicaid beneficiaries can be enrolled in a managed-care plan (either capitated or fee-for-service primary-care case management) or no managed-care plan (17). To examine the completeness of OAT data, we identified women who, at any point during pregnancy, had fee-for-service primary-care case management, no managed-care plan, or another managed-care plan that would not restrict prescription benefits. We also identified women who, at any point during pregnancy, had private health insurance or capitated Medicaid benefits that potentially would not submit prescription claims through Medicaid. Finally, we determined the proportion of women with prenatal prescriptions other than OAT. Infant outcomes The length of infant hospitalization was calculated as the difference between the starting and ending dates of the infant’s hospitalization for the first hospitalization that began in the infant’s birth month. Preterm birth (birth at <37 weeks’ gestation) (18) and low birth weight for gestational age were identified through International Classification of Diseases, Ninth Revision (ICD-9) codes in infant records (Web Table 1). Statistical approach Differences between women and infants who matched and those who did not were assessed statistically with the χ2, Fischer exact, or Kruskal-Wallis test, as appropriate. Risk ratios for preterm birth and low birth weight according to prenatal OAT were estimated using generalized linear models. Mean differences in the length of infant hospitalization were estimated with linear regression. We adjusted for potential confounding from maternal age (years; continuous), race/ethnicity (non-Hispanic white vs. other), year of delivery recorded in the MAX data (categorical, with 2006 as the referent), prenatal SSRI use before OAT (yes/no), and prenatal benzodiazepine use before OAT (yes/no). External cohort Study population Pregnant opioid-dependent women (n = 113) from a high-risk obstetrical and addiction recovery clinic in Boston, Massachusetts—the largest-volume clinic for pregnant women with opioid dependence in New England—were enrolled in the study (2015–2016) (19) to measure confounders unavailable in MAX. Women were recruited at OAT initiation or the first prenatal visit (for women on OAT at conception). The clinic operated in conjunction with methadone clinics, counseling centers, and residential treatment programs for pregnant women. Maternal criteria for OAT included being 18 years of age or older, having a Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, diagnosis of opioid dependence, having laboratory and/or radiographic documentation of pregnancy, and voluntarily consenting to engage in the clinic’s comprehensive treatment program. Per the clinical practice protocol, women were scheduled for prenatal care and relapse prevention visits every 1–3 weeks through delivery. Observed urine drug tests were carried out at each prenatal visit. The type of OAT initiated was selected considering patient choice, treatment history, disease severity, and medical and psychiatric comorbidity. The clinic protocol recommended methadone for women with more severe opioid dependence, women with intolerance to buprenorphine, and women for whom buprenorphine was medically contraindicated. In general, women for whom previous buprenorphine treatment was nonefficacious, who had a history of buprenorphine diversion within the last 6 months, who were unable to make weekly-to-biweekly office visits, or who were unwilling to engage in independent relapse prevention counseling were offered methadone (19). Data collection At cohort enrollment, a research assistant administered the Addiction Severity Index, 5th Edition Lite (20), to collect data on a wide range of potential confounders. The Addiction Severity Index has been widely employed among substance users and includes questions on 7 domains: medical status, employment and support, drug use, alcohol use, legal status, family/social status, and psychiatric status. Responses for the particular domain questions were summed to create a score that captured the degree of patient-reported problems in each domain, with higher scores indicating more severe problems (20). Prenatal OAT, pregnancy, substance use, and delivery information was routinely abstracted from maternal medical charts. Infant outcomes from birth to hospital discharge were abstracted from medical charts. Propensity scores Propensity scores for the probability of initiating buprenorphine versus methadone in the external cohort were estimated using logistic regression. A priori confounders included maternal demographic, clinical, and substance use characteristics based on substantive knowledge and prior studies. Characteristics retained in the final multivariate propensity score model were predictive of initial OAT and associated with an infant outcome of interest. They included: generally having unstable housing during the past 3 years (yes/no), gravidity (>1 pregnancy vs. ≤1), having ever regularly injected cocaine (yes/no), and Addiction Severity Index drug score (continuous, ranging from 0 to 1). No data were missing for the propensity score models. Associations between propensity scores and infant outcomes were restricted to infants for whom outcome data were available; some external cohort pregnancies ended through therapeutic abortion (n = 3) or miscarriage (n = 4) or had an unknown outcome, as some women did not continue prenatal care (n = 26). Bias analysis for unmeasured confounders in the MAX cohort Regression models were fitted with propensity scores as the independent variable and each infant outcome as the dependent variable using the external cohort data. The resultant risk ratio or mean difference quantified the amount of unmeasured confounding that would be expected in the crude MAX cohort results (the parameters RRγ and γ, described below). Mean values of the propensity score also were determined in the buprenorphine and methadone groups in the external cohort. Following the approach of Lin et al. (21), the MAX cohort risk ratio (RR) and mean difference (MD) were then corrected for unmeasured confounding (assuming a uniform effect of confounding by OAT) as follows:   RRcorrected=RR*[RRγ(μ1−μ0)]and   MDcorrected=MD*−[(μ1−μ0)γ)],where RR* and MD* are measures of association for buprenorphine versus methadone and the infant outcomes in the MAX cohort, adjusted for measured MAX confounders; RRγ and γ represent the effect of the unmeasured confounder (i.e., propensity score, which included confounders unavailable in MAX) on the infant outcome in the external cohort; and μ1 and μ0 are the mean values of the unmeasured confounder in the buprenorphine and methadone groups of the external cohort, respectively. The 95% confidence intervals for the bias-adjusted estimates were calculated using the standard errors of the MAX estimates (21). The small number of MAX infants with a missing inpatient hospitalization record or with insufficient data to evaluate length of stay were excluded from analyses of the relevant outcome. We also examined potential misclassification of MAX data. To evaluate the robustness of our cohort inclusion criterion, we determined the number of women with a diagnosis of opioid dependence within the year before delivery and assessed the OAT use distribution among these women. To examine bias from possible inclusion of postnatal OAT due to an unknown exact date of delivery (i.e., because birth dates were set to the first day of the month by the Centers for Medicare and Medicaid Services), we reran our infant-outcomes analyses with the pregnancy period ending on the first day of the month rather than using a window of 30 days after the infant’s birth date. RESULTS From 2006 to 2011, 5,348 women with 1 or more delivery windows and 5,755 infant records (4,357 infants with 5,755 different MSIS ID/family case identifier combinations) were available for linkage in the MAX cohort (Figure 1). A total of 5,010 infant records (87%) matched to a delivery record; all infants matched to only 1 woman. A total of 2,237 women (41.8%) with 1 or more delivery records did not match to an infant. There was no difference in the mean age (27.5 years vs. 27.8 years), race/ethnicity (70.8% white vs. 70.6%), or OAT status (14.9% treated with buprenorphine vs. 16.0% treated) of women who did and did not match to an infant. Women who did not match to an infant were more likely to deliver in earlier years (35.3% in 2006–2007 among those who did not match vs. 26.6% in 2006–2007 among those who matched). The 12 infants who did not match to a maternal record and the 4,347 infants who did match were similar in terms of year of birth, preterm birth (8.3% vs. 22.8%), and low birth weight (8.3% vs. 10.2%) and had similar mean lengths of hospital stay (22.9 days vs. 13.9 days). After removal of 1,578 duplicate records/multifetal births (1 infant per delivery was retained), 1 mother with an implausible maternal age at delivery, and women who did not meet our definition of opioid dependence (n = 962), there were 2,469 mother-infant pairs in the MAX cohort study population (Figure 1). Among these women, 2,012 women had 1 delivery, 204 had 2 deliveries, 15 had 3 deliveries, and 1 had 4. There was a 3-fold increase in the number of deliveries among opioid-dependent women covered by Medicaid in Massachusetts from 2006 (n = 205) to 2011 (n = 688). A total of 477 pregnant women were treated with methadone and 543 with buprenorphine; 1,449 did not have a record of OAT (unexposed; see Table 1). More unexposed and methadone-exposed women were black or Hispanic (Table 1). Almost all women had unrestricted Medicaid coverage at some point during pregnancy, and 7.5% of unexposed, 9.4% of buprenorphine-exposed, and 12.6% of methadone-exposed women had private health insurance at some point during pregnancy. Women treated with buprenorphine were more likely to have SSRI and benzodiazepine prescriptions than methadone-treated or unexposed women. Methadone-treated women, however, had the lowest prevalence of any prenatal prescription (77.6%). Unexposed mother-infant pairs were more likely to deliver in earlier years. Unexposed mother-infant pairs were excluded from analysis of the association of buprenorphine versus methadone with infant outcomes. Table 1. Characteristics and Infant Outcomes of Massachusetts Mother-Infant Pairs in a MAX Medicaid Cohort With a Diagnosis of Drug Dependence Within 2 Years of Delivery (2006–2011), by Prenatal OAT Statusa Characteristic or Outcome  OAT Status  Buprenorphine (n = 543)  Methadone (n = 477)  No OAT (1,449)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Maternal Characteristics  Year of delivery   2006  22  4.1  38  8.0  145  10.0   2007  53  9.8  57  12.0  208  14.4   2008  78  14.4  53  11.1  234  16.2   2009  95  17.5  74  15.5  273  18.8   2010  97  17.9  86  18.0  268  18.5   2011  198  36.5  169  35.4  321  22.2  Maternal age at delivery, yearsb  27.9 (4.5)  28.2 (4.9)  27.9 (5.0)  Maternal race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  60  4.1   Non-Hispanic white  426  78.5  366  76.7  1,105  76.3   Hispanic  15  2.8  23  4.8  77  5.3   Other/unknownc  100  18.4  75  15.7  207  14.3  Diagnosis of unspecified drug dependence only  24  4.4  0  0  0  0  Any unrestricted Medicaid coverage during pregnancy  541  99.6  475  99.6  1,445  99.7  Any private health insurance during pregnancy  51  9.4  60  12.6  109  7.5  No. of prenatal prescription medications other than OAT   0  57  10.5  107  22.4  273  18.8   1  19  3.5  31  6.5  55  3.8   2–3  54  9.9  81  17.0  141  9.7   ≥4  413  76.1  258  54.1  980  67.6  SSRI prescription  158  29.1  77  16.1  299  20.6  Benzodiazepine prescription  173  31.9  122  25.6  334  23.1  Cesarean delivery  187  34.3  179  37.5  494  34.1  Infant Outcomes  Infant sex   Female  260  47.9  234  49.1  692  47.8   Male  283  52.1  243  50.9  757  52.2  Preterm birth (<37 weeks)  99  18.4  155  32.9  351  24.4  Low birth weight for gestational age  54  10.9  61  13.0  153  10.7  Length of infant hospital stay, daysb  13.9 (12.6)  21.4 (15.7)  15.7 (15.0)  Characteristic or Outcome  OAT Status  Buprenorphine (n = 543)  Methadone (n = 477)  No OAT (1,449)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Maternal Characteristics  Year of delivery   2006  22  4.1  38  8.0  145  10.0   2007  53  9.8  57  12.0  208  14.4   2008  78  14.4  53  11.1  234  16.2   2009  95  17.5  74  15.5  273  18.8   2010  97  17.9  86  18.0  268  18.5   2011  198  36.5  169  35.4  321  22.2  Maternal age at delivery, yearsb  27.9 (4.5)  28.2 (4.9)  27.9 (5.0)  Maternal race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  60  4.1   Non-Hispanic white  426  78.5  366  76.7  1,105  76.3   Hispanic  15  2.8  23  4.8  77  5.3   Other/unknownc  100  18.4  75  15.7  207  14.3  Diagnosis of unspecified drug dependence only  24  4.4  0  0  0  0  Any unrestricted Medicaid coverage during pregnancy  541  99.6  475  99.6  1,445  99.7  Any private health insurance during pregnancy  51  9.4  60  12.6  109  7.5  No. of prenatal prescription medications other than OAT   0  57  10.5  107  22.4  273  18.8   1  19  3.5  31  6.5  55  3.8   2–3  54  9.9  81  17.0  141  9.7   ≥4  413  76.1  258  54.1  980  67.6  SSRI prescription  158  29.1  77  16.1  299  20.6  Benzodiazepine prescription  173  31.9  122  25.6  334  23.1  Cesarean delivery  187  34.3  179  37.5  494  34.1  Infant Outcomes  Infant sex   Female  260  47.9  234  49.1  692  47.8   Male  283  52.1  243  50.9  757  52.2  Preterm birth (<37 weeks)  99  18.4  155  32.9  351  24.4  Low birth weight for gestational age  54  10.9  61  13.0  153  10.7  Length of infant hospital stay, daysb  13.9 (12.6)  21.4 (15.7)  15.7 (15.0)  Abbreviations: MAX, Medicaid Analytic eXtract; OAT, opioid agonist therapy; SSRI, selective serotonin reuptake inhibitor. a Twenty-six women initiated treatment with buprenorphine and switched to methadone, and 1 woman initiated treatment with methadone and switched to buprenorphine. Five infants exposed to buprenorphine, 6 exposed to methadone, and 9 unexposed infants were missing information from inpatient hospitalization (preterm birth, low birth weight for gestational age); 25 infants exposed to buprenorphine, 44 infants exposed to methadone, and 109 unexposed infants were missing information on length of hospital stay. b Values are expressed as mean (standard deviation). c “Other” includes non-Hispanic women with unknown race/ethnicity. Table 1. Characteristics and Infant Outcomes of Massachusetts Mother-Infant Pairs in a MAX Medicaid Cohort With a Diagnosis of Drug Dependence Within 2 Years of Delivery (2006–2011), by Prenatal OAT Statusa Characteristic or Outcome  OAT Status  Buprenorphine (n = 543)  Methadone (n = 477)  No OAT (1,449)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Maternal Characteristics  Year of delivery   2006  22  4.1  38  8.0  145  10.0   2007  53  9.8  57  12.0  208  14.4   2008  78  14.4  53  11.1  234  16.2   2009  95  17.5  74  15.5  273  18.8   2010  97  17.9  86  18.0  268  18.5   2011  198  36.5  169  35.4  321  22.2  Maternal age at delivery, yearsb  27.9 (4.5)  28.2 (4.9)  27.9 (5.0)  Maternal race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  60  4.1   Non-Hispanic white  426  78.5  366  76.7  1,105  76.3   Hispanic  15  2.8  23  4.8  77  5.3   Other/unknownc  100  18.4  75  15.7  207  14.3  Diagnosis of unspecified drug dependence only  24  4.4  0  0  0  0  Any unrestricted Medicaid coverage during pregnancy  541  99.6  475  99.6  1,445  99.7  Any private health insurance during pregnancy  51  9.4  60  12.6  109  7.5  No. of prenatal prescription medications other than OAT   0  57  10.5  107  22.4  273  18.8   1  19  3.5  31  6.5  55  3.8   2–3  54  9.9  81  17.0  141  9.7   ≥4  413  76.1  258  54.1  980  67.6  SSRI prescription  158  29.1  77  16.1  299  20.6  Benzodiazepine prescription  173  31.9  122  25.6  334  23.1  Cesarean delivery  187  34.3  179  37.5  494  34.1  Infant Outcomes  Infant sex   Female  260  47.9  234  49.1  692  47.8   Male  283  52.1  243  50.9  757  52.2  Preterm birth (<37 weeks)  99  18.4  155  32.9  351  24.4  Low birth weight for gestational age  54  10.9  61  13.0  153  10.7  Length of infant hospital stay, daysb  13.9 (12.6)  21.4 (15.7)  15.7 (15.0)  Characteristic or Outcome  OAT Status  Buprenorphine (n = 543)  Methadone (n = 477)  No OAT (1,449)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Maternal Characteristics  Year of delivery   2006  22  4.1  38  8.0  145  10.0   2007  53  9.8  57  12.0  208  14.4   2008  78  14.4  53  11.1  234  16.2   2009  95  17.5  74  15.5  273  18.8   2010  97  17.9  86  18.0  268  18.5   2011  198  36.5  169  35.4  321  22.2  Maternal age at delivery, yearsb  27.9 (4.5)  28.2 (4.9)  27.9 (5.0)  Maternal race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  60  4.1   Non-Hispanic white  426  78.5  366  76.7  1,105  76.3   Hispanic  15  2.8  23  4.8  77  5.3   Other/unknownc  100  18.4  75  15.7  207  14.3  Diagnosis of unspecified drug dependence only  24  4.4  0  0  0  0  Any unrestricted Medicaid coverage during pregnancy  541  99.6  475  99.6  1,445  99.7  Any private health insurance during pregnancy  51  9.4  60  12.6  109  7.5  No. of prenatal prescription medications other than OAT   0  57  10.5  107  22.4  273  18.8   1  19  3.5  31  6.5  55  3.8   2–3  54  9.9  81  17.0  141  9.7   ≥4  413  76.1  258  54.1  980  67.6  SSRI prescription  158  29.1  77  16.1  299  20.6  Benzodiazepine prescription  173  31.9  122  25.6  334  23.1  Cesarean delivery  187  34.3  179  37.5  494  34.1  Infant Outcomes  Infant sex   Female  260  47.9  234  49.1  692  47.8   Male  283  52.1  243  50.9  757  52.2  Preterm birth (<37 weeks)  99  18.4  155  32.9  351  24.4  Low birth weight for gestational age  54  10.9  61  13.0  153  10.7  Length of infant hospital stay, daysb  13.9 (12.6)  21.4 (15.7)  15.7 (15.0)  Abbreviations: MAX, Medicaid Analytic eXtract; OAT, opioid agonist therapy; SSRI, selective serotonin reuptake inhibitor. a Twenty-six women initiated treatment with buprenorphine and switched to methadone, and 1 woman initiated treatment with methadone and switched to buprenorphine. Five infants exposed to buprenorphine, 6 exposed to methadone, and 9 unexposed infants were missing information from inpatient hospitalization (preterm birth, low birth weight for gestational age); 25 infants exposed to buprenorphine, 44 infants exposed to methadone, and 109 unexposed infants were missing information on length of hospital stay. b Values are expressed as mean (standard deviation). c “Other” includes non-Hispanic women with unknown race/ethnicity. The risks of preterm birth and low birth weight were higher among methadone-exposed infants (Table 2), as was the mean length of hospital stay. Adjustment for maternal age, race/ethnicity, year of delivery, and prenatal SSRI and benzodiazepine use in the MAX cohort did not change estimates, but incorporating data from the external cohort altered the associations. Crude risk ratios, risk ratios adjusted for MAX confounders only, and risk ratios adjusted for MAX confounders and bias analysis were as follows: for preterm birth, risk ratio (RR) = 0.46 (95% confidence interval (CI): 0.34, 0.61), RR = 0.45 (95% CI: 0.34, 0.61), and RR = 0.53 (95% CI: 0.39, 0.71), respectively; for low birth weight for gestational age, RR = 0.75 (95% CI: 0.51, 1.11), RR = 0.75 (95% CI: 0.51, 1.11), and RR = 1.14 (95% CI: 0.77, 1.69), respectively. Corresponding mean differences in the length of infant hospitalization were −7.45 days (95% CI: −9.25, −5.65), −7.35 days (95% CI: −9.16, −5.55), and −3.66 days (95% CI: −5.46, −1.87), respectively. Table 2. Unadjusted, Adjusted, and Externally Adjusted Associations Between Prenatal Buprenorphine Exposure Versus Methadone Exposure and Infant Outcomes Among Massachusetts Medicaid Beneficiaries Who Delivered in 2006–2011 Infant Outcome  Unadjusted  Adjusted Using MAX Confounder Dataa  Adjusted Using MAX Confounder Dataa and External Confounder Data With Bias Analysis  PE  95% CI  PE  95% CI  PE  95% CI  Preterm birth (<37 weeks)  0.46  0.34, 0.61  0.45  0.34, 0.61  0.53  0.39, 0.71  Low birth weight for gestational age  0.75  0.51, 1.11  0.75  0.51, 1.11  1.14  0.77, 1.69  Length of infant hospital stay, days  −7.45  −9.25, −5.65  −7.35  −9.16, −5.55  −3.66  −5.46, −1.87  Infant Outcome  Unadjusted  Adjusted Using MAX Confounder Dataa  Adjusted Using MAX Confounder Dataa and External Confounder Data With Bias Analysis  PE  95% CI  PE  95% CI  PE  95% CI  Preterm birth (<37 weeks)  0.46  0.34, 0.61  0.45  0.34, 0.61  0.53  0.39, 0.71  Low birth weight for gestational age  0.75  0.51, 1.11  0.75  0.51, 1.11  1.14  0.77, 1.69  Length of infant hospital stay, days  −7.45  −9.25, −5.65  −7.35  −9.16, −5.55  −3.66  −5.46, −1.87  Abbreviations: CI, confidence interval; MAX, Medicaid Analytic eXtract; PE, point estimate. a Adjusted for maternal age, maternal race/ethnicity, and year of delivery. Table 2. Unadjusted, Adjusted, and Externally Adjusted Associations Between Prenatal Buprenorphine Exposure Versus Methadone Exposure and Infant Outcomes Among Massachusetts Medicaid Beneficiaries Who Delivered in 2006–2011 Infant Outcome  Unadjusted  Adjusted Using MAX Confounder Dataa  Adjusted Using MAX Confounder Dataa and External Confounder Data With Bias Analysis  PE  95% CI  PE  95% CI  PE  95% CI  Preterm birth (<37 weeks)  0.46  0.34, 0.61  0.45  0.34, 0.61  0.53  0.39, 0.71  Low birth weight for gestational age  0.75  0.51, 1.11  0.75  0.51, 1.11  1.14  0.77, 1.69  Length of infant hospital stay, days  −7.45  −9.25, −5.65  −7.35  −9.16, −5.55  −3.66  −5.46, −1.87  Infant Outcome  Unadjusted  Adjusted Using MAX Confounder Dataa  Adjusted Using MAX Confounder Dataa and External Confounder Data With Bias Analysis  PE  95% CI  PE  95% CI  PE  95% CI  Preterm birth (<37 weeks)  0.46  0.34, 0.61  0.45  0.34, 0.61  0.53  0.39, 0.71  Low birth weight for gestational age  0.75  0.51, 1.11  0.75  0.51, 1.11  1.14  0.77, 1.69  Length of infant hospital stay, days  −7.45  −9.25, −5.65  −7.35  −9.16, −5.55  −3.66  −5.46, −1.87  Abbreviations: CI, confidence interval; MAX, Medicaid Analytic eXtract; PE, point estimate. a Adjusted for maternal age, maternal race/ethnicity, and year of delivery. The comparability of women in the MAX and external cohorts was assessed (Table 3). Similar proportions of both cohorts were treated with buprenorphine and methadone. The delivery periods differed, as previously noted. Most women in the MAX cohort and the external cohort were non-Hispanic white, but a higher percentage of women in the external cohort were black. No other differences were identified. Table 3. Comparison of Women in a Massachusetts MAX Medicaid Cohort Who Had a Diagnosis of Drug Dependence Within 2 Years of Delivery (2006–2011) and Women With Opioid Dependence in an External Cohort (2015–2016), by Prenatal OAT Status Characteristic or Outcome  MAX Medicaid Cohorta  External Cohortb  Buprenorphine (n = 543)  Methadone (n = 477)  Buprenorphine (n = 58)  Methadone (n = 55)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Year of delivery   2006–2011  543  100  477  100  0  0  0  0   2015–2016  0  0  0  0  58  100  55  100  Age at delivery, yearsc  27.9 (4.5)  28.2 (4.9)  28.2 (4.4)  28.3 (4.4)  Race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  4  6.9  7  12.7   Non-Hispanic white  426  78.5  366  76.7  50  86.2  41  74.6   Hispanic  15  2.8  23  4.8  4  6.9  7  12.7   Other/unknownd  100  18.4  75  15.7  0  0  0  0  SSRI prescription  158  29.1  77  16.1  14  24.1  9  16.4  Benzodiazepine prescription  173  31.9  122  25.6  15  25.9  12  21.8  Preterm birth (<37 weeks)  99  18.4  155  32.9  10  22.2  11  28.2  Cesarean delivery  187  34.3  179  37.5  20  47.6  16  43.2  Characteristic or Outcome  MAX Medicaid Cohorta  External Cohortb  Buprenorphine (n = 543)  Methadone (n = 477)  Buprenorphine (n = 58)  Methadone (n = 55)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Year of delivery   2006–2011  543  100  477  100  0  0  0  0   2015–2016  0  0  0  0  58  100  55  100  Age at delivery, yearsc  27.9 (4.5)  28.2 (4.9)  28.2 (4.4)  28.3 (4.4)  Race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  4  6.9  7  12.7   Non-Hispanic white  426  78.5  366  76.7  50  86.2  41  74.6   Hispanic  15  2.8  23  4.8  4  6.9  7  12.7   Other/unknownd  100  18.4  75  15.7  0  0  0  0  SSRI prescription  158  29.1  77  16.1  14  24.1  9  16.4  Benzodiazepine prescription  173  31.9  122  25.6  15  25.9  12  21.8  Preterm birth (<37 weeks)  99  18.4  155  32.9  10  22.2  11  28.2  Cesarean delivery  187  34.3  179  37.5  20  47.6  16  43.2  Abbreviations: MAX, Medicaid Analytic eXtract; OAT, opioid agonist therapy; SSRI, selective serotonin reuptake inhibitor. a Five infants exposed to buprenorphine, 6 infants exposed to methadone, and 20 unexposed infants were missing information from inpatient hospitalization (preterm birth, low birth weight for gestational age); 25 infants exposed to buprenorphine, 44 infants exposed to methadone, and 185 unexposed infants were missing information on length of hospital stay. b The “preterm birth” variable excludes 3 women who underwent therapeutic abortion (2 receiving buprenorphine, 1 receiving methadone) and 26 women who were lost to follow-up (11 receiving buprenorphine, 15 receiving methadone). The “cesarean delivery” variable excludes those excluded from the preterm birth variable, in addition to 3 women who delivered at another site (1 receiving buprenorphine, 2 receiving methadone). c Values are expressed as mean (standard deviation). d “Other” includes non-Hispanic women with unknown race/ethnicity. Table 3. Comparison of Women in a Massachusetts MAX Medicaid Cohort Who Had a Diagnosis of Drug Dependence Within 2 Years of Delivery (2006–2011) and Women With Opioid Dependence in an External Cohort (2015–2016), by Prenatal OAT Status Characteristic or Outcome  MAX Medicaid Cohorta  External Cohortb  Buprenorphine (n = 543)  Methadone (n = 477)  Buprenorphine (n = 58)  Methadone (n = 55)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Year of delivery   2006–2011  543  100  477  100  0  0  0  0   2015–2016  0  0  0  0  58  100  55  100  Age at delivery, yearsc  27.9 (4.5)  28.2 (4.9)  28.2 (4.4)  28.3 (4.4)  Race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  4  6.9  7  12.7   Non-Hispanic white  426  78.5  366  76.7  50  86.2  41  74.6   Hispanic  15  2.8  23  4.8  4  6.9  7  12.7   Other/unknownd  100  18.4  75  15.7  0  0  0  0  SSRI prescription  158  29.1  77  16.1  14  24.1  9  16.4  Benzodiazepine prescription  173  31.9  122  25.6  15  25.9  12  21.8  Preterm birth (<37 weeks)  99  18.4  155  32.9  10  22.2  11  28.2  Cesarean delivery  187  34.3  179  37.5  20  47.6  16  43.2  Characteristic or Outcome  MAX Medicaid Cohorta  External Cohortb  Buprenorphine (n = 543)  Methadone (n = 477)  Buprenorphine (n = 58)  Methadone (n = 55)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Year of delivery   2006–2011  543  100  477  100  0  0  0  0   2015–2016  0  0  0  0  58  100  55  100  Age at delivery, yearsc  27.9 (4.5)  28.2 (4.9)  28.2 (4.4)  28.3 (4.4)  Race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  4  6.9  7  12.7   Non-Hispanic white  426  78.5  366  76.7  50  86.2  41  74.6   Hispanic  15  2.8  23  4.8  4  6.9  7  12.7   Other/unknownd  100  18.4  75  15.7  0  0  0  0  SSRI prescription  158  29.1  77  16.1  14  24.1  9  16.4  Benzodiazepine prescription  173  31.9  122  25.6  15  25.9  12  21.8  Preterm birth (<37 weeks)  99  18.4  155  32.9  10  22.2  11  28.2  Cesarean delivery  187  34.3  179  37.5  20  47.6  16  43.2  Abbreviations: MAX, Medicaid Analytic eXtract; OAT, opioid agonist therapy; SSRI, selective serotonin reuptake inhibitor. a Five infants exposed to buprenorphine, 6 infants exposed to methadone, and 20 unexposed infants were missing information from inpatient hospitalization (preterm birth, low birth weight for gestational age); 25 infants exposed to buprenorphine, 44 infants exposed to methadone, and 185 unexposed infants were missing information on length of hospital stay. b The “preterm birth” variable excludes 3 women who underwent therapeutic abortion (2 receiving buprenorphine, 1 receiving methadone) and 26 women who were lost to follow-up (11 receiving buprenorphine, 15 receiving methadone). The “cesarean delivery” variable excludes those excluded from the preterm birth variable, in addition to 3 women who delivered at another site (1 receiving buprenorphine, 2 receiving methadone). c Values are expressed as mean (standard deviation). d “Other” includes non-Hispanic women with unknown race/ethnicity. In sensitivity analyses, when applying our definition of opioid dependence to the year before delivery (instead of 2 years prior), all women met the inclusion criteria. The interpretation of results was unchanged when analyses were conducted with the pregnancy period ending on the first day of the month rather than 30 days after the infant’s birth date. DISCUSSION In this paper, we report on a novel method for addressing potential uncontrolled confounding of the association between prenatal OAT exposure and infant outcomes in Massachusetts Medicaid data. The Massachusetts population is ideal for the study of OAT outcomes due to the high rates of opioid use and virtually unrestricted Medicaid coverage for buprenorphine and methadone in the state (13). Only 2 studies of the comparative safety of prenatal OAT have included more than 300 mother-infant pairs—1 from Denmark (n = 557) (4) and 1 from Vermont (n = 609) (5). To our knowledge, this MAX-based cohort study is the largest carried out to date, with 1,020 OAT-exposed mother-infant pairs. However, MAX data on potential confounders such as severity of addiction were unavailable. Careful consideration of confounding by indication is needed in assessing the comparative safety of buprenorphine and methadone with regard to infant outcomes (10, 22), because the systems of care differ substantially. In the United States, methadone treatment consists of structured observed daily dosing at a methadone clinic, while buprenorphine is self-administered through prescriptions. Some women benefit from the structured delivery and full-agonist effect of methadone, while others are better suited for the partial opioid agonist buprenorphine. Prior studies have suggested differential access to buprenorphine by race/ethnicity and social status (23), and in the MAX cohort, more women treated with methadone or without OAT prescriptions were Hispanic or black. Similar to another study that used administrative health-care data, MAX women treated with buprenorphine had more concomitant medication use, suggesting better engagement in clinical care and more stable drug dependence (24). In the external cohort used for confounding adjustment, women receiving buprenorphine had lower risk profiles, including more stable housing, more stable income, less severe drug dependence (as measured by the Addiction Severity Index), and less injection cocaine use. If women with these lower risk profiles deliver healthier babies, confounding by indication is likely. Our approach to controlling for confounding by indication did not fully attenuate the observed lower risk of preterm delivery or hospital stay length for buprenorphine as compared with methadone, but evidence of confounding was present. Shorter unadjusted hospital stays in buprenorphine-exposed infants have been observed in some but not all studies (3, 6, 25). The unadjusted risk ratio for preterm birth in buprenorphine- versus methadone-exposed infants in the Vermont study was 0.53 (95% CI: 0.33, 0.85), similar to our result. Our finding of somewhat attenuated but beneficial associations brings up the question of whether our methods fully corrected for confounding by indication. For example, clinical prescribing may have differed in the MAX cohort and the external cohort, resulting in misclassification in our propensity scores. There was little buprenorphine treatment failure—defined as transition from prenatal buprenorphine to methadone for nonadherence or continued illicit opioid use—in women from the external study clinic during 2006–2010 (5.7%) in comparison with women who delivered in 2015 and 2016 (22%) (19). This change probably reflects provider efforts at the study clinic to make buprenorphine treatment more accessible for women with more severe drug dependence. Because of the smaller size of the external cohort, we assumed that the effect of confounders was uniform by OAT, which may not have adequately captured the confounding. Nonetheless, our findings suggest that there may be a higher risk of preterm birth and longer hospital stays in methadone-exposed infants. Unlike preterm birth and length of hospital stay, there was no association between OAT exposure and low birth weight for gestational age after adjustment with bias analysis. Other investigators have reported higher mean birth weights in buprenorphine-exposed infants (3, 5, 7) and in opioid-unexposed compared with methadone-exposed infants (26, 27), but few have examined growth appropriate for gestational age (28, 29). A limitation of using administrative health-care data for research is misclassification of study exposures and outcomes. We included ICD-9 codes used in prior studies to identify deliveries, opioid dependence, methadone and buprenorphine exposure, and infant outcomes (13, 15, 30). The sensitivity and specificity of ICD-9 codes for preterm birth was quite high in a database validation study (18), while the use of ICD-9 codes 656.5x for infants born small for gestational age had low sensitivity (31). We expanded on these prior codes to include others that indicated low birth weight for gestational age; however, some nondifferential misclassification probably exists. The MAX cohort demonstrated successful linkage of deliveries to infants at 59%, a proportion similar to that of another study (12). Linkage is likely to be affected by level of claim completeness, inconsistent assignment of the same family case identifier to women and children, and use of delivery codes that did not identify viable pregnancies or pregnancies carried to term. Women were included in the MAX cohort regardless of consistent noncapitated Medicaid coverage in pregnancy, raising the possibility that some women classified as untreated may have received OAT for which Medicaid claims were not submitted. The number of such women is expected to have been small, given that few women in our study had private health insurance coverage at any time during pregnancy, almost all had unrestricted Medicaid coverage at some point during pregnancy, and a majority had at least 1 other prenatal medication claim. Restricting the study population to women with Medicaid coverage throughout pregnancy would have unnecessarily excluded women for whom Medicaid coverage, prenatal care, and OAT began after conception. Our finding that 59% of women with a diagnosis of opioid dependence in the MAX cohort did not have an OAT prescription in the year before delivery is concerning, especially given that Massachusetts has unrestricted Medicaid coverage for buprenorphine and methadone. Cash clinics, where patients pay out of pocket for buprenorphine prescriptions because of a lack of buprenorphine providers in the area, may have been a possible source of treatment for the unexposed women. It unlikely that this would account for OAT in the majority of the unexposed low-income women, given that comprehensive prenatal care in Medicaid hospital clinics would include OAT. Limited availability of treatment for substance users is well documented and has been exacerbated by the current opioid crisis (32). Possible barriers to care among pregnant drug-dependent women, including stigma that prevents women from disclosing drug dependence in pregnancy and/or seeking prenatal care, lack of transportation to care, and lack of referrals to addiction providers, should be investigated. High rates of drug dependence in pregnant women and of adverse outcomes in their infants are ongoing concerns (33, 34). It is feasible to use extant Medicaid data paired with detailed data from an external cohort to adjust for unmeasured confounding in order to study the comparative safety of prenatal OAT exposure, but limitations exist. Some but not all of the associations in this study were explained by confounding. Our study provides insight into the use Medicaid data for OAT outcomes, prescribing preferences in a Medicaid population, and the effect of unmeasured confounding bias from maternal characteristics. ACKNOWLEDGMENTS Author affiliations: Department of Surgery, Queen’s University, Kingston, Ontario, Canada (Susan B. Brogly); Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Sonia Hernández-Diaz); and Department of Epidemiology, School of Public Health, Boston University, Boston, Massachusetts (Emily Regan, Ela Fadli, Kristen A. Hahn, Martha M. Werler). This work was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development under grant 1R21HD081271-01 REVISED (S.B.B.). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Conflict of interest: none declared. Abbreviations CI confidence interval ICD-9 International Classification of Diseases, Ninth Revision MAX Medicaid Analytic eXtract MSIS ID Medicaid Statistical Information System Identifiers OAT opioid agonist therapy RR risk ratio SSRI selective serotonin reuptake inhibitor REFERENCES 1 Ko JY, Patrick SW, Tong VT, et al.  . Incidence of neonatal abstinence syndrome—28 states, 1999–2013. MMWR Morb Mortal Wkly Rep . 2016; 65( 31): 799– 802. Google Scholar CrossRef Search ADS PubMed  2 Substance Use and Mental Health Services Administration, US Department of Health and Human Services. Federal Guidelines for Opioid Treatment Programs. Rockville, MD: Substance Use and Mental Health Services Administration; 2015. (DHHS publication no. EP15-FEDGUIDEOTP). 3 Metz V, Jagsch R, Ebner N, et al.  . Impact of treatment approach on maternal and neonatal outcome in pregnant opioid-maintained women. Hum Psychopharmacol . 2011; 26( 6): 412– 421. Google Scholar PubMed  4 Nørgaard M, Nielsson MS, Heide-Jørgensen U. Birth and neonatal outcomes following opioid use in pregnancy: a Danish population-based study. Subst Abuse . 2015; 9( suppl 2): 5– 11. Google Scholar PubMed  5 Meyer MC, Johnston AM, Crocker AM, et al.  . Methadone and buprenorphine for opioid dependence during pregnancy: a retrospective cohort study. J Addict Med . 2015; 9( 2): 81– 86. Google Scholar CrossRef Search ADS PubMed  6 Jones HE, Kaltenbach K, Heil SH, et al.  . Neonatal abstinence syndrome after methadone or buprenorphine exposure. N Engl J Med . 2010; 363( 24): 2320– 2331. Google Scholar CrossRef Search ADS PubMed  7 Welle-Strand GK, Skurtveit S, Jones HE, et al.  . Neonatal outcomes following in utero exposure to methadone or buprenorphine: a national cohort study of opioid-agonist treatment of pregnant women in Norway from 1996 to 2009. Drug Alcohol Depend . 2013; 127( 1–3): 200– 206. Google Scholar CrossRef Search ADS PubMed  8 Pritham UA. Breastfeeding promotion for management of neonatal abstinence syndrome. J Obstet Gynecol Neonatal Nurs . 2013; 42( 5): 517– 526. Google Scholar CrossRef Search ADS PubMed  9 Brogly SB, Saia KA, Walley AY, et al.  . Prenatal buprenorphine versus methadone exposure and neonatal outcomes: systematic review and meta-analysis. Am J Epidemiol . 2014; 180( 7): 673– 686. Google Scholar CrossRef Search ADS PubMed  10 Krans EE, Bogen D, Richardson G, et al.  . Factors associated with buprenorphine versus methadone use in pregnancy. Subst Abus . 2016; 37( 4): 550– 557. Google Scholar CrossRef Search ADS PubMed  11 Palmsten K, Huybrechts KF, Michels KB, et al.  . Antidepressant use and risk for preeclampsia. Epidemiology . 2013; 24( 5): 682– 691. Google Scholar CrossRef Search ADS PubMed  12 Patorno E, Huybrechts KF, Bateman BT, et al.  . Lithium use in pregnancy and the risk of cardiac malformations. N Engl J Med . 2017; 376( 23): 2245– 2254. Google Scholar CrossRef Search ADS PubMed  13 Baxter JD, Clark RE, Samnaliev M, et al.  . Factors associated with Medicaid patients’ access to buprenorphine treatment. J Subst Abuse Treat . 2011; 41( 1): 88– 96. Google Scholar CrossRef Search ADS PubMed  14 Research Data Assistance Center, Centers for Medicare and Medicaid Services. CMS Data Request Center. http://www.resdac.org/cms-data/request/cms-data-request-center. Accessed November 9, 2016. 15 Palmsten K, Huybrechts KF, Mogun H, et al.  . Harnessing the Medicaid Analytic eXtract (MAX) to evaluate medications in pregnancy: design considerations. PLoS One . 2013; 8( 6): e67405. Google Scholar CrossRef Search ADS PubMed  16 Substance Abuse and Mental Health Services Administration, US Department of Health and Human Services. The Confidentiality of Alcohol and Drug Abuse Patient Records Regulation and the HIPAA Privacy Rule: Implications for Alcohol and Substance Abuse Programs. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2004. (Publication no. PHD1083). https://store.samhsa.gov/product/The-Confidentiality-of-Alcohol-and-Drug-Abuse-Patient-Records-Regulation-and-the-HIPAA-Privacy-Rule/PHD1083. Accessed April 2, 2018. 17 Freund DA, Neuschler E. Overview of Medicaid capitation and case-management initiatives. Health Care Financ Rev . 1986;( Spec. No.): 21– 30. 18 Margulis AV, Setoguchi S, Mittleman MA, et al.  . Algorithms to estimate the beginning of pregnancy in administrative databases. Pharmacoepidemiol Drug Saf . 2013; 22( 1): 16– 24. Google Scholar CrossRef Search ADS PubMed  19 Saia K, Bagley SM, Wachman EM, et al.  . Prenatal treatment for opioid dependency: observations from a large inner-city clinic. Addict Sci Clin Pract . 2017; 12( 1): 5. Google Scholar CrossRef Search ADS PubMed  20 McLellan AT, Luborsky L, Woody GE, et al.  . An improved diagnostic evaluation instrument for substance abuse patients. The Addiction Severity Index. J Nerv Ment Dis . 1980; 168( 1): 26– 33. Google Scholar CrossRef Search ADS PubMed  21 Lin DY, Psaty BM, Kronmal RA. Assessing the sensitivity of regression results to unmeasured confounders in observational studies. Biometrics . 1998; 54( 3): 948– 963. Google Scholar CrossRef Search ADS PubMed  22 Brogly SB, Hahn KA, Diaz SH, et al.  . Confounding of the comparative safety of prenatal opioid agonist therapy. J Addict Res Ther . 2015; 6( 4): 252. Google Scholar CrossRef Search ADS PubMed  23 Hansen HB, Siegel CE, Case BG, et al.  . Variation in use of buprenorphine and methadone treatment by racial, ethnic, and income characteristics of residential social areas in New York City. J Behav Health Serv Res . 2013; 40( 3): 367– 377. Google Scholar CrossRef Search ADS PubMed  24 Brogly SB, Turner SD, Lajkosz K, et al.  . Infants born to opioid-dependent women in Ontario, 2002–2014. J Obstet Gynaecol Can . 2017; 39( 3): 157– 165. Google Scholar CrossRef Search ADS PubMed  25 Wachman EM, Newby PK, Vreeland J, et al.  . The relationship between maternal opioid agonists and psychiatric medications on length of hospitalization for neonatal abstinence syndrome. J Addict Med . 2011; 5( 4): 293– 299. Google Scholar CrossRef Search ADS PubMed  26 Liu AJ, Sithamparanathan S, Jones MP, et al.  . Growth restriction in pregnancies of opioid-dependent mothers. Arch Dis Child Fetal Neonatal Ed . 2010; 95( 4): F258– F262. Google Scholar CrossRef Search ADS PubMed  27 Kandall SR, Albin S, Lowinson J, et al.  . Differential effects of maternal heroin and methadone use on birthweight. Pediatrics . 1976; 58( 5): 681– 685. Google Scholar PubMed  28 Pritham UA, Paul JA, Hayes MJ. Opioid dependency in pregnancy and length of stay for neonatal abstinence syndrome. J Obstet Gynecol Neonatal Nurs . 2012; 41( 2): 180– 190. Google Scholar CrossRef Search ADS PubMed  29 Kelty E, Hulse G. A retrospective cohort study of birth outcomes in neonates exposed to naltrexone in utero: a comparison with methadone, buprenorphine- and non-opioid-exposed neonates. Drugs . 2017; 77( 11): 1211– 1219. Google Scholar CrossRef Search ADS PubMed  30 Ailes EC, Dawson AL, Lind JN, et al.  . Opioid prescription claims among women of reproductive age—United States, 2008–2012. MMWR Morb Mortal Wkly Rep . 2015; 64( 2): 37– 41. Google Scholar PubMed  31 Phiri K, Hernandez-Diaz S, Tsen LC, et al.  . Accuracy of ICD-9-CM coding to identify small for gestational age newborns. Pharmacoepidemiol Drug Saf . 2015; 24( 4): 381– 388. Google Scholar CrossRef Search ADS PubMed  32 Saloner B, Karthikeyan S. Changes in substance abuse treatment use among individuals with opioid use disorders in the United States, 2004–2013. JAMA . 2015; 314( 14): 1515– 1517. Google Scholar CrossRef Search ADS PubMed  33 Patrick SW, Schumacher RE, Benneyworth BD, et al.  . Neonatal abstinence syndrome and associated health care expenditures: United States, 2000–2009. JAMA . 2012; 307( 18): 1934– 1940. Google Scholar CrossRef Search ADS PubMed  34 Tolia VN, Patrick SW, Bennett MM, et al.  . Increasing incidence of the neonatal abstinence syndrome in US neonatal ICUs. N Engl J Med . 2015; 372( 22): 2118– 2126. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Journal of Epidemiology Oxford University Press

Neonatal Outcomes in a Medicaid Population With Opioid Dependence

Loading next page...
 
/lp/ou_press/neonatal-outcomes-in-a-medicaid-population-with-opioid-dependence-CbB9SlutJz
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0002-9262
eISSN
1476-6256
D.O.I.
10.1093/aje/kwx341
Publisher site
See Article on Publisher Site

Abstract

Abstract Confounding may account for the apparently improved infant outcomes after prenatal exposure to buprenorphine versus methadone. We used Massachusetts Medicaid Analytic eXtract (MAX) data to identify a cohort of opioid-dependent mother-infant pairs (2006–2011), supplemented with confounder data from an external Boston, Massachusetts, cohort (2015–2016). Associations between prenatal buprenorphine exposure versus methadone exposure and infant outcomes in the MAX cohort were adjusted for measured MAX confounders and were additionally adjusted for unmeasured confounders with bias analysis using external cohort data. A total of 477 women in MAX were treated with methadone and 543 with buprenorphine. More buprenorphine users than methadone users were white and used psychotropic medications. After adjustment for MAX confounders, risk ratios among infants exposed to buprenorphine versus those exposed to methadone were 0.45 (95% confidence interval (CI): 0.34, 0.61) for preterm birth (birth at <37 weeks) and 0.75 (95% CI: 0.51, 1.11) for low birth weight for gestational age. The mean difference in infant hospitalization was −7.35 days (95% CI: −9.16, −5.55). After further adjustment with bias analysis, the risk ratios were 0.53 (95% CI: 0.39, 0.71) for preterm birth and 1.14 (95% CI: 0.77, 1.69) for low birth weight for gestational age, and the mean difference in infant hospitalization was −3.66 days (95% CI: −5.46, −1.87). External confounder data can be used to adjust for unmeasured confounding in studies of prenatal outcomes among women on opioid agonist therapy based on administrative databases. buprenorphine, confounding factors (epidemiology), methadone, neonatal abstinence syndrome, opioids, pregnancy Editor’s note: An invited commentary on this article appears on page 1162. High rates of opioid dependence in pregnant women and associated birth outcomes are major public health issues (1). Prenatal opioid agonist therapy (OAT) with buprenorphine, a partial opioid agonist, or methadone, a full opioid agonist, is recommended for pregnant opioid-dependent women (2). Some observational studies have suggested improved fetal growth for prenatal treatment with buprenorphine as compared with methadone, including higher gestational age at birth (3–6), higher birth weight (3, 5, 7), and fewer infants born small for gestational age (8). Pregnant opioid-dependent women with better clinical and psychosocial profiles are more likely to be treated with buprenorphine, raising the possibility that associations are confounded (9, 10). In a clinical trial, Jones et al. (6) found shorter hospital stays and less morphine treatment for gastrointestinal, respiratory, autonomic, and central nervous system disturbances from opioid withdrawal after delivery (neonatal abstinence syndrome) among infants prenatally exposed to buprenorphine as compared with infants prenatally exposed to methadone, but high dropout rates in that study raised concerns about selection bias. Administrative health-care databases can be an efficient data source for epidemiologic studies due to the reduced time and cost otherwise required to assemble large cohorts. Use of psychotropic medication during pregnancy has been studied in Medicaid beneficiaries (11, 12). Massachusetts Medicaid data are efficient for studying OAT, because OAT coverage is virtually unrestricted in the state (13). Most clinic-based studies of prenatal OAT have been small, with fewer than 300 mother-infant pairs (9). To address small sample sizes and the possibility of confounding in prior studies, we designed a study that used extant Medicaid data paired with data from a prospective cohort study and bias analysis methods to examine the safety of buprenorphine as compared with methadone in terms of infant outcomes. Herein, we describe the methods, limitations, and findings of this novel approach. METHODS Medicaid Analytic eXtract cohort Study population The study population was identified using Medicaid Analytic eXtract (MAX) data (14) from Massachusetts. Women were included if they were 14 years of age or older, delivered between 2006 (when prenatal buprenorphine became available in Massachusetts) and 2011 (most recent MAX data) (14), and 1) had a Medicaid claim with a diagnosis of opioid dependence within 2 years before delivery or 2) had a Medicaid claim with an unspecified type of drug dependence within 2 years before delivery and were on prenatal OAT (see Web Table 1, available at https://academic.oup.com/aje, for codes). Inpatient and outpatient deliveries were identified using previously identified codes (15), in addition to others (Web Table 1). Mothers were linked to infants by matching family case identifiers (Figure 1) and an infant birth date that fell within the delivery date window (15). Medicaid Statistical Information System Identifiers (MSIS ID) identify unique individuals within the Medicaid system (Figure 1). Multifetal deliveries may have a different MSIS ID for each infant, and some infants (singleton or multifetal) may have multiple MSIS IDs because of assignment of temporary MSIS IDs at birth followed by a permanent MSIS ID (15). We therefore could not distinguish repeat records of the same infant with different MSIS IDs and multiple births from the same delivery; thus, 1 infant per delivery was retained. These criteria were applied to MAX inpatient, other therapy, patient summary, and prescription databases provided by the Centers for Medicare and Medicaid Services. Per the Confidentiality of Alcohol and Drug Abuse Patient Records Regulations (16), the Centers for Medicare and Medicaid Services does not release postal codes and sets birth dates to the first day of the month for the given month and year of birth. Figure 1. View largeDownload slide Linkage of mother-infant pairs in Medicaid Analytic eXtract (MAX) data provided by the Centers for Medicare and Medicaid Services (CMS) for deliveries to opioid-dependent women in Massachusetts, 2006–2011. Case ID, family case identifier; MSIS ID, Medicaid Statistical Information System Identifiers; OAT, opioid agonist therapy; PS, personal summary. Figure 1. View largeDownload slide Linkage of mother-infant pairs in Medicaid Analytic eXtract (MAX) data provided by the Centers for Medicare and Medicaid Services (CMS) for deliveries to opioid-dependent women in Massachusetts, 2006–2011. Case ID, family case identifier; MSIS ID, Medicaid Statistical Information System Identifiers; OAT, opioid agonist therapy; PS, personal summary. Prenatal OAT exposure The prenatal OAT exposure period was defined as the period from 294 days before the matched infant’s birth date through 30 days after the matched infant’s birth date. A window of 30 days after the infant’s birth date was used because birth dates were set to the first day of the month by the Centers for Medicare and Medicaid Services. Prescriptions for prenatal buprenorphine, selective serotonin reuptake inhibitors (SSRIs), and benzodiazepines were identified in the Medicaid prescription claims database (National Drug Codes are provided in Web Table 1). Methadone was identified with procedure code H0020 in the other therapy database. A small number of women switched treatment from buprenorphine to methadone (n = 26) and vice versa (n = 1); an intent-to-treat approach with the initial OAT for classification was used. Medicaid beneficiaries can be enrolled in a managed-care plan (either capitated or fee-for-service primary-care case management) or no managed-care plan (17). To examine the completeness of OAT data, we identified women who, at any point during pregnancy, had fee-for-service primary-care case management, no managed-care plan, or another managed-care plan that would not restrict prescription benefits. We also identified women who, at any point during pregnancy, had private health insurance or capitated Medicaid benefits that potentially would not submit prescription claims through Medicaid. Finally, we determined the proportion of women with prenatal prescriptions other than OAT. Infant outcomes The length of infant hospitalization was calculated as the difference between the starting and ending dates of the infant’s hospitalization for the first hospitalization that began in the infant’s birth month. Preterm birth (birth at <37 weeks’ gestation) (18) and low birth weight for gestational age were identified through International Classification of Diseases, Ninth Revision (ICD-9) codes in infant records (Web Table 1). Statistical approach Differences between women and infants who matched and those who did not were assessed statistically with the χ2, Fischer exact, or Kruskal-Wallis test, as appropriate. Risk ratios for preterm birth and low birth weight according to prenatal OAT were estimated using generalized linear models. Mean differences in the length of infant hospitalization were estimated with linear regression. We adjusted for potential confounding from maternal age (years; continuous), race/ethnicity (non-Hispanic white vs. other), year of delivery recorded in the MAX data (categorical, with 2006 as the referent), prenatal SSRI use before OAT (yes/no), and prenatal benzodiazepine use before OAT (yes/no). External cohort Study population Pregnant opioid-dependent women (n = 113) from a high-risk obstetrical and addiction recovery clinic in Boston, Massachusetts—the largest-volume clinic for pregnant women with opioid dependence in New England—were enrolled in the study (2015–2016) (19) to measure confounders unavailable in MAX. Women were recruited at OAT initiation or the first prenatal visit (for women on OAT at conception). The clinic operated in conjunction with methadone clinics, counseling centers, and residential treatment programs for pregnant women. Maternal criteria for OAT included being 18 years of age or older, having a Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, diagnosis of opioid dependence, having laboratory and/or radiographic documentation of pregnancy, and voluntarily consenting to engage in the clinic’s comprehensive treatment program. Per the clinical practice protocol, women were scheduled for prenatal care and relapse prevention visits every 1–3 weeks through delivery. Observed urine drug tests were carried out at each prenatal visit. The type of OAT initiated was selected considering patient choice, treatment history, disease severity, and medical and psychiatric comorbidity. The clinic protocol recommended methadone for women with more severe opioid dependence, women with intolerance to buprenorphine, and women for whom buprenorphine was medically contraindicated. In general, women for whom previous buprenorphine treatment was nonefficacious, who had a history of buprenorphine diversion within the last 6 months, who were unable to make weekly-to-biweekly office visits, or who were unwilling to engage in independent relapse prevention counseling were offered methadone (19). Data collection At cohort enrollment, a research assistant administered the Addiction Severity Index, 5th Edition Lite (20), to collect data on a wide range of potential confounders. The Addiction Severity Index has been widely employed among substance users and includes questions on 7 domains: medical status, employment and support, drug use, alcohol use, legal status, family/social status, and psychiatric status. Responses for the particular domain questions were summed to create a score that captured the degree of patient-reported problems in each domain, with higher scores indicating more severe problems (20). Prenatal OAT, pregnancy, substance use, and delivery information was routinely abstracted from maternal medical charts. Infant outcomes from birth to hospital discharge were abstracted from medical charts. Propensity scores Propensity scores for the probability of initiating buprenorphine versus methadone in the external cohort were estimated using logistic regression. A priori confounders included maternal demographic, clinical, and substance use characteristics based on substantive knowledge and prior studies. Characteristics retained in the final multivariate propensity score model were predictive of initial OAT and associated with an infant outcome of interest. They included: generally having unstable housing during the past 3 years (yes/no), gravidity (>1 pregnancy vs. ≤1), having ever regularly injected cocaine (yes/no), and Addiction Severity Index drug score (continuous, ranging from 0 to 1). No data were missing for the propensity score models. Associations between propensity scores and infant outcomes were restricted to infants for whom outcome data were available; some external cohort pregnancies ended through therapeutic abortion (n = 3) or miscarriage (n = 4) or had an unknown outcome, as some women did not continue prenatal care (n = 26). Bias analysis for unmeasured confounders in the MAX cohort Regression models were fitted with propensity scores as the independent variable and each infant outcome as the dependent variable using the external cohort data. The resultant risk ratio or mean difference quantified the amount of unmeasured confounding that would be expected in the crude MAX cohort results (the parameters RRγ and γ, described below). Mean values of the propensity score also were determined in the buprenorphine and methadone groups in the external cohort. Following the approach of Lin et al. (21), the MAX cohort risk ratio (RR) and mean difference (MD) were then corrected for unmeasured confounding (assuming a uniform effect of confounding by OAT) as follows:   RRcorrected=RR*[RRγ(μ1−μ0)]and   MDcorrected=MD*−[(μ1−μ0)γ)],where RR* and MD* are measures of association for buprenorphine versus methadone and the infant outcomes in the MAX cohort, adjusted for measured MAX confounders; RRγ and γ represent the effect of the unmeasured confounder (i.e., propensity score, which included confounders unavailable in MAX) on the infant outcome in the external cohort; and μ1 and μ0 are the mean values of the unmeasured confounder in the buprenorphine and methadone groups of the external cohort, respectively. The 95% confidence intervals for the bias-adjusted estimates were calculated using the standard errors of the MAX estimates (21). The small number of MAX infants with a missing inpatient hospitalization record or with insufficient data to evaluate length of stay were excluded from analyses of the relevant outcome. We also examined potential misclassification of MAX data. To evaluate the robustness of our cohort inclusion criterion, we determined the number of women with a diagnosis of opioid dependence within the year before delivery and assessed the OAT use distribution among these women. To examine bias from possible inclusion of postnatal OAT due to an unknown exact date of delivery (i.e., because birth dates were set to the first day of the month by the Centers for Medicare and Medicaid Services), we reran our infant-outcomes analyses with the pregnancy period ending on the first day of the month rather than using a window of 30 days after the infant’s birth date. RESULTS From 2006 to 2011, 5,348 women with 1 or more delivery windows and 5,755 infant records (4,357 infants with 5,755 different MSIS ID/family case identifier combinations) were available for linkage in the MAX cohort (Figure 1). A total of 5,010 infant records (87%) matched to a delivery record; all infants matched to only 1 woman. A total of 2,237 women (41.8%) with 1 or more delivery records did not match to an infant. There was no difference in the mean age (27.5 years vs. 27.8 years), race/ethnicity (70.8% white vs. 70.6%), or OAT status (14.9% treated with buprenorphine vs. 16.0% treated) of women who did and did not match to an infant. Women who did not match to an infant were more likely to deliver in earlier years (35.3% in 2006–2007 among those who did not match vs. 26.6% in 2006–2007 among those who matched). The 12 infants who did not match to a maternal record and the 4,347 infants who did match were similar in terms of year of birth, preterm birth (8.3% vs. 22.8%), and low birth weight (8.3% vs. 10.2%) and had similar mean lengths of hospital stay (22.9 days vs. 13.9 days). After removal of 1,578 duplicate records/multifetal births (1 infant per delivery was retained), 1 mother with an implausible maternal age at delivery, and women who did not meet our definition of opioid dependence (n = 962), there were 2,469 mother-infant pairs in the MAX cohort study population (Figure 1). Among these women, 2,012 women had 1 delivery, 204 had 2 deliveries, 15 had 3 deliveries, and 1 had 4. There was a 3-fold increase in the number of deliveries among opioid-dependent women covered by Medicaid in Massachusetts from 2006 (n = 205) to 2011 (n = 688). A total of 477 pregnant women were treated with methadone and 543 with buprenorphine; 1,449 did not have a record of OAT (unexposed; see Table 1). More unexposed and methadone-exposed women were black or Hispanic (Table 1). Almost all women had unrestricted Medicaid coverage at some point during pregnancy, and 7.5% of unexposed, 9.4% of buprenorphine-exposed, and 12.6% of methadone-exposed women had private health insurance at some point during pregnancy. Women treated with buprenorphine were more likely to have SSRI and benzodiazepine prescriptions than methadone-treated or unexposed women. Methadone-treated women, however, had the lowest prevalence of any prenatal prescription (77.6%). Unexposed mother-infant pairs were more likely to deliver in earlier years. Unexposed mother-infant pairs were excluded from analysis of the association of buprenorphine versus methadone with infant outcomes. Table 1. Characteristics and Infant Outcomes of Massachusetts Mother-Infant Pairs in a MAX Medicaid Cohort With a Diagnosis of Drug Dependence Within 2 Years of Delivery (2006–2011), by Prenatal OAT Statusa Characteristic or Outcome  OAT Status  Buprenorphine (n = 543)  Methadone (n = 477)  No OAT (1,449)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Maternal Characteristics  Year of delivery   2006  22  4.1  38  8.0  145  10.0   2007  53  9.8  57  12.0  208  14.4   2008  78  14.4  53  11.1  234  16.2   2009  95  17.5  74  15.5  273  18.8   2010  97  17.9  86  18.0  268  18.5   2011  198  36.5  169  35.4  321  22.2  Maternal age at delivery, yearsb  27.9 (4.5)  28.2 (4.9)  27.9 (5.0)  Maternal race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  60  4.1   Non-Hispanic white  426  78.5  366  76.7  1,105  76.3   Hispanic  15  2.8  23  4.8  77  5.3   Other/unknownc  100  18.4  75  15.7  207  14.3  Diagnosis of unspecified drug dependence only  24  4.4  0  0  0  0  Any unrestricted Medicaid coverage during pregnancy  541  99.6  475  99.6  1,445  99.7  Any private health insurance during pregnancy  51  9.4  60  12.6  109  7.5  No. of prenatal prescription medications other than OAT   0  57  10.5  107  22.4  273  18.8   1  19  3.5  31  6.5  55  3.8   2–3  54  9.9  81  17.0  141  9.7   ≥4  413  76.1  258  54.1  980  67.6  SSRI prescription  158  29.1  77  16.1  299  20.6  Benzodiazepine prescription  173  31.9  122  25.6  334  23.1  Cesarean delivery  187  34.3  179  37.5  494  34.1  Infant Outcomes  Infant sex   Female  260  47.9  234  49.1  692  47.8   Male  283  52.1  243  50.9  757  52.2  Preterm birth (<37 weeks)  99  18.4  155  32.9  351  24.4  Low birth weight for gestational age  54  10.9  61  13.0  153  10.7  Length of infant hospital stay, daysb  13.9 (12.6)  21.4 (15.7)  15.7 (15.0)  Characteristic or Outcome  OAT Status  Buprenorphine (n = 543)  Methadone (n = 477)  No OAT (1,449)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Maternal Characteristics  Year of delivery   2006  22  4.1  38  8.0  145  10.0   2007  53  9.8  57  12.0  208  14.4   2008  78  14.4  53  11.1  234  16.2   2009  95  17.5  74  15.5  273  18.8   2010  97  17.9  86  18.0  268  18.5   2011  198  36.5  169  35.4  321  22.2  Maternal age at delivery, yearsb  27.9 (4.5)  28.2 (4.9)  27.9 (5.0)  Maternal race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  60  4.1   Non-Hispanic white  426  78.5  366  76.7  1,105  76.3   Hispanic  15  2.8  23  4.8  77  5.3   Other/unknownc  100  18.4  75  15.7  207  14.3  Diagnosis of unspecified drug dependence only  24  4.4  0  0  0  0  Any unrestricted Medicaid coverage during pregnancy  541  99.6  475  99.6  1,445  99.7  Any private health insurance during pregnancy  51  9.4  60  12.6  109  7.5  No. of prenatal prescription medications other than OAT   0  57  10.5  107  22.4  273  18.8   1  19  3.5  31  6.5  55  3.8   2–3  54  9.9  81  17.0  141  9.7   ≥4  413  76.1  258  54.1  980  67.6  SSRI prescription  158  29.1  77  16.1  299  20.6  Benzodiazepine prescription  173  31.9  122  25.6  334  23.1  Cesarean delivery  187  34.3  179  37.5  494  34.1  Infant Outcomes  Infant sex   Female  260  47.9  234  49.1  692  47.8   Male  283  52.1  243  50.9  757  52.2  Preterm birth (<37 weeks)  99  18.4  155  32.9  351  24.4  Low birth weight for gestational age  54  10.9  61  13.0  153  10.7  Length of infant hospital stay, daysb  13.9 (12.6)  21.4 (15.7)  15.7 (15.0)  Abbreviations: MAX, Medicaid Analytic eXtract; OAT, opioid agonist therapy; SSRI, selective serotonin reuptake inhibitor. a Twenty-six women initiated treatment with buprenorphine and switched to methadone, and 1 woman initiated treatment with methadone and switched to buprenorphine. Five infants exposed to buprenorphine, 6 exposed to methadone, and 9 unexposed infants were missing information from inpatient hospitalization (preterm birth, low birth weight for gestational age); 25 infants exposed to buprenorphine, 44 infants exposed to methadone, and 109 unexposed infants were missing information on length of hospital stay. b Values are expressed as mean (standard deviation). c “Other” includes non-Hispanic women with unknown race/ethnicity. Table 1. Characteristics and Infant Outcomes of Massachusetts Mother-Infant Pairs in a MAX Medicaid Cohort With a Diagnosis of Drug Dependence Within 2 Years of Delivery (2006–2011), by Prenatal OAT Statusa Characteristic or Outcome  OAT Status  Buprenorphine (n = 543)  Methadone (n = 477)  No OAT (1,449)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Maternal Characteristics  Year of delivery   2006  22  4.1  38  8.0  145  10.0   2007  53  9.8  57  12.0  208  14.4   2008  78  14.4  53  11.1  234  16.2   2009  95  17.5  74  15.5  273  18.8   2010  97  17.9  86  18.0  268  18.5   2011  198  36.5  169  35.4  321  22.2  Maternal age at delivery, yearsb  27.9 (4.5)  28.2 (4.9)  27.9 (5.0)  Maternal race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  60  4.1   Non-Hispanic white  426  78.5  366  76.7  1,105  76.3   Hispanic  15  2.8  23  4.8  77  5.3   Other/unknownc  100  18.4  75  15.7  207  14.3  Diagnosis of unspecified drug dependence only  24  4.4  0  0  0  0  Any unrestricted Medicaid coverage during pregnancy  541  99.6  475  99.6  1,445  99.7  Any private health insurance during pregnancy  51  9.4  60  12.6  109  7.5  No. of prenatal prescription medications other than OAT   0  57  10.5  107  22.4  273  18.8   1  19  3.5  31  6.5  55  3.8   2–3  54  9.9  81  17.0  141  9.7   ≥4  413  76.1  258  54.1  980  67.6  SSRI prescription  158  29.1  77  16.1  299  20.6  Benzodiazepine prescription  173  31.9  122  25.6  334  23.1  Cesarean delivery  187  34.3  179  37.5  494  34.1  Infant Outcomes  Infant sex   Female  260  47.9  234  49.1  692  47.8   Male  283  52.1  243  50.9  757  52.2  Preterm birth (<37 weeks)  99  18.4  155  32.9  351  24.4  Low birth weight for gestational age  54  10.9  61  13.0  153  10.7  Length of infant hospital stay, daysb  13.9 (12.6)  21.4 (15.7)  15.7 (15.0)  Characteristic or Outcome  OAT Status  Buprenorphine (n = 543)  Methadone (n = 477)  No OAT (1,449)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Maternal Characteristics  Year of delivery   2006  22  4.1  38  8.0  145  10.0   2007  53  9.8  57  12.0  208  14.4   2008  78  14.4  53  11.1  234  16.2   2009  95  17.5  74  15.5  273  18.8   2010  97  17.9  86  18.0  268  18.5   2011  198  36.5  169  35.4  321  22.2  Maternal age at delivery, yearsb  27.9 (4.5)  28.2 (4.9)  27.9 (5.0)  Maternal race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  60  4.1   Non-Hispanic white  426  78.5  366  76.7  1,105  76.3   Hispanic  15  2.8  23  4.8  77  5.3   Other/unknownc  100  18.4  75  15.7  207  14.3  Diagnosis of unspecified drug dependence only  24  4.4  0  0  0  0  Any unrestricted Medicaid coverage during pregnancy  541  99.6  475  99.6  1,445  99.7  Any private health insurance during pregnancy  51  9.4  60  12.6  109  7.5  No. of prenatal prescription medications other than OAT   0  57  10.5  107  22.4  273  18.8   1  19  3.5  31  6.5  55  3.8   2–3  54  9.9  81  17.0  141  9.7   ≥4  413  76.1  258  54.1  980  67.6  SSRI prescription  158  29.1  77  16.1  299  20.6  Benzodiazepine prescription  173  31.9  122  25.6  334  23.1  Cesarean delivery  187  34.3  179  37.5  494  34.1  Infant Outcomes  Infant sex   Female  260  47.9  234  49.1  692  47.8   Male  283  52.1  243  50.9  757  52.2  Preterm birth (<37 weeks)  99  18.4  155  32.9  351  24.4  Low birth weight for gestational age  54  10.9  61  13.0  153  10.7  Length of infant hospital stay, daysb  13.9 (12.6)  21.4 (15.7)  15.7 (15.0)  Abbreviations: MAX, Medicaid Analytic eXtract; OAT, opioid agonist therapy; SSRI, selective serotonin reuptake inhibitor. a Twenty-six women initiated treatment with buprenorphine and switched to methadone, and 1 woman initiated treatment with methadone and switched to buprenorphine. Five infants exposed to buprenorphine, 6 exposed to methadone, and 9 unexposed infants were missing information from inpatient hospitalization (preterm birth, low birth weight for gestational age); 25 infants exposed to buprenorphine, 44 infants exposed to methadone, and 109 unexposed infants were missing information on length of hospital stay. b Values are expressed as mean (standard deviation). c “Other” includes non-Hispanic women with unknown race/ethnicity. The risks of preterm birth and low birth weight were higher among methadone-exposed infants (Table 2), as was the mean length of hospital stay. Adjustment for maternal age, race/ethnicity, year of delivery, and prenatal SSRI and benzodiazepine use in the MAX cohort did not change estimates, but incorporating data from the external cohort altered the associations. Crude risk ratios, risk ratios adjusted for MAX confounders only, and risk ratios adjusted for MAX confounders and bias analysis were as follows: for preterm birth, risk ratio (RR) = 0.46 (95% confidence interval (CI): 0.34, 0.61), RR = 0.45 (95% CI: 0.34, 0.61), and RR = 0.53 (95% CI: 0.39, 0.71), respectively; for low birth weight for gestational age, RR = 0.75 (95% CI: 0.51, 1.11), RR = 0.75 (95% CI: 0.51, 1.11), and RR = 1.14 (95% CI: 0.77, 1.69), respectively. Corresponding mean differences in the length of infant hospitalization were −7.45 days (95% CI: −9.25, −5.65), −7.35 days (95% CI: −9.16, −5.55), and −3.66 days (95% CI: −5.46, −1.87), respectively. Table 2. Unadjusted, Adjusted, and Externally Adjusted Associations Between Prenatal Buprenorphine Exposure Versus Methadone Exposure and Infant Outcomes Among Massachusetts Medicaid Beneficiaries Who Delivered in 2006–2011 Infant Outcome  Unadjusted  Adjusted Using MAX Confounder Dataa  Adjusted Using MAX Confounder Dataa and External Confounder Data With Bias Analysis  PE  95% CI  PE  95% CI  PE  95% CI  Preterm birth (<37 weeks)  0.46  0.34, 0.61  0.45  0.34, 0.61  0.53  0.39, 0.71  Low birth weight for gestational age  0.75  0.51, 1.11  0.75  0.51, 1.11  1.14  0.77, 1.69  Length of infant hospital stay, days  −7.45  −9.25, −5.65  −7.35  −9.16, −5.55  −3.66  −5.46, −1.87  Infant Outcome  Unadjusted  Adjusted Using MAX Confounder Dataa  Adjusted Using MAX Confounder Dataa and External Confounder Data With Bias Analysis  PE  95% CI  PE  95% CI  PE  95% CI  Preterm birth (<37 weeks)  0.46  0.34, 0.61  0.45  0.34, 0.61  0.53  0.39, 0.71  Low birth weight for gestational age  0.75  0.51, 1.11  0.75  0.51, 1.11  1.14  0.77, 1.69  Length of infant hospital stay, days  −7.45  −9.25, −5.65  −7.35  −9.16, −5.55  −3.66  −5.46, −1.87  Abbreviations: CI, confidence interval; MAX, Medicaid Analytic eXtract; PE, point estimate. a Adjusted for maternal age, maternal race/ethnicity, and year of delivery. Table 2. Unadjusted, Adjusted, and Externally Adjusted Associations Between Prenatal Buprenorphine Exposure Versus Methadone Exposure and Infant Outcomes Among Massachusetts Medicaid Beneficiaries Who Delivered in 2006–2011 Infant Outcome  Unadjusted  Adjusted Using MAX Confounder Dataa  Adjusted Using MAX Confounder Dataa and External Confounder Data With Bias Analysis  PE  95% CI  PE  95% CI  PE  95% CI  Preterm birth (<37 weeks)  0.46  0.34, 0.61  0.45  0.34, 0.61  0.53  0.39, 0.71  Low birth weight for gestational age  0.75  0.51, 1.11  0.75  0.51, 1.11  1.14  0.77, 1.69  Length of infant hospital stay, days  −7.45  −9.25, −5.65  −7.35  −9.16, −5.55  −3.66  −5.46, −1.87  Infant Outcome  Unadjusted  Adjusted Using MAX Confounder Dataa  Adjusted Using MAX Confounder Dataa and External Confounder Data With Bias Analysis  PE  95% CI  PE  95% CI  PE  95% CI  Preterm birth (<37 weeks)  0.46  0.34, 0.61  0.45  0.34, 0.61  0.53  0.39, 0.71  Low birth weight for gestational age  0.75  0.51, 1.11  0.75  0.51, 1.11  1.14  0.77, 1.69  Length of infant hospital stay, days  −7.45  −9.25, −5.65  −7.35  −9.16, −5.55  −3.66  −5.46, −1.87  Abbreviations: CI, confidence interval; MAX, Medicaid Analytic eXtract; PE, point estimate. a Adjusted for maternal age, maternal race/ethnicity, and year of delivery. The comparability of women in the MAX and external cohorts was assessed (Table 3). Similar proportions of both cohorts were treated with buprenorphine and methadone. The delivery periods differed, as previously noted. Most women in the MAX cohort and the external cohort were non-Hispanic white, but a higher percentage of women in the external cohort were black. No other differences were identified. Table 3. Comparison of Women in a Massachusetts MAX Medicaid Cohort Who Had a Diagnosis of Drug Dependence Within 2 Years of Delivery (2006–2011) and Women With Opioid Dependence in an External Cohort (2015–2016), by Prenatal OAT Status Characteristic or Outcome  MAX Medicaid Cohorta  External Cohortb  Buprenorphine (n = 543)  Methadone (n = 477)  Buprenorphine (n = 58)  Methadone (n = 55)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Year of delivery   2006–2011  543  100  477  100  0  0  0  0   2015–2016  0  0  0  0  58  100  55  100  Age at delivery, yearsc  27.9 (4.5)  28.2 (4.9)  28.2 (4.4)  28.3 (4.4)  Race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  4  6.9  7  12.7   Non-Hispanic white  426  78.5  366  76.7  50  86.2  41  74.6   Hispanic  15  2.8  23  4.8  4  6.9  7  12.7   Other/unknownd  100  18.4  75  15.7  0  0  0  0  SSRI prescription  158  29.1  77  16.1  14  24.1  9  16.4  Benzodiazepine prescription  173  31.9  122  25.6  15  25.9  12  21.8  Preterm birth (<37 weeks)  99  18.4  155  32.9  10  22.2  11  28.2  Cesarean delivery  187  34.3  179  37.5  20  47.6  16  43.2  Characteristic or Outcome  MAX Medicaid Cohorta  External Cohortb  Buprenorphine (n = 543)  Methadone (n = 477)  Buprenorphine (n = 58)  Methadone (n = 55)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Year of delivery   2006–2011  543  100  477  100  0  0  0  0   2015–2016  0  0  0  0  58  100  55  100  Age at delivery, yearsc  27.9 (4.5)  28.2 (4.9)  28.2 (4.4)  28.3 (4.4)  Race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  4  6.9  7  12.7   Non-Hispanic white  426  78.5  366  76.7  50  86.2  41  74.6   Hispanic  15  2.8  23  4.8  4  6.9  7  12.7   Other/unknownd  100  18.4  75  15.7  0  0  0  0  SSRI prescription  158  29.1  77  16.1  14  24.1  9  16.4  Benzodiazepine prescription  173  31.9  122  25.6  15  25.9  12  21.8  Preterm birth (<37 weeks)  99  18.4  155  32.9  10  22.2  11  28.2  Cesarean delivery  187  34.3  179  37.5  20  47.6  16  43.2  Abbreviations: MAX, Medicaid Analytic eXtract; OAT, opioid agonist therapy; SSRI, selective serotonin reuptake inhibitor. a Five infants exposed to buprenorphine, 6 infants exposed to methadone, and 20 unexposed infants were missing information from inpatient hospitalization (preterm birth, low birth weight for gestational age); 25 infants exposed to buprenorphine, 44 infants exposed to methadone, and 185 unexposed infants were missing information on length of hospital stay. b The “preterm birth” variable excludes 3 women who underwent therapeutic abortion (2 receiving buprenorphine, 1 receiving methadone) and 26 women who were lost to follow-up (11 receiving buprenorphine, 15 receiving methadone). The “cesarean delivery” variable excludes those excluded from the preterm birth variable, in addition to 3 women who delivered at another site (1 receiving buprenorphine, 2 receiving methadone). c Values are expressed as mean (standard deviation). d “Other” includes non-Hispanic women with unknown race/ethnicity. Table 3. Comparison of Women in a Massachusetts MAX Medicaid Cohort Who Had a Diagnosis of Drug Dependence Within 2 Years of Delivery (2006–2011) and Women With Opioid Dependence in an External Cohort (2015–2016), by Prenatal OAT Status Characteristic or Outcome  MAX Medicaid Cohorta  External Cohortb  Buprenorphine (n = 543)  Methadone (n = 477)  Buprenorphine (n = 58)  Methadone (n = 55)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Year of delivery   2006–2011  543  100  477  100  0  0  0  0   2015–2016  0  0  0  0  58  100  55  100  Age at delivery, yearsc  27.9 (4.5)  28.2 (4.9)  28.2 (4.4)  28.3 (4.4)  Race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  4  6.9  7  12.7   Non-Hispanic white  426  78.5  366  76.7  50  86.2  41  74.6   Hispanic  15  2.8  23  4.8  4  6.9  7  12.7   Other/unknownd  100  18.4  75  15.7  0  0  0  0  SSRI prescription  158  29.1  77  16.1  14  24.1  9  16.4  Benzodiazepine prescription  173  31.9  122  25.6  15  25.9  12  21.8  Preterm birth (<37 weeks)  99  18.4  155  32.9  10  22.2  11  28.2  Cesarean delivery  187  34.3  179  37.5  20  47.6  16  43.2  Characteristic or Outcome  MAX Medicaid Cohorta  External Cohortb  Buprenorphine (n = 543)  Methadone (n = 477)  Buprenorphine (n = 58)  Methadone (n = 55)  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  No. of Pairs  %  Year of delivery   2006–2011  543  100  477  100  0  0  0  0   2015–2016  0  0  0  0  58  100  55  100  Age at delivery, yearsc  27.9 (4.5)  28.2 (4.9)  28.2 (4.4)  28.3 (4.4)  Race/ethnicity   Non-Hispanic black  2  0.4  13  2.7  4  6.9  7  12.7   Non-Hispanic white  426  78.5  366  76.7  50  86.2  41  74.6   Hispanic  15  2.8  23  4.8  4  6.9  7  12.7   Other/unknownd  100  18.4  75  15.7  0  0  0  0  SSRI prescription  158  29.1  77  16.1  14  24.1  9  16.4  Benzodiazepine prescription  173  31.9  122  25.6  15  25.9  12  21.8  Preterm birth (<37 weeks)  99  18.4  155  32.9  10  22.2  11  28.2  Cesarean delivery  187  34.3  179  37.5  20  47.6  16  43.2  Abbreviations: MAX, Medicaid Analytic eXtract; OAT, opioid agonist therapy; SSRI, selective serotonin reuptake inhibitor. a Five infants exposed to buprenorphine, 6 infants exposed to methadone, and 20 unexposed infants were missing information from inpatient hospitalization (preterm birth, low birth weight for gestational age); 25 infants exposed to buprenorphine, 44 infants exposed to methadone, and 185 unexposed infants were missing information on length of hospital stay. b The “preterm birth” variable excludes 3 women who underwent therapeutic abortion (2 receiving buprenorphine, 1 receiving methadone) and 26 women who were lost to follow-up (11 receiving buprenorphine, 15 receiving methadone). The “cesarean delivery” variable excludes those excluded from the preterm birth variable, in addition to 3 women who delivered at another site (1 receiving buprenorphine, 2 receiving methadone). c Values are expressed as mean (standard deviation). d “Other” includes non-Hispanic women with unknown race/ethnicity. In sensitivity analyses, when applying our definition of opioid dependence to the year before delivery (instead of 2 years prior), all women met the inclusion criteria. The interpretation of results was unchanged when analyses were conducted with the pregnancy period ending on the first day of the month rather than 30 days after the infant’s birth date. DISCUSSION In this paper, we report on a novel method for addressing potential uncontrolled confounding of the association between prenatal OAT exposure and infant outcomes in Massachusetts Medicaid data. The Massachusetts population is ideal for the study of OAT outcomes due to the high rates of opioid use and virtually unrestricted Medicaid coverage for buprenorphine and methadone in the state (13). Only 2 studies of the comparative safety of prenatal OAT have included more than 300 mother-infant pairs—1 from Denmark (n = 557) (4) and 1 from Vermont (n = 609) (5). To our knowledge, this MAX-based cohort study is the largest carried out to date, with 1,020 OAT-exposed mother-infant pairs. However, MAX data on potential confounders such as severity of addiction were unavailable. Careful consideration of confounding by indication is needed in assessing the comparative safety of buprenorphine and methadone with regard to infant outcomes (10, 22), because the systems of care differ substantially. In the United States, methadone treatment consists of structured observed daily dosing at a methadone clinic, while buprenorphine is self-administered through prescriptions. Some women benefit from the structured delivery and full-agonist effect of methadone, while others are better suited for the partial opioid agonist buprenorphine. Prior studies have suggested differential access to buprenorphine by race/ethnicity and social status (23), and in the MAX cohort, more women treated with methadone or without OAT prescriptions were Hispanic or black. Similar to another study that used administrative health-care data, MAX women treated with buprenorphine had more concomitant medication use, suggesting better engagement in clinical care and more stable drug dependence (24). In the external cohort used for confounding adjustment, women receiving buprenorphine had lower risk profiles, including more stable housing, more stable income, less severe drug dependence (as measured by the Addiction Severity Index), and less injection cocaine use. If women with these lower risk profiles deliver healthier babies, confounding by indication is likely. Our approach to controlling for confounding by indication did not fully attenuate the observed lower risk of preterm delivery or hospital stay length for buprenorphine as compared with methadone, but evidence of confounding was present. Shorter unadjusted hospital stays in buprenorphine-exposed infants have been observed in some but not all studies (3, 6, 25). The unadjusted risk ratio for preterm birth in buprenorphine- versus methadone-exposed infants in the Vermont study was 0.53 (95% CI: 0.33, 0.85), similar to our result. Our finding of somewhat attenuated but beneficial associations brings up the question of whether our methods fully corrected for confounding by indication. For example, clinical prescribing may have differed in the MAX cohort and the external cohort, resulting in misclassification in our propensity scores. There was little buprenorphine treatment failure—defined as transition from prenatal buprenorphine to methadone for nonadherence or continued illicit opioid use—in women from the external study clinic during 2006–2010 (5.7%) in comparison with women who delivered in 2015 and 2016 (22%) (19). This change probably reflects provider efforts at the study clinic to make buprenorphine treatment more accessible for women with more severe drug dependence. Because of the smaller size of the external cohort, we assumed that the effect of confounders was uniform by OAT, which may not have adequately captured the confounding. Nonetheless, our findings suggest that there may be a higher risk of preterm birth and longer hospital stays in methadone-exposed infants. Unlike preterm birth and length of hospital stay, there was no association between OAT exposure and low birth weight for gestational age after adjustment with bias analysis. Other investigators have reported higher mean birth weights in buprenorphine-exposed infants (3, 5, 7) and in opioid-unexposed compared with methadone-exposed infants (26, 27), but few have examined growth appropriate for gestational age (28, 29). A limitation of using administrative health-care data for research is misclassification of study exposures and outcomes. We included ICD-9 codes used in prior studies to identify deliveries, opioid dependence, methadone and buprenorphine exposure, and infant outcomes (13, 15, 30). The sensitivity and specificity of ICD-9 codes for preterm birth was quite high in a database validation study (18), while the use of ICD-9 codes 656.5x for infants born small for gestational age had low sensitivity (31). We expanded on these prior codes to include others that indicated low birth weight for gestational age; however, some nondifferential misclassification probably exists. The MAX cohort demonstrated successful linkage of deliveries to infants at 59%, a proportion similar to that of another study (12). Linkage is likely to be affected by level of claim completeness, inconsistent assignment of the same family case identifier to women and children, and use of delivery codes that did not identify viable pregnancies or pregnancies carried to term. Women were included in the MAX cohort regardless of consistent noncapitated Medicaid coverage in pregnancy, raising the possibility that some women classified as untreated may have received OAT for which Medicaid claims were not submitted. The number of such women is expected to have been small, given that few women in our study had private health insurance coverage at any time during pregnancy, almost all had unrestricted Medicaid coverage at some point during pregnancy, and a majority had at least 1 other prenatal medication claim. Restricting the study population to women with Medicaid coverage throughout pregnancy would have unnecessarily excluded women for whom Medicaid coverage, prenatal care, and OAT began after conception. Our finding that 59% of women with a diagnosis of opioid dependence in the MAX cohort did not have an OAT prescription in the year before delivery is concerning, especially given that Massachusetts has unrestricted Medicaid coverage for buprenorphine and methadone. Cash clinics, where patients pay out of pocket for buprenorphine prescriptions because of a lack of buprenorphine providers in the area, may have been a possible source of treatment for the unexposed women. It unlikely that this would account for OAT in the majority of the unexposed low-income women, given that comprehensive prenatal care in Medicaid hospital clinics would include OAT. Limited availability of treatment for substance users is well documented and has been exacerbated by the current opioid crisis (32). Possible barriers to care among pregnant drug-dependent women, including stigma that prevents women from disclosing drug dependence in pregnancy and/or seeking prenatal care, lack of transportation to care, and lack of referrals to addiction providers, should be investigated. High rates of drug dependence in pregnant women and of adverse outcomes in their infants are ongoing concerns (33, 34). It is feasible to use extant Medicaid data paired with detailed data from an external cohort to adjust for unmeasured confounding in order to study the comparative safety of prenatal OAT exposure, but limitations exist. Some but not all of the associations in this study were explained by confounding. Our study provides insight into the use Medicaid data for OAT outcomes, prescribing preferences in a Medicaid population, and the effect of unmeasured confounding bias from maternal characteristics. ACKNOWLEDGMENTS Author affiliations: Department of Surgery, Queen’s University, Kingston, Ontario, Canada (Susan B. Brogly); Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Sonia Hernández-Diaz); and Department of Epidemiology, School of Public Health, Boston University, Boston, Massachusetts (Emily Regan, Ela Fadli, Kristen A. Hahn, Martha M. Werler). This work was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development under grant 1R21HD081271-01 REVISED (S.B.B.). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Conflict of interest: none declared. Abbreviations CI confidence interval ICD-9 International Classification of Diseases, Ninth Revision MAX Medicaid Analytic eXtract MSIS ID Medicaid Statistical Information System Identifiers OAT opioid agonist therapy RR risk ratio SSRI selective serotonin reuptake inhibitor REFERENCES 1 Ko JY, Patrick SW, Tong VT, et al.  . Incidence of neonatal abstinence syndrome—28 states, 1999–2013. MMWR Morb Mortal Wkly Rep . 2016; 65( 31): 799– 802. Google Scholar CrossRef Search ADS PubMed  2 Substance Use and Mental Health Services Administration, US Department of Health and Human Services. Federal Guidelines for Opioid Treatment Programs. Rockville, MD: Substance Use and Mental Health Services Administration; 2015. (DHHS publication no. EP15-FEDGUIDEOTP). 3 Metz V, Jagsch R, Ebner N, et al.  . Impact of treatment approach on maternal and neonatal outcome in pregnant opioid-maintained women. Hum Psychopharmacol . 2011; 26( 6): 412– 421. Google Scholar PubMed  4 Nørgaard M, Nielsson MS, Heide-Jørgensen U. Birth and neonatal outcomes following opioid use in pregnancy: a Danish population-based study. Subst Abuse . 2015; 9( suppl 2): 5– 11. Google Scholar PubMed  5 Meyer MC, Johnston AM, Crocker AM, et al.  . Methadone and buprenorphine for opioid dependence during pregnancy: a retrospective cohort study. J Addict Med . 2015; 9( 2): 81– 86. Google Scholar CrossRef Search ADS PubMed  6 Jones HE, Kaltenbach K, Heil SH, et al.  . Neonatal abstinence syndrome after methadone or buprenorphine exposure. N Engl J Med . 2010; 363( 24): 2320– 2331. Google Scholar CrossRef Search ADS PubMed  7 Welle-Strand GK, Skurtveit S, Jones HE, et al.  . Neonatal outcomes following in utero exposure to methadone or buprenorphine: a national cohort study of opioid-agonist treatment of pregnant women in Norway from 1996 to 2009. Drug Alcohol Depend . 2013; 127( 1–3): 200– 206. Google Scholar CrossRef Search ADS PubMed  8 Pritham UA. Breastfeeding promotion for management of neonatal abstinence syndrome. J Obstet Gynecol Neonatal Nurs . 2013; 42( 5): 517– 526. Google Scholar CrossRef Search ADS PubMed  9 Brogly SB, Saia KA, Walley AY, et al.  . Prenatal buprenorphine versus methadone exposure and neonatal outcomes: systematic review and meta-analysis. Am J Epidemiol . 2014; 180( 7): 673– 686. Google Scholar CrossRef Search ADS PubMed  10 Krans EE, Bogen D, Richardson G, et al.  . Factors associated with buprenorphine versus methadone use in pregnancy. Subst Abus . 2016; 37( 4): 550– 557. Google Scholar CrossRef Search ADS PubMed  11 Palmsten K, Huybrechts KF, Michels KB, et al.  . Antidepressant use and risk for preeclampsia. Epidemiology . 2013; 24( 5): 682– 691. Google Scholar CrossRef Search ADS PubMed  12 Patorno E, Huybrechts KF, Bateman BT, et al.  . Lithium use in pregnancy and the risk of cardiac malformations. N Engl J Med . 2017; 376( 23): 2245– 2254. Google Scholar CrossRef Search ADS PubMed  13 Baxter JD, Clark RE, Samnaliev M, et al.  . Factors associated with Medicaid patients’ access to buprenorphine treatment. J Subst Abuse Treat . 2011; 41( 1): 88– 96. Google Scholar CrossRef Search ADS PubMed  14 Research Data Assistance Center, Centers for Medicare and Medicaid Services. CMS Data Request Center. http://www.resdac.org/cms-data/request/cms-data-request-center. Accessed November 9, 2016. 15 Palmsten K, Huybrechts KF, Mogun H, et al.  . Harnessing the Medicaid Analytic eXtract (MAX) to evaluate medications in pregnancy: design considerations. PLoS One . 2013; 8( 6): e67405. Google Scholar CrossRef Search ADS PubMed  16 Substance Abuse and Mental Health Services Administration, US Department of Health and Human Services. The Confidentiality of Alcohol and Drug Abuse Patient Records Regulation and the HIPAA Privacy Rule: Implications for Alcohol and Substance Abuse Programs. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2004. (Publication no. PHD1083). https://store.samhsa.gov/product/The-Confidentiality-of-Alcohol-and-Drug-Abuse-Patient-Records-Regulation-and-the-HIPAA-Privacy-Rule/PHD1083. Accessed April 2, 2018. 17 Freund DA, Neuschler E. Overview of Medicaid capitation and case-management initiatives. Health Care Financ Rev . 1986;( Spec. No.): 21– 30. 18 Margulis AV, Setoguchi S, Mittleman MA, et al.  . Algorithms to estimate the beginning of pregnancy in administrative databases. Pharmacoepidemiol Drug Saf . 2013; 22( 1): 16– 24. Google Scholar CrossRef Search ADS PubMed  19 Saia K, Bagley SM, Wachman EM, et al.  . Prenatal treatment for opioid dependency: observations from a large inner-city clinic. Addict Sci Clin Pract . 2017; 12( 1): 5. Google Scholar CrossRef Search ADS PubMed  20 McLellan AT, Luborsky L, Woody GE, et al.  . An improved diagnostic evaluation instrument for substance abuse patients. The Addiction Severity Index. J Nerv Ment Dis . 1980; 168( 1): 26– 33. Google Scholar CrossRef Search ADS PubMed  21 Lin DY, Psaty BM, Kronmal RA. Assessing the sensitivity of regression results to unmeasured confounders in observational studies. Biometrics . 1998; 54( 3): 948– 963. Google Scholar CrossRef Search ADS PubMed  22 Brogly SB, Hahn KA, Diaz SH, et al.  . Confounding of the comparative safety of prenatal opioid agonist therapy. J Addict Res Ther . 2015; 6( 4): 252. Google Scholar CrossRef Search ADS PubMed  23 Hansen HB, Siegel CE, Case BG, et al.  . Variation in use of buprenorphine and methadone treatment by racial, ethnic, and income characteristics of residential social areas in New York City. J Behav Health Serv Res . 2013; 40( 3): 367– 377. Google Scholar CrossRef Search ADS PubMed  24 Brogly SB, Turner SD, Lajkosz K, et al.  . Infants born to opioid-dependent women in Ontario, 2002–2014. J Obstet Gynaecol Can . 2017; 39( 3): 157– 165. Google Scholar CrossRef Search ADS PubMed  25 Wachman EM, Newby PK, Vreeland J, et al.  . The relationship between maternal opioid agonists and psychiatric medications on length of hospitalization for neonatal abstinence syndrome. J Addict Med . 2011; 5( 4): 293– 299. Google Scholar CrossRef Search ADS PubMed  26 Liu AJ, Sithamparanathan S, Jones MP, et al.  . Growth restriction in pregnancies of opioid-dependent mothers. Arch Dis Child Fetal Neonatal Ed . 2010; 95( 4): F258– F262. Google Scholar CrossRef Search ADS PubMed  27 Kandall SR, Albin S, Lowinson J, et al.  . Differential effects of maternal heroin and methadone use on birthweight. Pediatrics . 1976; 58( 5): 681– 685. Google Scholar PubMed  28 Pritham UA, Paul JA, Hayes MJ. Opioid dependency in pregnancy and length of stay for neonatal abstinence syndrome. J Obstet Gynecol Neonatal Nurs . 2012; 41( 2): 180– 190. Google Scholar CrossRef Search ADS PubMed  29 Kelty E, Hulse G. A retrospective cohort study of birth outcomes in neonates exposed to naltrexone in utero: a comparison with methadone, buprenorphine- and non-opioid-exposed neonates. Drugs . 2017; 77( 11): 1211– 1219. Google Scholar CrossRef Search ADS PubMed  30 Ailes EC, Dawson AL, Lind JN, et al.  . Opioid prescription claims among women of reproductive age—United States, 2008–2012. MMWR Morb Mortal Wkly Rep . 2015; 64( 2): 37– 41. Google Scholar PubMed  31 Phiri K, Hernandez-Diaz S, Tsen LC, et al.  . Accuracy of ICD-9-CM coding to identify small for gestational age newborns. Pharmacoepidemiol Drug Saf . 2015; 24( 4): 381– 388. Google Scholar CrossRef Search ADS PubMed  32 Saloner B, Karthikeyan S. Changes in substance abuse treatment use among individuals with opioid use disorders in the United States, 2004–2013. JAMA . 2015; 314( 14): 1515– 1517. Google Scholar CrossRef Search ADS PubMed  33 Patrick SW, Schumacher RE, Benneyworth BD, et al.  . Neonatal abstinence syndrome and associated health care expenditures: United States, 2000–2009. JAMA . 2012; 307( 18): 1934– 1940. Google Scholar CrossRef Search ADS PubMed  34 Tolia VN, Patrick SW, Bennett MM, et al.  . Increasing incidence of the neonatal abstinence syndrome in US neonatal ICUs. N Engl J Med . 2015; 372( 22): 2118– 2126. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

American Journal of EpidemiologyOxford University Press

Published: Nov 16, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off