Infant Mortality Associated With Prenatal Opioid Exposure

Key Points

Question  What is the risk of mortality in opioid-exposed infants, and is mortality associated with neonatal opioid withdrawal syndrome?

Findings  In this cohort study of more than 1 million maternal-infant dyads, the odds of mortality in infants with a history of neonatal opioid withdrawal syndrome was not significantly different from the reference population in adjusted analyses. In contrast, the odds of mortality in opioid-exposed infants not diagnosed with neonatal opioid withdrawal syndrome was 72% greater than the reference population.

Meaning  The findings of this study suggest that programs and policies to support women with opioid use disorder and their infants are warranted, regardless of perceived severity of neonatal opioid withdrawal.

Importance  Knowledge of health outcomes among opioid-exposed infants is limited, particularly for those not diagnosed with neonatal opioid withdrawal syndrome (NOWS).

Objectives  To describe infant mortality among opioid-exposed infants and identify how mortality risk differs in opioid-exposed infants with and without a diagnosis of NOWS compared with infants without opioid exposure.

Design, Setting, and Participants  A retrospective cohort study of maternal-infant dyads was conducted, linking health care claims with vital records for births from January 1, 2010, to December 31, 2014, with follow-up of infants until age 1 year (through 2015). Maternal-infant dyads were included if the infant was born in Texas at 22 to 43 weeks’ gestational age to a woman aged 15 to 44 years insured by Texas Medicaid. Data analysis was performed from May 2019 to October 2020.

Exposure  The primary exposure was prenatal opioid exposure, with infants stratified by the presence or absence of a diagnosis of NOWS during the birth hospitalization.

Main Outcomes and Measures  Risk of infant mortality (death at age <365 days) was examined using Kaplan-Meier and log-rank tests. A series of logistic regression models was estimated to determine associations between prenatal opioid exposure and mortality, adjusting for maternal and neonatal characteristics and clustering infants at the maternal level to account for statistical dependence owing to multiple births during the study period.

Results  Among 1 129 032 maternal-infant dyads, 7207 had prenatal opioid exposure, including 4238 diagnosed with NOWS (mean [SD] birth weight, 2851 [624] g) and 2969 not diagnosed with NOWS (mean [SD] birth weight, 2971 [639] g). Infant mortality was 20 per 1000 live births for opioid-exposed infants not diagnosed with NOWS, 11 per 1000 live births for infants with NOWS, and 6 per 1000 live births in the reference group (P < .001). After adjusting for maternal and neonatal characteristics, mortality in infants with a NOWS diagnosis was not significantly different from the reference population (odds ratio, 0.82; 95% CI, 0.58-1.14). In contrast, the odds of mortality in opioid-exposed infants not diagnosed with NOWS was 72% greater than the reference population (odds ratio, 1.72; 95% CI, 1.25-2.37).

Conclusions and Relevance  In this study, opioid-exposed infants appeared to be at increased risk of mortality, and the treatments and supports provided to those diagnosed with NOWS may be protective. Interventions to support opioid-exposed maternal-infant dyads are warranted, regardless of the perceived severity of neonatal opioid withdrawal.

In 2019, an estimated 9.5 million adults in the US misused opioids, and 1.6 million had an opioid use disorder (OUD).¹ The increasing prevalence of OUD among pregnant women has resulted in a corresponding increase in the number of infants exposed to opioids in utero and, in turn, the number who experience neonatal opioid withdrawal.^2-4 Neonatal opioid withdrawal syndrome (NOWS) is diagnosed in approximately 20 000 newborns in the US each year, with neurologic and gastrointestinal symptoms presenting along a continuum of disease severity.^5,6 Hospital stays for infants with NOWS average 16 to 19 days, with national health care costs exceeding $500 million annually.^2,3,5-7 Eighty percent of these costs are borne by Medicaid.^5,6

Recognizing this substantial burden of disease, several research and quality improvement initiatives have aimed to improve the hospital-based care of infants with NOWS and support maternal-infant dyads in their transitions home following hospital discharge.^8-11 However, 2 knowledge gaps persist. First, given our clinical focus on NOWS, we know little about birth outcomes in opioid-exposed infants who are not diagnosed with opioid withdrawal; outside of clinical trials, even knowledge about the incidence of NOWS following prenatal opioid exposure is lacking. Second, our understanding of infants’ health outcomes following hospital discharge is limited, a gap highlighted in 3 systematic reviews and a joint statement by the National Institute of Child Health and Human Development, the American College of Obstetrics and Gynecology, the American Academy of Pediatrics, and the Centers for Disease Control and Prevention.^2,12-14 Infant mortality is a widely accepted indicator of population health but has been the focus of very few studies in opioid-exposed infants.^15-17

To address these knowledge gaps, we investigated infant mortality in a large cohort of infants with prenatal opioid exposure, examining how survival differed in infants with and without a diagnosis of NOWS compared with infants without opioid exposure.

We conducted a population-based, retrospective cohort study using the Texas Neonatal Intensive Care Project data set developed by Texas Health and Human Services. Texas Medicaid is the primary payer for 1 of 18 births in the US each year, and more than 50% of all births in the state.¹⁸ This data set links (1) maternal and infant professional and facility health care claims and encounters, (2) enrollment files, (3) birth certificate data, and (4) infant mortality data for all Medicaid-insured newborns born between January 1, 2010, and December 31, 2014. The data set includes maternal health care claims/encounters during pregnancy and reflects infant mortality through December 31, 2015. We excluded infants born outside of Texas (n = 1151); infants missing a facility or professional claim within 2 days of birth unless they had died within 48 hours of birth, making claim generation less likely (n = 122 883); infants with a birth weight less than 400 g (n = 932); mothers younger than 15 years or older than 44 years (n = 3452); and infants born less than 22 weeks or greater than 43 weeks of gestational age (n = 957). The institutional review boards of Texas Health and Human Services and Dartmouth College provided study approval and granted a waiver of consent for research that involves no more than minimal risk.

Our primary exposure was prenatal opioid exposure, with infants stratified based on the presence or absence of a NOWS diagnosis during the birth hospitalization. We defined prenatal opioid exposure as (1) a maternal International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis of OUD documented in at least 1 health care encounter during pregnancy or the birth hospitalization using a previously developed algorithm (eTable 1 in the Supplement),^19-21 (2) a Current Procedural Terminology code for methadone delivery during the prenatal period (eTable 1 in the Supplement), or (3) an infant diagnosis of NOWS, documented in any claims position during the birth hospitalization. NOWS was defined using the ICD-9-CM designation of drug withdrawal syndrome in newborn (code 779.5), which has been shown to have a high positive predictive value.²² The reference population was all other infants meeting the above-described inclusion criteria.

From infant birth certificates we determined maternal age, maternal-reported race/ethnicity, educational attainment, marital status, parity (number of births ≥24 weeks’ gestational age), smoking during pregnancy, receipt of prenatal care (reported as a binary yes/no variable as well as by number of prenatal visits), mode of delivery (vaginal or cesarean delivery) and plurality (singleton or multiple birth). From enrollment files we determined the number of months of maternal Medicaid enrollment during pregnancy. From maternal health care claims we determined the presence of co-occurring substance abuse and mental illness using established ICD-9-CM–based algorithms,^19,23,24 and calculated the Obstetric Comorbidity Index, a validated index that identifies and enumerates 20 maternal health conditions in a single score.^25,26 Regarding birth outcomes, we examined (1) preterm birth, defined as birth at less than 37 weeks’ gestational age as recorded on the birth certificate; (2) small for gestational age, defined as a birth weight of less than the 10th percentile for gestational age and sex using the INTERGROWTH-21st standards^27,28; and (3) congenital abnormalities, identified using a previously established diagnostic algorithm.^29,30 To reduce the risk of ascertainment bias given differential birth hospitalization lengths of stay in infants with and without NOWS, diagnostic codes for congenital abnormalities were required to occur on 2 separate dates and at least once following discharge from the birth hospitalization.

Our primary outcome was infant mortality, defined as death before age 365 days as identified on health care claims or vital records. We classified the underlying cause of death using International Statistical Classification of Disease, 10th Revision (ICD-10) cause-of-death codes, applying the Lavista Ferres et al³¹ algorithm for sudden unexplained infant death, which includes code R95 (sudden infant death syndrome), code R99 (ill-defined and unknown cause), or code W75 (accidental suffocation or strangulation in bed), and the Nakamura et al³² classification scheme for all other causes of infant mortality (excluding sudden infant death syndrome from the Nakamura et al congenital abnormalities category to report mutually exclusive causes of death). To avoid reporting cell sizes less than 11, we report causes of death due to (1) sudden unexplained infant death, (2) prematurity or congenital abnormalities and related causes, and (3) other or unknown causes (including obstetric conditions, birth asphyxia, infections, and other external causes).

Statistical analysis was performed from May 2019 to October 2020. We first calculated individual-level summary statistics to examine differences in maternal and infant characteristics and birth outcomes across the 3 exposure groups, using χ² tests for categorical variables and analysis of variance for continuous variables. To examine differences in risk of infant mortality, we created Kaplan-Meier curves and tested for statistical differences in these curves using log-rank tests. Because NOWS is typically diagnosed between 2 and 5 days of life, opioid-exposed infants who died before age 5 days may have been less likely to receive this diagnosis (survivorship bias). Therefore, as a sensitivity analysis, we repeated these analyses contingent upon survival until 5 days.

To further examine the association between prenatal opioid exposure and infant mortality, we developed a series of logistic regression models that we estimated using generalized estimating equations to account for the repeated measures made among women with multiple infants born during the study period. We favored fixed time outcomes over time-to-event analyses owing to evidence of nonproportional hazards and no observable censoring of infant mortality during the follow-up period. In the first model we adjusted for maternal sociodemographic factors, in the second model we added factors related to maternal health and behaviors (excluding maternal substance use disorders, given the high rate of co-occurrence of these disorders with opioid exposure), and in the third model we added factors related to the infant and mode of delivery (Figure 1). All covariates were selected a priori based on established or hypothesized associations with infant mortality. All models included a variable to indicate those missing a maternal-infant link.

Three variables in our models that adjusted for maternal health and behaviors (mental health diagnoses, Obstetric Comorbidity Index, and duration of Medicaid enrollment) were missing for 185 140 infants whose claims could not be linked to their mothers (eTable 2 in the Supplement), and 17 873 maternal-infant dyads were missing data regarding receipt of prenatal care. To account for these missing data, we used multiple imputation to create 10 imputed data sets, replacing missing values with imputed values informed by maternal age at birth, race/ethnicity, educational attainment, and marital status.

We conducted several sensitivity analyses to evaluate the robustness of our findings using the multiply imputed data sets described above. First, to account for potential survivorship bias in NOWS diagnosis, we repeated our logistic regression models contingent upon survival until age 5 days. Second, we repeated these analyses using only maternal-infant dyads with linked infant and maternal claims. Third, recognizing that women with and without a documented OUD diagnosis may differ in observed and unobserved characteristics, we repeated our analysis limited to maternal-infant dyads with documented maternal OUD. Fourth, recognizing that approximately 1% of infants with NOWS develop withdrawal symptoms due to iatrogenic factors, we repeated our models excluding those with a birth weight less than 1500 g or with any of the following complex neonatal diagnoses: intraventricular hemorrhage, periventricular leukomalacia, necrotizing entercolitis, spontaneous intestinal perforation, or bronchopulmonary dysplasia.^22,33 Fifth, given that infant mortality may be associated with the hospital of birth, we estimated a companion series of regression models including a random effect for birth hospital instead of using generalized estimating equations. Sixth, as an alternative approach to account for confounding that is more robust to outcome model misspecification, we calculated a propensity score for opioid-exposed infants to measure the predicted probability of receiving a NOWS diagnosis using a logistic regression model, including characteristics shown in Figure 1 as predictors. We then accounted for differences in the propensity of a NOWS diagnosis using inverse probability weighting (trimming weights at the 90th percentile), thereby avoiding the need to adjust for these factors in the infant mortality models.³⁴

Analyses were conducted using SAS, version 9.3 (SAS Institute Inc) and Stata, version 16 (StataCorp LLC) software. All statistical testing was 2-sided, with P < .05 considered statistically significant.

Following application of the above-described eligibility criteria, our sample included 1 129 032 maternal-infant dyads. Of these, 7207 had prenatal opioid exposure, including 4238 diagnosed with NOWS (3.8 per 1000 live births) (mean [SD] birth weight, 2851 [624] g) and 2969 who did not receive a NOWS diagnosis (2.6 per 1000 live births) (mean [SD] birth weight, 2971 [639] g).

In total, 62.0% of the maternal cohort was Hispanic, 21.0% was non-Hispanic White, and 14.1% was non-Hispanic Black; mean (SD) age was 25.4 (5.8) years (Table 1). Opioid exposure was most frequently observed in women who were non-Hispanic White (3062 [42.5%]), were unmarried (5386 [74.7%]), and had smoked during pregnancy (2452 [34.0%]). Women with prenatal opioid exposure also had considerably higher rates of mental illness (5158 [81.2%]) and substance use disorders (4952 [77.9%]) than women in the reference group (mental illness, 125 639 [13.4%]; substance use disorders, 33 965 [3.6%]). In evaluating rates of preterm delivery, small for gestational age, and congenital abnormalities, we observed a stepwise pattern in which infants diagnosed with NOWS were most likely to have experienced these birth outcomes, followed by opioid-exposed infants without NOWS, with the lowest rates observed in the reference population. In contrast, the infant mortality rate was highest in opioid-exposed infants without NOWS (20 per 1000 live births), compared with 11 per 1000 live births in infants with a history of NOWS, and 6 per 1000 live births in the reference group infants (P < .001) (Table 2).

Figure 2 illustrates differences in survival over time; opioid-exposed infants had significantly lower estimated survival probabilities than the reference population (survival during infancy in the reference group: 99.4%; 95% CI, 99.4%-99.5%; in the NOWS group: 98.9%; 95% CI, 98.5%-99.2%; in the opioid-exposed without NOWS group: 98.0%; 95% CI, 97.4%-98.5%; log-rank P < .001). The divergence of the curves for the 2 opioid-exposed groups at age 2 days raised concerns for potential survivorship bias in the NOWS group. In our sensitivity analysis beginning at age 5 days, the differences between the estimated survival probabilities remained and were most marked during the postneonatal period (survival during infancy in the reference group: 99.6%; 95% CI, 99.6%-99.6%; in the NOW group: 98.9%; 95% CI, 98.6%-99.2%; in the opioid-exposed without NOW group: 98.5%; 95% CI, 98.0%-98.9%; log-rank P < .001).

In our unadjusted analysis, infants with NOWS had a 2.01 (95% CI, 1.51-2.68) greater odds of mortality compared with the reference population, and opioid-exposed infants without NOWS had an odds of mortality 3.58 (95% CI, 2.76-4.63) times greater than the reference population (Table 3). Adjustment for maternal sociodemographic characteristics attenuated these findings. After adjusting for maternal health and behavioral characteristics, the odds of mortality were no longer significantly elevated in infants with a history of NOWS (odds ratio, 1.17; 95% CI, 0.88-1.56) but remained elevated in opioid-exposed infants without NOWS (odds ratio, 2.08; 95% CI, 1.57-2.76). After adding neonatal and delivery characteristics to the model, the odds of mortality in opioid-exposed infants without NOWS remained 72% higher than the reference population. Our models directly comparing mortality in opioid-exposed infants with and without NOWS illustrate that infants diagnosed with NOWS had significantly decreased odds of mortality that were independent of observed maternal sociodemographic and health characteristics, as well as birth weight, gestational age, and congenital abnormalities.

Our sensitivity analyses yielded similar results, demonstrating the robustness of these findings in analyses beginning at age 5 days (eTable 3 and eTable 4 in the Supplement), when limiting opioid-exposed maternal-infant dyads to those with a maternal diagnosis of OUD (eTable 5 in the Supplement), when excluding infants with risk factors for iatrogenic NOWS (eTable 6 in the Supplement), when modeled using mixed-effect logistic regression including a random effect for birth hospital (eTable 7 in the Supplement), and when limiting our analysis to maternal-infant dyads for whom we had a successful linkage of maternal and infant claims (eTable 8 in the Supplement).

Opioid-exposed infants not diagnosed with NOWS represented more than 40% of the opioid-exposed cohort and appear to be at particularly increased risk of death despite having lower rates of prematurity, small for gestational age, and congenital abnormalities than infants diagnosed with NOWS. After adjusting for maternal and infant characteristics, the odds of mortality in infants diagnosed with NOWS were not greater than the reference population. In contrast, opioid-exposed infants not diagnosed with NOWS had a substantially increased odds of mortality unexplained by observable factors.

Although 3 systematic reviews indicate a substantial gap in our knowledge regarding outcomes following hospital discharge for opioid-exposed infants,^12-14 a small number of studies^15,16 have reported that infants with a history of neonatal abstinence syndrome are at increased risk of mortality. One population-based study in the US found that infants with a history of neonatal abstinence syndrome had 3.4 times the unadjusted risk of mortality relative to the reference population,¹⁵ and a Canadian study observed a mortality rate of 12.2 per 1000 live births among opioid-exposed infants—a rate almost 3 times their national average.¹⁶ Our findings add to this literature and provide insight into causes of death. In the present cohort, approximately 1 in 4 deaths among opioid-exposed infants were due to sudden unexplained infant death, consistent with a number of studies indicating relatively higher rates of sudden unexplained infant death in substance-exposed infants.^35-38

In contrast, previous studies have reported prenatal opioid exposure to be associated with improved survival during the neonatal period, despite higher rates of maternal and infant comorbidities.¹⁷ This pattern of improved neonatal survival has been hypothesized to be due to accelerated lung maturation in opioid-exposed infants, a hypothesis supported by animal studies.^39,40 By stratifying the study population to examine survival separately among opioid-exposed infants with and without a NOWS diagnosis, we observed differences in survival during the first few days of life, suggesting that survivorship bias may contribute to relatively improved neonatal survival in those diagnosed with NOWS. However, even excluding deaths earlier than age 5 days, by which time most NOWS diagnoses are made, opioid-exposed infants without NOWS demonstrated increased odds of mortality that were independent of prematurity, low birth weight, and congenital abnormalities.

Recent initiatives have focused on improving hospital-to-home transitions for infants with NOWS, including referrals to early intervention, health services, and other community supports.^9,41 Our finding that opioid-exposed infants diagnosed with NOWS had improved survival compared with those who did not receive this diagnosis raises questions about whether enhanced provision of health and social services may have contributed to their improved survival. The postnatal period is one of substantial risk to women with OUD, including increased risks of treatment discontinuation, overdose, and postpartum depression.^42-45 We hypothesize that a NOWS diagnosis and the accompanying care for maternal-infant dyads may support maternal stability in treatment for OUD and concomitant mental health conditions and thereby decrease mortality risk. Although the present analysis cannot test these hypotheses, the increased mortality rate observed in opioid-exposed infants provides justification for enhanced health and community supports for both women and their infants, regardless of whether the infant received a NOWS diagnosis.

The results of this study should be interpreted in light of its strengths and limitations. Most studies of prenatal opioid exposure conducted to date have focused on predominantly non-Hispanic White populations; the racial and ethnic diversity of our sample adds to this literature. Our linkage of maternal-infant health care claims with vital records provides data for risk adjustment that are typically not available in administrative data sets, including maternal sociodemographic and health characteristics, as well as precise gestational age and birth weight. Limitations of this study include the absence of pharmacy claims data, which prevented us from examining maternal use of medications for OUD or how stability of maternal recovery is associated with infant outcomes, both of which are potential confounding variables and relevant to further understanding of infant mortality. In addition, the identification of modifiable factors potentially associated with infant mortality, such as birth hospital length of stay, postdischarge health care use, and custody status, is beyond the scope of this analysis. Regarding generalizability, Texas had a relatively low NOWS incidence during the study period compared with other US regions.⁷ Relatedly, eligibility criteria for Medicaid coverage during pregnancy varies across states, which may influence the generalizability of study findings.⁴⁶ Although ICD-9-CM and ICD-10 codes for neonatal drug withdrawal have been shown to have high positive predictive value for clinician-diagnosed NOWS,²² we do not have data regarding the specific symptoms, toxicologic test results, or clinical scoring for opioid-exposed infants. As with all studies reliant upon health care claims, our findings may be affected by unmeasured confounding. However, even our unadjusted results have clinical and health policy implications.

The findings of our study suggest that opioid-exposed infants are at increased risk of mortality during infancy, and the constellation of treatments and supports provided to infants diagnosed with NOWS may be protective. Clinical interventions, public health programs, and health policy to support women with OUD and their infants appear to be warranted, regardless of the perceived severity of neonatal opioid withdrawal.

Accepted for Publication: October 30, 2020.

Published Online: April 12, 2021. doi:10.1001/jamapediatrics.2020.6364

Corresponding Author: JoAnna K. Leyenaar, MD, PhD, MPH, Department of Pediatrics, Children’s Hospital at Dartmouth-Hitchcock Medical Center, Williamson Building, Lebanon, NH 03756 (joanna.k.leyenaar@hitchcock.org).

Author Contributions: Dr Schaefer had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Leyenaar, O'Malley.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Leyenaar, Schaefer.

Critical revision of the manuscript for important intellectual content: Schaefer, Wasserman, Moen, O'Malley, Goodman.

Statistical analysis: Schaefer, Wasserman, O'Malley.

Obtained funding: O'Malley, Goodman.

Administrative, technical, or material support: Leyenaar, Wasserman, O'Malley.

Supervision: Leyenaar, O'Malley.

Conflict of Interest Disclosures: Dr Leyenaar reported receiving a grant from Dartmouth-Hitchcock Medical Center Hearst Foundation endowment award in Maternal and Child Health during the conduct of the study. Dr O'Malley reported receiving grants from the National Institutes of Health outside the submitted work. Dr Goodman reported receiving grants from the Texas Health and Human Services Commission during the conduct of the study. No other disclosures were reported.

Funding/Support: Funding for this study was received, in part, from the Hearst Foundation endowment award in Maternal and Child Health, Dartmouth-Hitchcock Medical Center. Data were provided by the State of Texas Health and Human Services Commission through the State of Texas Medicaid External Quality Review Organization at the University of Florida Institute for Child Health Policy.

Role of the Funder/Sponsor: The design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication were done entirely at The Dartmouth Institute for Health Policy & Clinical Practice and Dartmouth-Hitchcock Medical Center; the funders had no role in these study aspects. Although State of Texas staff members provided comments on a draft of the manuscript, decisions related to analyses and manuscript development and submission were entirely independent of the State of Texas and solely the responsibility of the authors.

Additional Contributions: David Lynch, MA, Dan Culica, MD, PhD, Jimmy Blanton, MPAff, and colleagues (Texas Health and Human Services) played instrumental roles in data sharing, facilitating institutional review board review, and engaging key stakeholders in discussion of this work; in addition, they provided feedback on the manuscript. We did not provide compensation for their contributions.