Long-term mortality in mothers of infants with neonatal abstinence syndrome: A population-based parallel-cohort study in England and Ontario, Canada.

BACKGROUND: Opioid addiction is a major public health threat to healthy life expectancy; however, little is known of long-term mortality for mothers with opioid use in pregnancy. Pregnancy and delivery care are opportunities to improve access to addiction and supportive services. Treating neonatal abstinence syndrome (NAS) as a marker of opioid use during pregnancy, this study reports long-term maternal mortality among mothers with a birth affected by NAS in relation to that of mothers without a NAS-affected birth in 2 high-prevalence jurisdictions, England and Ontario, Canada. METHODS AND
FINDINGS: We conducted a population-based study using linked administrative health data to develop parallel cohorts of mother-infant dyads in England and Ontario between 2002 and 2012. The study population comprised 13,577 and 4,966 mothers of infants with NAS and 4,205,675 and 929,985 control mothers in England and Ontario, respectively. Death records captured all-cause maternal mortality after delivery through March 31, 2016, and cause-specific maternal mortality to December 31, 2014. The primary exposure was a live birth of an infant with NAS, and the main outcome was all deaths among mothers following their date of delivery. We modelled the association between NAS and all-cause maternal mortality using Cox regression, and the cumulative incidence of cause-specific mortality within a competing risks framework. All-cause mortality rates, 10-year cumulative incidence risk of death, and crude and age-adjusted hazard ratios were calculated. Estimated crude 10-year mortality based on Kaplan-Meier curves in mothers of infants with NAS was 5.1% (95% CI 4.7%-5.6%) in England and 4.6% (95% CI 3.8%-5.5%) in Ontario versus 0.4% (95% CI 0.41%-0.42%) in England and 0.4% (95% CI 0.38%-0.41%) in Ontario for controls (p < 0.001 for all comparisons). Survival curves showed no clear inflection point or period of heightened risk. The crude hazard ratio for all-cause mortality was 12.1 (95% 11.1-13.2; p < 0.001) in England and 11.4 (9.7-13.4; p < 0.001) in Ontario; age adjustment did not reduce the hazard ratios. The cumulative incidence of death was higher among NAS mothers than controls for almost all causes of death. The majority of deaths were by avoidable causes, defined as those that are preventable, amenable to care, or both. Limitations included lack of direct measures of maternal opioid use, other substance misuse, and treatments or supports received.
CONCLUSIONS: In this study, we found that approximately 1 in 20 mothers of infants with NAS died within 10 years of delivery in both England and Canada-a mortality risk 11-12 times higher than for control mothers. Risk of death was not limited to the early postpartum period targeted by most public health programs. Policy responses to the current opioid epidemic require effective strategies for long-term support to improve the health and welfare of opioid-using mothers and their children.

Introduction

Opioid use is responsible for an important increase in premature mortality in young and middle-aged adults in the US [1] and Canada [2], 2 of the countries with the highest per capita prescription opioid consumption in Western industrialized nations [3]. Other countries such as England have seen similar rates of increase in prescription opioid use but not concomitant increases in mortality rates, likely related in part to better access to addiction treatment and more oversight of prescription opioids [4]. Across all of these jurisdictions, there is increasing opioid use by pregnant women, and while little is known about associated maternal mortality, a recent study using data from 22 US states and the District of Columbia reports a higher than 3-fold increase from 2007 to 2017 in opioid-related deaths in women during or within the first year after pregnancy [5].

Population-based surveillance of opioid use during pregnancy is difficult given the lack of prescription medication data in many jurisdictions and the challenges in measuring illicit use. Neonatal abstinence syndrome (NAS) is coded in the infant birth hospitalization record and offers a widely used but imperfect proxy measure of maternal opioid use during pregnancy. NAS manifests typically within hours to 1 to 2 days of delivery with autonomic, gastrointestinal, and neurologic symptoms of drug withdrawal in the infant that often require prolonged postnatal care [6]. Not all infants exposed to opioids in utero will experience NAS [7,8]. In trial and observational settings, approximately half of women receiving methadone or buprenorphine maintenance therapy in pregnancy gave birth to an infant with signs of NAS [7], and estimates of up to 91% have been reported in other groups of women with chronic opioid use [6]. The incidence of NAS rose dramatically between the early 2000s and 2014 in the US (from 28 to 144 per 10,000 births) [9] and Canada (from 18 to 54 per 10,000 live births) [10], but remained relatively stable in Australia and England [11,12].

Mothers of infants with NAS represent a heterogeneous group including those using prescription opioids and opioid agonists for analgesia, those on medically supervised maintenance treatment for dependence, and those using illicit opioids [12]. Overall social disadvantage and opioid use are inextricably linked, and many women who misuse these drugs are at greater risk of adversity, including deprivation, violence and abuse, and use of other substances [13]. All of these factors impact negatively on maternal health and may diminish parenting capacity [14]. Improving maternal health and preventing premature maternal mortality in this population is therefore critical for both mothers and their children. Pregnancy and delivery care are opportunities to improve access to addiction and supportive services. While public health programs such as nurse home visits tend to focus on pregnancy and the early postpartum period [15], it is largely unknown whether this is the only period of risk for poor outcomes for mothers who use opioids in pregnancy.

Evidence on mortality for mothers who use opioids in pregnancy is limited but consistently shows increased rates around the time of delivery (Table 1) [5,16-18]. Very high rates of longer-term maternal mortality have been reported in 2 older studies (1 Australian and 1 Finnish) [19,20], but these studies may not reflect the current opioid epidemic in North America or the UK [4]. Recent estimates of mortality from a meta-analysis of people with substance misuse disorder and homeless and prison populations reported a standardized mortality ratio for women of 11.9 (95% CI 10.4–13.3), which was higher than the equivalent figure for men (7.9; 95% CI 7.0–8.7) [21]. There is a dearth of information about long-term health outcomes for women—particularly mothers—with opioid use, which is an important knowledge gap given rising rates of prescription and illicit opioid use. Pregnancy can be seen as a window of opportunity for identifying and managing substance misuse and the implications for parenting capacity.

Table 1

Mortality rates from 6 population-based studies of women with substance misuse during pregnancy.

Study (country)	Study design (study years)	Study population	Duration of follow-up	Number of participants	Number of deaths	Mortality rate	Mortality rate ratio
Long-term mortality
Kahila et al. 2010 [20] (Finland)	Registry-based retrospective case–control (1992–2001)	Women living in Helsinki metropolitan area who gave birth between 1992 and 2001 and were referred to a specialist alcohol/substance misuse antenatal clinic	Mean of 9.4 (cases) and 10.1 (controls) years	2,316 (524 cases)	46 (42 cases)	8.52 (cases) and 0.22 (controls) per 1,000 person-years	OR: 38 (95% CI 14–108)
Hser et al. 2012 [22] (US)	Prospective cohort study (2000–2010)	Pregnant or parenting women assessed and admitted to 40 drug treatment programs in California between 2000 and 2002	8–10 years	4,447	194	4.47 per 1,000 person-years	SMR: 8.4 (95% CI 7.2–9.6)
Short-term mortality
Wolfe et al. 2005 [18] (US)	Linked discharge, birth, and death cohort (1991–1998)	Mother and newborn pairs between 1991 and 1998 in California with drug and/or alcohol use indication as discharge diagnostic codes during pregnancy	Death ≤ 72 hours after delivery	4,536,701 (54,290 cases)	1,944 (62 cases)	Unknown/not stated	RR: 2.7 (95% CI 2.1–3.5)
Whiteman et al. 2014 [17] (US)	Cross-sectional analysis (1998–2009)	Women with a pregnancy-related hospital discharge between 1998 and 2009 with indication of opioid use during pregnancy on the discharge record	Death during the delivery hospital stay	55,781,965 (138,224 cases)	Unknown/not stated	0.8 (cases) and 0.1 (controls) per 1,000 pregnancy-related discharges	OR: 5.9 (95% CI 3.7–9.3)
Maeda et al. 2014 [16] (US)	Cross-sectional analysis (1998–2011)	Women with a delivery admission between 2007 and 2011 with indication of opioid abuse or dependence during pregnancy on the discharge record	Death during the delivery hospital stay	20,517,479 (60,994 cases)	1,331 (20 cases)	0.03 (cases) and 0.006 (controls) per 100 delivery hospitalizations	Adjusted OR: 4.6 (95% CI 1.8–12.1)
Gemmill et al. 2019 [5] (US)	Retrospective cohort study (2007–2016)	Women aged 15–49 years with a pregnancy-associated death between 2007 and 2016 across 22 US states and the District of Columbia	Death while pregnant or within 1 year of end of pregnancy	Unknown/not stated	Unknown/not stated	Mortality per 100,000 live births: 31.7 in 2007 to 42.3 in 2016 (all-cause); 1.3 in 2007 to 4.2 in 2016 (opioid-related)	Unknown/not stated

Non-population-based studies [19,23–25] or studies that examined opioid use but not during pregnancy [26–29] were excluded.

OR, odds ratio; RR, relative risk; SMR, standardized mortality ratio.

In this study, we capitalize on linked population-based maternal–infant healthcare records and mortality files in 2 jurisdictions (England and Ontario, Canada). Both have similar healthcare systems, including universal access to healthcare and similar postnatal public health programs that focus predominantly on the year after birth. We hypothesized that NAS mothers would have significantly higher rates of mortality than control mothers. We aimed to quantify this excess mortality, investigating all-cause mortality as the primary outcome and cause-specific mortality as secondary outcomes. We present all-cause mortality in relation to some key maternal characteristics at the time of birth, to identify clinically useful sub-groups with poor prognostic outcomes and to guide opioid misuse policy and research.

Methods

An analytic plan was written and approved before starting statistical analyses (S1 Text). All analyses were undertaken as planned and reported, with the exception of the removal of comorbidity adjustment in the Cox proportional hazards model (see section on statistical analysis below) as a result of feedback from peer review. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (S2 Text).

Cohort

We derived whole-region population-based cohorts of mothers aged 12 to 49 years and their live born infants delivered between April 1, 2002, and December 31, 2012, in England and Ontario, Canada. This period coincides with a steep increase in opioid use, particularly in Canada. The cohorts used longitudinal hospital discharge records for mothers (back to April 1, 1997, for some covariates) linked to hospitalization records for the infant. Mortality registration records were linked to hospital records for the mother and infant. In England, de-identified data on inpatient admissions for all National Health Service (NHS) hospitals linked to Office for National Statistics mortality data were obtained from NHS Digital, linked between mother and infant using previously reported methods [30], and analyzed within the UCL Data Safe Haven, England [31]. In Ontario, hospital discharge records were obtained from the Canadian Institute for Health Information Discharge Abstract Database and the Ontario Mental Health Reporting System. Cause of death information was taken from the Office of the Registrar General’s vital statistics database (data available only until 2014), and demographic information captured from the 2006 Canadian Census and the Registered Persons Database (which includes date of death). Canadian datasets were linked using unique coded identifiers common across all aforementioned datasets and analyzed at ICES in Toronto, Ontario, Canada. We restricted the cohort to singleton births, and if a woman had more than 1 live birth delivery during the study period, 1 delivery was chosen at random as the focus of the study, i.e., a delivery date was selected at random and used as the entry point for the mother (referred to as the index delivery), and all other deliveries were ignored.

In both jurisdictions, infants with a diagnosis of NAS were identified using International Classification of Diseases and Related Health Problems—10th Revision (ICD-10) codes P96.1 (neonatal withdrawal symptoms from maternal use of drugs of addiction) or P04.4 (newborn [suspected to be] affected by maternal use of drugs of addiction) recorded during the delivery admission or subsequent readmission within 14 days of birth [11]. The ICD-9 equivalent to P96.1 has been shown to have high sensitivity (88.1%; 95% CI 83.3%–91.7%) and specificity (97.0%; 95% CI 93.8%–98.5%) and a positive predictive value of 91.2% (95% CI 86.8%–94.2%) for measuring NAS [32] and is used by the US Agency for Healthcare Research and Quality [33]. We included P04.4 as it is often used for opioid withdrawal in both jurisdictions. In a sample of all Ontario women giving birth in hospital from 2014 to 2017 (n = 464,400), during which time all prescription opioids were registered, over 56% of women whose infant had a diagnosis of P04.4 had either a prescription for opioids or opioid agonists, or a healthcare encounter related to opioid use during pregnancy (personal communication, A. Camden, University of Toronto, September 4, 2019). In our study, NAS was ascertained by P96.1 in 83% of cases in England and 65.7% in Ontario. The English maternal cohort was identified using linked data for singleton babies and mothers, corresponding to 96% of all live births in England within the study period. Mothers and babies were matched deterministically using data on hospital, general practitioner practice, maternal age, birthweight, gestation, birth order, and sex, or probabilistically using additional data including admission dates, ethnicity, and partial postal code [34]. Ontario mothers were identified using a unique number linking all newborn and maternal hospital records that is assigned at the delivery hospitalization.

Outcomes and covariates

The main outcome was all-cause mortality measured from April 1, 2002, to March 31, 2016, derived from linked death registrations. Cause-specific mortality was available only up until December 31, 2014, for Ontario. Cause-specific deaths were classified as avoidable (defined as preventable, amenable to care, or both), unavoidable (as specified by the UK’s Office for National Statistics) [35], or cancer (avoidable and unavoidable).

Longitudinal hospital records were used to derive baseline characteristics at the index delivery including maternal age (12–19 years, 20–34 years, and 35+ years); time since last birth (defined as no previous births [with lookback to April 1, 1997, only], <2 years, 2–5 years, and 6+ years); neighbourhood income quintile; urban or rural residence; pregnancy-related outcomes: cesarean delivery, gestational age at delivery (<34 weeks, 34–36 weeks, and 37+ weeks), gestational hypertension, pre-eclampsia/eclampsia, and gestational diabetes; neonatal mortality (death within 28 days of birth); and infant discharge from hospital to social services. We excluded observations with extreme values, including women aged <12 years or >49 years at the date of delivery. Missing values for neighbourhood income quintile (n = 3,850 [0.41%] for Ontario; n = 24,067 [0.57%] for England) were categorized into the lowest income quintile, and missing area of residence information (n = 151 [0.02%] for Ontario; n = 22,970 [0.54%] for England) was categorized into the urban area of residence. In Ontario, these missing data are suppressed for neighbourhoods with high rates of residential instability, which are predominantly low income and urban.

Baseline maternal morbidities were measured for a 5-year lookback period using all diagnostic codes in all hospital admissions within this period to assign the Charlson comorbidity index (Deyo version) (0, 1, or 2+ comorbid conditions) and to identify hospitalizations related to (i) any psychiatric condition, (ii) addiction-related conditions, and (iii) other mental health conditions. We chose the Deyo version of the Charlson comorbidity index as it has been most widely used in maternal mortality studies and, unlike other indices, has been validated on longer-term mortality, although not in pregnant women [36,37]. S1 and S2 Tables list diagnostic codes and provide definitions for neighbourhood income quintile and urban or rural residence.

Statistical analysis

Descriptive statistics

We compared mothers with an infant affected by NAS and controls within each jurisdiction. We compared baseline characteristics using Pearson’s chi-squared test (categorical variables) and ANOVA (continuous variables).

All-cause mortality

Our primary outcome was all-cause mortality of mothers after the birth of an infant with NAS, relative to control mothers. Survival analysis for time to all-cause mortality was modelled using multivariable Cox regression, with proportionality of hazards assessed by Schoenfeld residuals and log–log plots, and included all deaths to the end of the study period. Crude and adjusted hazard ratios were produced that were adjusted only for maternal age group at delivery to describe the extent of the mortality gap between NAS mothers and controls. We explicitly did not attempt to adjust models for other covariates as we do not attempt to infer causality. This is primarily because our data reflect maternal opioid use at a single time point (delivery). Thus we are unable to examine the direction of effect for key factors such as mental illness or socioeconomic status that may either confound the association between opioid use and death or lie on the causal pathway. We initially adjusted the model for the Charlson comorbidity index but as a result of peer review do not report these estimates because of our aim to not make causal inferences. Crude and adjusted survival curves were plotted to estimate the absolute risk of mortality at 5 and 10 years after birth. We derived 95% confidence intervals for mortality at 5 and 10 years after delivery through log–log transformation of the survival function and computed p-values using the z test [38]. Mothers with missing values for maternal age at delivery were excluded from the adjusted models. Individuals surviving beyond the end of the study period were censored at March 31, 2016.

Age-standardized all-cause mortality rates were estimated for England and Ontario using the direct method of standardization and the Canadian 2006 Census as the standard population. We present age-standardized all-cause mortality rates stratified by maternal age at delivery (excluding those with missing age information), neighbourhood income quintile (Q1 [most deprived] versus Q2–Q5 due to low numbers of deaths in the most affluent quintiles), urban/rural residence, previous addiction-related or other mental-health-related hospitalization, Charlson comorbidity index (0 versus 1+ due to low numbers of deaths in the 2+ category for NAS mothers), and infant discharge to social service out-of-home care.

Cause-specific mortality

Ten-year cumulative incidence of death was calculated for each cause of death category with consideration of other causes of death as competing risk events using Gray’s test for the homogeneity of 2 or more cumulative incidence functions. Individuals surviving beyond the end of the study period were censored at December 31, 2014, for this particular analysis.

Statistical analyses were performed using Stata version 15 for England data and SAS version 9.4 statistical software for Ontario data in a Unix environment. p-Values for age-standardized all-cause mortality and cause-specific mortality were derived using the z test.

ICES is a prescribed entity under section 45 of Ontario’s Personal Health Information Protection Act. Section 45 authorizes ICES to collect personal health information, without consent, for the purpose of analysis or compiling statistical information with respect to the management of, evaluation or monitoring of, the allocation of resources to, or planning for all or part of the health system. Projects conducted under section 45, by definition, do not require review by a research ethics board. This project was conducted under section 45, and approved by the ICES Privacy and Legal Office. Research ethics approval was granted by the Hospital for Sick Children Research Ethics Board for Ontario analyses. The English analyses were exempt from UK NHS Research Ethics Committee approval because it involved the analysis of de-identified administrative data.

Results

Cohort characteristics of mothers in England and Ontario

After applying our inclusion criteria, there were 13,577 and 4,966 mothers of infants with NAS and 4,205,675 and 929,985 control mothers in England and Ontario, respectively. Baseline characteristics are described in Table 2. In both jurisdictions, the majority of mothers had no previous recorded hospital birth (lookback to April 1, 1997) and lived in urban areas. Compared to controls, a larger proportion of mothers of infants with NAS lived in neighbourhoods in the lowest income quintile (37.3% versus 23.5% in England and 44.9% versus 22.9% in Ontario; p < 0.001 for both jurisdictions); mothers of infants with NAS were on average younger than controls, and, in Ontario, a higher proportion were teenage mothers (9.8% versus 3.7%; p < 0.001). Mothers of infants with NAS were more likely to have higher comorbidity scores (1 or 2+ on the Charlson comorbidity index) than controls (21.2% versus 6.9% in England and 7.6% versus 1.7% in Ontario; p < 0.001 for both jurisdictions), and a greater proportion also had a previous psychiatric hospitalization (13.5% versus 0.6% in England and 15.3% versus 1.1% in Ontario; p < 0.001 for both jurisdictions). Infant discharge to care by social services was much more common among infants with NAS than among control infants in both England (9.7% versus 0.1%; p < 0.001) and Ontario (15.2% versus 0.1%; p < 0.001). We found no differences in neonatal mortality between infants with NAS and controls in Ontario, but a marginally higher rate among infants with NAS in England (0.3% versus 0.2% p = 0.01).

Table 2

Characteristics of mothers and infants at baseline, April 1, 2002 to December 31, 2012.

Characteristic	England		Ontario
	NAS cases(n = 13,577)	Controls(n = 4,205,675)	NAS cases(n = 4,966)	Controls(n = 929,985)
Mean (SD) maternal age at delivery, years*	28.5 (5.6)	29.8 (6.1)	27.0 (6.1)	30.2 (5.6)
Maternal age at delivery, categorized*
≤19 years	775 (5.7)	258,200 (6.1)	488 (9.8)	34,414 (3.7)
20–34 years	10,712 (78.9)	3,041,112 (72.3)	3,815 (76.8)	683,910 (73.5)
35+ years	1,869 (13.8)	898,110 (21.4)	663 (13.4)	211,661 (22.8)
Missing	221 (1.6)	8,253 (0.2)	0 (0.0)	0 (0.0)
Time since last birth*
No previous birth since 1997	6,944 (51.1)	2,692,743 (64.0)	2,322 (46.8)	568,109 (61.1)
<2 years	1,504 (11.1)	315,454 (7.5)	697 (14.0)	78,540 (8.4)
2 to 5 years	2,761 (20.3)	840,237 (20.0)	1,366 (27.5)	238,227 (25.6)
6+ years	2,368 (17.4)	357,240 (8.5)	581 (11.7)	45,109 (4.9)
Neighbourhood income quintile*†
Q1 (lowest)	5,059 (37.3)	988,493 (23.5)	2,231 (44.9)	212,869 (22.9)
Q2	3,487 (25.7)	898,448 (21.4)	1,043 (21.0)	187,607 (20.2)
Q3	2,304 (17.0)	792,989 (18.9)	668 (13.5)	188,884 (20.3)
Q4	1,622 (11.9)	747,224 (17.8)	599 (12.1)	189,978 (20.4)
Q5 (highest)	1,105 (8.1)	778,521 (18.5)	425 (8.6)	150,647 (16.2)
Q2–Q5	8,518 (62.7)	3,217,182 (76.5)	2,735 (55.1)	717,116 (77.1)
Area of residence*‡
Urban	12,217 (90.0)	3,573,814 (85.0)	4,164 (83.9)	841,133 (90.4)
Rural	1,360 (10.0)	631,861 (15.0)	802 (16.1)	88,852 (9.6)
Charlson comorbidity index*
0	10,695 (78.8)	3,915,844 (93.1)	4,586 (92.3)	914,048 (98.3)
1	2,606 (19.2)	272,330 (6.5)	304 (6.1)	12,325 (1.3)
2+	276 (2.0)	17,501 (0.4)	76 (1.5)	3,612 (0.4)
Any psychiatric condition*	1,832 (13.5)	26,108 (0.6)	760 (15.3)	10,440 (1.1)
Addiction-related*	1,396 (10.3)	8,922 (0.2)	318 (6.4)	1,411 (0.2)
Other mental health*	566 (4.2)	18,055 (0.4)	572 (11.5)	9,490 (1.0)
Cesarean delivery	3,121 (23.0)*	1,044,111 (24.8)	1,374 (27.7)NS	268,553 (28.9)
Pre-eclampsia/eclampsia	290 (2.1)*	127,753 (3.0)	60 (1.2)NS	11,390 (1.2)
Gestational diabetes*	48 (0.4)	36,276 (0.9)	181 (3.6)	48,893 (5.3)
Gestational hypertension	220 (1.6)*	123,908 (3.0)	193 (3.9)NS	40,505 (4.4)
Gestational age at delivery*
<34 weeks	671 (4.9)	92,155 (2.2)	318 (6.4)	18,825 (2.0)
34 to 36 weeks	1,224 (9.0)	132,599 (3.2)	721 (14.5)	50,529 (5.4)
37+ weeks	7,538 (55.5)	2,962,970 (70.5)	3,875 (78.0)	859,353 (92.4)
Missing§	4,144 (30.5)	1,018,010 (24.2)	52 (1.1)	1,278 (0.1)
Infant discharged to social services*	1,316 (9.7)	4,446 (0.1)	753 (15.2)	1,265 (0.1)
Neonatal mortality	37 (0.3)	7,497 (0.2)	18 (0.4)NS	2,724 (0.3)

Data are number (percent) unless otherwise indicated. Neonatal mortality was statistically significantly different between cases and controls for England, at p = 0.01. p-Values for continuous variables were derived from ANOVA, while p-values for categorical variables were derived from Pearson’s chi-squared test.

*There was a statistically significant difference between cases and controls (p < 0.001).

†Missing values for neighbourhood income quintile for England (0.82% and 0.57% of cases and controls, respectively) and Ontario (2.11% and 0.40% of cases and controls, respectively) were included in the lowest quintile (Q1).

‡Missing values for area of residence for England (1.1% of cases and 0.54% of controls) and Ontario (0.08% and 0.02% of cases and controls, respectively) were included in the urban residence category.

NSThere was no statistically significant difference between cases and controls.

§Gestational age is not a mandatory field reported to NHS Digital in England.

NAS, neonatal abstinence syndrome.

Risk of death among mothers of infants born with NAS

From 2002 to 2016, there were 112,890 total person-years of follow-up for mothers of infants with NAS in England (34.9 million for controls) and 35,740 total person-years of follow-up for mothers of infants with NAS in Ontario (7.9 million for controls). The mean duration of follow-up was 8.4 years (England) and 7.2 years (Ontario) for NAS mothers and 8.2 years (England) and 8.6 years (Ontario) for controls. Crude mortality rate for mothers of infants with NAS was 5.01 (95% CI 4.62–5.44) per 1,000 person-years in England and 4.28 (95% CI 3.63–5.02) per 1,000 person-years in Ontario. In both jurisdictions, the crude cumulative mortality incidence (superimposed on the survival curves in Fig 1) was significantly different between mothers of infants with NAS and controls. In England, 5- and 10-year mortality (95% CI) was 1.81% (1.59%–2.05%) and 5.13% (4.69–5.62%), respectively, for mothers of infants with NAS and 0.15% (0.15%–0.16%) and 0.42% (0.41–0.43%), respectively, for controls (p < 0.001 for 5- and 10-year mortality for NAS mothers versus controls); in Ontario, it was 1.85% (1.51%–2.28%) and 4.58% (3.81%–5.49%), respectively, for mothers of infants with NAS and 0.15% (0.15%–0.16%) and 0.40% (0.38%–0.41%), respectively, for controls (p < 0.001 for 5- and 10-year mortality for NAS mothers versus controls). The decline in survival of mothers of infants with NAS over time was steady in both jurisdictions, with no clear inflection point or distinct period of risk.

Fig 1

Survival curves for all-cause mortality—Crude (Kaplan–Meier curve) and adjusted (derived from Cox model) for maternal age at delivery, 2002 to 2016.

Crude curves for England (A) and Ontario (C); adjusted curves for England (B) and Ontario (D). Red, neonatal abstinence syndrome (NAS) mothers; blue, controls. The adjusted survival curve for the England controls is a 10% sample of the full control population. p < 0.001 for a difference in all-cause mortality between NAS mothers and controls for the overall study period (crude and adjusted), 5-year mortality (crude), and 10-year mortality (crude) for England and Ontario. Mortality rates superimposed on figures are accompanied by 95% confidence limits presented in parentheses.

Table 3 outlines age-standardized mortality rates overall and stratified by clinical and demographic characteristics for the whole study period. In both jurisdictions, rates of death were higher among mothers of infants with NAS compared to controls across most risk groups. Differences in rates were not significant (95% confidence intervals overlapped) for NAS mothers and controls for those with a history of addiction-related admission (England and Ontario) or who had an infant discharged to social services (Ontario only) or who were 19 years old or under at delivery (Ontario only).

Table 3

Age-standardized all-cause mortality rates per 1,000 women, April 1, 2002, to March 31, 2016.

Characteristic	England					Ontario
	NAS mothers		Controls		p-Value	NAS mothers		Controls		p-Value
	Number	Rate(95% CI)	Number	Rate(95% CI)		Number	Rate(95% CI)	Number	Rate(95% CI)
Overall	566	41.7(37.9–45.4)	14,356	3.5(3.4–3.6)	<0.001	153	33.3(27.6–39.9)	3,194	3.6(3.5–3.8)	<0.001
Maternal age at delivery, years
≤19	11	15.5(5.1–25.8)	682	2.7(2.5–2.9)	<0.001	7	11.9(3.5–29.1)	167	5.1(4.3–6.1)	0.3
20–34	439	40.5(36.7–44.2)	8,796	2.8(2.8–2.9)	<0.001	119	31.9(26.3–38.3)	2,067	3.0(2.9–3.2)	<0.001
35+	116	65.4(50.5–80.2)	4,878	6.3(5.8–6.2)	<0.001	27	39.5(22.7–64.1)	960	5.2(4.8–5.6)	<0.001
Neighbourhood income quintile
Q1 (lowest)	228	47.6(40.8–54.3)	3,801	4.5(4.3–4.6)	<0.001	71	37.7(28.2–49.4)	932	4.6(4.3–5.0)	<0.001
Q2–Q5	328	38.3(33.8–42.9)	10,555	3.3(3.2–3.4)	<0.001	82	30.7(23.7–39.1)	2,262	3.4(3.2–3.5)	<0.001
Area of residence
Urban	526	43.1(39.0–47.2)	12,176	3.5(3.4–3.6)	<0.001	130	33.6(27.5–40.8)	2,728	3.4(3.3–3.6)	<0.001
Rural	40	28.5(18.7–38.3)	2,180	3.4(3.2–3.5)	<0.001	23	30.0(16.7–49.8)	466	5.4 (4.9–6.0)	<0.001
History of psychiatric hospitalizations
Addiction-related	109	78.6(62.5–94.7)	327	56.8(50.2–63.4)	0.01	16	58.0(31.4–98.1)	46	39.0(27.5–53.6)	0.3
Other mental health	41	64.7(43.9–85.6)	358	20.1(17.9–22.2)	<0.001	29	55.3(35.5–82.2)	139	16.5(13.5–20.1)	<0.001
Infant discharge to social services	77	58.3(44.2–72.3)	83	22.5(17.3–27.6)	<0.001	22	27.8(16.0–45.0)	30	32.9(20.9–49.1)	0.6
Charlson comorbidity index
0	400	37.1(33.0–41.1)	11,933	3.1(3.1–3.2)	<0.001	123	27.8(22.4–34.1)	2,881	3.4(3.2–3.5)	<0.001
1+	166	56.3(46.8–65.8)	2,423	8.7(8.3–9.1)	<0.001	30	80.8(52.6–118.8)	313	19.3(16.9–22.0)	<0.001

Estimates were age-standardized to the 2006 Canadian population using the age groups 12–18, 19, 20–29, 30–34, 35–37, and 38–49 years for maternal age at delivery, and 12–21, 22–29, and 30+ years for the remaining characteristics.

NAS, neonatal abstinence syndrome.

The crude hazard ratio for all-cause mortality among mothers of infants with NAS was 12.1 (95% CI 11.1–13.2; p < 0.001) in England and 11.4 (95% CI 9.7–13.4; p < 0.001) in Ontario (Table 4). After adjustment for age, the hazard ratio increased for England and was unchanged for Ontario (England adjusted HR 13.0; 95% CI 11.9–14.1; p < 0.001; and Ontario adjusted HR 11.4; 95% CI 9.7–13.4; p < 0.001).

Table 4

All-cause mortality risks for mothers with infants with neonatal abstinence syndrome, crude and adjusted for maternal age at delivery, 2002 to 2016.

Model	England		Ontario
	Hazard ratio(95% CI)	p-Value	Hazard ratio(95% CI)	p-Value
Crude	12.1 (11.1–13.2)	<0.001	11.4 (9.7–13.4)	<0.001
Adjusted for maternal age at delivery	13.0 (11.9–14.1)	<0.001	11.4 (9.7–13.4)	<0.001

Table 5 presents results for cause-specific mortality between 2002 and 2014 (cause-specific data for 2015–2016 were not available in Ontario). Avoidable deaths were the most common cause of death among mothers of infants with NAS in both jurisdictions (accounting for >85% in England and 75% in Ontario), with a 10-year cumulative incidence risk of 42.9 deaths per 1,000 population (95% CI 38.4–47.9) among English mothers and 30.8 deaths per 1,000 population (95% CI 24.1–38.8) among Ontario mothers. Intentional and unintentional injuries (e.g., transport injuries, unintentional falls) made up the majority of avoidable mortality in the mothers of infants with NAS in both jurisdictions.

Table 5

Cause-specific mortality per 1,000 population among mothers (10-year cumulative incidence risk), April 1, 2002, to December 31, 2014.

Type of deaths	England					Ontario
	NAS mothers		Controls		p-Value	NAS mothers		Controls		p-Value
	Number	Cumulative incidence(95% CI)	Number	Cumulative incidence(95% CI)		Number	Cumulative incidence(95% CI)	Number	Cumulative incidence(95% CI)
Avoidable (excluding cancer)	393	42.9(38.4–47.9)	5,758	2.1(2.0–2.1)	<0.001	93	30.8(24.1–38.8)	956	1.4(1.3–1.5)	<0.001
Unintentional injuries	177	19.6(16.6–23.1)	1,132	0.4(0.4–0.4)	<0.001	46	13.9(9.7–19.3)	285	0.4(0.4–0.5)	<0.001
Intentional injuries	63	6.5(4.9–8.7)	1,298	0.5(0.4–0.5)	<0.001	20	6.4(3.8–10.4)	312	0.5(0.4–0.5)	<0.001
Drug use disorders	90	9.7(7.6–12.3)	701	0.3(0.3–0.3)	<0.001	15	5.7(3.1–9.9)	45	0.1(0.1–0.1)	<0.001
All other avoidable deaths	63	7.8(5.9–10.2)	2,627	0.9(0.9–1.0)	<0.001	12	4.8(2.3–9.1)	314	0.5(0.4–0.5)	<0.001
Unavoidable (excluding cancer)	42	4.2(3.0–5.9)	1,649	0.6(0.5–0.6)	<0.001	19	8.5(4.6–14.7)	417	0.6(0.6–0.7)	<0.001
Cancer (avoidable and unavoidable)	16	1.6(1.5–1.7)	4,233	1.6(0.9–2.8)	1.00	10	3.3(1.6–6.3)	1,113	1.8(1.7–1.9)	0.2
Missing causea	≤10	—	≤10	—	—	≤ 5	—	139	0.2(0.2–0.2)	—

aPrivacy legislation requires suppression of cell sizes <11 in England and <6 in Ontario.

NAS, neonatal abstinence syndrome.

Discussion

In this large population-based study across 2 countries, 1 in 20 mothers of infants with NAS died within 10 years of delivery—a mortality risk that was 11–12 times higher than for control mothers. Findings were consistent across both jurisdictions. For virtually all causes of death, mortality rates were substantially higher for mothers of infants with NAS than for controls, with the majority of deaths attributable to avoidable causes such as intentional and unintentional injuries. We also identified universally high mortality rates among mothers who had a history of hospitalization for addiction, irrespective of whether or not their infant had NAS. We found no evidence of a high risk period in the 1–2 years after birth (corresponding to the period typically targeted by public health nursing or other support for high-risk families) for maternal deaths in the NAS group.

Other population-based studies report increased perinatal maternal mortality in mothers using opioids [5,16,17] and longer-term risk in mothers with alcohol or drug misuse during pregnancy [20,22] (Table 1). None of these studies address long-term all-cause mortality for mothers of NAS-affected infants as we have done. High mortality rates have been reported for non-pregnant populations of female opioid users. Notably, 3 large-scale studies present crude mortality rates per 1,000 person-years of 6.5 (95% CI 6.1–6.9) in New South Wales [39], between 7.5 and 13.9 in opioid-using women aged 15–44 years in California [29], and 12.2 (95% CI 10.3–14.4) and 19.7 (95% CI 15–25.8), respectively, for female users of heroin and other opioids in Denmark [40]. Our crude mortality rates per 1,000 person-years for mothers with infants with NAS of 5.01 (95% CI 4.62–5.44) in England and 4.28 (95% CI 3.63–5.02) in Ontario are comparatively lower, which is likely driven in part by a relatively younger age distribution in our study population. For example, in the New South Wales study, Degenhardt et al. included all ages and reported increasing mortality rates with older age. The higher risk of all-cause mortality we report is not surprising and has been shown in other marginalized groups or those with mental health problems [21,41]. In particular, unintentional injury deaths (which include those related to victimization) predominate and may result from social vulnerability or misclassification of injuries that are intentional [42-46].

In our study, the high rate of premature mortality among the mothers of infants with NAS was mirrored by high rates for the mothers of control infants discharged to social services and for mothers with a history of hospitalization for addiction. Other studies have demonstrated maternal well-being declining in association with the loss of a child to foster care [42,47], whereas retaining care of the child may help facilitate treatment [48]. An estimated 7%–20% of NAS-affected infants do not return home with their mother at the time of postnatal discharge from hospital [49-52], which is similar to the percentage in our study (10%–15%). However, these figures may not reflect the much higher cumulative risk of foster care placements occurring later in childhood [53]. Our findings also mirror the growing body of literature describing the constellation of psychosocial risk factors linking mental illness, addiction, and social adversity [43,54] and suggest the need for multifaceted support for these mothers irrespective of whether their children are living with them.

The longitudinal and population-based nature of this study, its size, and comparison of similar universal healthcare systems are strengths. Limitations include potential linkage error, misclassification of mothers using opioids whose babies did not develop NAS, and lack of direct measures of maternal opioid use or treatment, or other substance misuse, which may underestimate the burden of mortality in mothers with opioid use within our study or make findings less generalizable to opioid-using mothers whose infants do not have NAS or who live in other jurisdictions with other approaches to treatment and available supportive services. In the extract of English hospital data, 96% of live births were matched to maternal records, but linkage was lower (88%) for infants with NAS. Comparison of the characteristics of these English infants with NAS by linkage status indicated an association between non-linkage and both longer hospital stays and greater risk of placement in out-of-home care [55], with the implication that our results may underestimate the risk of death among women with opioid use. Our results reflect the subset of mothers with prenatal opioid exposure who gave birth to an infant with NAS, and as such they likely underestimate the true risk of maternal mortality associated with prenatal opioid use. However, we also describe mortality in mothers without an infant with NAS, but who received care for mental illness and addictions, thus broadening the scope for generalizability. Our study cohort may also include the rare cases of NAS related to withdrawal from other substances [56] or from postnatal opioid use (“iatrogenic NAS”) [57], but data on these other exposures were not available.

Our study has implications for research, practice, and policy to improve maternal and, arguably, child outcomes related to prenatal opioid use. Enhanced treatment programs for opioid dependence that integrate maintenance therapy, psychotherapy, reproductive health, and obstetric care have been found to be effective in reducing substance misuse, unplanned pregnancies, and obstetric complications during the perinatal period [58]. Some evaluations of programs supporting mothers with opioid use and their children suggest that multifaceted services addressing health, addiction, housing, and parenting needs can improve parenting capacity and attachment and reduce child apprehension [56,59-64]. However, rigorous evidence on interventions promoting long-term support is limited and should be a research and policy priority. New funding for child welfare agencies in the US to provide services related to mental health, addictions, and parenting in response to the growing numbers of mothers using opioids is an opportunity for evaluation of different models of support [65]. Most current home-visiting programs target only families with children and only for a short period of time. Our findings suggest that interventions need to extend past the early postpartum period and include mothers whose children may not return home. Finally, the findings from our study also indicate that studies and surveillance focused only on deaths directly attributable to opioid overdose will miss the full extent of the problem, given the importance of deaths due to unintentional and intentional injuries, not all of which involve opioids.

In conclusion, while much attention and research on NAS has focused on infant and child outcomes in isolation, our study is the first population-based analysis to our knowledge of long-term maternal mortality following the birth of an infant with NAS. The findings provide a stark reminder of the vulnerability and sustained poor outcomes of these mothers. Policy responses to the current opioid epidemic will require effective strategies for risk mitigation and ongoing support for families affected by opioid use. Large-scale linkage of health and social care administrative data would facilitate ongoing research, program evaluation, and surveillance.

Study diagnostic codes and description of sociodemographic characteristics.

(DOCX)

Click here for additional data file.

Description of baseline sociodemographic characteristics for neighbourhood income quintile and urban and rural area of residence.

(DOCX)

Click here for additional data file.

Dataset creation plan and analytic plan: Long-term mortality in mothers of infants with neonatal abstinence syndrome: A population-based parallel-cohort study in England and Ontario, Canada.

(DOCX)

Click here for additional data file.

STROBE Statement—Checklist of items that should be included in reports of cohort studies.

(DOCX)

Click here for additional data file.

22 Aug 2019

Dear Dr. Guttmann,

Thank you very much for submitting your manuscript "Long-term mortality in mothers of infants with neonatal abstinence syndrome: Findings from a population-based parallel-cohort study in England and Ontario, Canada" (PMEDICINE-D-19-02772) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

We consider that your paper may be suitable for inclusion in our upcoming Special Issue on Substance Use, Misuse and Dependence, and we may be able to include your paper in this, all being well. Please indicate in your rebuttal letter whether this would be of interest to you.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Sep 05 2019 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Louise Gaynor, MBBS PhD

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

Title

Please revise your title according to PLOS Medicine's style. ‘Findings from’ can be omitted.

Abstract

Abstract Background:

Please remove the hyphens between opioid and addiction / use.

Abstract Methods and Findings:

Please add summary demographic details for study participants early in the "methods and findings" subsection of your abstract.

Please clarify why death records only extended to 10 years after delivery when health data was assessed from 2002 and death records assessed up until 2016

Please ensure that all numbers presented in the abstract are present and identical to numbers presented in the main manuscript text.

Please clarify what is meant by ‘index delivery’

Please include 95% CIs and p values for 10-year mortality data.

Please clarify which of the data presented are observed or modelled

Please include the important dependent variables that are adjusted for in the analyses

Please expand upon what was considered an ‘avoidable cause’ of death

Abstract Conclusions:

Please add the main finding from the control group for comparison. You state in the first line of your discussion that ‘a mortality risk that is 11-12 times higher than for control mothers’, for example.

Please interpret the study based on the results presented in the abstract, emphasizing what is new without overstating your conclusions.

Please begin the first sentence of the "conclusions" subsection of your abstract with "In this study, we found that ..." or similar.

Author Summary

At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

Introduction

Please explain the need for and potential importance of your study.

If there has been a systematic review of the evidence related to your study (or you have conducted one), please refer to and reference that review and indicate whether it supports the need for your study.

Please conclude the Introduction with a clear description of the study question or hypothesis, including what is measured as primary and secondary outcomes.

Methods

Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section. If available, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript. If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place. In either case, changes in the analysis-- including those made in response to peer review comments-- should be identified as such in the Methods section of the paper, with rationale.

Please indicate: (1) the specific hypotheses you intended to test, (2) the analytical methods by which you planned to test them, (3) the analyses you actually performed, and (4) when reported analyses differ from those that were planned, transparent explanations for differences that affect the reliability of the study's results.

Please clarify what is meant by ‘Mothers and infants and these different datasets were linked’

Please clarify what is meant by ‘randomly selected one delivery date’

Please clarify what is meant by ‘index delivery’

Please ensure that the study is reported according to the STROBE guideline. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist).”

Results

Where results are presented in the cohort characteristics, please quantify the differences between mothers of infants with NAS and controls with % and p values. (page 9)

On page 9, please clarify what you mean by "significant". If statistical significance is intended, please indicate that.

When a p value is given, please specify the statistical test used to determine it.

Please avoid general terms like ‘similar’ and ‘generally’

Please include 95% CIs and p values for 5-year and 10-year mortality data.

Please indicate which factors are adjusted for in the main text (page 12)

The Supporting Information file Table S3 is central to the understanding of the paper. Please incorporate it into the main paper (page 13).

Please clarify why data not available up to 2016 (page 14).

Please highlight in your results section the data for which you infer that ‘We also identified universally high mortality rates among mothers who had a history of hospitalization for mental health or addictions, irrespective of whether or not their infant had NAS.’ in your discussion.

Discussion

Please present and organize the Discussion as follows: a short, clear summary of the article's findings; what the study adds to existing research and where and why the results may differ from previous research; strengths and limitations of the study; implications and next steps for research, clinical practice, and/or public policy; one-paragraph conclusion.

Please clarify what is considered a ‘ early high risk period’ (p.16)

PLOS does not permit "data not shown” (page 17). Please remove this claim, or do one of the following:

a) If you are the owner of the data relevant to this claim, please provide the data in accordance with the PLOS data policy, and update your Data Availability Statement as needed.

b) If the data not shown refer to a study from another group that has not been published, please cite personal communication in your manuscript text (it should not be included in the reference section). Please provide the name of the individual, the affiliation, and date of communication. The individual must provide PLOS Medicine written permission to be named for this purpose.

c) For any other circumstance, please contact me ASAP.

References

In your reference list, please ensure that journal names are abbreviated consistently, and add spaces where needed (e.g. reference 32 "BrMedBull").

Tables

Table 2 - please indicate which differences between NAS cases and controls are statistically significant (or are not, if this is easier) for ease of interpretation. The table legend is lengthy and some text could be replaced by * and NS annotations within the table itself.

Figures

In Figure 1, please provide 95% CIs, and please indicate whether differences between 5-year mortality are statistically significant.

Comments from the reviewers:

Reviewer #1: This article is about the long term mortality of mothers likely to have taken opioids during pregnancy as indicated by neonatal abstinence syndrome of the newborn children. In general, the article is well written, uses a large and detailed data set, and uses methods suitable for analysis throughout. Other reviewers will be better placed to comment on the novelty of the findings and how this could affect any clinical practice, but there are nonetheless some comments that I think could be addressed by the authors:

(1) On page 4 of introduction (no line numbers) it is stated that "risks may differ given patterns of opioid use in the current global epidemic" but I don't think the patterns of use is tested within this study, and I don't believe that there is evidence within report that the effect of opioid use would differ depending on the population prevalence/patterns being assessed as opposed to the effects on an individual.

(2) The study population of mothers seems to be relevant only in terms of the children providing a proxy measure of the opioid status, and that as stated on page 10 there was "no difference in neonatal mortality between infants with NAS and controls". As such, it may be relevant to expand further on how comparable the results from this study are in terms of mortality rates to other studies that exist and have taken a more general population but with different issues of accurately recording opioid use as suggested in introduction (eg lack of prescription data, measuring illicit use).

(3) Further detail is in my opinion necessary to justify the non-consideration of other covariates in the statement of statistical analysis section that "..other covariates may be on the causal pathway". While this could be a justification for excluding terms as adjustments in the model, there is no evidence within the report of the proposed directions of causality (eg in graph), and I would consider that some of the other available covariates (eg deprivation status) could potentially be a confounding factor as opposed to being caused in the first instance by opioid use.

(4) Given the categorisation of age and charlson score in some of the descriptive tables, it would be useful to confirm if in fact that continuous measures are used for the models that have been fitted as this could provide a more appropriate model than if they have been categorised in a relatively arbitrary way.

(5) Individuals would technically not be "censored at death" as for all cause-mortality death is the outcome/event of interest and there would be no time after death to censor (as there would be with the end of follow up)

(6) To asses rates by factors not in the models stratified analysis has been done, but this is without first having tested interaction terms of the terms under consideration and opioid group which would provide more robust interpretation of whether there are any genuine sub-group differences.

(7) In figure 1 results from adjusted models are shown in Kaplan Meier plots, but it is unclear exactly what results this is showing. Potentially it is for the average age and average charlson score across the entire study, and if so this could be made clear within the figure caption.

(8) The grouping of Q2-Q5 for neighborhood income quarintile in table 3 seems somewhat arbitrary instead of using the quintiles directly, particularly as missing data has been recoded into the lowest level which could also have an effect on how this comparison would be interpreted. Although the assumptions for recoding the missing income-quintile or rural/urban in general seems reasonable, the robustness of the results to these assumptions and whether there is any systematic reasons for why there would be missing data if not just random would be useful to comment on.

(9) In the discussion section on page 16, it is noted that there is "potential for impact to improve maternal and child outcomes", and similarly on page 18 "potential implications for their children", but within the analysis done as part of this article it is only stated that there is no difference in mortality between NAS and control children in terms of mortality and no other child outcomes were considered or summaries shown. Based on the data from this study alone the statements on child outcomes are therefore not supported. However, there are also summaries in the discussion section relating to children not presented in the main results of the study ("..10-15% NAS affected infants do not return home with mother..", "greater risk of placement in out of home care") suggesting that there potentially is data available that could be shown to support the statement and give greater clarity on the outcomes for children.

Reviewer #2: Thank you for the opportunity to review this interesting manuscript.

1. I was a little confused about the study period and the look-back time, when reading through the first time. Was there linked hospital data back to 1 April 1997 for all? Was the look-back a consistent 5 years for all, or was it back to 1997 for all? All-cause mortality had data up to March 31 2016, so only those from 2002 to 2006 had a full 10 years of follow-up, is that correct? Was 10 years the maximum follow-up time for each individual? Please add further details in the methods. Please specify the average follow-up time per person (split by jurisdiction as well as NAS and controls) in the Results section.

2. The abstract background discusses the pregnancy and delivery settings as an opportunity to access support, and potentially assist new mothers with opioid addiction. However, I couldn't see mention of this in the Introduction. If this is an important theme of the paper, then I think it needs further exploration. What, if any, studies have shown an impact of targeting pregnant women or mothers with opioid addiction? Are women on methadone or other opioid-replacement therapy better off than those who are not?

3. Please state the number and proportion of all the missing data noted in the methods, e.g. income, area of residence and age. The authors state that missing values for neighbourhood income quintile were put in the lowest income quintile, and that missing area of residence was categorised into urban. Including the missing data in one category will bias your findings towards the null, as the missing presumably include a proportion of values in all categories. I would suggest either removing the missing, if the proportion is very small, or using principled missing data methods such as multiple imputation or full information maximum likelihood, if the missing is more substantial. And why is the missing data for gestational week of birth so high in the English cohort?

4. The Charlson Comorbidity index is a weighted score based on the relative risk of mortality. The index is not a continuous variable, nor a count of comorbidities (please revise the footnote on Table 2). Using it as a continuous variable (e.g. reporting mean and including it in the Cox regression as a continuous variable) suggests that the relative impact on mortality from 0 to 1 comorbidity score is the same as that from 4 to 5, for example. What is the justification for using it as a continuous variable? Has the functional form been tested and shown to be linear?

5. I agree that it is important to control for other underlying causes of death, not related to the possible opioid addiction, but there are some aspects of the Charlson score that are probably on the causal pathway, like HIV/AIDS and liver disease. I think a directed acyclic graph of the theoretical relationships between the variables and the mortality outcome would make the assumptions about the causal pathways explicit.

Minor comments:

Page 7, para 3, line 7: Suggest removing "young", as the extreme values are young and old.

Page 7, para 3, line 8: "Missing values for neighbourhood income quintile" should be followed by "were" not "was".

Page 12, para 1, line 3: I understand the intent of this sentence but on first read, it sounds like the cumulative mortality was similar for those with NAS and controls. Suggest rewording.

Table 2 - The percentage of primiparous women in the control groups looks very high. A quick look for other studies using Ontario or Toronto data, shows the proportion of primips is between 39 to 47%, not close to 60%. (E.g. Park et al, 2015, JOGC; Booth et al, 2017, CMAJ). I did not search for English studies.

Table 2 - I would be interested in whether the preterm births are planned or spontaneous. Is it possible to split this out?

Page 17, para 2, line 2: What is the expected impact of these limitations on the mortality outcomes?

Page 18, para 1, line 6: Are there differences between England and Ontario really a source of 'bias', if the data are never combined?

Reviewer #3: This is a very interesting and important study looking at the relationship of NAS (using it as a marker of OUD in pregnancy) in the infant to all-cause maternal mortality in the years following delivery. It is well-executed. The strengths and limitations are well-delineated. The abstract is clear and reflects the findings in the paper. The discussion and implications to public health were well-stated.

I appreciated the summary of the introductory literature in table 1

Suggestions for revision

1. Table 2 would be strengthened by the addition of p-values to each variable.

2. Supplementary table 3 is so important. I would move it to the body of the paper.

3. Also Tables 3 and 4 would be strengthened by aOR.

Any attachments provided with reviews can be seen via the following link:

[LINK]

17 Sep 2019

Submitted filename: 00_Response to Reviewers_PLOS Med NAS Paper_FINAL.docx

Click here for additional data file.

9 Oct 2019

Dear Dr. Guttmann,

Thank you very much for re-submitting your manuscript "Long-term mortality in mothers of infants with neonatal abstinence syndrome: A population-based parallel-cohort study in England and Ontario, Canada" (PMEDICINE-D-19-02772R1) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by one of the reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

*PLEASE NOTE: we have a very close deadline for inclusion in the substance abuse special issue, which publishes throughout November and so please do submit as soon as possible to allow every opportunity for this to make the deadline, but not before you hear from my colleagues in production who will be in touch in the next couple of days with formatting edits they require - please do not resubmit until you here from them too and combine all edits into one resubmission.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

Our publications team (plosmedicine@plos.org) will be in touch shortly about the production requirements for your paper, and the link and deadline for resubmission. DO NOT RESUBMIT BEFORE YOU'VE RECEIVED THE PRODUCTION REQUIREMENTS.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Oct 16 2019 11:59PM.

Sincerely,

Clare Stone

for

Louise Gaynor, MBBS PhD

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

Abstract - Please add p values to all quantifiable data and where 95% Cis are provided (also throughout main text and tables)

Data statement – Ontario date – we need a point of contact that isn’t an author, per PLOS data policy.

Author summary –please add it into the main text immediately after the abstract and remove the Supp file containing it. In addition, it’s quite long and so I suggest you remove:

• Only one recent study from the United States has explored death rates in women who took opioids in pregnancy, but solely in the first year and focused only on opioid-related deaths

And

• Surveillance of the harm of the opioid crisis should measure all types of mortality, not just that related to opioid overdoses, and jurisdictions should capitalize on available sources of health, social and demographic administrative data to monitor and evaluate programs

In your author summary, you say “Mothers with infants with versus without NAS were more likely to be teenagers” yet table shows that iin fact age 20-34is the largest number of NAS cases for both cohorts and fololowed by age 35+ and then te4enagers. Have I misunderstood table 2? If not, please remove declamatory language from the author summary.

Page 21 (our study is the first population-based analysis) – to our knowledge..

Comments from Reviewers:

Reviewer #2: Thank you to the authors for responding to my queries and comments. I only have one minor additional comment.

I think that the current description of the random selection of one birth for each mother still needs further clarification. As a suggestion, perhaps:

"We restricted the cohort to singleton births and if a woman had more than one livebirth delivery during the study period, one delivery was chosen at random as the focus of the study. Thus, a delivery date was selected at random and used as the study entry point for the mother (referred to as index delivery) and subsequent deliveries were ignored."

Any attachments provided with reviews can be seen via the following link:

[LINK]

17 Oct 2019

Submitted filename: 00_Response to Reviewers2_PLOS Med NAS Paper_FINAL.docx

Click here for additional data file.

21 Oct 2019

Dear Dr. Guttmann,

On behalf of my colleagues and the academic editor, Dr. Louisa Degenhardt, I am delighted to inform you that your manuscript entitled "Long-term mortality in mothers of infants with neonatal abstinence syndrome: A population-based parallel-cohort study in England and Ontario, Canada" (PMEDICINE-D-19-02772R2) has been accepted for publication in PLOS Medicine.

PRODUCTION PROCESS

Before publication you will see the copyedited word document (in around 1-2 weeks from now) and a PDF galley proof shortly after that. The copyeditor will be in touch shortly before sending you the copyedited Word document. We will make some revisions at the copyediting stage to conform to our general style, and for clarification. When you receive this version you should check and revise it very carefully, including figures, tables, references, and supporting information, because corrections at the next stage (proofs) will be strictly limited to (1) errors in author names or affiliations, (2) errors of scientific fact that would cause misunderstandings to readers, and (3) printer's (introduced) errors.

If you are likely to be away when either this document or the proof is sent, please ensure we have contact information of a second person, as we will need you to respond quickly at each point.

PRESS

A selection of our articles each week are press released by the journal. You will be contacted nearer the time if we are press releasing your article in order to approve the content and check the contact information for journalists is correct. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact.

PROFILE INFORMATION

Now that your manuscript has been accepted, please log into EM and update your profile. Go to https://www.editorialmanager.com/pmedicine, log in, and click on the "Update My Information" link at the top of the page. Please update your user information to ensure an efficient production and billing process.

Thank you again for submitting the manuscript to PLOS Medicine. We look forward to publishing it.

Best wishes,

Louise Gaynor, MBBS PhD

Associate Editor

PLOS Medicine

plosmedicine.org