Severe Maternal Morbidity and Infant Mortality in Canada

A population-based cohort study was completed by using provincial health administrative data collected within a universal health care system. These data sets were linked by using unique encoded identifiers and analyzed at ICES (the Institute for Clinical Evaluative Sciences), Ontario, Canada. Linked maternal-infant pairs were identified in the Institute for Clinical Evaluative Sciences MOMBABY database, and their characteristics and outcomes were captured in the following data sources: the Canadian Institute for Health Information’s (CIHI) Discharge Abstract Database for all hospital live birth and stillbirth details, with up to 25 diagnostic and procedural codes arising within a hospitalization. Diagnostic codes are based on the International Classification of Diseases, 10th Revision, Canada (ICD-10-CA), and procedural codes are based on the Canadian Classification of Health Interventions (CCI). The Ontario Health Insurance Plan (OHIP) database identifies preexisting maternal health conditions by using an International Classification of Diseases, Ninth Revision, Canada diagnostic code for every outpatient visit. The Ministry of Health’s Registered Persons Database (RPDB) contains vital status and sociodemographic information for all individuals ever eligible for OHIP. The Better Outcomes Registry and Network (BORN) database contains additional information and maternal characteristics. The Immigration, Refugees and Citizenship Canada’s permanent resident database includes the country of origin for immigrants who landed in Ontario. Importantly, these data sets have been well validated and studied in a number of previous studies to capture and describe the study participants.^2,13,16,17 

We identified all singleton hospital live births and stillbirths born at ≥ 23 weeks’ gestation within the province of Ontario between April 1, 2002, and March 31, 2017. We excluded multifetal births, non–Ontario resident mothers, mothers aged <10 or >55 years, and, for live births, missing gestational age at birth, birth weight, or sex (Fig 1). All inclusion and exclusion criteria and variable definitions are described in Supplemental Table 5. The study was limited to singleton births because they constitute >96% of all births and enable a more straightforward 1-to-1 comparison with each mother. Additionally, multifetal newborns are more likely to experience death from factors not necessarily related to maternal health, such as twin-twin transfusion syndrome or a major congenital anomaly.¹⁸ 

SMM was defined by using a modified version of that developed by the Canadian Perinatal Surveillance System,⁶  as described elsewhere,²  and similar to that developed by the US Centers for Disease Control and Prevention.^2,19  SMM includes >40 unique indicators that are based on ICD-10-CA diagnostic codes and CCI procedural codes, such as ICU admission, eclampsia, and use of assisted ventilation, for example. The composite of SMM has been validated against maternal mortality and hospital length of stay and in different countries.^20–22  The quality of the variables collected within the Canadian hospital data sets is generally high.²³ 

SMM was captured anytime between the index delivery hospitalization admission date and up to 42 days after the delivery even if there was a readmission to hospital within that time frame. The primary study exposure was the presence versus absence of SMM as an overall construct, further broken down by the number of SMM indicators, as outlined below.

The primary study outcome was infant death from birth to 365 days thereafter. Secondary outcomes included stillbirth at ≥23 weeks’ gestation, neonatal death from birth to 27 days, and postneonatal death from 28 to 365 days. Infant deaths related to sepsis were also analyzed, as described below. Infant deaths were identified on the basis of a discharge disposition from the birth or subsequent inpatient hospital admission or from the RPDB in the case of out-of-hospital deaths. Stillbirths are recorded on the maternal record.

Means and SDs were used to describe normally distributed continuous variables, whereas categorical variables were expressed as counts and proportions. Standardized differences compared women with and without SMM, with a value <0.10 indicating no important difference.²⁴ 

In the main model, the relative risk (RR) and 95% confidence interval (CI) of the association between SMM and infant death was estimated by using modified Poisson regression with a robust error variance.²⁵  Generalized estimating equations with an exchangeable correlation structure accounted for correlated errors in the case of multiple pregnancies within the same woman.²⁶  RRs were adjusted for maternal age (continuous); residential income quintile (quintile 1 or unknown, quintiles 2–5); rural residence; world region (Canada or long-term resident, Caribbean or sub-Saharan Africa, South Asia, East Asia or Pacific, other or unknown), each at the time of the index birth; as well as diabetes mellitus and chronic hypertension within 2 years before the index birth (Supplemental Table 5).

The main model of SMM and infant death was further stratified by maternal factors (age [≥40 or <40 years], parity [nulliparous or parous], world region of origin [Afro-Caribbean origin or other], BMI [≥25 or <25], preexisting type 1 or 2 diabetes mellitus, preexisting chronic hypertension, spontaneous or provider-initiated preterm birth <37 weeks’ gestation, and vaginal or cesarean delivery) and by infant factors (sex, timing of birth, birth weight percentile, 5-minute Apgar score, and any congenital or chromosomal anomaly diagnosed in the first year of life). When applicable, a covariate was removed from its respective regression model when it formed a stratification variable. The temporal relation between SMM and infant death was also explored by assessing the timing of onset of SMM: from 23 weeks’ gestation up to, but not including, the hospitalization for the index birth or any time thereafter; from 23 weeks’ gestation up to, and including, the index birth hospitalization; as well as from 23 weeks’ gestation up to 42 days after the index birth date.

The same modeling approach used in the main model was applied in the analysis of the number of SMM indicators (0, 1, 2, 3, 4, ≥5) and infant death from birth to 365 days, as well as for SMM and stillbirth, SMM and neonatal death from birth to 27 days, and SMM and postneonatal death from 28 to 365 days after birth.

The association between sepsis-related SMM (ICD-10-CA code O75.3) and infant death with a diagnosis of bacterial sepsis (ICD-10-CA code P36) at the time of or preceding death was analyzed by using multinomial logistic regression to generate odds ratios (ORs) and corresponding 95% CIs.²⁷  ORs can be interpreted as risk ratios when the event rate is rare.²⁸  The odds of an infant death with and without sepsis was evaluated in relation to SMM with sepsis and SMM without sepsis, each relative to no SMM, and adjusted for the same covariates as in the main model.

All analyses were performed by using SAS statistical software, version 9.4 (SAS Institute, Inc, Cary, NC). The use of data in this project was authorized under section 45 of Ontario’s Personal Health Information Protection Act, which does not require review by a research ethics board.

There were 2 007 880 maternal-infant live-born or stillborn pairs during the study period, of which 115 023 (5.7%) were excluded predominantly because of multifetal birth (Fig 1). Of the eligible 1 892 857 singleton pregnancies ending in a live birth or stillbirth from 23 weeks’ gestation onward, 20 107 were affected by SMM, corresponding to a rate of 10.6 per 1000 births (Table 1). There were 20 164 pregnancies affected by SMM or maternal death (10.7 per 1000 births). Women with SMM were more likely than those without SMM to be of Caribbean or sub-Saharan African origin, have chronic hypertension (27.3% vs 14.7%), a cesarean delivery (44.6% vs 19.8%), a provider-initiated preterm birth (12.0% vs 3.1%), and a low Apgar score (2.2% vs 0.7%) (Table 1).

There were 5463 infant deaths within 365 days after a birth. Infant death occurred more often in mothers with SMM (8.9 per 1000 live births) than in those without SMM (2.8 per 1000 live births) (an unadjusted RR of 3.12 [95% CI 2.68–3.64] and an adjusted relative risk [aRR] of 2.93 [95% CI 2.51–3.41]; Table 2).

The greater risk of infant mortality with SMM was robustly seen across various maternal characteristics (Fig 2A). The exception was among women with SMM and a prepregnancy BMI ≥25. A greater risk of infant death in the presence of SMM remained after stratification by infant factors, except for extreme preterm births <32 weeks’ gestation and infants affected by an anomaly (Fig 2B). The risk of infant mortality was higher regardless of when SMM occurred (Supplemental Table 6). The rate of infant mortality generally increased with the number of SMM indicators: 0 (2.8 per 1000 births), 1 (7.6 per 1000 births), 2 (11.4 per 1000 births), 3 (9.2 per 1000 births), 4 (13.9 per 1000 births), and ≥5 (34.9 per 1000 births), the latter corresponding to an aRR of 12.13 (95% CI 6.85–21.48; Table 3).

Out of 19 587 pregnancies with SMM, 4523 (23.1%) had maternal sepsis. Relative to births without SMM, the adjusted OR for infant death from sepsis was 1.95 (95% CI 1.10–3.45) associated with non–sepsis-related SMM and 6.36 (95% CI 3.50–11.55) with sepsis-related SMM (Table 4).

Relative to the absence of SMM, the risks of stillbirth (aRR 6.22; 95% CI 5.69–6.81), neonatal death (aRR 3.33; 95% CI 2.79–3.97), and postneonatal death (aRR 2.12; 95% CI 1.56–2.89) were all higher in those whose mother had SMM (Supplemental Table 7).

In this large population-based study of 1.9 million singleton maternal-infant pairs, SMM was strongly associated with infant mortality as well as with stillbirth. As the number of SMM indicators rose, so did the risk of infant death. A coupling tendency was observed between a woman having SMM marked by sepsis and her infant dying from sepsis.

Recognized risk factors for neonatal mortality at birth include extreme preterm weight, severe growth restriction, and a low 5-minute Apgar score.^29,30  Sepsis, lethal congenital or chromosomal anomalies, and sudden infant death syndrome are causes of postneonatal death.³¹  Recent discussion³²  and data³³  have emerged about an association between stillbirth and SMM, as well as the relation between stillbirth and maternal death after delivery. For example, among 3.2 million births in Ontario, the maternal mortality rate in the first year delivery was 206 per 100 000 stillbirths compared to 23 per 100 000 live births.³³  Among 97 095 women who gave birth in 12 Latin American countries, early neonatal death was associated with SMM, with an OR of 4.77 (95% CI 3.74–6.07).³⁴ 

To our knowledge, this is the first study linking SMM with infant mortality, including adjusting for sociodemographic factors like maternal age, income status, and world region of origin, and further stratifying by conventional risk factors for perinatal mortality. As in a previous report of a curvilinear rise in the risk of maternal death with the number of SMM indicators,²  the same pattern for infant mortality was seen in the current study.

This study included almost 2 million singleton maternal-infant pairs within a universal health care system, including 99% of all eligible hospital births over a 15-year period,^2,13,16,17  thereby avoiding selection bias. It is unlikely that infrequent outward migration of infants materially altered the current findings, such as the undercapture of infant deaths. The completeness of these population-based data sets enabled precise and generalizable estimates of the rates and risks of infant death. Only singleton births were included herein, so the relation between SMM and infant mortality might differ for multiple births. For example, twins and triplets experience greater neonatal mortality from preterm birth, fetal growth restriction, twin-twin transfusion syndrome, and major congenital abnormalities,¹⁸  and their mothers tend to experience greater SMM (eg, preeclampsia).³⁵  Hence, further study is needed about SMM and infant mortality in multifetal pregnancies. Stratifying by gestational age of <32 weeks neutralized any apparent relation (Fig 2B) likely because of collider stratification bias, as described elsewhere.³⁶ 

There is no single international definition of SMM, but the Canadian version employed herein has been validated against maternal mortality and hospital length of stay, is similar to that used by the US Centers for Disease Control and Prevention,^2,5,6,19  and aligns with the World Health Organization’s definition of severe acute maternal morbidity (maternal near miss).³⁷  Although SMM may have been based on inaccurate or incorrect use of ICD-10-CA and CCI codes, CIHI periodically completes data quality checks to assess coding accuracy. Although some SMM events may have occurred postpartum, the majority arise antepartum and intrapartum largely because of severe preeclampsia or eclampsia, major obstetric hemorrhage, and sepsis.³²  Not all SMM indicators represent a sentinel event, defined as “a patient safety event (not primarily related to the natural course of the patient’s illness or underlying condition) that reaches a patient and results in any of the following: death, permanent harm, or severe temporary harm.”³⁸  Only sepsis from an overall SMM construct was additionally investigated as an exposure. However, other indicators of SMM (eg, placenta previa with hemorrhage) may also affect fetal and neonatal outcomes such as prematurity and low birth weight.³⁹  Hence, it is worth conducting further studies to explore each indicator of SMM and perinatal, neonatal, and infant outcomes. Some common preexisting or chronic conditions (eg, prepregnancy diabetes mellitus and/or chronic hypertension) can adversely influence placental and fetal growth and well-being over time and then evolve into an acute SMM event (eg, severe preeclampsia). Although the current analysis did not distinguish between such chronic and acute factors contributing to SMM, antecedent maternal diabetes mellitus and chronic hypertension were adjusted for.

In North America, the incidence of SMM has increased over the past 2 decades.^5,7  Although some SMM indicators (eg, amniotic fluid embolism) are rare, hard to predict, and even tougher to prevent, other SMM indicators are seemingly more preventable and manageable, creating potential opportunities to not only reduce maternal mortality but infant mortality as well. In New Zealand, for example, authors of a multidisciplinary review of SMM suggested that at least one-third of cases are preventable.⁴⁰  In the United States, a nationwide hospital-level implementation of the Maternal Early Warning Trigger tool significantly reduced the incidence of SMM.¹⁹  The effect such a tool has on perinatal mortality remains unknown and requires further investigation. For infant mortality and stillbirth, because the risk of these outcomes may differ by the nature of the SMM, application to each SMM indicator of something like a number needed to prevent might provide insight about which indicator(s) warrants the greatest focus, are most common, or most linked to both maternal and a perinatal death.¹³  Certainly, one starting point example is the use of low-dose acetylsalicylic acid for the prevention of preeclampsia,⁴¹  fetal growth restriction,⁴²  and preterm birth,⁴³  which are all risk factors for stillbirth,⁴⁴  neonatal death, and prolonged hospital length of stay.⁴⁵ 

It is also of interest that we observed a coupling tendency between sepsis-related SMM and sepsis-related infant death, which is consistent with emerging evidence.¹⁵  The Global Maternal and Neonatal Sepsis Initiative aims to “accelerate the reduction of preventable maternal and neonatal deaths related to sepsis.”¹⁴  The timing of maternal infection during pregnancy and the postpartum period was not analyzed herein nor was the onset of infant infection. The responsible pathogen(s) for maternal and infant infection was also not documented. Even so, reducing maternal sepsis should, in theory, reduce newborn sepsis as well.

Risk of infant death is greater in the presence of SMM, including a dose-response effect, as is the risk of stillbirth. A coupling tendency exists between sepsis-related SMM and sepsis-related infant death. There now exists a promising opportunity to identify preventable SMM and facilitate care that may limit both the progress of SMM and reduce perinatal mortality.

aRR

adjusted relative risk

BORN

Better Outcomes Registry and Network

CCI

Canadian Classification of Health Interventions

CI

confidence interval

CIHI

Canadian Institute for Health Information

ICD-10-CA

International Classification of Diseases 10th Revision Canada

OHIP

Ontario Health Insurance Plan

OR

odds ratio

RPDB

Ontario Ministry of Health’s Registered Persons Database

RR

relative risk

SMM

severe maternal morbidity