Study design and setting

This was a population-based retrospective cohort study in Ontario, Canada. With 14 million residents and over 140,000 obstetrical deliveries every year, Ontario is Canada’s largest province [17]. Essential health care services, including all primary and acute care, obstetrical care, and mental health care with a physician, are delivered at no direct cost to residents.

The current study cohort considered women aged 15 to 49 years with a live birth conceived between April 1, 2005 and March 31, 2015. Women were followed for 365 days post partum to ascertain study outcomes, to a maximum date of December 31, 2016. Excluded were women with diagnosed mental illness in the 2-year period before conception in the index pregnancy. A mental illness diagnosis was based on one or more physician or emergency department visits or a hospitalization for any psychotic disorder, mood or anxiety disorder, substance use disorder, self-harm event, or any other mental illness.

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist). A prospective analysis plan was used to design the study (S1 Text). The use of data was authorized under section 45 of Ontario’s Personal Health Information Protection Act, which does not require review by a research ethics board.

Data sources

We accessed and analysed health administrative data at ICES, in Toronto, Ontario. ICES captures sociodemographic and diagnostic data, from the healthcare encounters of Ontario residents, that are linked at the individual level using a unique, encoded identifier. Women with a singleton livebirth were identified in the MOMBABY data set, which links >98% of maternal and newborn records for the delivery hospitalization [17]. The date of conception was determined by subtracting the recorded gestational age—generally based on dating from a first trimester ultrasound [18]—from the recorded date of birth. We also used the Canadian Institute for Health Information Discharge Abstract Database for hospitalizations, the Ontario Mental Health Reporting System for psychiatric hospitalizations, the National Ambulatory Care Reporting System for emergency department visits, the Ontario Health Insurance Plan (OHIP) Database for outpatient visits, the Client Agency Program Enrolment Database for a registry of patients enrolled in primary care groups, the ICES Physician Database for physician specialties, and the Registered Persons Database for sociodemographic characteristics. Information in the OHIP Database is recorded using physician billing claims; information in hospital databases is recorded using the Canadian Coding Standards for the International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10) or Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV). The information in these databases has been shown to be complete and reliable; sociodemographic data, physician billing claims, and primary diagnoses in hospital data sets have excellent completeness and accuracy [19].

Exposure

Women in the exposed group were those with one or more chronic physical conditions. Our conceptualization of chronic physical conditions was based on the Agency for Healthcare Research and Quality Chronic Condition Indicator for the ICD-10 [20]. Conditions were defined as chronic according to whether they last 12 months or longer and (1) limit independent living, social interactions, or self-care or (2) require ongoing intervention with medical services, products, or equipment [20]. Chronic physical conditions were grouped by body system, according to ICD-10 chapters: endocrine, nutritional, and metabolic diseases and immunity disorders; diseases of the circulatory system, respiratory system, musculoskeletal system, nervous system and sense organs, digestive system, genitourinary system, skin and subcutaneous tissue, and blood and blood-forming organs; neoplasms; infections and parasitic disease; congenital anomalies; and injuries (S1 Table).

We considered a woman to have a chronic physical condition if any ICD-10 code included in this definition was present in an acute healthcare encounter in the 24 months prior to conception, denoted either as an emergency department visit without a hospitalization or a hospitalization. Women in the unexposed group were those with no acute healthcare encounter for such a condition during the 24-month period preceding conception. Although women with mild or well-controlled conditions may be more likely to present to a primary care provider than an acute care setting, this conceptualization captures those with the most severe conditions, which would be expected to carry the greatest risk for perinatal mental illness.

Outcomes

Perinatal mental illness was defined as at least 1 mental illness diagnosis between conception and 365 days post partum, captured by one of the following encounters: (1) a visit to a general practitioner or family physician with a mental illness diagnosis, (2) a visit to a psychiatrist, (3) an emergency department visit with a mental illness diagnosis, or (4) a hospitalization with a mental illness diagnosis [21,22]. Perinatal mental illness was further categorized by (a) its timing (i.e., prenatally, with the first encounter occurring during pregnancy; or post partum, with the first encounter occurring within 365 days of the delivery date) and (b) the diagnosis, namely, a psychotic disorder, mood or anxiety disorder, substance use disorder, another mental illness, or self-harm. Classification of the mental illness diagnosis was determined based on all healthcare encounters in the follow-up period, because women could have more than one perinatal mental illness diagnosis during this time (S2 Table) [21,22]. Because women with a recent history of mental illness in the 2 years before conception were excluded, it was reasonable to assume that perinatal mental illness arising in the index pregnancy largely represented a new episode of mental illness.

Covariates

Covariates were social and health characteristics hypothesized to distinguish women with and without chronic physical conditions and that were available within the administrative data sets: maternal age, parity, neighbourhood income quintile, rural residence, and remote history of mental healthcare. Neighbourhood income quintile and rural residence are derived by linking residential postal code with census data [23]. Rural residences are those in communities of <10,000 individuals. A remote history of mental healthcare was defined as having 1 or more physician visits, emergency department visits, or hospitalizations for a psychotic disorder, mood or anxiety disorder, substance use disorder, self-harm, or another mental illness more than 2 years before the date of conception in the index pregnancy.

Data analyses

Baseline characteristics of women with and without chronic physical conditions were described using means and proportions. Differences were expressed using standardized differences, with a value more than 0.10 determined to be clinically meaningful [24].

Unadjusted relative risks (RRs) and adjusted RRs (aRRs), and 95% confidence intervals (CIs), contrasted the risk of perinatal mental illness between women with versus without a prepregnancy chronic physical condition, using modified Poisson regression [25]. Because women may have delivered more than once during the study period, we accounted for potential clustering of births within mothers using generalized estimating equations [26]. All multivariable models were adjusted for maternal age, parity, rural residence, neighbourhood income quintile, and remote history of mental healthcare.

We conducted several prespecified additional analyses. One additional analysis examined the relationship between chronic physical conditions and perinatal mental illness by (a) the timing of its diagnosis (i.e., prenatally or post partum, as described above) and (b) the specific diagnosis of mental illness (as described above and in S2 Table). A second additional analysis explored the relationship between chronic physical conditions and perinatal mental illness by the number and type of body systems affected by chronic physical conditions (S1 Table). We also explored the relationship between different types of body systems affected by chronic physical conditions and (a) timing and (b) specific diagnosis of perinatal mental illness. For these analyses, also included in the multivariable models was the presence of other chronic physical conditions. Third, a more rigorous definition of the study outcome was instituted, requiring at least 2 physician visits for a mental health reason within pregnancy or 365 days post partum.

All data analyses used SAS version 9.4 (SAS Institute, Cary, North Carolina).

Limitations

The Agency for Healthcare Research and Quality Chronic Condition Indicator for the ICD-10 is a comprehensive method by which to identify chronic physical conditions in administrative data. However, in Ontario, it can only be applied to acute healthcare encounters, because physician visits are coded using a 3-digit version of the ICD-9, which cannot distinguish between an acute versus a chronic disease. In our definition of chronic physical conditions, we likely captured those women with the most severe, complicated, or poorly managed conditions. Such individuals could be at greatest risk for perinatal mental illness. It is also possible that women with chronic physical conditions presenting in acute healthcare settings differ from those with chronic physical conditions more broadly on factors other than condition severity. For example, they could be more anxious or depressed but less likely to seek care for mental illness. The impact of misclassification on our RRs is therefore difficult to predict. Our definition may explain the higher proportion of women with a chronic physical condition living in rural areas, where emergency departments tend to be accessed for all levels of care. The generalizability of our cohort of women with chronic physical conditions to the target population should therefore be interpreted with caution. Finally, the algorithm classified chronic physical conditions by body system and not by each disease diagnosis, so specific details about how a condition influences mental health cannot be described.

Our definition of diagnosed perinatal mental illness required that a woman sought healthcare. Yet a significant proportion of women with perinatal mental illness never receive a diagnosis because of barriers to accessing care [42]. Hence, the current study may have underestimated the true rate of perinatal mental illness, especially in marginalized groups. Because of the truncated codes used for physician visits, we could not distinguish between depression and anxiety. However, research suggests that there is increased risk for both in association with chronic physical conditions [14]. We used liberal definitions of recent and remote history of mental health care, resulting in higher rates of these variables (25.7% and 47.9%, respectively) than in general-population self-reported surveys [43]. Finally, because our cohort was restricted to women without a recent history of mental illness, our results pertain to the development of new-onset mental illness or relapse in the perinatal period, with implications for prevention interventions. Future research could include all women, regardless of recent mental healthcare, to examine patterns of ongoing mental illness in the perinatal period associated with chronic physical conditions.

Finally, residual confounding may be a threat to the validity of our findings. Health administrative data were not collected for research purposes, and, as such, data on important confounders were missing. We did not have information on demographic variables such as ethnicity or individual-level measures of socioeconomic status such as income and education. We also did not have information on BMI, lifestyle factors such as diet and smoking, or social factors such as stressful life events or social support. These are all important confounders in the association between chronic physical conditions and perinatal mental illness [6]; however, we were unable to adjust for these variables in our multivariable models. Similarly, we were unable to adjust for shared genetic influences underlying both chronic physical conditions and perinatal mental illness [36]; this is an important area for future study.

Conclusion

Findings from this large population-based study suggest that women with a chronic physical condition predating pregnancy are at heightened risk of mental illness in pregnancy and post partum. This risk remained higher for all specific mental illness diagnoses except self-harm and was largely consistent across chronic physical conditions affecting different body systems. These women may require targeted efforts to lower the severity of their condition and improve their coping strategies and supports in pregnancy and the postpartum period.