Mortality Among Mothers Whose Children Were Taken Into Care by Child Protection Services: A Discordant Sibling Analysis

Mortality Among Mothers Whose Children Were Taken Into Care by Child Protection Services: A... Abstract This study examines whether mothers who had a child taken into care by child protection services have higher mortality rates compared with rates seen in their biological sisters who did not have a child taken into care. We conducted this retrospective cohort study using linkable administrative data from 3,948 mothers whose oldest child was born in Manitoba, Canada, between April 1, 1992, and March 31, 2015. These mothers were from 1,974 families in which one sister had a child taken into care and one sister did not. We computed rate differences and hazard ratios of all-cause, avoidable, and unavoidable mortality. There were an additional 24 deaths per 10,000 person-years among mothers who had had a child taken into care. Mothers who had a child taken into care had higher rates of mortality due to avoidable causes (hazard ratio = 3.46; 95% confidence interval: 1.41, 8.48) and unavoidable causes (hazard ratio = 2.92; 95% confidence interval: 1.01, 8.44). The number of children taken into care did not affect mortality rates among mothers with at least 1 child taken into care. The higher mortality rates—particularly avoidable mortality—among mothers who had a child taken into care indicate a need for more specific interventions for these mothers. avoidable mortality, child protection services, linkable administrative data, mortality among mothers, unavoidable mortality Editor’s note:An invited commentary on this article appears on page 1189. When children are taken into care by child protection services, the safety and well-being of the child are the highest priority. This process often overlooks the health and well-being of the mother. We focused on mothers in this work because most children are not living with their fathers when they go into care (1). Previous studies have found that mothers who had a child taken into care often have more health issues and social instability than mothers in the general population; these challenges worsen after their child is taken (2, 3). The distress that a mother faces after a different type of loss—the death of a child—is publicly acknowledged and has been linked with many health consequences, such as increased mental illness and heightened mortality (4, 5). Recent findings indicate that mothers who had lost custody of a child to child protection services have higher rates of mental illness following separation from their child than mothers who experienced the death of a child (6). While mothers who had a child taken into care have higher rates of suicide attempts and completions, it is not known whether there is a higher rate of mortality among mothers from other causes after losing custody of a child (7). This study aims to fill this knowledge gap by examining all-cause mortality rates among mothers who had a child taken into care, and comparing these rates with those of their sisters who were also mothers but did not have a child taken. We compared mortality rates in discordant sisters because mortality has been found to cluster in families due to environmental and genetic factors; discordant sibling analyses account for these stable family characteristics (8, 9). Mortality rates are often classified into avoidable causes of mortality and unavoidable causes; avoidable mortality is a death from a cause that could have been prevented or treated (10). Avoidable mortality include deaths due to infections, certain cancers, and injuries (unintentional or intentional). Avoidable mortality rates are particularly high among individuals from low socioeconomic backgrounds and those with mental disorders; mothers involved with child protection services often have one or both of these characteristics (11–13). Thus, we also examined whether mothers with a child taken into care had higher rates of avoidable and unavoidable mortality than did mothers not having a child taken into care. METHODS Setting and data Data for this study came from Manitoba, Canada, which is a central Canadian province with approximately 1.2 million residents (14). Manitoba residents receive universal healthcare coverage (physician visits and hospitalizations) and financial assistance for pharmaceuticals if high prescription-drug costs seriously affect their income (15). Rates of children in care in Manitoba are among the highest in the world, with approximately 3% of children in care (16). Manitoba has among the highest rates of premature mortality among Canadian provinces, with 246.3 deaths per 100,000 people in 2008 (17). This is higher than the premature mortality rate throughout Canada (203.9 per 100,000) and in the United Kingdom (225 per 100,000) but lower than in the United States (271 per 100,000) (10, 17). This study used the linkable administrative data in the Population Research Data Repository housed at the Manitoba Centre for Health Policy. The population registry is linked at the individual level with physician claims, hospital discharge abstracts, pharmaceutical claims, Child and Family Services case reports, Employment and Income Assistance case reports, vital statistics, and the Canadian Census. A scrambled personal health number was used to link these deidentified data sets. Information on linkage methods, confidentiality/privacy, and validity can be found elsewhere (18). Cohort formation The cohort consisted of biological sisters whose first child was born between April 1, 1992, and March 31, 2015. Sisters were defined as having the same biological mother. This cohort was divided into two groups. Group 1 consisted of mothers whose children were taken into care before March 31, 2015. These mothers were identified through the child protection case files; child protection is delivered through Child and Family Services in Manitoba. The index date for these mothers was selected as the date that a child was first taken into care. Group 2 included mothers not having a child taken into care before March 31, 2015, and with a biological sister in group 1. The oldest child was selected for mothers in group 2 as the index child, and her index date was defined as the date the index child turned the age her sister’s child had been when that child was taken into care. Mothers who did not live in Manitoba for at least 2 years before the index date were excluded. Each mother was followed from the index date until the end of coverage (date of death or move out of province) or December 31, 2015, whichever came first. For families with more than 1 sister in group 1 or group 2, 1 sister was randomly selected. Figure 1 details the cohort selection process. Figure 1. View largeDownload slide Cohort formation for a study of mortality among mothers with children who were taken into protective care in Manitoba, Canada, 1992–2015. Figure 1. View largeDownload slide Cohort formation for a study of mortality among mothers with children who were taken into protective care in Manitoba, Canada, 1992–2015. Mortality among mothers Date of death and underlying cause of death were obtained from the Vital Statistics data set. Cause of death is coded using the International Classification of Diseases (ICD), Ninth Revision, for deaths before 2000 and Tenth Revision for deaths in 2000 or later. The ICD-9 codes for primary cause of death for deaths occurring before 2000 were converted to ICD-10 codes (19). We examined all-cause mortality (any death between April 1, 1992, and December 31, 2015) and avoidable and unavoidable mortality. Avoidable mortality generally refers to deaths that could have been avoided through proper medical care, through intersectoral health policy interventions, or through public health programs (20). A list of causes of death for avoidable mortality created by the Canadian Institute for Health Information was used to identify avoidable mortality (see Web Table 1, available at https://academic.oup.com/aje); causes of deaths not included in this list are noted as unavoidable mortality (10). Statistical analysis Mortality rates (overall, avoidable, and unavoidable) per 10,000 person-years were first examined for each group of mothers, and rate differences are presented. Next, we investigated differences in mortality rates using crude and adjusted Cox proportional hazards regression models. Mothers who have had a child taken into care differ from those not having a child taken; these individual and family-level differences may contribute to mortality among mothers. To account for differences, the Cox proportional hazards regression models adjusted in 3 ways. First, individual-level differences (health and sociodemographic) were balanced for using inverse probability of treatment weights. Second, stable family covariates (such as genetics and environment) were accounted for by including a family fixed effect in the regression model. Finally, both individual and family-level variables were accounted for using an inverse probability of treatment–weighted fixed-effects model. Inverse probability of treatment weights were obtained from high-dimensional propensity scores (hdPSs). Unlike traditional propensity score methods, which are limited to covariates specified by the investigator, the multistep algorithm used to derive hdPSs identifies potential confounders from a database by selecting variables correlated to both the exposure and outcome, prioritizing covariates by prevalence and potential for bias (21). Covariates drawn from the 2 years before the index date were based on: 1) medical service tariff codes; 2) physician diagnostic codes; 3) hospital procedure codes; 4) hospital diagnostic codes. These data describe the health status of the women in our study (21). In our cohort, more than 97% of mothers had contact with a primary care physician in the 2 years before the index date. In addition to the health variables obtained from these data, a series of sociodemographic variables were included in the models. In the 2 years before the index date, we noted whether the mother had moved and whether the mother had received Employment and Income Assistance (analogous to welfare). We also included the mother’s age, neighborhood socioeconomic status, and neighborhood location (urban/rural) at the index date. The 4 data sets used to assess the health status in the 2 years before the index date are extremely large. For example, physician diagnoses are claimed as ICD-9-CM (clinical modification) codes; these include approximately 17,000 different diagnostic codes. The hdPSs were constructed including all sociodemographic variables and the top 500 health covariates. The unweighted and weighted standardized differences for the top 20 covariates are presented in Table 1; we considered a covariate to be balanced if the standardized difference was less than 0.25 (22). Information on all 500 covariates included in the hdPS is available on request. The kernel density plot of the distribution of propensity scores showed sufficient common support for this analysis (see Web Figure 1). Table 1. Comparison of Baseline Characteristics of Mothers in a Study of Maternal Mortality, Manitoba, Canada, 1992–2015 Maternal Characteristic Had a Child Taken Into Care (n = 1,974) Did Not Have a Child Taken Into Care (n = 1,974) Unweighted Standardized Differences IPT Weighted Standardized Differences No. of Mothers % No. of Mothers % At the index date  Age category of mother, years   <20 547 27.71 452 22.90 0.111 0.082   20–29 1,019 51.62 1,116 56.53 0.099 0.178   ≥30 408 20.67 406 20.57 0.003 0.131  Socioeconomic status of neighborhood 1.17 (1.13)a 1.12 (1.23)a 0.045 0.014  Urban neighborhood 1,011 51.22 896 45.39 0.117 0.096 Event in the 2 years before the index date  Changed residence 954 48.33 767 38.86 0.192 0.011  Received welfare 1,076 54.51 683 34.60 0.409 0.045 Health in the 2 years before the index date  Office visits, complete history, and physical examination 373 18.90 212 10.74 0.231 0.006  Hospital care, subsequent care, first and second week 57 2.89 26 1.32 0.110 0.062  Depressive disorder 172 8.71 66 3.34 0.227 0.080  Disorders of function of stomach 38 1.93 13 0.66 0.112 0.035  Wounds, simple repair any location single 87 4.41 32 1.62 0.164 0.008  Psychotherapy by nonpsychiatrist 70 3.55 23 1.17 0.157 0.182  Neurotic disorders 155 7.85 87 4.41 0.144 0.121  Radiology, chest, PA, and lateral 85 4.31 52 2.63 0.091 0.038  Drug dependence 56 2.84 11 0.56 0.177 0.096  Other venereal diseases 270 13.68 155 7.85 0.189 0.028  Office visits, regional of subsequent visit or well-baby care 417 21.12 271 13.73 0.196 0.018  Nondependent abuse of drugs 77 3.90 26 1.32 0.163 0.024  Electrocardiogram 85 4.31 37 1.87 0.141 0.041 Maternal Characteristic Had a Child Taken Into Care (n = 1,974) Did Not Have a Child Taken Into Care (n = 1,974) Unweighted Standardized Differences IPT Weighted Standardized Differences No. of Mothers % No. of Mothers % At the index date  Age category of mother, years   <20 547 27.71 452 22.90 0.111 0.082   20–29 1,019 51.62 1,116 56.53 0.099 0.178   ≥30 408 20.67 406 20.57 0.003 0.131  Socioeconomic status of neighborhood 1.17 (1.13)a 1.12 (1.23)a 0.045 0.014  Urban neighborhood 1,011 51.22 896 45.39 0.117 0.096 Event in the 2 years before the index date  Changed residence 954 48.33 767 38.86 0.192 0.011  Received welfare 1,076 54.51 683 34.60 0.409 0.045 Health in the 2 years before the index date  Office visits, complete history, and physical examination 373 18.90 212 10.74 0.231 0.006  Hospital care, subsequent care, first and second week 57 2.89 26 1.32 0.110 0.062  Depressive disorder 172 8.71 66 3.34 0.227 0.080  Disorders of function of stomach 38 1.93 13 0.66 0.112 0.035  Wounds, simple repair any location single 87 4.41 32 1.62 0.164 0.008  Psychotherapy by nonpsychiatrist 70 3.55 23 1.17 0.157 0.182  Neurotic disorders 155 7.85 87 4.41 0.144 0.121  Radiology, chest, PA, and lateral 85 4.31 52 2.63 0.091 0.038  Drug dependence 56 2.84 11 0.56 0.177 0.096  Other venereal diseases 270 13.68 155 7.85 0.189 0.028  Office visits, regional of subsequent visit or well-baby care 417 21.12 271 13.73 0.196 0.018  Nondependent abuse of drugs 77 3.90 26 1.32 0.163 0.024  Electrocardiogram 85 4.31 37 1.87 0.141 0.041 Abbreviations: IPT, inverse probability of treatment; PA, posteroanterior. a Values are expressed as mean (standard deviation). Table 1. Comparison of Baseline Characteristics of Mothers in a Study of Maternal Mortality, Manitoba, Canada, 1992–2015 Maternal Characteristic Had a Child Taken Into Care (n = 1,974) Did Not Have a Child Taken Into Care (n = 1,974) Unweighted Standardized Differences IPT Weighted Standardized Differences No. of Mothers % No. of Mothers % At the index date  Age category of mother, years   <20 547 27.71 452 22.90 0.111 0.082   20–29 1,019 51.62 1,116 56.53 0.099 0.178   ≥30 408 20.67 406 20.57 0.003 0.131  Socioeconomic status of neighborhood 1.17 (1.13)a 1.12 (1.23)a 0.045 0.014  Urban neighborhood 1,011 51.22 896 45.39 0.117 0.096 Event in the 2 years before the index date  Changed residence 954 48.33 767 38.86 0.192 0.011  Received welfare 1,076 54.51 683 34.60 0.409 0.045 Health in the 2 years before the index date  Office visits, complete history, and physical examination 373 18.90 212 10.74 0.231 0.006  Hospital care, subsequent care, first and second week 57 2.89 26 1.32 0.110 0.062  Depressive disorder 172 8.71 66 3.34 0.227 0.080  Disorders of function of stomach 38 1.93 13 0.66 0.112 0.035  Wounds, simple repair any location single 87 4.41 32 1.62 0.164 0.008  Psychotherapy by nonpsychiatrist 70 3.55 23 1.17 0.157 0.182  Neurotic disorders 155 7.85 87 4.41 0.144 0.121  Radiology, chest, PA, and lateral 85 4.31 52 2.63 0.091 0.038  Drug dependence 56 2.84 11 0.56 0.177 0.096  Other venereal diseases 270 13.68 155 7.85 0.189 0.028  Office visits, regional of subsequent visit or well-baby care 417 21.12 271 13.73 0.196 0.018  Nondependent abuse of drugs 77 3.90 26 1.32 0.163 0.024  Electrocardiogram 85 4.31 37 1.87 0.141 0.041 Maternal Characteristic Had a Child Taken Into Care (n = 1,974) Did Not Have a Child Taken Into Care (n = 1,974) Unweighted Standardized Differences IPT Weighted Standardized Differences No. of Mothers % No. of Mothers % At the index date  Age category of mother, years   <20 547 27.71 452 22.90 0.111 0.082   20–29 1,019 51.62 1,116 56.53 0.099 0.178   ≥30 408 20.67 406 20.57 0.003 0.131  Socioeconomic status of neighborhood 1.17 (1.13)a 1.12 (1.23)a 0.045 0.014  Urban neighborhood 1,011 51.22 896 45.39 0.117 0.096 Event in the 2 years before the index date  Changed residence 954 48.33 767 38.86 0.192 0.011  Received welfare 1,076 54.51 683 34.60 0.409 0.045 Health in the 2 years before the index date  Office visits, complete history, and physical examination 373 18.90 212 10.74 0.231 0.006  Hospital care, subsequent care, first and second week 57 2.89 26 1.32 0.110 0.062  Depressive disorder 172 8.71 66 3.34 0.227 0.080  Disorders of function of stomach 38 1.93 13 0.66 0.112 0.035  Wounds, simple repair any location single 87 4.41 32 1.62 0.164 0.008  Psychotherapy by nonpsychiatrist 70 3.55 23 1.17 0.157 0.182  Neurotic disorders 155 7.85 87 4.41 0.144 0.121  Radiology, chest, PA, and lateral 85 4.31 52 2.63 0.091 0.038  Drug dependence 56 2.84 11 0.56 0.177 0.096  Other venereal diseases 270 13.68 155 7.85 0.189 0.028  Office visits, regional of subsequent visit or well-baby care 417 21.12 271 13.73 0.196 0.018  Nondependent abuse of drugs 77 3.90 26 1.32 0.163 0.024  Electrocardiogram 85 4.31 37 1.87 0.141 0.041 Abbreviations: IPT, inverse probability of treatment; PA, posteroanterior. a Values are expressed as mean (standard deviation). Two sensitivity analyses were conducted. The first examined mortality rates among mothers who had a child taken into care, comparing rates among mothers who had 1, 2, or 3 or more children taken into care. The second compared mortality rates between mothers who had a child taken into care and mothers who experienced the death of a child. Both sensitivity analyses adjusted for mother’s age, neighborhood location and socioeconomic status at the index date, and whether she received welfare or moved in the 2 years before the index date. All data management, programming, and analyses were performed using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina). RESULTS Table 1 displays the sociodemographic covariates and the top health covariates of mothers in each group. Mothers whose children were taken into care were more likely to have received welfare and to have worse health in the 2 years before the index date than their sisters who did not have a child taken. Mothers in the cohort were followed for a total of 33,624 person-years; mothers in group 1 were followed for 17,487 person-years (average of 8.9 years per person), and group 2 mothers for 16,137 person-years (average of 8.2 years per person). A total of 86 deaths occurred in the cohort before December 31, 2015. There were 24 additional deaths per 10,000 person-years among mothers who had a child taken into care (Table 2). Of the mothers who died in the follow-up period, those having a child taken into care were aged from 18 to 52 years (average = 30.8 years) at the time of death, and mothers who did not have a child taken into care were aged from 19 to 54 years (average = 31.3 years) at the time of death. For both mothers whose children were taken into care and those whose children were not taken into care, the most common cause of death was unintentional injury; this cause of death accounted for approximately 25% of deaths (see Web Table 2). Table 2. Mortality Rates and Mortality Rate Difference Among Mothers, Manitoba, Canada, 1992–2015 Mortality Had a Child Taken Into Care Did Not Have a Child Taken Into Care Rate Difference per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years Rate 95% CI Rate 95% CI Rate Difference 95% CI All-cause 65 37 29, 47 21 13 9, 20 24 14, 35 Avoidable 45 26 19, 35 15 9 6, 15 16 8, 25 Unavoidable 20 11 7, 18 6 4 2, 8 8 2, 14 Mortality Had a Child Taken Into Care Did Not Have a Child Taken Into Care Rate Difference per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years Rate 95% CI Rate 95% CI Rate Difference 95% CI All-cause 65 37 29, 47 21 13 9, 20 24 14, 35 Avoidable 45 26 19, 35 15 9 6, 15 16 8, 25 Unavoidable 20 11 7, 18 6 4 2, 8 8 2, 14 Abbreviation: CI, confidence interval. Table 2. Mortality Rates and Mortality Rate Difference Among Mothers, Manitoba, Canada, 1992–2015 Mortality Had a Child Taken Into Care Did Not Have a Child Taken Into Care Rate Difference per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years Rate 95% CI Rate 95% CI Rate Difference 95% CI All-cause 65 37 29, 47 21 13 9, 20 24 14, 35 Avoidable 45 26 19, 35 15 9 6, 15 16 8, 25 Unavoidable 20 11 7, 18 6 4 2, 8 8 2, 14 Mortality Had a Child Taken Into Care Did Not Have a Child Taken Into Care Rate Difference per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years Rate 95% CI Rate 95% CI Rate Difference 95% CI All-cause 65 37 29, 47 21 13 9, 20 24 14, 35 Avoidable 45 26 19, 35 15 9 6, 15 16 8, 25 Unavoidable 20 11 7, 18 6 4 2, 8 8 2, 14 Abbreviation: CI, confidence interval. The unadjusted hazard ratios were almost 3 times greater among mothers who had a child taken into care (hazard ratio (HR) = 2.90) (Table 3). For both avoidable and unavoidable causes, mothers who had a child taken into care were about 3 times more likely to die during the follow-up period (for avoidable causes, HR = 2.82, 95% confidence interval: 1.57, 5.06; for unavoidable causes, HR = 3.08, 95% confidence interval: 1.24, 7.68). Table 3. Unadjusted and Adjusted Risk of Mortality for Mothers Who Had a Child Taken Into Care, Manitoba, Canada, 1992-2015 Mortality Crude Model IPTW-Adjusted Crude Model With Family Fixed Effect IPTW-Adjusted With Family Fixed Effect HR 95% CI HR 95% CI HR 95% CI HR 95% CI All-cause 2.90 1.77, 4.74 3.04 1.80, 5.14 3.46 1.87, 6.42 3.23 1.62, 6.41 Avoidable 2.82 1.57, 5.06 2.99 1.62, 5.52 3.86 1.68, 8.86 3.46 1.41, 8.48 Unavoidable 3.08 1.24, 7.68 3.19 1.16, 8.78 3.00 1.19, 7.56 2.92 1.01, 8.44 Mortality Crude Model IPTW-Adjusted Crude Model With Family Fixed Effect IPTW-Adjusted With Family Fixed Effect HR 95% CI HR 95% CI HR 95% CI HR 95% CI All-cause 2.90 1.77, 4.74 3.04 1.80, 5.14 3.46 1.87, 6.42 3.23 1.62, 6.41 Avoidable 2.82 1.57, 5.06 2.99 1.62, 5.52 3.86 1.68, 8.86 3.46 1.41, 8.48 Unavoidable 3.08 1.24, 7.68 3.19 1.16, 8.78 3.00 1.19, 7.56 2.92 1.01, 8.44 Abbreviations: CI, confidence interval; HR, hazard ratio; IPTW, inverse probability of treatment weight. Table 3. Unadjusted and Adjusted Risk of Mortality for Mothers Who Had a Child Taken Into Care, Manitoba, Canada, 1992-2015 Mortality Crude Model IPTW-Adjusted Crude Model With Family Fixed Effect IPTW-Adjusted With Family Fixed Effect HR 95% CI HR 95% CI HR 95% CI HR 95% CI All-cause 2.90 1.77, 4.74 3.04 1.80, 5.14 3.46 1.87, 6.42 3.23 1.62, 6.41 Avoidable 2.82 1.57, 5.06 2.99 1.62, 5.52 3.86 1.68, 8.86 3.46 1.41, 8.48 Unavoidable 3.08 1.24, 7.68 3.19 1.16, 8.78 3.00 1.19, 7.56 2.92 1.01, 8.44 Mortality Crude Model IPTW-Adjusted Crude Model With Family Fixed Effect IPTW-Adjusted With Family Fixed Effect HR 95% CI HR 95% CI HR 95% CI HR 95% CI All-cause 2.90 1.77, 4.74 3.04 1.80, 5.14 3.46 1.87, 6.42 3.23 1.62, 6.41 Avoidable 2.82 1.57, 5.06 2.99 1.62, 5.52 3.86 1.68, 8.86 3.46 1.41, 8.48 Unavoidable 3.08 1.24, 7.68 3.19 1.16, 8.78 3.00 1.19, 7.56 2.92 1.01, 8.44 Abbreviations: CI, confidence interval; HR, hazard ratio; IPTW, inverse probability of treatment weight. When accounting for individual differences (the inverse probability of treatment weight–adjusted HRs), mortality rates among mothers remained similar to those seen in the crude analysis (Table 3). Accounting for stable family characteristics (such as genetics, ethnicity, and environment), mothers whose children were taken into care were 3.46 times more likely to die in the follow-up period, with avoidable mortality rates 3.86 times greater for mothers in this group. The final adjustment, which accounted for both individual differences (using inverse probability of treatment weights) and family characteristics (by including a fixed effect) showed that mothers whose children were taken into care had 3.23 times greater mortality rates, with greater hazard ratios for avoidable mortality (adjusted HR = 3.46) than unavoidable mortality (adjusted HR = 2.92). Tests of proportionality of hazard indicated that this increased risk is constant throughout the follow-up period (see Web Appendix 1, Web Figures 2–4). Most of the adjusted hazard ratios were larger than the unadjusted hazard ratios, indicating negative confounding. This means that failing to account for individual and family factors was underestimating the true association between having a child taken into care and mortality among mothers. Number of children taken into care The number of children taken into care may affect mortality rates. The final analysis examined all-cause mortality rates among mothers who had a child taken into care, comparing rates between mothers who had 1 child taken, 2 children taken, or 3 or more (range, 3–11) children taken into care. Among the 1,974 mothers with a child taken into care, 896 (45.4%) had 1 child taken, 453 (22.9%) had 2 children taken, and 625 (31.7%) had 3 or more children taken. Hazard ratios did not differ significantly between the 3 groups of mothers, suggesting that the number of children taken into care did not affect mortality rates among mothers with at least 1 child taken into care (Table 4). Table 4. All-Cause Mortality Rates and Results From Cox Regression Model for Mothers Who Had a Child Taken Into Care, According to Number of Children Taken Into Care, Manitoba, Canada, 1992–2015 No. of Children Taken Into Care No. of Deaths Rate per 10,000 Person-Years Crude Model Adjusteda Rate 95% CI HR 95% CI HR 95% CI 1 27 37 26, 54 1.00 Referent 1.00 Referent 2 20 50 32, 77 1.35 0.76, 2.42 1.18 0.65, 2.16 ≥3 18 29 18, 46 0.78 0.43, 1.42 0.73 0.40, 1.35 No. of Children Taken Into Care No. of Deaths Rate per 10,000 Person-Years Crude Model Adjusteda Rate 95% CI HR 95% CI HR 95% CI 1 27 37 26, 54 1.00 Referent 1.00 Referent 2 20 50 32, 77 1.35 0.76, 2.42 1.18 0.65, 2.16 ≥3 18 29 18, 46 0.78 0.43, 1.42 0.73 0.40, 1.35 Abbreviations: CI, confidence interval; HR, hazard ratio. a Adjusted for mother’s age, socioeconomic status, location of neighborhood at index date, and whether mother received welfare or moved in the 2 years before the index date. Table 4. All-Cause Mortality Rates and Results From Cox Regression Model for Mothers Who Had a Child Taken Into Care, According to Number of Children Taken Into Care, Manitoba, Canada, 1992–2015 No. of Children Taken Into Care No. of Deaths Rate per 10,000 Person-Years Crude Model Adjusteda Rate 95% CI HR 95% CI HR 95% CI 1 27 37 26, 54 1.00 Referent 1.00 Referent 2 20 50 32, 77 1.35 0.76, 2.42 1.18 0.65, 2.16 ≥3 18 29 18, 46 0.78 0.43, 1.42 0.73 0.40, 1.35 No. of Children Taken Into Care No. of Deaths Rate per 10,000 Person-Years Crude Model Adjusteda Rate 95% CI HR 95% CI HR 95% CI 1 27 37 26, 54 1.00 Referent 1.00 Referent 2 20 50 32, 77 1.35 0.76, 2.42 1.18 0.65, 2.16 ≥3 18 29 18, 46 0.78 0.43, 1.42 0.73 0.40, 1.35 Abbreviations: CI, confidence interval; HR, hazard ratio. a Adjusted for mother’s age, socioeconomic status, location of neighborhood at index date, and whether mother received welfare or moved in the 2 years before the index date. Comparison with mothers who experienced the death of a child An additional analysis was conducted to compare mortality rates of mothers who lost custody of a child with mothers who experienced the death of a child. This comparison was made because both of these groups of mothers have experienced a loss; however, the type of loss, and how that loss is publicly acknowledged, differs a great deal. We found that mothers whose children were taken into care had 2.71 times greater hazard ratios for avoidable mortality and significantly lower hazard ratios for unavoidable mortality (adjusted HR = 0.36) (See Web Appendix 2: Web Figure 5, Web Tables 3 and 4). DISCUSSION To our knowledge, this is the first study to examine mortality among mothers who had a child taken into care. This research identifies mothers who had a child taken into care as a portion of the population with much higher rates of unavoidable and avoidable mortality. Two-thirds of the excess deaths were due to avoidable causes. The higher rates of avoidable mortality indicate that their health attainment could potentially be influenced by health and social systems. These mothers have a great deal of contact with the health-care system and the social services system, however, the treatment and services received did not sufficiently address their health and social issues, leading to premature mortality. Possible reasons for the increased mortality rate among mothers whose children are taken into care are the health risks associated with identified coping mechanisms (e.g., substance use) and a deterioration of mental health after custody loss (2, 3). We do not take into account any changes to maternal health and well-being after the index date because these changes are on the causal pathway leading to the mortality outcome. Children who spent time in care have been found to have much higher rates of mortality, specifically avoidable mortality (23, 24). Other than work showing that mothers whose children were taken into care had higher rates of suicide attempts and completions, mortality rates for these mothers have not, to our knowledge, been examined (7). Previous research has found that mothers who have a child taken into care have higher rates of mental illness than do mothers dealing with the death of a child (6). Our study found higher rates of mortality for mothers who have a child taken into care (HR = 3.23) than a previous study examining mortality among mothers after the death of a child (HR = 1.43) (25). In our supplemental analysis, we saw that avoidable mortality was much higher (HR = 2.71) and unavoidable mortality was much lower (HR = 0.36) among mothers whose children were taken into care than among mothers who experienced the death of a child. The population-based repository at the Manitoba Centre for Health Policy (the Centre) has some significant strengths, such as a large sample size, minimal attrition, and potential for adjusting for a wide range of covariates. Mortality has been linked with both individual and family-level factors. The data housed the Centre facilitated the use of novel statistical techniques to account for differences seen both at the individual and at the family level. The rich longitudinal data allowed us to follow mothers whose children were born between 1992 and 2015, resulting in a sample of 1,974 families with discordant sisters (where one had a child taken into care and one did not). Comparisons of discordant siblings excludes confounding of having a child taken into care and the environmental factors shared by the sisters, and can also help rule out genetic confounding (9). The availability of information on all physician visits (diagnoses and tariff codes) and hospitalizations (diagnoses and procedures) in the 2 years before the mother’s index date, as well as information on important sociodemographic characteristics, allowed us to balance baseline covariates using hdPS adjustments. Using hdPS adjustments instead of standard covariate adjustment results in effect estimates closer to randomized control trial findings (26, 27). Combining the family-based approach with hdPS adjustments allowed for robust estimates of the risk of mortality among mothers. This study’s primary limitations concern the availability of variables. We do not know the specific reason that a child was taken into care—this reason (such as intimate partner violence or involvement with the criminal justice system) could also be associated with mortality. In this study, we used data about children born between 1992 and 2015. Information on child protection service involvement is available starting in 1992; however, not all agencies recorded case data in the provincial database in the early years. Data quality become more reliable starting in 1998. Given that we also do not know the reason a child was placed in care, it is possible that poor health led to both the child being placed in care and increased rates of mortality. This is unlikely to be the reason for our results because we adjusted for all hospitalizations and physician visits in the 2 years prior to the index date using hdPS. Additionally, if mothers had very poor health prior to their children being taken into care, we would expect to see elevated mortality shortly after the removal, with diminishing hazard ratios over time. We tested the proportional hazards assumption and found that this did not occur (Web Figures 2–4). Finally, these findings need to be replicated in other settings to ensure generalizability. This study provides evidence that mothers whose children are taken into care have greater rates of mortality, specifically avoidable mortality. This excess mortality is not only a health-care problem but also a societal problem. These mothers often have specific health challenges, such as intellectual disability and mental health conditions, requiring more specific public health interventions (28, 29). Current population-based strategies focus on cardiovascular risk and lifestyle factors, which may also be important for these mothers but do not address their specific health-care needs. Mothers involved with child protection services often face stigma; many have been accused of abuse or neglect and have not met society’s ideal of what constitutes good parenting (30). Public health interventions that provide more stability and address the unique health-care challenges of individuals (both mothers and children) involved with the child protection services could reduce rates of premature mortality. ACKNOWLEDGMENTS Author affiliations: Department of Community Health Sciences, Max Rady College of Medicine, University of Manitoba, Winnipeg, Manitoba, Canada (Elizabeth Wall-Wieler, Leslie L. Roos, Nathan C. Nickel, Dan Chateau, Marni Brownell); and Manitoba Centre for Health Policy, Winnipeg, Manitoba, Canada (Leslie L. Roos, Nathan C. Nickel, Dan Chateau, Marni Brownell). This work was supported by a Social Sciences and Humanities Research Council of Canada Joseph-Armand Bombardier Canada Doctoral Scholarship to E.W.-W., a Graduate Enhancement of Tri-Council Stipend to E.W.-W., and a Women’s Health Research Foundation of Canada Full Time Scholarship to E.W.-W. Data used in this study are from the Population Research Data Repository housed at the Manitoba Centre for Health Policy, University of Manitoba, and were derived from data provided by Manitoba Health, Seniors and Active Living, Manitoba Families, Vital Statistics (project 2016/2017-09). The results and conclusions are those of the authors, and no official endorsement by the Manitoba Centre for Health Policy, Manitoba Health, Seniors and Active Living, or other data providers is intended or should be inferred. Conflict of interest: none declared. Abbreviations IPT inverse probability of treatment PA posteroanterior CI confidence interval HR hazard ratio IPTW inverse probability of treatment weight REFERENCES 1 Berger LM , Paxson C , Waldfogel J . Mothers, men, and child protective services involvement . Child Maltreat . 2009 ; 14 ( 3 ): 263 – 276 . Google Scholar CrossRef Search ADS PubMed 2 Wall-Wieler E , Roos LL , Bolton J , et al. . Maternal health and social outcomes after having a child taken into care: population-based longitudinal cohort study using linkable administrative data . J Epidemiol Community Health . 2017 ; 71 ( 12 ): 1145 – 1151 . Google Scholar PubMed 3 Kenny KS , Barrington C , Green SL . “I felt for a long time like everything beautiful in me had been taken out”: women’s suffering, remembering, and survival following the loss of child custody . Int J Drug Policy . 2015 ; 26 ( 11 ): 1158 – 1166 . Google Scholar CrossRef Search ADS PubMed 4 Li J , Laursen TM , Precht DH , et al. . Hospitalization for mental illness among parents after the death of a child . N Engl J Med . 2005 ; 352 ( 12 ): 1190 – 1196 . Google Scholar CrossRef Search ADS PubMed 5 Espinosa J , Evans WN . Maternal bereavement: the heightened mortality of mothers after the death of a child . Econ Hum Biol . 2013 ; 11 ( 3 ): 371 – 381 . Google Scholar CrossRef Search ADS PubMed 6 Wall-Wieler E , Roos LL , Bolton J , et al. . Maternal mental health after custody loss and death of child: a retrospective cohort study using linkable administrative data [published online ahead of print October 29, 2017]. Can J Psychiatry . (doi: 10.1177/0706743717738494 ). 7 Wall-Wieler E , Roos LL , Brownell M , et al. . Suicide attempts and completions among mothers whose children were taken into care by child protection services: a cohort study using linkable administrative data . Can J Psychiatry . 2018 ; 63 ( 3 ): 170 – 177 . Google Scholar CrossRef Search ADS PubMed 8 Sørensen TI , Nielsen GG , Andersen PK , et al. . Genetic and environmental influences on premature death in adult adoptees . N Engl J Med . 1988 ; 318 ( 12 ): 727 – 732 . Google Scholar CrossRef Search ADS PubMed 9 D’Onofrio BM , Lahey BB , Turkheimer E , et al. . Critical need for family-based, quasi-experimental designs in integrating genetic and social science research . Am J Public Health . 2013 ; 103 ( suppl 1 ): S46 – S56 . Google Scholar CrossRef Search ADS PubMed 10 Canadian Institute for Health Information . Health Indicators 2012 Our Vision . Ottawa, ON : Canadian Institute for Health Infromation ; 2012 11 Kinge JM , Vallejo-Torres L , Morris S . Income related inequalities in avoidable mortality in Norway: a population-based study using data from 1994–2011 . Health Policy . 2015 ; 119 ( 7 ): 889 – 898 . Google Scholar CrossRef Search ADS PubMed 12 Tidemalm D , Waern M , Stefansson CG , et al. . Excess mortality in persons with severe mental disorder in Sweden: a cohort study of 12 103 individuals with and without contact with psychiatric services . Clin Pract Epidemiol Ment Health . 2008 ; 4 : 23 . Google Scholar PubMed 13 Wall-Wieler E , Roos LL , Brownell M , et al. . Predictors of having a first child taken into care at birth : a population-based retrospective cohort study . Child Abuse Negl . 2018 ; 76 : 1 – 9 . Google Scholar CrossRef Search ADS PubMed 14 Statistics Canada . Focus on Geography Series, 2011 Census. 2014 . http://www12.statcan.gc.ca/census-recensement/2011/as-sa/fogs-spg/Facts-pr-eng.cfm?Lang=Eng&GC=46. Accessed December 4, 2017. 15 Manitoba Health Seniors and Active Living . About the Manitoba Pharmacare Program. https://www.gov.mb.ca/health/pharmacare/index.html. Accessed March 15, 2017. 16 Gilbert R , Fluke J , O’Donnell M , et al. . Child maltreatment: variation in trends and policies in six developed countries . Lancet . 2012 ; 379 ( 9817 ): 758 – 772 . Google Scholar CrossRef Search ADS PubMed 17 Statistics Canada . Premature and potentially avoidable mortality, three-year average, Canada, provinces, territories, health regions, and peer groups (Table 102-4315). http://www5.statcan.gc.ca/cansim/a05?lang=eng&id=01024315. Accessed October 15, 2017. 18 Roos LL , Gupta S , Soodeen RA , et al. . Data quality in an information-rich environment: Canada as an example . Can J Aging . 2005 ; 24 ( suppl 1 ): 153 – 170 . Google Scholar CrossRef Search ADS PubMed 19 Manitoba Centre for Health Policy . Concept: ICD-10 to ICD-9-CM Conversion. 2015 . http://mchp-appserv.cpe.umanitoba.ca/viewConcept.php?conceptID=1157. Accessed September 24, 2016. 20 Pérez G , Rodríguez-Sanz M , Cirera E , et al. . Commentary: approaches, strengths, and limitations of avoidable mortality . J Public Health Policy . 2014 ; 35 ( 2 ): 171 – 184 . Google Scholar CrossRef Search ADS PubMed 21 Schneeweiss S , Eddings W , Glynn RJ , et al. . Variable selection for confounding adjustment in high-dimensional covariate spaces when analyzing healthcare databases . Epidemiology . 2017 ; 28 ( 2 ): 237 – 248 . Google Scholar CrossRef Search ADS PubMed 22 Harder VS , Stuart EA , Anthony JC . Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research . Psychol Methods . 2010 ; 15 ( 3 ): 234 – 249 . Google Scholar CrossRef Search ADS PubMed 23 Hjern A , Vinnerljung B , Lindblad F . Avoidable mortality among child welfare recipients and intercountry adoptees: a national cohort study . J Epidemiol Community Health . 2004 ; 58 ( 5 ): 412 – 417 . Google Scholar CrossRef Search ADS PubMed 24 Kalland M , Pensola TH , Meriläinen J , et al. . Mortality in children registered in the Finnish child welfare registry: population based study . BMJ . 2001 ; 323 ( 7306 ): 207 – 208 . Google Scholar CrossRef Search ADS PubMed 25 Li J , Precht DH , Mortensen PB , et al. . Mortality in parents after death of a child in Denmark: a nationwide follow-up study . Lancet . 2003 ; 361 ( 9355 ): 363 – 367 . Google Scholar CrossRef Search ADS PubMed 26 Schneeweiss S , Rassen JA , Glynn RJ , et al. . High-dimensional propensity score adjustment in studies of treatment effects using health care claims data . Epidemiology . 2009 ; 20 ( 4 ): 512 – 522 . Google Scholar CrossRef Search ADS PubMed 27 Guertin JR , Rahme E , Dormuth CR , et al. . Head to head comparison of the propensity score and the high-dimensional propensity score matching methods . BMC Med Res Methodol . 2016 ; 16 : 22 . Google Scholar CrossRef Search ADS PubMed 28 McConnell D , Llewellyn G . Stereotypes, parents with intellectual disability and child protection . J Soc Welf Fam Law . 2002 ; 24 ( 3 ): 297 – 317 . Google Scholar CrossRef Search ADS 29 Hoffman D , Rosenheck R . Homeless mothers with severe mental illnesses and their children: predictors of family reunification . Psychiatr Rehabil J . 2001 ; 25 ( 2 ): 163 – 169 . Google Scholar CrossRef Search ADS PubMed 30 McKegney S . Silenced Suffering: the Disenfranchised Grief of Birthmothers Compulsorily Seperated From Their Children [thesis]. Montreal, Canada: McGill University; 2003 . © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Journal of Epidemiology Oxford University Press

Mortality Among Mothers Whose Children Were Taken Into Care by Child Protection Services: A Discordant Sibling Analysis

Loading next page...
 
/lp/ou_press/mortality-among-mothers-whose-children-were-taken-into-care-by-child-iYclK5L4XR
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0002-9262
eISSN
1476-6256
D.O.I.
10.1093/aje/kwy062
Publisher site
See Article on Publisher Site

Abstract

Abstract This study examines whether mothers who had a child taken into care by child protection services have higher mortality rates compared with rates seen in their biological sisters who did not have a child taken into care. We conducted this retrospective cohort study using linkable administrative data from 3,948 mothers whose oldest child was born in Manitoba, Canada, between April 1, 1992, and March 31, 2015. These mothers were from 1,974 families in which one sister had a child taken into care and one sister did not. We computed rate differences and hazard ratios of all-cause, avoidable, and unavoidable mortality. There were an additional 24 deaths per 10,000 person-years among mothers who had had a child taken into care. Mothers who had a child taken into care had higher rates of mortality due to avoidable causes (hazard ratio = 3.46; 95% confidence interval: 1.41, 8.48) and unavoidable causes (hazard ratio = 2.92; 95% confidence interval: 1.01, 8.44). The number of children taken into care did not affect mortality rates among mothers with at least 1 child taken into care. The higher mortality rates—particularly avoidable mortality—among mothers who had a child taken into care indicate a need for more specific interventions for these mothers. avoidable mortality, child protection services, linkable administrative data, mortality among mothers, unavoidable mortality Editor’s note:An invited commentary on this article appears on page 1189. When children are taken into care by child protection services, the safety and well-being of the child are the highest priority. This process often overlooks the health and well-being of the mother. We focused on mothers in this work because most children are not living with their fathers when they go into care (1). Previous studies have found that mothers who had a child taken into care often have more health issues and social instability than mothers in the general population; these challenges worsen after their child is taken (2, 3). The distress that a mother faces after a different type of loss—the death of a child—is publicly acknowledged and has been linked with many health consequences, such as increased mental illness and heightened mortality (4, 5). Recent findings indicate that mothers who had lost custody of a child to child protection services have higher rates of mental illness following separation from their child than mothers who experienced the death of a child (6). While mothers who had a child taken into care have higher rates of suicide attempts and completions, it is not known whether there is a higher rate of mortality among mothers from other causes after losing custody of a child (7). This study aims to fill this knowledge gap by examining all-cause mortality rates among mothers who had a child taken into care, and comparing these rates with those of their sisters who were also mothers but did not have a child taken. We compared mortality rates in discordant sisters because mortality has been found to cluster in families due to environmental and genetic factors; discordant sibling analyses account for these stable family characteristics (8, 9). Mortality rates are often classified into avoidable causes of mortality and unavoidable causes; avoidable mortality is a death from a cause that could have been prevented or treated (10). Avoidable mortality include deaths due to infections, certain cancers, and injuries (unintentional or intentional). Avoidable mortality rates are particularly high among individuals from low socioeconomic backgrounds and those with mental disorders; mothers involved with child protection services often have one or both of these characteristics (11–13). Thus, we also examined whether mothers with a child taken into care had higher rates of avoidable and unavoidable mortality than did mothers not having a child taken into care. METHODS Setting and data Data for this study came from Manitoba, Canada, which is a central Canadian province with approximately 1.2 million residents (14). Manitoba residents receive universal healthcare coverage (physician visits and hospitalizations) and financial assistance for pharmaceuticals if high prescription-drug costs seriously affect their income (15). Rates of children in care in Manitoba are among the highest in the world, with approximately 3% of children in care (16). Manitoba has among the highest rates of premature mortality among Canadian provinces, with 246.3 deaths per 100,000 people in 2008 (17). This is higher than the premature mortality rate throughout Canada (203.9 per 100,000) and in the United Kingdom (225 per 100,000) but lower than in the United States (271 per 100,000) (10, 17). This study used the linkable administrative data in the Population Research Data Repository housed at the Manitoba Centre for Health Policy. The population registry is linked at the individual level with physician claims, hospital discharge abstracts, pharmaceutical claims, Child and Family Services case reports, Employment and Income Assistance case reports, vital statistics, and the Canadian Census. A scrambled personal health number was used to link these deidentified data sets. Information on linkage methods, confidentiality/privacy, and validity can be found elsewhere (18). Cohort formation The cohort consisted of biological sisters whose first child was born between April 1, 1992, and March 31, 2015. Sisters were defined as having the same biological mother. This cohort was divided into two groups. Group 1 consisted of mothers whose children were taken into care before March 31, 2015. These mothers were identified through the child protection case files; child protection is delivered through Child and Family Services in Manitoba. The index date for these mothers was selected as the date that a child was first taken into care. Group 2 included mothers not having a child taken into care before March 31, 2015, and with a biological sister in group 1. The oldest child was selected for mothers in group 2 as the index child, and her index date was defined as the date the index child turned the age her sister’s child had been when that child was taken into care. Mothers who did not live in Manitoba for at least 2 years before the index date were excluded. Each mother was followed from the index date until the end of coverage (date of death or move out of province) or December 31, 2015, whichever came first. For families with more than 1 sister in group 1 or group 2, 1 sister was randomly selected. Figure 1 details the cohort selection process. Figure 1. View largeDownload slide Cohort formation for a study of mortality among mothers with children who were taken into protective care in Manitoba, Canada, 1992–2015. Figure 1. View largeDownload slide Cohort formation for a study of mortality among mothers with children who were taken into protective care in Manitoba, Canada, 1992–2015. Mortality among mothers Date of death and underlying cause of death were obtained from the Vital Statistics data set. Cause of death is coded using the International Classification of Diseases (ICD), Ninth Revision, for deaths before 2000 and Tenth Revision for deaths in 2000 or later. The ICD-9 codes for primary cause of death for deaths occurring before 2000 were converted to ICD-10 codes (19). We examined all-cause mortality (any death between April 1, 1992, and December 31, 2015) and avoidable and unavoidable mortality. Avoidable mortality generally refers to deaths that could have been avoided through proper medical care, through intersectoral health policy interventions, or through public health programs (20). A list of causes of death for avoidable mortality created by the Canadian Institute for Health Information was used to identify avoidable mortality (see Web Table 1, available at https://academic.oup.com/aje); causes of deaths not included in this list are noted as unavoidable mortality (10). Statistical analysis Mortality rates (overall, avoidable, and unavoidable) per 10,000 person-years were first examined for each group of mothers, and rate differences are presented. Next, we investigated differences in mortality rates using crude and adjusted Cox proportional hazards regression models. Mothers who have had a child taken into care differ from those not having a child taken; these individual and family-level differences may contribute to mortality among mothers. To account for differences, the Cox proportional hazards regression models adjusted in 3 ways. First, individual-level differences (health and sociodemographic) were balanced for using inverse probability of treatment weights. Second, stable family covariates (such as genetics and environment) were accounted for by including a family fixed effect in the regression model. Finally, both individual and family-level variables were accounted for using an inverse probability of treatment–weighted fixed-effects model. Inverse probability of treatment weights were obtained from high-dimensional propensity scores (hdPSs). Unlike traditional propensity score methods, which are limited to covariates specified by the investigator, the multistep algorithm used to derive hdPSs identifies potential confounders from a database by selecting variables correlated to both the exposure and outcome, prioritizing covariates by prevalence and potential for bias (21). Covariates drawn from the 2 years before the index date were based on: 1) medical service tariff codes; 2) physician diagnostic codes; 3) hospital procedure codes; 4) hospital diagnostic codes. These data describe the health status of the women in our study (21). In our cohort, more than 97% of mothers had contact with a primary care physician in the 2 years before the index date. In addition to the health variables obtained from these data, a series of sociodemographic variables were included in the models. In the 2 years before the index date, we noted whether the mother had moved and whether the mother had received Employment and Income Assistance (analogous to welfare). We also included the mother’s age, neighborhood socioeconomic status, and neighborhood location (urban/rural) at the index date. The 4 data sets used to assess the health status in the 2 years before the index date are extremely large. For example, physician diagnoses are claimed as ICD-9-CM (clinical modification) codes; these include approximately 17,000 different diagnostic codes. The hdPSs were constructed including all sociodemographic variables and the top 500 health covariates. The unweighted and weighted standardized differences for the top 20 covariates are presented in Table 1; we considered a covariate to be balanced if the standardized difference was less than 0.25 (22). Information on all 500 covariates included in the hdPS is available on request. The kernel density plot of the distribution of propensity scores showed sufficient common support for this analysis (see Web Figure 1). Table 1. Comparison of Baseline Characteristics of Mothers in a Study of Maternal Mortality, Manitoba, Canada, 1992–2015 Maternal Characteristic Had a Child Taken Into Care (n = 1,974) Did Not Have a Child Taken Into Care (n = 1,974) Unweighted Standardized Differences IPT Weighted Standardized Differences No. of Mothers % No. of Mothers % At the index date  Age category of mother, years   <20 547 27.71 452 22.90 0.111 0.082   20–29 1,019 51.62 1,116 56.53 0.099 0.178   ≥30 408 20.67 406 20.57 0.003 0.131  Socioeconomic status of neighborhood 1.17 (1.13)a 1.12 (1.23)a 0.045 0.014  Urban neighborhood 1,011 51.22 896 45.39 0.117 0.096 Event in the 2 years before the index date  Changed residence 954 48.33 767 38.86 0.192 0.011  Received welfare 1,076 54.51 683 34.60 0.409 0.045 Health in the 2 years before the index date  Office visits, complete history, and physical examination 373 18.90 212 10.74 0.231 0.006  Hospital care, subsequent care, first and second week 57 2.89 26 1.32 0.110 0.062  Depressive disorder 172 8.71 66 3.34 0.227 0.080  Disorders of function of stomach 38 1.93 13 0.66 0.112 0.035  Wounds, simple repair any location single 87 4.41 32 1.62 0.164 0.008  Psychotherapy by nonpsychiatrist 70 3.55 23 1.17 0.157 0.182  Neurotic disorders 155 7.85 87 4.41 0.144 0.121  Radiology, chest, PA, and lateral 85 4.31 52 2.63 0.091 0.038  Drug dependence 56 2.84 11 0.56 0.177 0.096  Other venereal diseases 270 13.68 155 7.85 0.189 0.028  Office visits, regional of subsequent visit or well-baby care 417 21.12 271 13.73 0.196 0.018  Nondependent abuse of drugs 77 3.90 26 1.32 0.163 0.024  Electrocardiogram 85 4.31 37 1.87 0.141 0.041 Maternal Characteristic Had a Child Taken Into Care (n = 1,974) Did Not Have a Child Taken Into Care (n = 1,974) Unweighted Standardized Differences IPT Weighted Standardized Differences No. of Mothers % No. of Mothers % At the index date  Age category of mother, years   <20 547 27.71 452 22.90 0.111 0.082   20–29 1,019 51.62 1,116 56.53 0.099 0.178   ≥30 408 20.67 406 20.57 0.003 0.131  Socioeconomic status of neighborhood 1.17 (1.13)a 1.12 (1.23)a 0.045 0.014  Urban neighborhood 1,011 51.22 896 45.39 0.117 0.096 Event in the 2 years before the index date  Changed residence 954 48.33 767 38.86 0.192 0.011  Received welfare 1,076 54.51 683 34.60 0.409 0.045 Health in the 2 years before the index date  Office visits, complete history, and physical examination 373 18.90 212 10.74 0.231 0.006  Hospital care, subsequent care, first and second week 57 2.89 26 1.32 0.110 0.062  Depressive disorder 172 8.71 66 3.34 0.227 0.080  Disorders of function of stomach 38 1.93 13 0.66 0.112 0.035  Wounds, simple repair any location single 87 4.41 32 1.62 0.164 0.008  Psychotherapy by nonpsychiatrist 70 3.55 23 1.17 0.157 0.182  Neurotic disorders 155 7.85 87 4.41 0.144 0.121  Radiology, chest, PA, and lateral 85 4.31 52 2.63 0.091 0.038  Drug dependence 56 2.84 11 0.56 0.177 0.096  Other venereal diseases 270 13.68 155 7.85 0.189 0.028  Office visits, regional of subsequent visit or well-baby care 417 21.12 271 13.73 0.196 0.018  Nondependent abuse of drugs 77 3.90 26 1.32 0.163 0.024  Electrocardiogram 85 4.31 37 1.87 0.141 0.041 Abbreviations: IPT, inverse probability of treatment; PA, posteroanterior. a Values are expressed as mean (standard deviation). Table 1. Comparison of Baseline Characteristics of Mothers in a Study of Maternal Mortality, Manitoba, Canada, 1992–2015 Maternal Characteristic Had a Child Taken Into Care (n = 1,974) Did Not Have a Child Taken Into Care (n = 1,974) Unweighted Standardized Differences IPT Weighted Standardized Differences No. of Mothers % No. of Mothers % At the index date  Age category of mother, years   <20 547 27.71 452 22.90 0.111 0.082   20–29 1,019 51.62 1,116 56.53 0.099 0.178   ≥30 408 20.67 406 20.57 0.003 0.131  Socioeconomic status of neighborhood 1.17 (1.13)a 1.12 (1.23)a 0.045 0.014  Urban neighborhood 1,011 51.22 896 45.39 0.117 0.096 Event in the 2 years before the index date  Changed residence 954 48.33 767 38.86 0.192 0.011  Received welfare 1,076 54.51 683 34.60 0.409 0.045 Health in the 2 years before the index date  Office visits, complete history, and physical examination 373 18.90 212 10.74 0.231 0.006  Hospital care, subsequent care, first and second week 57 2.89 26 1.32 0.110 0.062  Depressive disorder 172 8.71 66 3.34 0.227 0.080  Disorders of function of stomach 38 1.93 13 0.66 0.112 0.035  Wounds, simple repair any location single 87 4.41 32 1.62 0.164 0.008  Psychotherapy by nonpsychiatrist 70 3.55 23 1.17 0.157 0.182  Neurotic disorders 155 7.85 87 4.41 0.144 0.121  Radiology, chest, PA, and lateral 85 4.31 52 2.63 0.091 0.038  Drug dependence 56 2.84 11 0.56 0.177 0.096  Other venereal diseases 270 13.68 155 7.85 0.189 0.028  Office visits, regional of subsequent visit or well-baby care 417 21.12 271 13.73 0.196 0.018  Nondependent abuse of drugs 77 3.90 26 1.32 0.163 0.024  Electrocardiogram 85 4.31 37 1.87 0.141 0.041 Maternal Characteristic Had a Child Taken Into Care (n = 1,974) Did Not Have a Child Taken Into Care (n = 1,974) Unweighted Standardized Differences IPT Weighted Standardized Differences No. of Mothers % No. of Mothers % At the index date  Age category of mother, years   <20 547 27.71 452 22.90 0.111 0.082   20–29 1,019 51.62 1,116 56.53 0.099 0.178   ≥30 408 20.67 406 20.57 0.003 0.131  Socioeconomic status of neighborhood 1.17 (1.13)a 1.12 (1.23)a 0.045 0.014  Urban neighborhood 1,011 51.22 896 45.39 0.117 0.096 Event in the 2 years before the index date  Changed residence 954 48.33 767 38.86 0.192 0.011  Received welfare 1,076 54.51 683 34.60 0.409 0.045 Health in the 2 years before the index date  Office visits, complete history, and physical examination 373 18.90 212 10.74 0.231 0.006  Hospital care, subsequent care, first and second week 57 2.89 26 1.32 0.110 0.062  Depressive disorder 172 8.71 66 3.34 0.227 0.080  Disorders of function of stomach 38 1.93 13 0.66 0.112 0.035  Wounds, simple repair any location single 87 4.41 32 1.62 0.164 0.008  Psychotherapy by nonpsychiatrist 70 3.55 23 1.17 0.157 0.182  Neurotic disorders 155 7.85 87 4.41 0.144 0.121  Radiology, chest, PA, and lateral 85 4.31 52 2.63 0.091 0.038  Drug dependence 56 2.84 11 0.56 0.177 0.096  Other venereal diseases 270 13.68 155 7.85 0.189 0.028  Office visits, regional of subsequent visit or well-baby care 417 21.12 271 13.73 0.196 0.018  Nondependent abuse of drugs 77 3.90 26 1.32 0.163 0.024  Electrocardiogram 85 4.31 37 1.87 0.141 0.041 Abbreviations: IPT, inverse probability of treatment; PA, posteroanterior. a Values are expressed as mean (standard deviation). Two sensitivity analyses were conducted. The first examined mortality rates among mothers who had a child taken into care, comparing rates among mothers who had 1, 2, or 3 or more children taken into care. The second compared mortality rates between mothers who had a child taken into care and mothers who experienced the death of a child. Both sensitivity analyses adjusted for mother’s age, neighborhood location and socioeconomic status at the index date, and whether she received welfare or moved in the 2 years before the index date. All data management, programming, and analyses were performed using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina). RESULTS Table 1 displays the sociodemographic covariates and the top health covariates of mothers in each group. Mothers whose children were taken into care were more likely to have received welfare and to have worse health in the 2 years before the index date than their sisters who did not have a child taken. Mothers in the cohort were followed for a total of 33,624 person-years; mothers in group 1 were followed for 17,487 person-years (average of 8.9 years per person), and group 2 mothers for 16,137 person-years (average of 8.2 years per person). A total of 86 deaths occurred in the cohort before December 31, 2015. There were 24 additional deaths per 10,000 person-years among mothers who had a child taken into care (Table 2). Of the mothers who died in the follow-up period, those having a child taken into care were aged from 18 to 52 years (average = 30.8 years) at the time of death, and mothers who did not have a child taken into care were aged from 19 to 54 years (average = 31.3 years) at the time of death. For both mothers whose children were taken into care and those whose children were not taken into care, the most common cause of death was unintentional injury; this cause of death accounted for approximately 25% of deaths (see Web Table 2). Table 2. Mortality Rates and Mortality Rate Difference Among Mothers, Manitoba, Canada, 1992–2015 Mortality Had a Child Taken Into Care Did Not Have a Child Taken Into Care Rate Difference per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years Rate 95% CI Rate 95% CI Rate Difference 95% CI All-cause 65 37 29, 47 21 13 9, 20 24 14, 35 Avoidable 45 26 19, 35 15 9 6, 15 16 8, 25 Unavoidable 20 11 7, 18 6 4 2, 8 8 2, 14 Mortality Had a Child Taken Into Care Did Not Have a Child Taken Into Care Rate Difference per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years Rate 95% CI Rate 95% CI Rate Difference 95% CI All-cause 65 37 29, 47 21 13 9, 20 24 14, 35 Avoidable 45 26 19, 35 15 9 6, 15 16 8, 25 Unavoidable 20 11 7, 18 6 4 2, 8 8 2, 14 Abbreviation: CI, confidence interval. Table 2. Mortality Rates and Mortality Rate Difference Among Mothers, Manitoba, Canada, 1992–2015 Mortality Had a Child Taken Into Care Did Not Have a Child Taken Into Care Rate Difference per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years Rate 95% CI Rate 95% CI Rate Difference 95% CI All-cause 65 37 29, 47 21 13 9, 20 24 14, 35 Avoidable 45 26 19, 35 15 9 6, 15 16 8, 25 Unavoidable 20 11 7, 18 6 4 2, 8 8 2, 14 Mortality Had a Child Taken Into Care Did Not Have a Child Taken Into Care Rate Difference per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years No. of Deaths Rate per 10,000 Person-Years Rate 95% CI Rate 95% CI Rate Difference 95% CI All-cause 65 37 29, 47 21 13 9, 20 24 14, 35 Avoidable 45 26 19, 35 15 9 6, 15 16 8, 25 Unavoidable 20 11 7, 18 6 4 2, 8 8 2, 14 Abbreviation: CI, confidence interval. The unadjusted hazard ratios were almost 3 times greater among mothers who had a child taken into care (hazard ratio (HR) = 2.90) (Table 3). For both avoidable and unavoidable causes, mothers who had a child taken into care were about 3 times more likely to die during the follow-up period (for avoidable causes, HR = 2.82, 95% confidence interval: 1.57, 5.06; for unavoidable causes, HR = 3.08, 95% confidence interval: 1.24, 7.68). Table 3. Unadjusted and Adjusted Risk of Mortality for Mothers Who Had a Child Taken Into Care, Manitoba, Canada, 1992-2015 Mortality Crude Model IPTW-Adjusted Crude Model With Family Fixed Effect IPTW-Adjusted With Family Fixed Effect HR 95% CI HR 95% CI HR 95% CI HR 95% CI All-cause 2.90 1.77, 4.74 3.04 1.80, 5.14 3.46 1.87, 6.42 3.23 1.62, 6.41 Avoidable 2.82 1.57, 5.06 2.99 1.62, 5.52 3.86 1.68, 8.86 3.46 1.41, 8.48 Unavoidable 3.08 1.24, 7.68 3.19 1.16, 8.78 3.00 1.19, 7.56 2.92 1.01, 8.44 Mortality Crude Model IPTW-Adjusted Crude Model With Family Fixed Effect IPTW-Adjusted With Family Fixed Effect HR 95% CI HR 95% CI HR 95% CI HR 95% CI All-cause 2.90 1.77, 4.74 3.04 1.80, 5.14 3.46 1.87, 6.42 3.23 1.62, 6.41 Avoidable 2.82 1.57, 5.06 2.99 1.62, 5.52 3.86 1.68, 8.86 3.46 1.41, 8.48 Unavoidable 3.08 1.24, 7.68 3.19 1.16, 8.78 3.00 1.19, 7.56 2.92 1.01, 8.44 Abbreviations: CI, confidence interval; HR, hazard ratio; IPTW, inverse probability of treatment weight. Table 3. Unadjusted and Adjusted Risk of Mortality for Mothers Who Had a Child Taken Into Care, Manitoba, Canada, 1992-2015 Mortality Crude Model IPTW-Adjusted Crude Model With Family Fixed Effect IPTW-Adjusted With Family Fixed Effect HR 95% CI HR 95% CI HR 95% CI HR 95% CI All-cause 2.90 1.77, 4.74 3.04 1.80, 5.14 3.46 1.87, 6.42 3.23 1.62, 6.41 Avoidable 2.82 1.57, 5.06 2.99 1.62, 5.52 3.86 1.68, 8.86 3.46 1.41, 8.48 Unavoidable 3.08 1.24, 7.68 3.19 1.16, 8.78 3.00 1.19, 7.56 2.92 1.01, 8.44 Mortality Crude Model IPTW-Adjusted Crude Model With Family Fixed Effect IPTW-Adjusted With Family Fixed Effect HR 95% CI HR 95% CI HR 95% CI HR 95% CI All-cause 2.90 1.77, 4.74 3.04 1.80, 5.14 3.46 1.87, 6.42 3.23 1.62, 6.41 Avoidable 2.82 1.57, 5.06 2.99 1.62, 5.52 3.86 1.68, 8.86 3.46 1.41, 8.48 Unavoidable 3.08 1.24, 7.68 3.19 1.16, 8.78 3.00 1.19, 7.56 2.92 1.01, 8.44 Abbreviations: CI, confidence interval; HR, hazard ratio; IPTW, inverse probability of treatment weight. When accounting for individual differences (the inverse probability of treatment weight–adjusted HRs), mortality rates among mothers remained similar to those seen in the crude analysis (Table 3). Accounting for stable family characteristics (such as genetics, ethnicity, and environment), mothers whose children were taken into care were 3.46 times more likely to die in the follow-up period, with avoidable mortality rates 3.86 times greater for mothers in this group. The final adjustment, which accounted for both individual differences (using inverse probability of treatment weights) and family characteristics (by including a fixed effect) showed that mothers whose children were taken into care had 3.23 times greater mortality rates, with greater hazard ratios for avoidable mortality (adjusted HR = 3.46) than unavoidable mortality (adjusted HR = 2.92). Tests of proportionality of hazard indicated that this increased risk is constant throughout the follow-up period (see Web Appendix 1, Web Figures 2–4). Most of the adjusted hazard ratios were larger than the unadjusted hazard ratios, indicating negative confounding. This means that failing to account for individual and family factors was underestimating the true association between having a child taken into care and mortality among mothers. Number of children taken into care The number of children taken into care may affect mortality rates. The final analysis examined all-cause mortality rates among mothers who had a child taken into care, comparing rates between mothers who had 1 child taken, 2 children taken, or 3 or more (range, 3–11) children taken into care. Among the 1,974 mothers with a child taken into care, 896 (45.4%) had 1 child taken, 453 (22.9%) had 2 children taken, and 625 (31.7%) had 3 or more children taken. Hazard ratios did not differ significantly between the 3 groups of mothers, suggesting that the number of children taken into care did not affect mortality rates among mothers with at least 1 child taken into care (Table 4). Table 4. All-Cause Mortality Rates and Results From Cox Regression Model for Mothers Who Had a Child Taken Into Care, According to Number of Children Taken Into Care, Manitoba, Canada, 1992–2015 No. of Children Taken Into Care No. of Deaths Rate per 10,000 Person-Years Crude Model Adjusteda Rate 95% CI HR 95% CI HR 95% CI 1 27 37 26, 54 1.00 Referent 1.00 Referent 2 20 50 32, 77 1.35 0.76, 2.42 1.18 0.65, 2.16 ≥3 18 29 18, 46 0.78 0.43, 1.42 0.73 0.40, 1.35 No. of Children Taken Into Care No. of Deaths Rate per 10,000 Person-Years Crude Model Adjusteda Rate 95% CI HR 95% CI HR 95% CI 1 27 37 26, 54 1.00 Referent 1.00 Referent 2 20 50 32, 77 1.35 0.76, 2.42 1.18 0.65, 2.16 ≥3 18 29 18, 46 0.78 0.43, 1.42 0.73 0.40, 1.35 Abbreviations: CI, confidence interval; HR, hazard ratio. a Adjusted for mother’s age, socioeconomic status, location of neighborhood at index date, and whether mother received welfare or moved in the 2 years before the index date. Table 4. All-Cause Mortality Rates and Results From Cox Regression Model for Mothers Who Had a Child Taken Into Care, According to Number of Children Taken Into Care, Manitoba, Canada, 1992–2015 No. of Children Taken Into Care No. of Deaths Rate per 10,000 Person-Years Crude Model Adjusteda Rate 95% CI HR 95% CI HR 95% CI 1 27 37 26, 54 1.00 Referent 1.00 Referent 2 20 50 32, 77 1.35 0.76, 2.42 1.18 0.65, 2.16 ≥3 18 29 18, 46 0.78 0.43, 1.42 0.73 0.40, 1.35 No. of Children Taken Into Care No. of Deaths Rate per 10,000 Person-Years Crude Model Adjusteda Rate 95% CI HR 95% CI HR 95% CI 1 27 37 26, 54 1.00 Referent 1.00 Referent 2 20 50 32, 77 1.35 0.76, 2.42 1.18 0.65, 2.16 ≥3 18 29 18, 46 0.78 0.43, 1.42 0.73 0.40, 1.35 Abbreviations: CI, confidence interval; HR, hazard ratio. a Adjusted for mother’s age, socioeconomic status, location of neighborhood at index date, and whether mother received welfare or moved in the 2 years before the index date. Comparison with mothers who experienced the death of a child An additional analysis was conducted to compare mortality rates of mothers who lost custody of a child with mothers who experienced the death of a child. This comparison was made because both of these groups of mothers have experienced a loss; however, the type of loss, and how that loss is publicly acknowledged, differs a great deal. We found that mothers whose children were taken into care had 2.71 times greater hazard ratios for avoidable mortality and significantly lower hazard ratios for unavoidable mortality (adjusted HR = 0.36) (See Web Appendix 2: Web Figure 5, Web Tables 3 and 4). DISCUSSION To our knowledge, this is the first study to examine mortality among mothers who had a child taken into care. This research identifies mothers who had a child taken into care as a portion of the population with much higher rates of unavoidable and avoidable mortality. Two-thirds of the excess deaths were due to avoidable causes. The higher rates of avoidable mortality indicate that their health attainment could potentially be influenced by health and social systems. These mothers have a great deal of contact with the health-care system and the social services system, however, the treatment and services received did not sufficiently address their health and social issues, leading to premature mortality. Possible reasons for the increased mortality rate among mothers whose children are taken into care are the health risks associated with identified coping mechanisms (e.g., substance use) and a deterioration of mental health after custody loss (2, 3). We do not take into account any changes to maternal health and well-being after the index date because these changes are on the causal pathway leading to the mortality outcome. Children who spent time in care have been found to have much higher rates of mortality, specifically avoidable mortality (23, 24). Other than work showing that mothers whose children were taken into care had higher rates of suicide attempts and completions, mortality rates for these mothers have not, to our knowledge, been examined (7). Previous research has found that mothers who have a child taken into care have higher rates of mental illness than do mothers dealing with the death of a child (6). Our study found higher rates of mortality for mothers who have a child taken into care (HR = 3.23) than a previous study examining mortality among mothers after the death of a child (HR = 1.43) (25). In our supplemental analysis, we saw that avoidable mortality was much higher (HR = 2.71) and unavoidable mortality was much lower (HR = 0.36) among mothers whose children were taken into care than among mothers who experienced the death of a child. The population-based repository at the Manitoba Centre for Health Policy (the Centre) has some significant strengths, such as a large sample size, minimal attrition, and potential for adjusting for a wide range of covariates. Mortality has been linked with both individual and family-level factors. The data housed the Centre facilitated the use of novel statistical techniques to account for differences seen both at the individual and at the family level. The rich longitudinal data allowed us to follow mothers whose children were born between 1992 and 2015, resulting in a sample of 1,974 families with discordant sisters (where one had a child taken into care and one did not). Comparisons of discordant siblings excludes confounding of having a child taken into care and the environmental factors shared by the sisters, and can also help rule out genetic confounding (9). The availability of information on all physician visits (diagnoses and tariff codes) and hospitalizations (diagnoses and procedures) in the 2 years before the mother’s index date, as well as information on important sociodemographic characteristics, allowed us to balance baseline covariates using hdPS adjustments. Using hdPS adjustments instead of standard covariate adjustment results in effect estimates closer to randomized control trial findings (26, 27). Combining the family-based approach with hdPS adjustments allowed for robust estimates of the risk of mortality among mothers. This study’s primary limitations concern the availability of variables. We do not know the specific reason that a child was taken into care—this reason (such as intimate partner violence or involvement with the criminal justice system) could also be associated with mortality. In this study, we used data about children born between 1992 and 2015. Information on child protection service involvement is available starting in 1992; however, not all agencies recorded case data in the provincial database in the early years. Data quality become more reliable starting in 1998. Given that we also do not know the reason a child was placed in care, it is possible that poor health led to both the child being placed in care and increased rates of mortality. This is unlikely to be the reason for our results because we adjusted for all hospitalizations and physician visits in the 2 years prior to the index date using hdPS. Additionally, if mothers had very poor health prior to their children being taken into care, we would expect to see elevated mortality shortly after the removal, with diminishing hazard ratios over time. We tested the proportional hazards assumption and found that this did not occur (Web Figures 2–4). Finally, these findings need to be replicated in other settings to ensure generalizability. This study provides evidence that mothers whose children are taken into care have greater rates of mortality, specifically avoidable mortality. This excess mortality is not only a health-care problem but also a societal problem. These mothers often have specific health challenges, such as intellectual disability and mental health conditions, requiring more specific public health interventions (28, 29). Current population-based strategies focus on cardiovascular risk and lifestyle factors, which may also be important for these mothers but do not address their specific health-care needs. Mothers involved with child protection services often face stigma; many have been accused of abuse or neglect and have not met society’s ideal of what constitutes good parenting (30). Public health interventions that provide more stability and address the unique health-care challenges of individuals (both mothers and children) involved with the child protection services could reduce rates of premature mortality. ACKNOWLEDGMENTS Author affiliations: Department of Community Health Sciences, Max Rady College of Medicine, University of Manitoba, Winnipeg, Manitoba, Canada (Elizabeth Wall-Wieler, Leslie L. Roos, Nathan C. Nickel, Dan Chateau, Marni Brownell); and Manitoba Centre for Health Policy, Winnipeg, Manitoba, Canada (Leslie L. Roos, Nathan C. Nickel, Dan Chateau, Marni Brownell). This work was supported by a Social Sciences and Humanities Research Council of Canada Joseph-Armand Bombardier Canada Doctoral Scholarship to E.W.-W., a Graduate Enhancement of Tri-Council Stipend to E.W.-W., and a Women’s Health Research Foundation of Canada Full Time Scholarship to E.W.-W. Data used in this study are from the Population Research Data Repository housed at the Manitoba Centre for Health Policy, University of Manitoba, and were derived from data provided by Manitoba Health, Seniors and Active Living, Manitoba Families, Vital Statistics (project 2016/2017-09). The results and conclusions are those of the authors, and no official endorsement by the Manitoba Centre for Health Policy, Manitoba Health, Seniors and Active Living, or other data providers is intended or should be inferred. Conflict of interest: none declared. Abbreviations IPT inverse probability of treatment PA posteroanterior CI confidence interval HR hazard ratio IPTW inverse probability of treatment weight REFERENCES 1 Berger LM , Paxson C , Waldfogel J . Mothers, men, and child protective services involvement . Child Maltreat . 2009 ; 14 ( 3 ): 263 – 276 . Google Scholar CrossRef Search ADS PubMed 2 Wall-Wieler E , Roos LL , Bolton J , et al. . Maternal health and social outcomes after having a child taken into care: population-based longitudinal cohort study using linkable administrative data . J Epidemiol Community Health . 2017 ; 71 ( 12 ): 1145 – 1151 . Google Scholar PubMed 3 Kenny KS , Barrington C , Green SL . “I felt for a long time like everything beautiful in me had been taken out”: women’s suffering, remembering, and survival following the loss of child custody . Int J Drug Policy . 2015 ; 26 ( 11 ): 1158 – 1166 . Google Scholar CrossRef Search ADS PubMed 4 Li J , Laursen TM , Precht DH , et al. . Hospitalization for mental illness among parents after the death of a child . N Engl J Med . 2005 ; 352 ( 12 ): 1190 – 1196 . Google Scholar CrossRef Search ADS PubMed 5 Espinosa J , Evans WN . Maternal bereavement: the heightened mortality of mothers after the death of a child . Econ Hum Biol . 2013 ; 11 ( 3 ): 371 – 381 . Google Scholar CrossRef Search ADS PubMed 6 Wall-Wieler E , Roos LL , Bolton J , et al. . Maternal mental health after custody loss and death of child: a retrospective cohort study using linkable administrative data [published online ahead of print October 29, 2017]. Can J Psychiatry . (doi: 10.1177/0706743717738494 ). 7 Wall-Wieler E , Roos LL , Brownell M , et al. . Suicide attempts and completions among mothers whose children were taken into care by child protection services: a cohort study using linkable administrative data . Can J Psychiatry . 2018 ; 63 ( 3 ): 170 – 177 . Google Scholar CrossRef Search ADS PubMed 8 Sørensen TI , Nielsen GG , Andersen PK , et al. . Genetic and environmental influences on premature death in adult adoptees . N Engl J Med . 1988 ; 318 ( 12 ): 727 – 732 . Google Scholar CrossRef Search ADS PubMed 9 D’Onofrio BM , Lahey BB , Turkheimer E , et al. . Critical need for family-based, quasi-experimental designs in integrating genetic and social science research . Am J Public Health . 2013 ; 103 ( suppl 1 ): S46 – S56 . Google Scholar CrossRef Search ADS PubMed 10 Canadian Institute for Health Information . Health Indicators 2012 Our Vision . Ottawa, ON : Canadian Institute for Health Infromation ; 2012 11 Kinge JM , Vallejo-Torres L , Morris S . Income related inequalities in avoidable mortality in Norway: a population-based study using data from 1994–2011 . Health Policy . 2015 ; 119 ( 7 ): 889 – 898 . Google Scholar CrossRef Search ADS PubMed 12 Tidemalm D , Waern M , Stefansson CG , et al. . Excess mortality in persons with severe mental disorder in Sweden: a cohort study of 12 103 individuals with and without contact with psychiatric services . Clin Pract Epidemiol Ment Health . 2008 ; 4 : 23 . Google Scholar PubMed 13 Wall-Wieler E , Roos LL , Brownell M , et al. . Predictors of having a first child taken into care at birth : a population-based retrospective cohort study . Child Abuse Negl . 2018 ; 76 : 1 – 9 . Google Scholar CrossRef Search ADS PubMed 14 Statistics Canada . Focus on Geography Series, 2011 Census. 2014 . http://www12.statcan.gc.ca/census-recensement/2011/as-sa/fogs-spg/Facts-pr-eng.cfm?Lang=Eng&GC=46. Accessed December 4, 2017. 15 Manitoba Health Seniors and Active Living . About the Manitoba Pharmacare Program. https://www.gov.mb.ca/health/pharmacare/index.html. Accessed March 15, 2017. 16 Gilbert R , Fluke J , O’Donnell M , et al. . Child maltreatment: variation in trends and policies in six developed countries . Lancet . 2012 ; 379 ( 9817 ): 758 – 772 . Google Scholar CrossRef Search ADS PubMed 17 Statistics Canada . Premature and potentially avoidable mortality, three-year average, Canada, provinces, territories, health regions, and peer groups (Table 102-4315). http://www5.statcan.gc.ca/cansim/a05?lang=eng&id=01024315. Accessed October 15, 2017. 18 Roos LL , Gupta S , Soodeen RA , et al. . Data quality in an information-rich environment: Canada as an example . Can J Aging . 2005 ; 24 ( suppl 1 ): 153 – 170 . Google Scholar CrossRef Search ADS PubMed 19 Manitoba Centre for Health Policy . Concept: ICD-10 to ICD-9-CM Conversion. 2015 . http://mchp-appserv.cpe.umanitoba.ca/viewConcept.php?conceptID=1157. Accessed September 24, 2016. 20 Pérez G , Rodríguez-Sanz M , Cirera E , et al. . Commentary: approaches, strengths, and limitations of avoidable mortality . J Public Health Policy . 2014 ; 35 ( 2 ): 171 – 184 . Google Scholar CrossRef Search ADS PubMed 21 Schneeweiss S , Eddings W , Glynn RJ , et al. . Variable selection for confounding adjustment in high-dimensional covariate spaces when analyzing healthcare databases . Epidemiology . 2017 ; 28 ( 2 ): 237 – 248 . Google Scholar CrossRef Search ADS PubMed 22 Harder VS , Stuart EA , Anthony JC . Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research . Psychol Methods . 2010 ; 15 ( 3 ): 234 – 249 . Google Scholar CrossRef Search ADS PubMed 23 Hjern A , Vinnerljung B , Lindblad F . Avoidable mortality among child welfare recipients and intercountry adoptees: a national cohort study . J Epidemiol Community Health . 2004 ; 58 ( 5 ): 412 – 417 . Google Scholar CrossRef Search ADS PubMed 24 Kalland M , Pensola TH , Meriläinen J , et al. . Mortality in children registered in the Finnish child welfare registry: population based study . BMJ . 2001 ; 323 ( 7306 ): 207 – 208 . Google Scholar CrossRef Search ADS PubMed 25 Li J , Precht DH , Mortensen PB , et al. . Mortality in parents after death of a child in Denmark: a nationwide follow-up study . Lancet . 2003 ; 361 ( 9355 ): 363 – 367 . Google Scholar CrossRef Search ADS PubMed 26 Schneeweiss S , Rassen JA , Glynn RJ , et al. . High-dimensional propensity score adjustment in studies of treatment effects using health care claims data . Epidemiology . 2009 ; 20 ( 4 ): 512 – 522 . Google Scholar CrossRef Search ADS PubMed 27 Guertin JR , Rahme E , Dormuth CR , et al. . Head to head comparison of the propensity score and the high-dimensional propensity score matching methods . BMC Med Res Methodol . 2016 ; 16 : 22 . Google Scholar CrossRef Search ADS PubMed 28 McConnell D , Llewellyn G . Stereotypes, parents with intellectual disability and child protection . J Soc Welf Fam Law . 2002 ; 24 ( 3 ): 297 – 317 . Google Scholar CrossRef Search ADS 29 Hoffman D , Rosenheck R . Homeless mothers with severe mental illnesses and their children: predictors of family reunification . Psychiatr Rehabil J . 2001 ; 25 ( 2 ): 163 – 169 . Google Scholar CrossRef Search ADS PubMed 30 McKegney S . Silenced Suffering: the Disenfranchised Grief of Birthmothers Compulsorily Seperated From Their Children [thesis]. Montreal, Canada: McGill University; 2003 . © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

American Journal of EpidemiologyOxford University Press

Published: Mar 28, 2018

There are no references for this article.

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


DeepDyve is your
personal research library

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

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

All for just $49/month

Explore the DeepDyve Library

Search

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

Organize

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

Access

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

Your journals are on DeepDyve

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

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off