Maternal Alcohol-Use Disorder and Child Outcomes

This project made use of routinely linked administrative data from Western Australia to identify preterm birth, being small for gestational age, cerebral palsy, intellectual disability, birth defects, education outcomes, and contact with the justice and child protection systems in the offspring of women with and without a record of an alcohol-related diagnosis. The linkage of these data sets has previously been described in detail.¹⁹ 

The cohort was selected from women who had a live birth recorded in the Western Australia Midwives Notification System (MNS) between 1983 and 2007.^18,19  The exposed cohort was defined as mothers who had an alcohol-related diagnosis, which is a proxy for heavy maternal alcohol use (recorded on ≥1 of the following: hospital morbidity, mental health inpatient or outpatient, or drug and alcohol services data set) and their offspring. An alcohol-related diagnosis (or alcohol diagnosis) was based on International Classification of Diseases, Ninth Revision (ICD-9) or International Classification of Diseases, 10th Revision (ICD-10) codes and included alcohol-related mental and behavioral disorders that could be entirely attributed to alcohol and “other” alcohol codes.¹⁸ 

The timing of an alcohol diagnosis was hierarchically coded, with those recorded during pregnancy given the highest priority.¹⁸  The following categories were used: (1) any alcohol diagnosis during pregnancy (this included women who may also have had an alcohol diagnosis at another time point), (2) an alcohol diagnosis ≤1 year before pregnancy (this group included women who may have also had an alcohol diagnosis recorded >1 year before pregnancy or after pregnancy), (3) an alcohol diagnosis ≤1 year after pregnancy (this included women who may have had an alcohol diagnosis >1 year before or after pregnancy), (4) an alcohol diagnosis >1 year before pregnancy (this could include those with a diagnosis >1 year after pregnancy), and (5) an alcohol diagnosis >1 year after pregnancy.

The comparison cohort included mothers with no alcohol-related diagnosis recorded on any of the relevant linked administrative data sets and their offspring. The exposed cohort was frequency matched with comparison mothers on maternal age within the Indigenous status and the year of the child’s birth at a ratio of 1:3 for non-Indigenous and 1:2 for Indigenous mothers. The final population cohort included 84 439 births. The exposed cohort included 13 969 non-Indigenous and 9635 Indigenous offspring and the comparison cohort included 40 302 non-Indigenous and 20 533 Indigenous offspring.

Births were identified through the MNS, a population-based surveillance system, which provided data on maternal demographic and pregnancy factors and data on small-for-gestational-age (bottom 10th percentile of optimal birth weight)²⁴  and preterm (<37 weeks’ gestation) births. The Western Australian Register of Developmental Anomalies provided data on birth defects, which are coded by using the British International Classification of Diseases System (ninth revision).²⁵  Defects due to a chromosomal cause were excluded. The alcohol-related birth defect codes included in this study and additional information about the registry have been previously published.¹⁹  Intellectual disability data were obtained from the Intellectual Disability Exploring Answers Database.²⁶  For this analysis, a binary outcome (presence or absence) of intellectual disability, defined as an IQ of <70, identified from either source was used. The Western Australian Cerebral Palsy Register²⁷  provided data on all cases of prenatal, perinatal, and postnatal cerebral palsy.¹⁸ 

Educational data were obtained from the Western Australia Department of Education for government public schools; details have been published previously.¹⁶  Briefly, achievement data were obtained from the Western Australia Literacy and Numeracy Assessment for the years 1999 through 2007, which included assessments for numeracy, reading, writing, and spelling for children in years 3, 5, and 7. From 2005 to 2007, children in year 9 were also assessed in reading, writing, and numeracy. In addition, National Assessment Program: Literacy and Numeracy results were obtained from 2008. The primary outcome of interest was failure to meet the minimum standards on any testing domain in any of the recorded school years with valid records for 45 382 children.¹⁶  Attendance was classified as 1 of the following: regular (90%–100% attendance), indicated (80%–89% attendance), moderate (60%–79% attendance), and severe (<60% attendance). The small number of students who had an attendance of <30% were excluded because there are some cases in which students change schools, sometimes enrolling under a different name, without informing the previous school.¹⁴  For the purpose of this analysis, we focused on those children who fell into the severe attendance category (equating to 13 486 children), referred to in this article as “having poor school attendance.”

Department of Corrective Services (now called Department of Justice) custodial and community corrections data for juveniles and adults from 1985 to 2011 were available for 68 598 children. In Western Australia, the minimum age for such an offense to be recorded on this data set is 10 years, so children born after 2002 were not included. A justice contact, for the purposes of this analysis, was defined as a custodial or community corrections record within the justice system.

The Department of Communities, Child Protection and Family Support data set included cases of substantiated maltreatment (substantiation) and out-of-home care from 1990 to 2007.¹⁷  Contact with child protection was defined as either a substantiated maltreatment allegation or a record of a period of out-of-home care for ≥1 day.

Ethics approval for the conduct of the study was granted by the Princess Margaret Hospital Human Research Ethics Committee (no. 1244/EP), the Western Australia Department of Health Human Research Ethics Committee (no. 2011/34), and the Western Australian Aboriginal Health Ethics Committee (no. 134-04/06).

All analysis was conducted by using SAS 9.3 (SAS Institute, Inc, Cary, NC). All outcomes were considered binary (presence or absence).

To identify if academic outcomes, justice outcomes, and child protection outcomes predominantly co-occurred in the same group of children or individually affecting a larger group of children, we cross-tabulated these outcomes.

To identify cumulative impacts, the number of negative outcomes were summed, and results were categorized as 1, 2, or ≥3. Because some outcome variables were likely to be negatively associated and not captured in a summed scale, the ϕ coefficient was calculated to estimate the association between outcome variables included in the analysis. This was displayed graphically to demonstrate both positive and negative associations between the outcome variables of interest in the study.

Logistic regression was used to examine the relationship between an alcohol diagnosis, the timing of diagnosis relative to pregnancy, and the odds of a negative outcome. Multinomial logistic regression was used to quantify the risk of multiple negative outcomes for children of mothers with an alcohol diagnosis.

All models included the frequency-matching variables of maternal age and child's year of birth, and non-Indigenous and Indigenous data were analyzed separately. The Indigenous status of the mother is self-reported and recorded in the MNS. Multivariate models were adjusted for other maternal and child characteristics associated with adverse child outcomes. These covariates included the mother’s marital status at the time of birth, maternal mental health or illicit drug record, socioeconomic status, health region (metropolitan or rural), child’s sex, and parity. To produce the most parsimonious model, covariates were eliminated from the multivariate model if they were not significantly associated with the outcome (α =.05).

The attributable fraction refers to the risk that can be attributed to exposure, whereas the population attributable fraction (PAF) takes into account the prevalence of exposure within the population. The attributable fraction and PAF were calculated to allow for estimation of the impact of heavy maternal alcohol use on any of the child outcomes included in the analyses. The SAS STDRATE procedure was used to calculate the PAF and attributable fraction.

Exposed non-Indigenous mothers were more likely than comparison mothers to be unmarried (26.5% vs 13.6%), be in the lowest 10% of socioeconomic status (14.5% vs 8.7%), have a mental health diagnosis (52.8% vs 9.2%), and have an illicit drug diagnosis (38.3% vs 1.5%) (Table 1). In the Indigenous mother cohort, there was little difference between exposed and comparison mothers in the distribution across socioeconomic groups. However, a greater proportion of exposed than of comparison mothers had a mental health diagnosis (29.5% vs 11.8%) or an illicit drug diagnosis (22.4% vs 4.7%).

A greater proportion of Indigenous children than of non-Indigenous children had a record of any negative outcome irrespective of exposure status (Table 2). A greater proportion of children of mothers with an alcohol diagnosis had 1 or more negative outcomes (non-Indigenous: 53.4%; Indigenous: 84.0%) than did comparison children (non-Indigenous: 31.2%; Indigenous: 68.0%). The relationship between the number of negative outcomes identified for both non-Indigenous and Indigenous comparison and exposed children is displayed in Figs 1 and 2.

The proportion of children with poor social outcomes was higher in the exposed than in the comparison cohort, with pronounced differences observed for child protection contact (non-Indigenous 15.3% vs 2.4% and Indigenous 28.9% vs 8.9%, respectively), academic outcomes (non-Indigenous 15.8% vs 10.3% and Indigenous 36.8% vs 32.0%, respectively), and contact with the justice system (non-Indigenous 6.2% vs 2.3% and Indigenous 22.4% vs 13.6%, respectively; Table 2). The same pattern was observed for each of the biological outcomes with the exception of cerebral palsy in Indigenous children, in which the proportion was lower for the exposed than for the comparison cohort (0.4% vs 0.5%, respectively).

The proportion of children having a justice contact who also failed to reach academic benchmarks is shown in Table 3. Approximately 35% of exposed and comparison non-Indigenous children in contact with the justice system failed to reach minimum academic standards (Table 3). Failure to reach minimum benchmarks was highest for Indigenous children in contact with the justice system, with 80% of the exposed and 71% of the comparison cohort failing to reach a minimum academic standard. For non-Indigenous exposed children in contact with the justice system, school attendance was also poorer (63.6%) than for comparison children (52.9%). However, this was not evident for Indigenous children, with the proportion being lower for exposed (81.4%) than for comparison children (89.6%; Table 4). Of note, numbers are small, so the results should be interpreted within this context.

The percentage of children failing to meet academic standards was higher in Indigenous than in non-Indigenous children irrespective of child protection contact. For children who had child protection contact, the proportion failing to meet minimum academic standards was similar between exposed and comparison cohorts (non-Indigenous 40.4% vs 40.8% and Indigenous 70.5% vs 75.2%, respectively; Table 5). The percentage of exposed and comparison children with poor school attendance showed a pattern similar to the results for academic failure (non-Indigenous 39.7% vs 41.9% and Indigenous 66.0% vs 72.9%, respectively; Table 6).

The adjusted odds of a child whose mother had an alcohol-related diagnosis having any negative outcome was >2.5 times greater than for the comparison cohort (non-Indigenous: odds ratio [OR] = 2.67 [95% confidence interval (CI) = 2.56–2.78]; Indigenous: OR = 2.67 [95% CI = 2.50–2.85]; Table 7). The highest odds were observed when an alcohol-related diagnosis was recorded during pregnancy (non-Indigenous: OR = 4.65 [95% CI = 3.87–5.59]; Indigenous: OR = 5.18 [95% CI = 4.10–6.55]).

The association between each of the outcome variables for children in the exposed cohort is shown in Supplemental Figs 3 and 4. These figures indicate that there are both positive and negative associations between the outcomes included in the study. For example, there were positive associations between child protection contact, school failure, poor school attendance, and justice contact. These outcomes were negatively associated with the biological outcomes. Given the negative associations between some outcomes, analysis that makes use of summed outcomes should be interpreted within this context.

The adjusted odds of exposed children having 1, 2, or ≥3 negative outcomes by the timing of an alcohol diagnosis in relation to pregnancy are shown in Table 8. The odds of exposed non-Indigenous children having ≥3 negative outcomes were 3.79 (95% CI = 3.3–4.35) and were 4.47 for Indigenous children (95% CI = 4.07–4.92) for “any” alcohol-related diagnosis. The highest odds were observed when an alcohol-related diagnosis was recorded during pregnancy (non-Indigenous: OR = 9.62 [95% CI = 6.78–13.64]; Indigenous: OR = 14.12 [95% CI = 10.67–18.70]).

The PAF for any alcohol-related diagnosis in the non-Indigenous cohort was 1.2% (95% CI = 1.1%–1.3%) and was 4.9% for the Indigenous cohort (95% CI = 4.5%–5.2%; Table 9).

This study adds new and important information about the magnitude and breadth of the negative impact of maternal alcohol-use disorders on children. Over half (55%) of non-Indigenous and 84% of Indigenous children whose mothers had an alcohol-related diagnosis had ≥1 of the negative outcomes examined in this study. Both Indigenous and non-Indigenous children whose mothers had an alcohol-related diagnosis were 2.67 times more likely than children in the respective comparison cohorts to have a negative outcome (non-Indigenous: OR = 2.67 [95% CI = 2.56–2.78]; Indigenous: OR = 2.67 [95% CI = 2.50–2.85]).

Of particular concern is the likelihood of exposed children having ≥3 negative outcomes. The odds of experiencing ≥3 negative outcomes when any alcohol-related diagnosis was the exposure variable were 3.8 for non-Indigenous and 4.5 for Indigenous children. When a mother had an alcohol-related diagnosis recorded during pregnancy, the ORs increased to 9.6 and 14.1, respectively.

The findings of this study revealed that children of mothers with an alcohol-related diagnosis are at risk of having multiple negative social outcomes, which supports the findings of previous research.^28,29  A higher proportion of children in contact with the justice system and whose mothers had an alcohol-related diagnosis had poor school attendance than comparison children did. In contrast, the proportion of children in contact with the justice system who failed to meet academic standards was similar between exposed and comparison cohorts for both non-Indigenous and Indigenous children, a finding that was replicated for school outcomes in children with a child protection contact. This may indicate that the relationship between exposure and multiple outcomes is not driven by school factors alone. These results highlight that multiple agencies need to work together to provide a coordinated and collaborative program of prevention and early intervention strategies to improve long-term outcomes for children of mothers with an alcohol-related diagnosis.

In this study, Indigenous children were more likely than non-Indigenous children to experience negative outcomes, particularly social outcomes, and this was more pronounced in the exposed cohort. A higher proportion of Indigenous Australians have deep and persistent disadvantage, unemployment, housing overcrowding, imprisonment, poor health outcomes, and early mortality than non-Indigenous Australians do,^30,31  and reducing this disparity is an ongoing government priority.³²  Comprehensive Indigenous-specific programs designed to address maternal alcohol-use disorder, prevent prenatal alcohol exposure, and improve the biological and social outcomes of Indigenous children need to be well funded and delivered consistently across time to improve the long-term outcomes of Indigenous people.

Over half of the exposed children had ≥1 negative outcome, highlighting that preventing and reducing the prevalence of maternal alcohol-use disorder should be a public health priority. The low PAF reflects the relatively low prevalence of mothers with an alcohol-related diagnosis recorded in health and mental health data sets, particularly among non-Indigenous women. In Western Australia, asking women who are pregnant and women of childbearing age about their alcohol consumption is not routinely done and urgently needs to be implemented and accompanied by brief interventions and treatment when required.³³ 

The use of linked, population-based, administrative data sets providing maternal and child data from the child’s birth through their life span is a major strength of this study and enables examination of a range of biological and social outcomes, overcoming many of the issues around longitudinal follow-up of high-risk populations.

An alcohol-related diagnosis was used as a proxy for heavy maternal alcohol use. The Western Australia data-linkage system is a valid means of identifying patients who had been admitted to hospitals for health-related conditions; however, to ensure that these are given the correct ICD-9 or ICD-10 classification by the clinical coders, it is reliant on medical officers accurately recording the alcohol diagnosis in the medical records.²¹  A limitation of this methodology is that it does not allow for direct estimation of the quantity, frequency, and timing of the mother’s drinking.

Although we are confident that mothers in the exposed cohort were consuming significant amounts of alcohol, we have previously identified that some women in the comparison cohort had a child diagnosed with FASD and were therefore drinking but did not have an alcohol diagnosis recorded.¹⁹  The outcomes in this study are based on administrative data sets and can be biased in reflecting the use of services. However, these limitations are likely to bias the results toward the null.

Information on attendance and achievement at school was available only for students attending public schools, which represents approximately two-thirds of Western Australia school students and 84% of Aboriginal students. Therefore, the results for educational outcomes may not be generalizable to students attending private schools.

The majority of children of mothers with an alcohol-related diagnosis are at risk of poor outcomes, which are due to both direct biological and indirect social and environmental effects of maternal alcohol misuse. Prevention, early identification, and treatment of alcohol-related problems in women of childbearing age or who are pregnant and follow-up of their children is required to prevent fetal and child harm. This is a public health priority requiring a coordinated, interagency approach to improve the outcomes for children of mothers with an alcohol-use disorder across their life course.

We acknowledge the Developmental Pathways Project and the industry partners whose data are used in this project (Western Australia Department of Health, Department of Education, Department for Child Protection and Family Support, and Department of Justice [formerly Department of the Attorney General and Department of Corrective Services]). Any material published or made publicly available by the researchers cannot be considered as either endorsed by the Western Australia Department of Justice or an expression of the policies or views of the Western Australia Department of Justice. Any errors of omission or commission are the responsibility of the researchers.

CI

confidence interval

FASD

fetal alcohol spectrum disorder

ICD-9

International Classification of Diseases, Ninth Revision

ICD-10

International Classification of Diseases, 10th Revision

MNS

Midwives Notification System

OR

odds ratio

PAF

population attributable fraction