Health of Newborns and Infants Born to Women With Disabilities: A Meta-analysis

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines²³  to identify studies published from database inception to January 2, 2020, in the Cumulative Index to Nursing and Allied Health Literature, Embase, Medline, and PsycINFO databases. We used an adapted version of a validated search strategy for disability,²⁴  which includes keywords and subject headings for disability generally (eg, “activity limitation”) and physical, sensory, and intellectual and/or developmental disabilities specifically. In consultation with a research librarian, we added keywords and subject headings for indicators of neonatal health (eg, “preterm birth”) and health care use (eg, “neonatal intensive care unit”) and infant health (eg, “infant mortality”) and health care use (eg, “emergency department visit”) (Supplemental Table 1). To identify studies that may have been missed in database searches, we hand searched reference lists of original articles selected for full-text review and review articles.

To be included in the review, studies were required to (1) report original data on the association between maternal physical (eg, arthritis, spinal cord injury), sensory (eg, hearing loss), or intellectual and/or developmental disabilities (eg, autism spectrum disorder, Down syndrome) and neonatal or infant health; (2) use an observational design with a referent group of women with no disabilities (eg, cohort, case-control, or cross-sectional studies); (3) be peer-reviewed journal articles or theses; and (4) be written in English. Studies in which authors (1) examined conditions that were not clearly disabilities (eg, diabetes without evidence of functional limitations) or (2) reported on the health of newborns or infants of women with only psychiatric disorders were excluded. Although women with psychiatric disorders experience many of the same risk factors for adverse outcomes as women with other disabilities (eg, poverty), we did not include this group because there is a sizeable body of literature on the health outcomes of their newborns and infants,^25–27  and the focus on physical, sensory, and intellectual and/or developmental disabilities in our review aligns with the National Institutes of Health’s call for research in this area.²⁸  (Studies in which authors included psychiatric disorders within their definition of “any” disability were included, however, and the impact of this decision on the results was tested in additional analyses.) All eligible studies were included in the qualitative synthesis.

Three authors (L.A.T., F.M., and D.S.) independently screened titles and abstracts. After agreement, the same authors then independently retrieved and reviewed relevant full-text articles in detail. They independently extracted data from studies that met the inclusion criteria using a standardized data extraction form created a priori on the basis of the Strengthening the Reporting of Observational Studies in Epidemiology statement.²⁹  The following data were extracted: study period and location, study design and data sources, sample size, exclusion criteria, disability definition and measurement, outcome definition(s) and measurement, and confounders. To calculate pooled effect estimates, frequencies and percentages as well as odds ratios (ORs) and 95% confidence intervals (CIs) were extracted. Discrepancies in data extraction were resolved through discussion with the senior author (H.K.B.).

Three authors (L.A.T., F.M., and D.S.) independently assessed the quality of each study using an adapted version of the Effective Public Health Practice Project Quality Assessment tool, a validated^30,31  and widely used tool in health research.^17,22  Studies were rated as weak, moderate, or strong overall on the basis of the following criteria: (1) study design, (2) selection bias (sample representativeness, response rate), (3) confounding (percentage of confounders controlled for), (4) detection bias (reliability and validity of the data sources), and (5) attrition bias (loss to follow-up, missing data) (Supplemental Table 2). Informed by existing literature regarding risks for adverse neonatal and infant outcomes,^32–35  we identified the following confounders a priori: demographics (eg, maternal age, race and ethnicity), socioeconomic status (eg, education, employment), comorbidities (eg, chronic health conditions, mental illness), lifestyle behaviors (eg, smoking), and social support (eg, marital status). Discrepancies in ratings were resolved through discussion with the senior author (H.K.B.).

DerSimonian and Laird³⁶  random effects models were used to calculate pooled ORs and 95% CIs for outcomes examined by at least 3 studies. When there were multiple studies published by using the same data sources with overlapping samples and study periods, we included the study with the largest cohort in the meta-analysis. Analyses were conducted for any disability (as defined by the study authors) and by disability type, as numbers allowed. The source of variance was examined across studies by calculating the Q statistic (ie, the weighted sum of squared differences between the effects of individual studies and the pooled effect across studies) and I² statistic (ie, the percent of variation across studies that results from real heterogeneity, not chance).³⁷  Although heterogeneity is to be expected,³⁸  to address potential bias introduced by significant heterogeneity, we (1) conservatively excluded meta-analyses with I² values ≥75% (corresponding to the Cochrane Handbook for Systematic Reviews of Interventions classification of “considerable” heterogeneity)³⁹  and (2) calculated 95% prediction intervals for the remaining meta-analyses to show the range of true effects in similar studies (for analyses with at least 5 studies).^40,41  In sensitivity analyses, fixed effects models were used to re-estimate pooled ORs for studies with a nonsignificant Q statistic and small I² value, assuming the true effect was the same for all studies.³⁷  The effect of removing individual studies from the meta-analysis was also tested to identify influential studies. Although we planned to generate a funnel plot to test for publication bias, we had an insufficient number of studies for any given analysis to do so.⁴²  Two additional analyses were undertaken to test the effect of analysis decisions on results. First, we tested the impact of substituting alternative studies into the meta-analysis from among those not included, having used the same data sources. Second, we tested the impact of excluding studies that included psychiatric disorders in their definition of any disability. For our analyses, we used Comprehensive Meta-Analysis version 3.0 (Biostats Inc, Englewood, NJ), and we generated figures using R version 3.6.0.

The article selection process is outlined in Fig 1. Database searches returned 15 737 articles, and 13 articles were identified from other sources. After removal of duplicates and title and abstract screening, we identified 115 articles to review in detail. After full-text review, 84 studies were removed because they had no comparison group (n = 43); were reviews, editorials, or commentaries (n = 17); had ineligible exposure definitions (n = 9); were not written in English (n = 6); were qualitative studies (n = 4); did not examine neonatal or infant outcomes (n = 3); or were abstracts (n = 2). Thirty-one studies,^6,43–72  representing >13 million unique women, including 30 139 with any disability, 10 198 with physical disabilities, 4436 with sensory disabilities, and 9535 with intellectual and/or developmental disabilities in 20 distinct cohorts (including 1 US study in which authors examined 2 states separately⁵⁰  and 1 Australian study in which authors examined Aboriginal and non-Aboriginal women separately⁴⁹ ), met our criteria. In 20 studies, authors used the same data sources and study periods (representing 7 distinct cohorts): 5 used the US (MA) Pregnancy to Early Life Longitudinal database,^{46,47,59,61,62}  3 used the US Nationwide Inpatient Sample of the Healthcare Cost and Utilization Project,^43,60,67  3 used California administrative data,^6,48,55  2 used the Pregnancy Risk Assessment Monitoring System for Rhode Island,^65,68  2 used Washington State administrative data,^64,70  3 used Swedish administrative data,^53,54,72  and 2 used the UK Millennium Cohort Study.^52,71 

Study characteristics are summarized in Supplemental Table 3. Nineteen studies were conducted in the United States,^* 5 in the United Kingdom,^{51,52,56,69,71}  3 in Sweden,^53,54,72  2 in Australia,^49,57  and 1 each in Canada⁴⁵  and Israel.⁶⁷  Twenty studies were retrospective cohort studies,^† 2 were prospective cohort studies,^44,57  and 9 were cross-sectional studies.^‡ Studies had between 62⁴⁴  and 6 453 012 participants,⁵⁵  with all but 4^44,57,63,64  having a total sample size >1500. In 11 studies, authors examined the impact of maternal disability overall,^§ 14 examined intellectual and/or developmental disabilities only,^‖ 4 examined sensory disabilities only,^60,62,67,70  and only 2 examined physical disabilities only.^63,66  Of the studies in which authors examined the impact of maternal disability overall, 10 compared an “any disability” group with a “no disability” referent group, and 7 also reported outcomes by disability type.^{6,45–48,55,56,69}  In 7 studies, authors included psychiatric disorders within their definition of any disability.^{46,47,56,58,65,69,71}  In 23 studies, authors used health administrative data (eg, hospital discharge data, insurance claims) to identify disability status,^¶ 7 used self-report measures of disability,^{52,56–58,65,69,71}  and 1 used a clinical measure.⁴⁴  Authors of most studies examined outcomes related to maternal and neonatal health (see Tarasoff et al¹⁷  for a review of studies on maternal outcomes). Among neonatal and infant outcomes, the most commonly reported were preterm birth (primarily defined as <37 weeks’ gestation; n = 29)^{6,43,45–60,62–72}  and low birth weight (n = 14).^{43,50,52,57–64,66,69,70}  Other neonatal outcomes were very preterm birth (primarily defined as <32 weeks’ gestation; n = 8), postterm birth (n = 6), very low birth weight (n = 6), high birth weight (n = 4), small for gestational age (n = 11), large for gestational age (n = 7), low Apgar scores (n = 9), NICU admission (n = 6), neonatal mortality (n = 5), and neonatal morbidity (n = 1). In few studies did authors examined infant outcomes (eg, infant mortality, n = 2; hospital admission, n = 1). Outcomes were primarily ascertained from health administrative data; in 4 studies, authors used maternal self-report data.^52,56,69,71  Studies varied with respect to their control for confounding factors.

Study quality is described in Supplemental Table 4. Overall, studies were rated as strong (n = 5),^{45–47,49,62}  moderate (n = 9),^# or weak (n = 17).^** Retrospective and prospective cohort studies (n = 22) were rated as having moderate risk of bias for study design, and cross-sectional studies (n = 9) were rated as having high risk. Concerning selection bias, many studies were population based and had good generalizability. Studies with moderate or high selection bias ratings were those using Medicaid data,⁵⁰  with urban clinic samples,^44,57,63  and/or with response rates that were <80% or unreported.^{44,52,56,58,65,69,71}  Authors of 14 studies did not control for confounders.^†† Just more than half (n = 17) controlled for demographics,^‡‡ 12 for socioeconomic status,^§§ 10 for comorbidities,^‖‖ 9 for lifestyle behaviors, ^¶¶ and 6 for social support.^{46,50,51,53,54,62}  Only 5 studies were rated as having low risk of detection bias^{6,44,57,63,67} ; most were rated as having moderate or high risk of bias because the authors provided limited or no data on the validity and/or reliability of their data sources. Finally, 15 studies^{44,50–52,55–59,63,65–69}  were rated as having high risk of attrition bias, primarily because the authors did not report on the frequency of missing data.

In the studies, authors primarily examined indicators of fetal growth and birth timing; fewer studies examined indicators of neonatal or infant morbidity and mortality. The results for outcomes that could be pooled, as well as those reported descriptively, on the basis of available data, are as follows.

Figure 2 shows the pooled unadjusted ORs for preterm birth in newborns of women with sensory (Fig 2A) and intellectual and/or developmental disabilities (Fig 2B). These were 1.37 for sensory (95% CI 1.27–1.48; 7 studies, n = 11 191 691) and 1.76 for intellectual and/or developmental disabilities (95% CI 1.59–1.96; 12 studies with 13 cohorts; n = 8 783 563). Heterogeneity was nonsignificant. Figure 3 shows the pooled adjusted ORs of preterm birth in newborns of women with sensory (Fig 3A) and intellectual and/or developmental disabilities (Fig 3B), which were 1.21 (95% CI 1.08–1.36; 4 studies; n = 4 728 092) and 1.56 (95% CI 1.45–1.68; 6 studies with 7 cohorts; n = 5 178 838), respectively. Heterogeneity was nonsignificant, but the 95% prediction interval crossed the null value for the first analysis. Findings were robust to use of fixed effects models (Supplemental Table 5) and removal of individual studies from the meta-analysis (Supplemental Table 6).

Notably, meta-analyses of 7 studies (n = 6 863 019) in which authors examined any disability and preterm birth and 5 studies (n = 6 639 560) in which authors examined physical disability and preterm birth were not presented because of considerable heterogeneity (I² values ≥75%); however, these studies indicated increased risk for preterm birth in these groups (Supplemental Table 7). Unadjusted pooled analyses for very preterm birth in newborns of women with intellectual and/or developmental disabilities (Supplemental Fig 6) and findings from studies on birth timing that could not be pooled (Supplemental Table 7) generally supported increased risk for very preterm birth but not postterm birth in newborns of women with disabilities compared with those without disabilities.

Figure 4 shows the pooled unadjusted ORs for low birth weight in newborns of women with any disability (Fig 4A) and physical (Fig 4B), sensory (Fig 4C), and intellectual and/or developmental disabilities (Fig 4D). These were 1.88 for any (95% CI 1.63–2.17; 3 studies with 4 cohorts; n = 167 684), 1.81 for physical (95% CI 1.47–2.23; 3 studies; n = 173 031), 1.36 for sensory (95% CI 1.18–1.57; 4 studies; n = 4 485 221), and 1.98 for intellectual and/or developmental disabilities (95% CI 1.50–2.61; 5 studies; n = 1 151 920). Heterogeneity was nonsignificant, but the 95% prediction interval crossed the null value for the last analysis. Figure 5 shows the pooled adjusted ORs for low birth weight in newborns of women with sensory (Fig 5A) and intellectual and/or developmental disabilities (Fig 5B), which were 1.24 (95% CI 1.06–1.45; 3 studies; n = 4 462 351) and 1.50 (95% CI 1.25–1.81; 3 studies; n = 4 442 979), respectively. Heterogeneity was nonsignificant. Findings were generally robust to the use of fixed effects models (Supplemental Table 5) and removal of individual studies from the meta-analysis (Supplemental Table 6).

Pooled analyses for very low birth weight in newborns of women with sensory disabilities (Supplemental Fig 7), small for gestational age (Supplemental Fig 8) in newborns of women with intellectual and/or developmental disabilities, as well as findings from studies on fetal growth that could not be pooled (Supplemental Table 7) generally supported elevated risk of restricted, but not excessive, fetal growth among newborns of women with disabilities compared with those without disabilities.

In examining other neonatal outcomes, we found that newborns of women with sensory (Supplemental Fig 9A) but not intellectual and/or developmental disabilities (Supplemental Fig 9 B and C) were at increased risk for 5-minute Apgar scores <7. However, the latter group was at increased risk of 1-minute Apgar scores <7 (Supplemental Fig 10), NICU admission (Supplemental Fig 11), and neonatal mortality (Supplemental Fig 12). Findings from studies that could not be pooled (Supplemental Table 7) generally supported increased risk for low Apgar scores, NICU admission, longer hospital stay, and neonatal morbidity and mortality but not congenital anomalies among newborns of women with disabilities. In examining infant outcomes, we found that infants of women with intellectual and/or developmental disabilities were not at increased risk of hospitalization (Supplemental Fig 13). Findings from studies that could not be pooled (Supplemental Table 7) were nonsignificant for hospitalizations and problems requiring a doctor and were significant for infant mortality. However, authors of few studies contributed data to these analyses.

Results were mostly unchanged when we substituted different studies into our analyses from among those in which authors used the same data sources (Supplemental Table 8). When we removed studies in which authors included psychiatric disorders in their overall disability definition, the association with low birth weight could no longer be meta-analyzed, but results of the remaining study⁵⁰  (which included 2 cohorts) suggested increased risk.

In our systematic review of 31 studies, representing >13 million unique women, we found that newborns of women with physical, sensory, and intellectual and/or developmental disabilities may be at elevated risk for adverse health outcomes compared with those of women without these disabilities, with particularly strong evidence of elevated risk for low birth weight across disability groups. Studies in which authors examined preterm birth as well as other neonatal outcomes were also consistent with elevated risk. However, findings from the small number of studies in which authors examined infant outcomes were more inconsistent, with some evidence of increased risk of infant mortality but no other adverse infant health outcomes. Findings were generally robust in sensitivity analyses, although some heterogeneity across studies should be acknowledged.

This is the first systematic review and meta-analysis to examine risk for adverse health outcomes among newborns and infants of women with disabilities compared with those of women without disabilities. Our findings are generally consistent with a 2011 summary of the literature, in which authors found elevated rates of preterm birth and other adverse neonatal outcomes among women with physical disabilities,⁷³  and with other studies focused on specific disabilities such as spinal cord injury,⁷⁴  muscular dystrophy,⁷⁵  lupus,⁷⁶  and autism spectrum disorder.⁷⁷  Other studies on specific disabilities, such as multiple sclerosis,^78,79  have yielded conflicting results, suggesting a need for further research to understand how social, health, and health care factors associated with different disabilities may influence risk.

Several factors may explain the adverse health outcomes observed in offspring of women with disabilities. Reproductive-aged women with disabilities experience significant socioeconomic and health disparities, including high rates of poverty, chronic health conditions, and mental illness,^11,12  with many disabilities or secondary conditions requiring use of potentially teratogenic medications.^11,80  Women with disabilities are also less likely than their peers to be married or partnered^10,12  and more likely to experience violence and isolation.^81–83  Numerous studies indicate that women with disabilities encounter barriers accessing and navigating perinatal care^16,55,56  and are at greater risk for pregnancy, delivery, and postpartum complications.¹⁷  In our review, women with intellectual and/or developmental disabilities were an especially vulnerable group. Along with the disparities described above, these women experience a lack of adequate sexual and reproductive health education, high rates of unintended pregnancy,⁸⁴  short interpregnancy intervals,⁸⁵  and difficulty understanding and following medical advice.^83,86,87  Finally, women with disabilities may be hesitant to ask for help or avoid accessing services because of fears of stigma, their parenting being scrutinized, and child welfare involvement.^88,89  Together, these disparities may contribute to adverse outcomes for their offspring. This hypothesis is consistent with research revealing that much of the variance in adverse health and developmental outcomes among older children of women with disabilities is driven by the mother’s circumstances (eg, low socioeconomic status, lack of adequate supports), not her disability per se.^88,90,91  However, few studies have isolated the effects of disability from pathologizing assumptions and disability-based discrimination or marginalization.⁸⁸ In our review, few studies comprehensively accounted for social, health, and health care factors that could explain disparities. Further research is needed to understand the contribution of these factors to adverse neonatal health outcomes because many are modifiable and could be valuable targets for supportive interventions for women with disabilities and their offspring.

Our methods and the included studies are not without limitations. Our protocol was not registered a priori. We excluded studies not written in English and identified no studies from low- and middle-income countries. Given the socioeconomic, health, and health care disparities experienced by women with disabilities in low- and middle-income countries,^92,93  it is possible that their offspring fare even worse. Thus, the generalizability of our findings may be limited. Not all studies defined disability in the same way. For example, several studies included psychiatric disorders within their definition of any disability; however, this heterogeneity did not appear to affect results. Authors of several studies based their definition of disability on diagnostic codes, which may not accurately reflect functional limitations or lived experiences of disability and may miss less obvious disabilities.^55,94  Moreover, most of the included studies examined disability overall (n = 11) or intellectual and/or developmental disabilities only (n = 14). As well, most studies examined neonatal outcomes, with few studies examining infant outcomes. Results of these smaller analyses should be interpreted with caution, particularly those with significant heterogeneity. Finally, just more than half (n = 17) of the studies were rated as weak in quality, with lack of control for confounding being a common limitation. There is a need for more high-quality research in this area.

Although women with disabilities are becoming pregnant at increasing rates,⁵  research and clinical practice have not kept pace. To our knowledge, there are no preconception health interventions tailored to women with disabilities, few clinical practice guidelines specific to this population in pregnancy,⁹⁵  and limited supports for disabled parents with infants and older children.⁸  This lack of practical support is concerning^88,96 : From their summary of the literature on parents with disabilities, McConnell and Hahn found that, “antipathy to parents and parenting with disabilities affects women and mothers and, by extension, their children, in profound ways.”⁸⁸ (p3) To address the disparities found in our review, in future work, researchers should explore how tailored preconception and prenatal care could benefit women with disabilities.¹¹ 

In addition, future research could be used to explore the role of family-centered pediatric care in better supporting neonatal and infant health.⁴⁷  As noted in the American Academy of Pediatrics 2003 Report of the Task Force on the Family, “The health and well-being of children are inextricably linked to their parents’ physical, emotional and social health, social circumstances, and child-rearing practices.”⁹⁷ (p1541) Because pediatricians are often the main point of contact for women in the immediate postpartum period, in addition to providing breastfeeding support and screening for postpartum depression, they could potentially support women with disabilities and in turn their infants by connecting them with peer support, community-based parenting programs, and other services they may need (eg, nutrition, housing). To facilitate pediatricians in caring for families of women with disabilities, there would be a need for provider education and training on the social circumstances, medical risks, and stigma faced by mothers with disabilities, as well as their health care–related needs, such as longer and more frequent appointments, accessibility and communication requirements, and other accommodations such as home or virtual visits. Such efforts could make a positive impact on the health of offspring of women with disabilities.

In our systematic review and meta-analysis, we found that offspring of women with physical, sensory, and intellectual and/or developmental disabilities are at elevated risk for low birth weight, with studies of preterm birth and other adverse neonatal outcomes also suggesting elevated risk. Notably, there were fewer studies in which authors examined infant health outcomes, with mixed results, highlighting the need for more research in this area. In future studies, researchers should explore the social, health, and health care factors explaining the disparities experienced by women with disabilities and their offspring. Interventions to address these disparities should also be explored, with disability-competent preconception and perinatal care and family-centered approaches to pediatric care potentially having a positive impact on health outcomes. Future research and interventions regarding the health of women with disabilities and their offspring should be attentive to the multiple social and health challenges women with disabilities face.

We thank Mary-Rose Faulkner for her assistance with data collection.

CI

confidence interval

OR

odds ratio