The current investigation utilized data from the larger, ongoing All Our Families (AoF) cohort study (37, 38) in Alberta, Canada. From 2008 to 2010, a total of 4,011 individuals responded to community advertisements or researchers at primary health care offices and laboratory services recruiting “pregnant women” (referred to as birthing parents or participants throughout since gender identity information was not collected), of which 3,387 met eligibility criteria (understand English, >18 years old, <25 weeks' gestation, receiving prenatal care near Calgary, Canada) and were enrolled in the study. All participants provided informed written consent and ethical approval was obtained from the Conjoint Health Research Ethics Board (CHREB) at the University of Calgary for both the original study (REB13-0868) and secondary analysis (REB19-1417). Participants were asked to complete questionnaires twice during pregnancy (<25 and 34–36 weeks' gestation) and were followed up at 4 months and 1, 2, 3, and 5 years postpartum.

Descriptive statistics and correlation analyses were conducted using SPSS version 25.0 (IBM, USA). Pearson correlation coefficients were estimated between the neighbourhood characteristics (deprivation, disorder) during pregnancy, lifestyle factors (physical activity, parent-child reading, and community resource use) at 3 years postpartum, and child development (ASQ-3 total score) at age 5. Multilevel modelling, in Mplus version 8.1 (49), was used to test the relationship between neighbourhood characteristics and child development, as well as potential interactions between neighbourhood characteristics and lifestyle factors. As participants in our sample were nested within neighbourhoods, a two-level random model with Bayes estimation (50) was conducted to account for within neighbourhood (level 1) and between-neighbourhood (level 2) effects. Moderation was tested using cross-level interactions by estimating the slope of each lifestyle factor on the ASQ-3 total score at level 1, and then regressing the neighbourhood factors on these slopes at level 2. Any non-significant moderators were removed from the final model. Additionally, several sociodemographic characteristics were considered as control variables including ethnicity, education, household income, preterm status, child sex, and moving (25, 26, 51), and included as level 1 covariates in the final model if they were significantly correlated with the ASQ-3 total score. Missing data were handled using Full Information Maximum Likelihood (FIML), which produces unbiased model parameters (52). Significant effects were determined by a 95% Bayesian credibility interval (BCI) that did not cross zero (53).

After removing those who gave birth to twins (n = 36) and those who could not be geocoded (n = 907) because they did not provide postal codes or lived outside the city of Calgary boundaries, the final sample consisted of 2,444 participants. Participants were distributed across 192 neighbourhoods, with an average of 12.73 participants each. Overall, 15.17% of the data were missing and covariance coverage ranged from .507–1.00. The sample mostly consisted of pregnant individuals who were married or in common law¹ relationships (94.9%), identified as European Canadian (77.2%), had attained post-secondary education (76.0%), and an annual household income of greater than $60,000 (82.3%). At 3 years postpartum, most participants had not moved since their child's birth (62.2%). The mean age of participants at <25 weeks gestation was 30.8 years (SD = 4.5). Most participants’ children were not born preterm (92.8%) and slightly above half were male (52.4%). Table 1 displays detailed demographic information for the sample.

The sociodemographic characteristics of the current sample are consistent with the local Calgary population, where for example the median household income is approximately $97,000, 69.9% have attained post-secondary education, and 78% identify their ethnic origin as North American or European (55). Among participants who could not be geocoded, slightly more identified as European Canadian (80.6%), less were married or in common law¹ relationships (92.9%), less attained post-secondary education (29.1%), more used ≥3 community resources (75,2%), and were on average older (M = 30.78 years).

The mean neighbourhood VANDIX score was −6.67, suggesting that, on average, most participants resided in socioeconomically advantaged neighbourhoods. The number of neighbourhood disorder reports varied considerably across neighbourhoods, ranging from 1 to 7000. At three years postpartum, slightly over half of participants read with their children for greater than 20 min per day (53%), most had accessed 3 or more community resources in the past year (66%), and most children met the CSEP daily physical activity recommendation of 3 h (63%). In terms of development at 5 years of age, children in the sample scored relatively high on the ASQ-3 (M = 274.55), given that the maximum possible total score is 300. Few children in the sample scored in either the “monitoring zone” (≤1 SD below the mean of the ASQ normative data) or the “referral zone” (≤2 SD below the mean of the ASQ normative data) on any of the individual ASQ-3 subscales: communication (7.9%), gross motor (7.7%), fine motor (7.1%), problem solving (2.1%), and personal-social (7.8%). See Table 1 for descriptive statistics for key study variables.

Correlations between neighbourhood variables, lifestyle factors, and child development are reported in Table 2. Neighbourhood deprivation was significantly negatively associated with parent-child reading and community resource use, and positively associated with physical activity. Additionally, neighbourhood disorder was significantly positively associated with physical activity. There was a significant positive relationship between the ASQ-3 total score and parent-child reading. However, the associations between ASQ-3 total score and community resource use and physical activity were not significant.

Correlations between the ASQ-3 total score and potential covariates revealed significant associations with ethnicity (r = .078, p < .01), education (r = .074, p < .01), household income (r = .091, p < .01), child sex (r = .175, p < .01), and preterm status (r = −.127, p < .01). These variables were therefore included as covariates in the multilevel analyses. Whether participants had moved or not between birth and 3 years was not associated with ASQ total score (r = −.016, p = .572) and therefore was not included as a covariate in the multilevel analyses.

The intraclass correlation (ICC) for child development was 0.024, indicating that 2.4% of the variance in total ASQ-3 score was due to variation between neighbourhoods, while 97.6% of the variance was attributable to variation across individuals within neighbourhoods. Parameter estimates from the final multilevel model considering neighbourhood deprivation and disorder during pregnancy as predictors of child development at age 5 are displayed in Table 3. After controlling for level 1 covariates, there was a statistically significant negative association between the VANDIX and ASQ-3 total score, such that higher levels of neighbourhood deprivation during pregnancy were associated with poorer overall child development at age 5. Neighbourhood disorder during pregnancy was not directly associated with total ASQ-3 score.

In terms of individual- and family- level factors, child sex was found to be significantly associated with child development, with girls having higher ASQ-3 total scores than boys. Preterm birth status was significantly associated with child development, where children who were born preterm had lower ASQ-3 scores than those who were born at <37 weeks gestational age. None of the other individual or family-level variables (ethnicity, education, household income, community resource use, physical activity) were significantly associated with the ASQ-3 total score.

In terms of moderation effects, there was a statistically significant cross-level interaction where neighbourhood disorder was found to predict a positive slope between parent-child reading and overall child development (see Figure 1). That is, as the amount of neighbourhood disorder increases (i.e., the number of crime and disorder reports), the slope between the ASQ-3 total score and parent-child reading increases. The cross-level interaction of neighborhood deprivation with parent-child reading on development slope was not significant. Additionally, no significant moderation effects were found for community resource use (b_deprivation= .385, 95% CI: −.880, 1.716; b_disorder= .003, 95% CI: −.002,.009) or physical activity (b_deprivation= −.686, 95% CI: −1.970, 0.642; b_disorder= −.003, 95% CI: −.009,.003), therefore these slopes were not included in the final model.

The current investigation utilized data from a large Canadian cohort to elucidate the influence of neighbourhood characteristics (deprivation and disorder) during pregnancy on child development, as well as the potential moderation by lifestyle factors (physical activity, parent-child reading, and community resource use). Multilevel analyses indicated that neighbourhood deprivation during pregnancy was associated with poorer child development at age 5 and that parent-child reading may function as a protective factor for child development in the presence of higher neighbourhood disorder. 

The observed association between neighbourhood deprivation and child development, even after controlling for child sex, preterm status, maternal ethnicity, maternal education, and household income, is consistent with extant findings that neighbourhood deprivation is independently associated with developmental delays in preschool- and school-aged children. For example, preschool-aged children living in more deprived neighbourhoods were reported to be 3.15 times more likely to have a speech, language, or communication concern as measured by the ASQ-3 (56). Similarly, after accounting for family-level SES, children in deprived neighbourhoods were found to have concurrently higher levels of behavioural problems and lower cognitive test scores (57). Previous literature has proposed various potential mechanisms for this association, which include reduced access to programs and institutions that promote healthy childhood development, less exposure to highly educated role models in one's neighbourhood, and lower levels of social support and control (58). The current findings are the first to demonstrate associations between prenatal exposure to neighbourhood deprivation and child development, providing further support for the DOHaD hypothesis and pointing to the need for early intervention. Although intervening during pregnancy is proposed to have the largest returns on investments (59), the multilevel nature of social determinants needs to be taken into account. As such, socioeconomic inequalities between neighbourhoods should be addressed as part of policies and programs that promote child and family well-being.

The lack of direct effect of neighbourhood disorder on overall child development at 5 years may reflect a function of age and domain. Neighbourhood disorder has been more frequently linked with development in later childhood and adolescence for conduct problems and mental health outcomes (60–62). Since the ASQ-3 total score captures more motor and cognitive domains and to a lesser extent personal-social outcomes, future research could examine follow-up of specific conduct and mental health outcomes at later ages in the AoF cohort. Interestingly however, there was a moderation effect where as neighbourhood disorder increased during pregnancy, the association between parent-child reading and overall development got stronger. This finding suggests that parents reading with their children more often may buffer the negative impact of prenatal exposure to a neighbourhood with higher disorder. Parent-child reading is proposed to promote child development through various mechanisms such as improved linguistic, interactive, and parental functioning including reduced stress and increased sense of control (63). Parent-child reading interventions are also associated with improved relationship quality (64), which could create a safe space to learn. It is possible that parent-child reading and the related improvement in relationship quality may help to mitigate the impact of prenatal exposure to neighbourhood disorder, such as less safety and more stress (65). Together with previous evidence of the positive psychosocial effects of parent-child reading interventions (64), our results point to this as an important target for public health and community programming (e.g., awareness campaigns, access to books and inclusive reading spaces), particularly for young children living in neighbourhoods with higher levels of disorder.

At the within neighbourhood level two individual characteristics, child sex and preterm birth, were significantly associated with ASQ-3 total scores at five years of age. Consistent with previous research, children reported as male or born before 37 weeks gestation had poorer overall development outcomes. For example, younger gestational age was uniquely associated with increased odds of speech, language, and communication concerns in a statistical model with neighbourhood deprivation (56). Male children have also been observed to have more behavioural problems in a statistical model with neighbourhood social conditions (51). Further research is needed to examine potential moderating effects of individual characteristics to identify who is more vulnerable or resilient to the impact of neighbourhood characteristics on early development (31).

Contrary to expectations, community resource use was not found to have a significant buffering effect against neighbourhood deprivation or disorder, nor was it directly associated with overall child development. These findings may be due to our measure of community resource use mainly capturing participants' own use of community resources rather than their children's direct participation (32, 33). Similarly, no significant moderation effects were observed for physical activity, nor was it significantly associated with overall child development. Although a systematic review indicated that physical activity was associated with improved motor skills and cognitive development in children aged 4–6 years (35), some of the included studies found mixed or no significant effects. Given that parent-report was used to measure children's physical activity and intensity was not distinguished (66), it is possible that we may not have captured the full extent of children's physical activity participation. Future studies in this area could consider using more objective measurements of physical activity, such as standardized questionnaires or accelerometers, as well as capturing variation in the intensity of physical activity that children participate in.

The current investigation utilized a large Canadian cohort study and advanced multilevel analyses to identify early social determinants of individual well-being, and represents a novel examination of prenatal exposure to neighbourhood deprivation on child development. Moreover, potential modifiable protective factors were explored including community resource use, physical activity, and parent-child reading. However, the results of the current investigation should be interpreted within the context of several limitations. Although the use of parent reported measures of child development is common and feasible in population-based birth cohorts (67, 68), potential bias could be mitigated by using multiple informants as well as observational or experimental methods. While the ASQ-3 is a well-validated screening tool for developmental delays, further assessment would be required to examine clinical diagnoses. Given neighborhood deprivation and disorder data were only available for pregnancy, it was not possible to isolate the impact of exposure during this period while accounting for the first five years postpartum. Future research examining cumulative effects of neighbourhood characteristics is warranted. Although several potential confounding variables were included as covariates, it is possible that there are others for which we did not have data such as physical health of children (35). Finally, there was a lack of variability in our sample in terms of sociodemographic and neighbourhood characteristics. In general, the socioeconomic deprivation of neighbourhoods in Calgary is relatively low and residents are more advantaged compared to the larger Canadian population (55). Thus, our findings may not be generalizable to more vulnerable families, or to areas where there are larger differences in socioeconomic deprivation between neighbourhoods.