Socio-Environmental Typologies and Incidence of Dental Caries from Childhood to Adolescence among Youth in Montreal: The QUALITY Cohort Open Access

Dental caries, the most prevalent oral and noncommunicable chronic disease worldwide [1], can occur as soon as the first tooth erupts and accumulates over time, posing a significant oral health threat to youth [2]. In 2019, 46.2% and 53.8% of the global paediatric population experienced dental caries in deciduous and permanent teeth, respectively [3]. More importantly, despite the implementation of various prevention strategies – such as health education and sealants – inequities in dental caries persist [4‒6]. That is, children in lower socio-economic positions (SEPs), typically characterized by lower parental education, lower household income, and lower parental occupational prestige [7], experience a higher risk of dental caries. This disparity can be partially attributable to the lack of consideration of the influence from “upstream” factors, such as poverty and food insecurity, which are the social and economic manifestations of the unequal and unjust allocation of power and resources across social classes beyond individual control [8]. However, they constitute the social environment that profoundly influences individual behaviours and health outcomes [8].

Previous studies have identified several socio-environmental factors that may protect against dental caries among youth, including higher SEP, higher social capital (resources and social norms shared within a social group), strong social interaction, and enhanced community safety [9]. However, few have adopted a longitudinal design. In addition, a holistic approach integrating multiple social-environmental domains is rarely considered despite the likely complex interactions between these factors. Leveraging the robust infrastructure of an ongoing cohort study among Quebec youth with a family history of obesity, this study aimed to comprehensively examine how school and residential neighbourhood social environments could influence dental caries development from childhood to adolescence. Specifically, we will (i) identify socio-environmental typologies by combining exposures from multiple domains of the neighbourhood social environment in childhood and (ii) estimate the association between the identified socio-environmental typologies and the incidence of dental caries in adolescence.

We used data from an ongoing cohort study, the Quebec Adipose and Lifestyle Investigation in Youth (QUALITY) Study [10], which aimed to investigate the natural history of childhood obesity and its cardiometabolic consequences, including oral health.

To minimize the influence of genetic heterogeneity, children of Western European ancestry aged 8–10 years with at least one biological parent with obesity were recruited from schools located in three major urban centres (i.e., Montreal, Quebec City, Sherbrooke) of Quebec, Canada. This specific age group was selected to ensure follow-up occurred before the onset of puberty while also facilitating reliable data collection from children. Children with a history of diabetes, serious illness, or conditions that hindered participation, or receiving treatment with hypertensive medication or steroids were excluded. In addition, families planned to move out of the province, or with the biological mother being pregnant or breastfeeding during baseline assessment, were excluded.

A total of 630 families completed baseline assessments during 2005–2008; 564 completed the first follow-up during 2007–2010; and 377 completed the second follow-up (FU2) during 2012–2015. Participants dropped out because of children’s fear of venipuncture, families’ lack of time for data collection procedures, and families’ moving out of the study area. Because key environmental exposures were collected in the context of a Montreal-based ancillary study, we limited our study to participants from Montreal (N = 512 at baseline).

Data collection was performed at the Unité de recherche clinique du Centre Hospitalier Universitaire (CHU) Sainte-Justine in Montreal, where participating families spent a full-day visit.

At each visit, interviewer-administered questionnaires collected information on demographics (e.g., age, sex), lifestyle behaviours (e.g., toothbrushing, diet), and general health conditions (e.g., oral health) from children. Parents completed self-administered questionnaires to provide information on sociodemographic, economic (e.g., income), and behavioural information.

At baseline, independent pairs of trained research assistants conducted on-site audits to characterize the residential neighbourhoods using an observation grid adapted from existing validated tools [11]. The grid included a checklist of 60 items assessing street-level social and built environmental features (e.g., land use and design) and the general impression (e.g., presence of green space and signs of social disorders, which are explained below). Audits covered up to ten street segments surrounding each participant’s residence, representing a 500-m road network. The ten segments included one street segment where the family was located and up to nine first- and second-degree connecting streets. Disagreements between research assistants were resolved through additional audits by and discussions with a third observer on another occasion.

Data from the 2006 Canadian Census were also used in this study. The smallest unit for which the census data were available is the “dissemination area” (DA), defined as a spatial subunit generally consisting of one or more neighbouring blocks of houses with a population of 400–700 persons [12]. Aligning with the scope of on-site audits, data were collected for all DAs within the 500-m road network buffers surrounding residences and schools.

At each visit, dental examinations on children were performed by several dentists who had been well trained and calibrated, using the 2003 Children’s Dental Health Survey diagnostic criteria [13]. To evaluate the degree of dental caries, the Decayed, Missing, Filled Surfaces (DMFS) index was calculated, which counts the number of tooth surfaces with decay, filling, or missing due to decay. Higher DMFS scores indicate more affected tooth surfaces. Details of the evaluation criteria are provided in online supplementary material 1 (for all online suppl. material, see https://doi.org/10.1159/000546747). Briefly, decay included those with or without cavities if visual colour change (white or brown spot) or opacity change on the tooth surface was detected after air-drying the tooth surface for 5 s; recurrent decay with or without cavity around a filling was also considered. Missing permanent tooth (surfaces) at baseline (when children aged 8–10 years) was mostly due to non-eruption, which we considered as caries-free; for other conditions (i.e., trauma, orthodontic treatment, extraction due to serious decay), questions like “was the tooth removed to make space for others” or “did this front tooth fall out or was it removed after an injury or a blow?” were asked to confirm the reason with the parents and children, and the tooth (surfaces) was not counted in subsequent analyses.

Typically, children spend a significant portion of their daily lives within and around schools and homes. Thus, our exposure was the socio-environmental typologies combining both residential and school neighbourhoods, which were characterized by three indicators: social disorder, social deprivation index, and material deprivation index. The selection of these indicators was based on a combination of several reasons: (i) they have been widely investigated in literature and have established associations with health; (ii) they captured neighbourhood socio-environmental features at either street- or area-level, and (iii) data availability. Details of how these indicators were measured can be found in previous studies [14] as well as in online supplementary material 2. Briefly, social disorder was only measured in residential neighbourhoods using on-site audit data, based on the presence of any of the four signs: graffiti, vandalism, litter, and abandoned buildings/lots. The social and material deprivation indices were measured in both residential and school neighbourhoods, using the 2006 Canadian Census data [15]. The social deprivation index was based on the proportions of single-parent families, people living alone, and separated, divorced, or widowed individuals. The material deprivation index was based on the proportions of people without a high school diploma, unemployed, and low-income households. DAs were then ranked based on the quintiles of the two deprivation indices, with the first quintile being the most privileged and the last quintile being the least.

The outcome variable was the incidence of dental caries, calculated as the difference in DMFS scores between baseline and FU2 (∆DMFS).

Using a directed acyclic graph [16] (DAG, see online suppl. material 3, Fig. S1), we identified the minimal set of confounders for adjustment. All confounders were measured at baseline, including the child’s age, sex, DMFS, oral health behaviours, and parental SEP.

Oral health behaviours were evaluated through two variables: daily consumption of sugar-sweetened beverages (SSBs) and toothbrushing frequency. Average daily SSB consumption, measured in millilitres (mL), was calculated using information from a 24-h diet recall collected on three non-consecutive days (two weekdays and a weekend day) as described before [17]. Toothbrushing frequency, collected from questionnaires, was initially measured in four categories (≥three times/day, twice/day, once/day, and <once/day). Based on the recommendation to brush teeth at least twice per day [18], we further consolidated the four categories into two: ≥twice/day and <twice/day.

Parental SEP was indicated by annual household income, which was collected from questionnaires based on 12 categories, ranging from less than CAD 10,000 to more than CAD 140,000. Then, the variable was adjusted for the number of people living in the same household [19].

The analyses comprised two parts, addressing our two objectives. First, we used latent class analysis (LCA), a model-based method with a probabilistic approach [20], to identify socio-environmental typologies of residential and school neighbourhoods based on the observed indicators. Participants with complete data were included (N = 499). Model selection was guided by the Akaike information criterion and Bayesian information criterion.

Second, we estimated the associations between identified typologies and the incidence of dental caries using negative binomial regressions, which is suitable for modelling count variables (in our case, ∆DMFS) while not requiring equal mean and variance of the data (the assumption of Poisson regression). The total number of caries-free tooth surfaces at baseline was calculated and included as an offset term. Three models were fitted: model 1 (crude estimates), model 2 (adjusted for baseline annual household income), and model 3 (further adjusted for child age, sex, DMFS, toothbrushing frequency, and SSB consumption at baseline). Incidence ratios (IRs) with corresponding 95% confidence intervals (CIs) were computed. Only participants with complete data (N = 281) were included. A flowchart detailing data attrition throughout the study is provided in online supplementary material 3, Figure S2.

To examine possible biases introduced by missing data, sensitivity analyses using multiple imputations with chained equations were performed [21], using the “mice” R package (v3.16.0) (30 iterations, 20 imputed datasets) [22]. Estimates were pooled together to compare with the results from complete-case analyses. In addition, the inverse probability of censoring weighting was applied after multiple imputations to assess bias from loss to follow-up [23]. Weights of censoring were estimated using the covariate balancing propensity score method [24] via the “weightthem” function from the “MatchThem” R package (v1.2.1) [25], based on socio-environmental typologies, annual household income, child age, sex, DMFS, toothbrushing frequency, and SSB consumption measured at baseline (online suppl. material 3 Fig. S3). All statistical analyses were performed in RStudio (4.0.2).

Online supplementary material 3 Table S1 displays the results of LCA on socio-environmental features of residential and school neighbourhoods. The model with three typologies was retained based on the lowest Bayesian information criterion observed, the parsimony rule, and model interpretation. Figure 1 presents the conditional probabilities of different responses to all five indicator variables across three typologies. That is, the probability of observing a specific response on the corresponding indicator given that a subject belongs to a particular latent class (in this case, a particular typology). Accordingly, the typologies are summarized below:

Type 1 (N = 147, 29.4%): high social disorder, high social deprivation, but low material deprivation (least favourable). Type 2 (N = 193, 38.7%): median social disorder, median social deprivation, and median material deprivation. Type 3 (N = 159, 31.9%): low social disorder, low social deprivation, but high material deprivation (most favourable).

Table 1 describes the distribution of variables across the three typologies among the 499 participants (271 boys, 228 girls) included in LCA. Type 3 neighbourhoods had the highest annual household income (48.1 ± 17.5 × CAD 1,000), the lowest daily consumption of SSB (128.9 ± 131.3 mL), and the lowest DMFS at both baseline (0.4 ± 1.0) and FU2 (4.8 ± 4.9).

Table 2 shows the estimates of the associations between identified socio-environmental typologies and dental caries incidence. After adjusting for both family- and individual-level covariates, compared to type 1, the least favourable neighbourhoods, children living in both type 2 (IR: 0.9; 95% CI: 0.6∼1.2) and type 3 (IR = 0.8, 95% CI: 0.6–1.1) neighbourhoods had a lower incidence of dental caries at FU2, though the estimates were not statistically significant.

The results of the sensitivity analyses are presented in online supplementary material 3, Figure S3, and Table S2. Similar estimates were obtained after using multiple imputations and inverse probability of censoring weighting, suggesting limited influence from missing data and loss to follow-up on our study results.

In this study, by combining multiple domains of socio-environmental factors measured in residential and school neighbourhoods, we identified three distinct socio-environmental typologies. In addition, we estimated the associations between the identified socio-environmental typologies and the incidence of dental caries among youth. Interestingly, we observed that children residing in neighbourhoods with lower social disorder, lower social deprivation, but higher material deprivation may experience a lower incidence of dental caries, compared to those in neighbourhoods with higher social disorder, higher social deprivation, but lower material deprivation. While these findings are consistent and suggest a potential buffering of low social disorder and low social deprivation against the adverse influence of high material deprivation, the statistical power was limited. Thus, further research with a larger sample size and other populations is warranted to validate the association between a favourable social environment and dental caries.

Deprivation, as a multidimensional concept, encompasses various unmet needs such as food, housing, social interactions, and social support [15, 26], all of which act as “upstream” socio-environmental factors shaping health conditions across populations [8]. Many studies used single-item indicators, such as income and female-headed household [27, 28], which may be insufficient to capture the multidimensional nature of deprivation. In this study, we addressed this challenge by including the measurements of social disorder, as well as two well-validated indices, which measured the overall social and material deprivations by incorporating various indicators of deprivation that are known to be associated with health [15].

The classification of social and material deprivation was proposed by Townsend [26] in 1987, which provides a straightforward framework to summarize and interpret the diverse dimensions of deprivation, thereby being widely used in the health literature. It is noteworthy that people may live in areas that are materially deprived but that are not necessarily socially deprived, and vice versa. The health impacts can vary significantly between those experiencing only one type of deprivation and those experiencing both. Therefore, focusing solely on one aspect of deprivation offers a limited perspective on how the social environment affects health outcomes [15]. Our study further confirmed this argument with findings of associations, though not statistically significant, between neighbourhoods with lower social disorder and social deprivation, but higher material deprivation and dental caries incidence. These suggest a possible buffering effect of low social disorder and social deprivation on the negative influence of high material deprivation in terms of dental caries. Such a phenomenon has also been reported in other studies, although they only used single-item indicators of deprivation [27, 28].

A possible explanation for this buffering effect is that for people with limited access to resources such as food and healthcare, or with a higher socio-economic burden (e.g., low income, loss of job), enhanced social interactions may provide them with stronger social support in the form of instruments (e.g., cash, food, labour), information (e.g., advice, guidance), or emotion (e.g., trust, empathy) [29], which is a coping strategy to increase their resilience and self-efficacy (one’s confidence to be able to perform a specific behaviour) [30], thereby mitigating the influence of low SEP on health [31]. Instead, people with higher SEP will have higher self-reliance and countless on support from others [31]. Accordingly, they may also present a lower interest in connecting with others, thereby benefiting less from social interactions [31]. This may also partially explain the opposite levels of material and social deprivation we observed across the three neighbourhood types.

Another paradox was observed in type 1 and type 3 neighbourhoods where opposite levels of material deprivation and average household incomes, i.e., high material deprivation but low household incomes or vice versa, were found. Material deprivation was an ecological variable derived from consensus data collected by DA, while household income was aggregated based on data collected from individuals. Since the QUALITY cohort was of higher SEP than the general Quebec population at baseline [10], there could be a discrepancy in the distributions of the two variables. One may question that the observed lower dental caries incidence in type 3 neighbourhoods was due to a higher average household income (similar for type 1 neighbourhoods). Indeed, higher SEP alone, either measured at the individual or environmental level, has been associated with a lower risk of dental caries [6, 9]. However, a DAG was used to help mitigate the potential confounding effect of individual SEP on dental caries.

Our findings provide some implications for policy-making, underpinning the need for a comprehensive approach to develop new effective prevention programs. Indeed, strategies focusing on a solitary factor or dimension of the social environment are inadequate to address the high prevalence of dental caries and its associated inequities. For example, in Canada, despite universal access to healthcare, dental coverage is only available for children under 10–12 years old (varied across provinces) [32]. This may partially explain the big increase in DMFS scores at FU2 of the QUALITY cohort. In the Netherlands, where children receive full dental coverage until the age of 18 years, those from lower SEP families still experienced a greater incidence of dental caries [33]. This inequity may stem from the fact that people with lower SEP are more likely to encounter challenges in accessing, comprehending, evaluating, and implementing healthcare information, due to inadequate time, material resources, and available social support [34].

Most current interventions for dental caries are community-based or school-based. A meta-analysis reported greater effectiveness in reducing caries through an inclusive and holistic approach that involves wider social circles, e.g., families and communities [35]. Our study reinforces this finding and further implies the potential importance of a neighbourhood with lower social disorder and social deprivation in preventing dental caries. Thus, future interventions targeting the paediatric population could consider a community-based comprehensive approach that can facilitate social interactions and social stability within neighbourhoods.

The results of this study should be interpreted in light of some limitations. First, loss to follow-up and missing values are inherent challenges in cohort studies, potentially introducing selection bias [36]. However, the results of the sensitivity analyses suggested a negligible effect of missingness. A DAG was used to identify confounders to help mitigate potential confounding bias. Nevertheless, influence from unknown and unmeasured confounders cannot be ruled out. Some factors were excluded from our analyses as we negotiated model simplicity, sample size, and result robustness. For example, while we recognize the influence of dental insurance coverage on dental caries, 22% of participants were unsure of their dental health insurance status at FU2. Among those with known status, the distribution of insurance coverage (yes or no) did not differ across neighbourhood typologies. This remained the same even when we assumed that the uncertain cases were either all covered or all not covered by insurance. Additionally, while nearly 70% of participants reported being partially or fully covered by private dental health insurance, the wide variation in coverage levels posed extra difficulties in quantifying its influence on our outcomes. For dietary habits, only the consumption of SSB was included in this study because SSB is rich in free sugars, which is one of the most well-established risk factors for caries [37]. Instead, intrinsic sugars contained in natural products, such as grains, whole fruits, milk, and vegetables, are much less important in caries development [37]. Some factors may affect the generalizability of our results. First, the QUALITY cohort included only participants of Western European ancestry to minimize genetic heterogeneity [10]. Also, children in this cohort had a family history of obesity, suggesting a potential difference in their underlying health attributes compared to the general population. Nevertheless, our study provides some insights into caries development among children who are more prone to obesity. Additionally, the cohort exhibited a more favourable baseline SEP; this may explain why we did not identify a neighbourhood with high social disorder, high social deprivation, and high material deprivation [10]. Moreover, we limited our study population to those in Montreal, a culturally diverse metropolitan city compared to other cities in Quebec. Yet, the baseline prevalence of dental caries in the permanent teeth of our study population (23.8%) closely parallelled that reported in the Canadian Health Measures Survey among children aged 6–11 years (24%) between 2007–2009 [38], indicating a fair representation of our study population. Lastly, our study only focused on the social environment in neighbourhoods, while previous studies have emphasized the effect of the physical environment, such as the food environment, on human health [39, 40]. Future research is needed to investigate the interactions between social and physical environments and their influence on children’s oral health.

There are also several strengths in this study. The cohort study design enabled us to investigate how the social environment longitudinally affects youths’ oral health from childhood to adolescence. In addition, we adopted a holistic analytical approach, integrating multi-level (i.e., individual, family, and neighbourhood) risk factors of dental caries. Furthermore, our investigation encompassed the social milieu of both school and residential neighbourhoods, which are the primary spaces where children typically spend a significant portion of their daily lives. Finally, we characterized the neighbourhood social environment by creating typologies integrating diverse factors, providing a comprehensive understanding of the complex interactions between various domains of the social environment.

Compared to neighbourhoods characterized by high social disorder and high social deprivation, neighbourhoods with lower social disorder and lower social deprivation may be associated with a lower incidence of dental caries in adolescence.

We would like to thank all participants in the QUALITY cohort. The QUALITY cohort research was conducted by members of Team QUALITY, an inter-university research team including Université de Montréal, Concordia University, Université Laval, McGill University, and INRS-Institute Armand-Frappier.

The QUALITY cohort study protocol was received and approved by the review board of the Centre Hospitalier Universitaire Sainte-Justine (#MP-21-2005-79, 2040) and performed in accordance with the Declaration of Helsinki. All participating children and their parents provided written informed assent and consent, respectively.

The authors have no conflicts of interest to declare.

The QUALITY cohort was funded by the Canadian Institutes of Health Research (#OHF-69442, #NMD-94067, #MOP-97853, and #MOP-119512), the Heart and Stroke Foundation of Canada (#PG-040291), and Fonds de Recherche du Québec-Santé.

Y.Y. contributed to conception, analysis, interpretation, and manuscript drafting. B.N. contributed to conception, acquisition, and critical review of the manuscript. A.H. contributed to analysis and critical review of the manuscript. M.H. contributed to acquisition and critical review of the manuscript. S.A.M. contributed to analysis, interpretation, and critical review of the manuscript. T.A.B. contributed to conception and critical review of the manuscript.

The data that support the findings of this study are not publicly available due to privacy concerns. Datasets may be available from M.H. ([email protected]) upon reasonable request.