The PROTECT databank a population based linked administrative resource on child maltreatment and intellectual disability with early findings

Abstract

Children with intellectual disability (ID) experience disproportionate social and health adversities; however, population-level evidence integrating child protection and healthcare data remains scarce. We introduce the Child Development and Protection Pathways in Intellectual Disability (PROTECT) databank a province-wide linked administrative infrastructure in Québec, Canada, and report initial findings from a population-based retrospective longitudinal cohort study (N = 18,095). Using linked child protection records, physician billing claims, and hospitalization data, we identified children born between 2001 and 2016 and residing in six administrative regions. ID was defined using outpatient and inpatient diagnostic codes, and maltreatment was defined as at least one child protection report retained for evaluation before age 18. Children were classified into four mutually exclusive cohorts based on the presence or absence of ID and child maltreatment. We estimated the prevalence and cumulative burden of mental and physical health service use and compared cohorts using negative binomial regression models. Children with co-occurring ID and maltreatment experienced the highest cumulative burden of service use, whereas those with neither exposure had the lowest. Across outcomes, the presence of ID was the primary correlate of elevated burden, with the highest mental health burden among children with dual exposure. PROTECT establishes a scalable population-level resource to examine intersecting developmental and protection-related risks and provides actionable evidence to inform strategies aimed at reducing health inequities among children with developmental vulnerabilities.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-61063-y.

Introduction

Child maltreatment represents a major public health concern in Canada, with persistently high prevalence and incidence rates^1,2. Recent population-based survey data indicate that more than 34% of young adults aged 18 to 27 report exposure to at least one form of child maltreatment during childhood¹. Complementing these retrospective self-report estimates, the Canadian Incidence Study of Reported Child Abuse and Neglect documented an estimated 48 maltreatment-related investigations per 1,000 children in 2019, with neglect and exposure to domestic violence among the most frequently identified concerns³. Together, these data highlight both the widespread lifetime burden of maltreatment and the substantial number of children who come to the attention of child protection authorities annually. In this study, child maltreatment is defined as physical, sexual, or psychological abuse, exposure to domestic violence, or neglect occurring before the age of 18 and likely to compromise a child’s health, safety, or development^4,5.

In Québec, child protection data similarly highlight the magnitude of the issue. Approximately 10% of children will experience at least one substantiated report of maltreatment under provincial child protection legislation during childhood⁶. However, these aggregate estimates mask important inequalities in maltreatment exposure, as some populations face disproportionately higher risks. Among these, children with neurodevelopmental conditions are consistently identified as having higher rates of child maltreatment^7–9. Within this group, children with intellectual disabilities (ID), a neurodevelopmental condition with onset before age 22, characterized by significant limitations in intellectual functioning and in adaptive functioning across conceptual, social, and practical domains¹⁰, appear to be at particularly elevated risk of maltreatment^11–13.

The elevated risk of abuse among children with ID is another major public health issue^14,15, with available evidence suggesting that these children may be several times more likely to be maltreated before adulthood than their peers without disabilities¹¹. Estimates of the magnitude of this excess risk vary considerably across studies, however, reflecting differences in study design, case ascertainment, and population, as well as the substantial heterogeneity and moderate methodological quality that characterize much of this literature. A systematic review and meta-analysis by Fang et al¹¹. estimated that children with disabilities had approximately twice the odds of experiencing violence compared with non-disabled children, while noting that heterogeneity across estimates remained high. Notably, this review also found that children with cognitive or learning disorders were among those at greatest risk, with more than twice the odds of violence victimization relative to non-disabled children. Children with ID may also be at higher risk of experiencing certain forms of abuse, particularly sexual abuse, than children with other types of disabilities (e.g., physical or sensory), although such comparisons should be interpreted with caution given the methodological limitations noted above. Consequently, children with ID account for up to 15% of abused children in Québec and Canada, depending on the type of abuse examined¹⁶, even though the overall prevalence of ID is 1%^17,18. These findings underscore both the emerging nature of this research field and the methodological and population-level gaps that persist.

In general, children with ID face a three- to five-fold higher risk of mental health problems¹⁹. Developmental psychopathology theory and research in general suggesting the central role of child maltreatment for subsequent mental health difficulties^20,21. Although short- and long-term adverse physical, cognitive, and mental health consequences have been associated with child maltreatment among children with ID^22–24, very few studies have examined population-based comparative cohorts capable of disentangling developmental vulnerability from maltreatment exposure or identifying risk and protective factors specific to this population. Identifying the role of maltreatment in explaining mental health disparities among children with ID is essential for promoting approaches that will reduce their mental and physical health burden. Moreover, the field remains constrained by persistent methodological limitations, including small or clinically derived samples, limited consideration of developmental comorbidity, and a limited number of longitudinal population-based designs capable of examining cumulative and comparative trajectories.

Québec is Canada’s second most populous province, accounting for approximately 23% of the Canadian population, with an estimated 1.4–1.6 million children under 18. The overall prevalence of ID in the child population is approximately 1%^17,18. Regarding child protection, Esposito et al.⁶ established that over 18% of Québec children were reported to child protection at least once before age 18, and that approximately 10% had a report retained with a finding that their security or development was compromised. In Québec, the absence of large-scale, integrated, population-based data limits the ability to document the prevalence of maltreatment among children with ID, its associations with mental and physical health, and their trajectories through health and social service systems. Fragmentation across administrative sectors further constrains efforts to generate evidence capable of informing equitable clinical, organizational, and policy responses^25,26. To address these gaps, population-based linked administrative databases integrating child protection data and health records could be considered an essential and emerging research tool.

This study introduces the Child Development and Protection Pathways in Intellectual Disability (PROTECT) databank, a population-based linked administrative infrastructure designed to document health, neurodevelopmental, and child protection trajectories among children with and without ID and with and without exposure to maltreatment. We pursue two objectives. First, we describe the design, data sources, linkage procedures, and cohort structure underpinning PROTECT. Second, we present initial descriptive analyses across four mutually exclusive cohorts defined by the presence or absence of ID and maltreatment, focusing on the prevalence and cumulative burden of mental and physical health service use.

Methods

We conducted a population-based retrospective longitudinal cohort study using linked administrative data from Québec, Canada. The PROTECT databank integrates child protection and health records to enable longitudinal examination of health and service trajectories across childhood and adolescence.

Data sources

Data were drawn from provincial administrative databases, including physician billing claims from the Québec Health Insurance Board (RAMQ), hospitalization discharge abstracts from the Québec Ministry of Health and Social Services (MSSS), injury records from the Québec Trauma Registry Information System, and mortality data from the provincial death registry. These databases were deterministically linked to the Child Protection data using encrypted identifiers within the secure research environment of the Institut de la statistique du Québec (ISQ).

Québec health administrative databases

Québec Health Administrative Databases contain demographic and eligibility information for all individuals covered under Québec’s universal health insurance system. Variables extracted included encrypted personal identifiers, year and month of birth, sex, region of residence, partial postal code, and date of death (when applicable). Eligibility data from the public drug insurance plan were also used to determine periods of medication coverage. Physician billing claims provided information on outpatient medical services, including service dates, diagnostic codes using the International Classification of Diseases, 9th and10th Revision (ICD-9/10), provider specialty, and act codes. Pharmaceutical services data included dispensed medications and pharmacist services. Hospitalization databases include hospital discharge data that capture all hospitalizations and day surgeries occurring in Québec hospitals. Extracted variables included admission and discharge dates, length of stay, type of care, intensive care unit stays, hospital service units, medical diagnoses, and procedural interventions. These data are used to identify ID, physical health conditions, mental health diagnoses, hospitalizations, and procedures.

Trauma and death registries

Data were drawn from two complementary provincial administrative sources. The Québec Trauma Registry Information System contains detailed information on traumatic events and injury characteristics for patients treated in designated trauma centers. Extracted variables included event characteristics such as dates, mechanism of injury, and clinical injury information. The Québec Death Registry captures all deaths occurring in Québec, documenting the date and place of death, and underlying cause of death.

Child protection data

Child protection data were obtained from the provincial administrative records of youth protection services. These administrative databases document all reports made under the Youth Protection Act of Québec, including information on report intake, screening decisions, substantiation and type of maltreatment (e.g., neglect, physical abuse, sexual abuse, psychological maltreatment, exposure to domestic violence), court orders, and service trajectories which include detailed service information, such as application of protective measures, duration of services, placement episodes, number of placement changes, reunification attempts, and cumulative duration under protection. All reports reaching the investigation stage were retained regardless of substantiation status, as unsubstantiated reports should not be interpreted as evidence that maltreatment did not occur. Non-substantiation may instead reflect limitations in evidence gathering, caseworker discretion, or variation in substantiation thresholds across cases. Whether outcomes differ between children with substantiated versus unsubstantiated reports represents an important avenue for future investigation.

Linkage procedure and selection criteria

ISQ, the provincial statistical authority responsible for secure administrative data integration, conducted data linkage and cohort selection. ISQ linked child protection records from child and youth protection centres with provincial health administrative data from the Québec health insurance registry. Children were required to (1) be born between April 1, 2001, and December 31, 2016; (2) be eligible for provincial health insurance coverage for at least one day before age 6; and (3) reside, before age 18, in one of six participating health regions, selected to capture geographically and demographically diverse territories in Québec (regions masked for review).

Cohort definition and extraction period

Children were classified into four mutually exclusive cohorts based on two exposures: intellectual disability (ID) and documented child maltreatment. The cohorts included children with both ID and maltreatment, children with ID only, children with maltreatment only, and children with neither ID nor maltreatment. Cohorts 1 and 2 cover all eligible children with an ID diagnosis in the six selected regions, identified through RAMQ physician billing codes and hospitalization records. In contrast, Cohorts 3 and 4 (children without ID) were drawn from randomly selected samples of the same underlying population. As a result, children with ID are overrepresented in the analytic sample relative to population prevalence. This design was implemented to ensure adequate statistical power for comparisons involving ID.

Consent statement, declaration of ethics & data storage

The PROTECT databank adheres to the principles of the Declaration of Helsinki for research involving human participants. This study relied exclusively on linked administrative health and social service records and did not involve direct contact with participants, biological samples, or the collection of primary data. As the study consisted of a retrospective analysis of existing administrative databases, the requirement for individual informed consent was formally waived by the Institutional Review Board (Comité d’éthique à la recherche of the Centre intégré universitaire de santé et de services sociaux de l’Estrie–Centre hospitalier universitaire de Sherbrooke (CÉR CIUSSS de l’Estrie-CHUS). Ethics approval from the CÉR CIUSSS de l’Estrie-CHUS was obtained prior to requesting access to the data (Project # MP-31–2021-3726). Authorization to access and link the administrative datasets was subsequently granted by the Institut de la statistique du Québec (ISQ). The use and disclosure of personal information for research purposes were conducted in accordance with Québec’s Act respecting Access to documents held by public bodies and the Protection of personal information (CQLR, c. A-2.1). All analyses were performed within the ISQ secure research environment in compliance with provincial data governance and privacy protection regulations.

All data were anonymized before researcher access in accordance with provincial privacy regulations. Individual-level records were linked using encrypted unique identifiers derived from the provincial health insurance registry. Only individuals with a valid and active provincial health insurance number were eligible for inclusion, ensuring data completeness and consistency across linked databases. Residential information was obtained from the provincial health insurance registry.

Measures

Although the PROTECT databank contains an extensive set of variables (see Supplementary Table S1), the present study focuses on the following key measures.

Child maltreatment

Child maltreatment was defined as any report to the Québec Child Protection Services (CPS) recorded in administrative child protection data during the observation period. Maltreatment status was based on official CPS records and included reports that were retained for evaluation under the Youth Protection Act. Since a report, whether substantiated or not, is an indicator of a situation at risk of maltreatment, children were classified as maltreated if they had at least one documented report meeting these criteria before age 18.

Intellectual disability

Children with ID were identified using physician billing claims and hospitalization discharge records. Individuals were classified as having ID if they had received at least one predefined ICD-9 or ICD-10 diagnosis corresponding to ID before age 18 between April 1, 2001, and December 31, 2022. Four mutually exclusive cohorts were then defined according to the presence or absence of ID and documented child maltreatment.

Mental health

Mental health diagnoses were identified using physician billing claims and hospitalization records. Diagnoses were coded according to the International Classification of Diseases, Ninth Revision (ICD-9; 67.2%) and Tenth Revision (ICD-10; 21.4%), and harmonized across classification systems. Mental disorders were defined based on ICD-10 (F00-F99) and corresponding ICD-9 codes, excluding ID diagnoses used to define the study cohorts (See Supplementary Table S2.3). Children were classified as having a given diagnostic category if at least one corresponding ICD code was recorded in outpatient data between April 1, 2001, and December 31, 2022, before age 18. The total number of presentations with a mental health condition per child was computed by summing all recorded diagnostic codes across encounters. The same code may appear across multiple visits and is counted each time, such that this variable reflects the cumulative burden of health service use rather than the number of unique diagnoses received. A complete list of ICD-9 and ICD-10 codes in each diagnostic category is provided in Supplementary Table S2.1.

Physical health

Physical health disease diagnoses were identified using physician billing claims and hospitalization records. Diagnoses were coded using ICD-9 and ICD-10 and harmonized across classification systems to ensure comparability across the study period. Physical conditions were defined using ICD-10 chapters, and children were classified as having a given physical health disease if at least one corresponding ICD code was recorded. The total number of presentations with a physical health condition per child was computed by summing all recorded codes across outpatient (medical visits and consultations) and inpatient encounters. As with the mental health outcome, repeated recording of the same code across visits is counted cumulatively, reflecting service use burden. A complete list of ICD-9 and ICD-10 codes in each category is provided in Supplementary Table S2.2.

Material deprivation

Area-level material deprivation was measured using the Material Deprivation Index developed by the Institut national de santé publique du Québec (INSPQ).The index is derived from census data on income, education, and employment within each individual’s residential dissemination area, linked through the 6-digit postal code, and is expressed as quintiles ranging from Q1 (least deprived) to Q5 (most deprived). For each participant, we used the earliest available quintile.

Statistical analyses

Descriptive analyses were conducted to characterize the burden of mental and physical health diagnoses across the four cohorts. For each domain, we calculated the proportion of individuals with at least one diagnosis, as well as the total number, mean, and standard deviation of diagnoses per individual. Demographic differences between cohorts (sex, age, material deprivation) were assessed using standardized mean differences (SMDs) rather than null-hypothesis significance tests. Given the large sample size, conventional tests would yield statistically significant differences even for substantively trivial discrepancies. Because diagnostic counts were highly right-skewed and overdispersed, exploratory between-cohort comparisons were performed using negative binomial regression models. Overdispersion parameters and Akaike Information Criterion (AIC) values were examined to confirm model adequacy. The total number of presentations of a mental or physical health condition per individual served as the dependent variable, and cohort membership as a categorical predictor, with Cohort 4 (neither intellectual disability nor maltreatment) as the reference group. Adjusted models included age, sex and material deprivation quintile as covariates, selected based on SMDs exceeding 0.10 in the descriptive comparison. Results are reported as incidence rate ratios (IRR) with 95% confidence intervals.

Results

Sample and sociodemographic description

The PROTECT databank includes 18,095 children born between April 1, 2001, and December 31, 2016, and that were observed till December 2022. Children were classified into four mutually exclusive cohorts based on two exposures: intellectual disability (ID) and documented child maltreatment identified through child protection records. The cohorts included children with both ID and maltreatment (n = 4,007), children with ID only (n = 6,409), children with maltreatment only (n = 4,007; randomly selected representative sample), and children with neither ID nor maltreatment (n = 3,672). Overall, 56% of the sample were boys (n = 10,133). Participants were drawn from six administrative regions of Québec. The mean age at cohort entry for the total sample was 13.9 years (SD = 4.5). Sociodemographic data across cohorts are presented in Table 1. All children were eligible for provincial health insurance coverage, ensuring consistency across linked administrative data sources.

Table 1.

Sociodemographic characteristics by cohort, with standardized mean differences relative to cohort 4 (N = 18,095).

Characteristic	Total	Cohort 1	Cohort 2	Cohort 3	Cohort 4	SMD vs. Cohort 4
						Cohort 1	Cohort 2	Cohort 3
n	18,095	4,007	6,409	4,007	3,672	—	—	—
Male sex, %	56.0	60.6	58.1	52.4	51.5	0.19	0.13	0.02
Age in years, M (SD)	13.9 (4.5)	14.8 (4.2)	13.8 (4.7)	13.6 (4.3)	13.1 (4.5)	0.41	0.17	0.12
Material deprivation quintile, %						0.33	0.11	0.27
Q1 (least deprived)	16.4	10.0	19.9	12.5	21.6	-	-	-
Q2	17.6	14.0	19.2	15.4	21.2	-	-	-
Q3	20.2	19.7	19.9	20.6	20.6	-	-	-
Q4	20.9	24.2	19.0	22.3	18.9	-	-	-
Q5 (most deprived)	25.0	32.0	22.0	29.2	17.8	-	-	-

Note. Values are percentages unless otherwise indicated. Cohort 1 = intellectual disability and maltreatment; Cohort 2 = intellectual disability only; Cohort 3 = maltreatment only; Cohort 4 = neither condition (reference group). The total sample includes 10,133 males (56%). SMD = standardized mean difference, computed for each exposed cohort relative to Cohort 4; values greater than 0.10 (in bold) indicate a meaningful imbalance warranting covariate adjustment. For material deprivation quintile, the value shown is the maximum SMD across the five quintile categories. Material deprivation quintiles are derived from the INSPQ Index of Material and Social Deprivation, matched to each individual’s earliest available observation year (Q1 = least deprived, Q5 = most deprived). M = mean; SD = standard deviation. Percentages may not total 100 because of rounding.

Burden of mental health service use

Across the four cohorts (N = 18,095), 63.8% of individuals had at least one recorded presentation for a mental health condition during the observation period. The mean number of presentations for a mental health condition was 7.27 (SD = 15.4) per individual. The most frequent diagnostic categories among these presentations were disorders of childhood and adolescence (46%), developmental disorders excluding ID (21%), and neurotic and stress-related disorders (18%). Presentations for behavioral conditions were also common (8%). The burden of mental health service use varied markedly across cohorts (Table 2). Children with ID and maltreatment (Cohort 1) presented the highest burden, with the greatest total and mean number of presentations, followed by children with ID without maltreatment (Cohort 2). Lower burdens were observed among children without ID (Cohorts 3 and 4), with those without maltreatment (Cohort 4) showing the least burden overall.

Table 2.

Number of presentations with a mental health condition per individual by cohort.

Cohort	n individuals	Total	M (SD)
Cohort 1	4,007	58,543	14.61 (24)
Cohort 2	6,409	48,132	7.51 (13)
Cohort 3	4,007	17,802	4.44 (11)
Cohort 4	3,672	6,438	1.71 (5)

Note. Values represent the number of recorded mental health diagnoses per individual over the observation period. Cohort definitions were as follows: Cohort 1: children with intellectual disability (ID) and documented child maltreatment; Cohort 2: children with ID without documented child maltreatment; Cohort 3: children without ID with documented child maltreatment; and Cohort 4: children without ID and without documented child maltreatment.

Exploratory comparison of mental health service use between cohorts

Differences in the number of presentations for a mental health condition across cohorts were examined using count regression models (Table 3). A Poisson regression model was initially estimated, but substantial overdispersion was observed, and model-fit comparisons based on the Akaike Information Criterion (AIC) favored a negative binomial model. To account for demographic differences between cohorts, the negative binomial model was adjusted for sex, age, and material deprivation quintile, with Cohort 4 (neither intellectual disability nor maltreatment) as the reference group. After adjustment, all three exposed cohorts showed significantly higher rates of presentations than Cohort 4, with the largest difference in Cohort 1 (IRR = 6.73), followed by Cohort 2 (IRR = 3.98) and Cohort 3 (IRR = 2.36), all p <.001. Among the covariates, male sex and older age were each independently associated with a greater number of presentations, whereas material deprivation showed only a modest association that reached significance at the third quintile alone (see Table 3).

Table 3.

Negative binomial regression of the number of f presentations with a mental health condition by cohort (N = 18,095).

Variable	IRR	95% CI	p
Cohort
Cohort 1	6.73	[6.22, 7.28]	< 0.001
Cohort 2	3.98	[3.71, 4.28]	< 0.001
Cohort 3	2.36	[2.18, 2.55]	< 0.001
Cohort 4 (reference)	1.00	-	-
Sex
Female (reference)	1.00	-	-
Male	1.38	[1.31, 1.45]	< 0.001
Age (per year)	1.12	[1.12, 1.13]	< 0.001
Material deprivation quintile
Q1 (reference)	1.00	-	-
Q2	1.05	[0.97, 1.14]	0.240
Q3	1.12	[1.03, 1.21]	0.009
Q4	1.07	[0.99, 1.16]	0.086
Q5	0.99	[0.92, 1.07]	0.860

Note. IRR = incidence rate ratio; CI = confidence interval. Estimates were obtained from a negative binomial regression model with a log link, adjusting simultaneously for cohort, sex, age, and material deprivation quintile. A negative binomial specification was used to accommodate overdispersion in the count of mental health diagnoses, which a Poisson model would not adequately capture. Cohort 4 (neither intellectual disability nor maltreatment) served as the reference category; female sex and the least deprived quintile (Q1) were the reference categories for the respective covariates. Age was modeled continuously, with the IRR expressed per one-year increase.

Burden of physical health service use

Among the 18,095 individuals included in the study, 97.8% had at least one recorded presentation for a physical health condition during the observation period. Percentages refer to recorded presentations, not individuals. The most frequent categories were respiratory disorders (26%), ear disorders (15%), and neurological conditions (9%). Other categories included digestive and congenital conditions (6% each), dermatological and infectious diseases (5% each), and endocrine, ophthalmological, musculoskeletal, and genitourinary disorders (4% each). Less frequent categories were circulatory disorders (3%), neoplasms and perinatal conditions (2%), and blood or immune disorders (1%), while pregnancy-related presentations accounted for < 1%. The distribution of presentations per individual was markedly right skewed. The mean number of presentations for a physical health condition per individual was 36.93 (SD = 53.2). Marked differences in burden were observed across cohorts (Table 4). Children with ID exhibited substantially higher burdens than children without ID. The highest mean number of presentations for a physical health condition was observed in children with ID (Cohorts 1 and 2). In contrast, children without ID presented markedly lower burdens, relatively similar among those with (Cohort 3) and without (Cohort 4) maltreatment.

Table 4.

Number of presentations with a physical health condition per individual by cohort.

Cohort	n individuals	Total	M (SD)
Cohort 1	4,007	186,348	46.5 (59)
Cohort 2	6,409	321,461	50.2 (68)
Cohort 3	4,007	83,986	21.0 (24)
Cohort 4	3,672	73,068	19.4 (22)

Note. Values represent the number of recorded physical health diagnoses per individual over the observation period. Cohort definitions were as follows: Cohort 1: children with intellectual disability (ID) and documented child maltreatment; Cohort 2: children with ID without documented child maltreatment; Cohort 3: children without ID with documented child maltreatment; and Cohort 4: children without ID and without documented child maltreatment.

Exploratory comparison of physical health service use between cohorts

Differences in the number of presentations for a physical health condition across cohorts were examined using a negative binomial regression model, chosen to account for the substantial overdispersion in the count data. The model was adjusted for sex, age, and material deprivation quintile, with Cohort 4 (neither intellectual disability nor maltreatment) as the reference group. After adjustment, children with ID, whether with maltreatment (Cohort 1, IRR = 2.12) or without (Cohort 2, IRR = 2.40), had significantly higher rates of presentations than Cohort 4 (both p <.001), whereas children with maltreatment but without ID (Cohort 3) did not differ from Cohort 4 (IRR = 1.00, p =.905). Among the covariates, older age was independently associated with a greater number of presentations (p <.001), male sex was not significantly associated (p =.255), and material deprivation showed no consistent gradient, with only the most deprived quintile differing modestly from the least deprived (IRR = 0.95, p =.012; see Table 5).

Table 5.

Negative binomial regression of the number of presentations with a physical health condition by cohort (N = 18,095).

Variable	IRR	95% CI	p
Cohort
Cohort 1	2.12	[2.03, 2.22]	< 0.001
Cohort 2	2.40	[2.31, 2.49]	< 0.001
Cohort 3	1.00	[0.96, 1.04]	0.905
Cohort 4 (reference)	1.00	-	-
Sex
Female (reference)	1.00	-	-
Male	0.98	[0.96, 1.01]	0.255
Age (per year)	1.05	[1.05, 1.06]	< 0.001
Material deprivation quintile
Q1 (reference)	1.00	-	-
Q2	1.01	[0.96, 1.06]	0.768
Q3	0.97	[0.92, 1.01]	0.164
Q4	0.99	[0.95, 1.04]	0.749
Q5	0.95	[0.90, 0.99]	0.012

Note. IRR = incidence rate ratio; CI = confidence interval. Estimates were obtained from a negative binomial regression model with a log link, adjusting simultaneously for cohort, sex, age, and material deprivation quintile. A negative binomial specification was used to the presentations for physical health conditions, which a Poisson model would not adequately capture. Cohort 4 (neither intellectual disability nor maltreatment) served as the reference category; female sex and the least deprived quintile (Q1) were the reference categories for the respective covariates. Age was modeled continuously, with the IRR expressed per one-year increase.

Across both mental and physical health outcomes, the presence of intellectual disability was associated with a markedly higher burden of recorded service use relative to children with neither condition (Cohort 4). For mental health, all exposed cohorts showed elevated rates, with the highest burden among children with both ID and maltreatment (Cohort 1), exceeding that observed among children with ID alone (Cohort 2). For physical health, by contrast, elevated rates were confined to the two cohorts with ID (Cohorts 1 and 2), whereas children with maltreatment alone (Cohort 3) did not differ from Cohort 4. These cohort differences persisted after adjustment for sex, age, and material deprivation, indicating that they are not explained by differences in these characteristics. Overall, the findings reveal marked, domain-specific disparities in health burden across the four mutually exclusive cohorts.

Discussion

This study introduced and characterized the burden of physical and mental health service use captured in the PROTECT databank, a population-based linked administrative resource integrating child protection and health records across Québec. We detailed its design, data sources, and the four-cohort structure defined by ID and child maltreatment exposure. Descriptive analyses revealed different levels of service use burden, with the highest mental and physical health counts observed among children with ID and maltreatment, ID without maltreatment, followed by the two other cohorts, one exposed to maltreatment without ID and the other with neither ID nor maltreatment. The elevated physical service use burden observed among children with ID is consistent with evidence documenting increased rates of chronic medical conditions and complex care needs in this population^27–29. Similarly, exposure to child maltreatment has been associated with long-term somatic morbidity through stress-related biological pathways³⁰.

These results highlight disparities in health service use, which may reflect differences in underlying health burden and need among children facing developmental and child maltreatment exposure. Beyond documenting service use, the PROTECT databank provides a reliable infrastructure for examining the longitudinal interplay among neurodevelopmental conditions such as ID, child protection involvement, and health trajectories across population-level cohorts. Very few cohort studies have been published on the health burden associated with maltreatment³¹, and even fewer on children with ID^12,22, given their small demographic weight, which limits analytical power.

The elevated burden among youth with ID aligns with prior evidence demonstrating their overrepresentation in child protection investigations⁷. Furthermore, several studies based on cross-sectionally associated outcomes from child protection services have shown that maltreated children with ID present more problems than maltreated children without ID, such as running away, and self-harm^7,9,32. Moreover, exposure to sexual abuse in this population has been linked to significantly increased risks of subsequent physical and mental health disorders^22,24. However, most prior studies relied on clinical or protection-service samples and did not include population-based comparison groups without ID, limiting the ability to disentangle contributions of developmental and maltreatment-related risk. Our findings extend this work by quantifying these associations within a population-based, health administrative framework and by explicitly comparing mutually exclusive exposure groups. This design allows clearer differentiation between the independent and combined effects of ID and maltreatment, an analytical distinction that remains under-investigated. The heightened mental health burden observed among maltreated youth is also consistent with established literature linking childhood adversity to later psychiatric morbidity³³. The particularly elevated burden observed among children with dual exposure is consistent with models of cumulative biological and psychosocial load, in which developmental susceptibility and environmental stressors interact across time³⁴. Furthermore, the magnitude of the mental health service use may appear high but reflect two features of the data: the outcome captures cumulative service contacts associated with a diagnostic code over up to 21 years of follow-up rather than distinct diagnoses, so repeated encounters for the same condition each contribute to the count; and Québec’s universal health system enables near-complete administrative ascertainment, yielding higher recorded counts than survey-based or non-universal systems. The observed levels are nonetheless consistent with comparable administrative-data studies. In Canada, approximately one in five children meet criteria for a mental disorder, although administrative data capture only a portion of community cases^35,36. Using linked administrative data in Manitoba, 41% of children with special health needs received a mental health diagnosis by age 16³⁷, and international estimates place the prevalence of co-occurring mental disorders among children with ID at roughly 40%, about double the rate in children without ID³⁸. Registry-based figures from Scandinavia are similarly high, with Danish lifetime cumulative incidence estimated at 45–49% in the general population³⁹. Those comparisons should nonetheless be made cautiously, given differences in observation period, coding systems, and health-system structure.

Strengths

First, the databank is population-based, drawing on provincial administrative records, which reduces volunteer and sampling bias and supports generalizability within the Québec healthcare and child protection systems. Second, it links child protection data with longitudinal healthcare records, enabling cross-sector analyses that have not previously been undertaken to isolate the impact of maltreatment on the health of children with ID. Furthermore, research with ID and child maltreatment remains methodologically fragmented and often underpowered, with limited population-based longitudinal evidence. By integrating protection and health data at scale, the PROTECT databank builds a solid path to advance research on intersecting ID and child maltreatment. Third, its longitudinal design permits the study of developmental trajectories, cumulative exposures, and patterns of healthcare utilization from childhood to adolescence over up to 20 years of observation. Fourth, the deliberate oversampling of children with ID, including an entire population from six geographically and demographically diverse regions of Québec, enhances statistical power to examine a frequently underrepresented group in large-scale research, while allowing analytic strategies to account for the sampling design.

Methodological considerations

Findings should be interpreted considering limitations inherent to administrative data, which are primarily used for billing services rather than research. This means that some measures, such as diagnostic codes, reflect recorded service use rather than standardized clinical assessments, and may be subject to misclassification or differential detection. Children with ID typically have more frequent contact with healthcare services, which may increase the likelihood of diagnostic detection relative to children without ID. Consequently, part of the observed gradient in diagnostic burden may reflect differential healthcare engagement. ID identification relied on ICD-coded diagnoses and may preferentially capture individuals engaged with specialized services. Additionally, certain sociodemographic variables (e.g., residential information) depend on administrative registration and may not fully reflect mobility or contextual instability. Within child protection data, a maltreatment report does not precisely capture the timing of the event, and the determination of its outcome (substantiated vs. unsubstantiated) is largely contingent on professional judgment which may be influenced by workload-related factors^40,41 and by biases associated with ID^42,43. Despite these constraints, integrating multiple administrative databases mitigates some limitations typical of single-source administrative studies.

Conclusion: proposed research agenda

PROTECT databank provides population-based data to characterize disparities in mental and physical health service use among children with ID and child protection involvement in Québec. The graded patterns observed across cohorts underscore the concentration of clinical complexity among children with dual exposure to developmental and childhood adversities. Based on our results, we propose four priority research directions.

First, longitudinal modeling of exposure to child maltreatment and ID and multimorbidity trajectories should examine how timing, duration, and recurrence of child protection involvement shape cumulative diagnostic burden and healthcare service use across childhood and adolescence. Such analyses will clarify whether disparities widen, stabilize, or attenuate over time. Second, sequence-based analyses (e.g., per year) of healthcare use, defined by medical emergency visits, hospitalizations, outpatient care, and medical consultations, can identify specific patterns of service use, intensity, and (dis)continuity, particularly among children with dual exposure to ID and maltreatment. Third, future work should extend analyses beyond ID to include neurodivergence (or neurodiversity) more broadly, including autism spectrum disorder, attention-deficit/hyperactivity disorder, and specific learning or language disorders. Given the substantial representation of neurodevelopmental profiles in the PROTECT databank, this expansion would enhance inclusivity and enable a more comprehensive examination of heterogeneity in risk, service utilization, and health outcomes. Lastly, from a service perspective, the graded burden observed across cohorts underscores the need for integrated, developmentally informed service responses. Children with dual exposure to ID and maltreatment represent a subgroup with marked clinical complexity and cumulative multimorbidity, suggesting that fragmented service models may be insufficient to address their needs. Coordinated case management approaches that integrate child protection, primary care, mental health, and specialized developmental services could reduce service discontinuities and prevent escalation of cumulative health burden. By identifying a population subgroup with disproportionate health burden, the PROTECT databank provides an empirical foundation for proportionate, integrated service planning aimed at reducing avoidable health disparities.

In conclusion, these directions leverage the longitudinal and multidisciplinary structure of the PROTECT databank to move beyond descriptive risk estimates toward the examination of developmental trajectories and the impact of child maltreatment and ID on health burden and service use among youth in Québec.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

The first two authors (S.A.C and J.D.) share co-first authorship, contributed equally to this manuscript, and are listed in alphabetical order by last name. S.A.C., J.D., and G.P. conceptualized and designed the study. S.A.C and J.D. wrote the main manuscript text. S.A.C. prepared the tables and supplementary materials. S.A.C. and M.R. performed the statistical analyses. J.D., G.P., and I.D. supervised the study and contributed to data interpretation. All authors reviewed and approved the manuscript.

Funding

This work was supported by the Canadian Institutes of Health Research (CIHR). S.A.C. was supported by a CIHR fellowship.

Data availability

The data used in this study are derived from linked administrative health and social service records and are subject to strict confidentiality regulations under Québec law. These data are securely held at the Institut de la statistique du Québec (ISQ) and cannot be made publicly available. Access to these data is restricted to authorized members of the research team who have sworn confidentiality and obtained approval from the Principal Investigator. Researchers wishing to request access should contact the Principal Investigator (genevieve.paquette@usherbrooke.ca), who will facilitate the authorization process through the ISQ ([https://statistique.quebec.ca/en/services-recherche/data](https:/statistique.quebec.ca/en/services-recherche/data)) secure research environment.

Declarations

Competing interests

The authors declare no competing interests.