To date, few population-based studies have examined inequities in respite care maintenance using linked administrative data. A population-based approach using linked administrative data is a robust method as the same individuals and their family members can be tracked over time across various services [31, 32], allowing for the identification of family member outcomes (e.g., siblings) [33], and gaps and inequities in the receipt of respite care.

This study, which is a part of a larger mixed methods study [30, 34], sought to determine the sociodemographic and health characteristics associated with receipt of respite services. We used receipt as a proxy for access. We hypothesized that respite care is not accessible and equitably distributed to all Manitoba families of CYSHCN. The data only reflects respite services received from and funded through the publicly funded provincial government disability services programs and does not include private respite services paid for out-of-pocket by families.

This study used administrative data in the Manitoba Population Research Data Repository (‘the Repository’) at the Manitoba Centre for Health Policy (MCHP) [35]. Although administrative datasets for research exist in other jurisdictions across Canada, the Repository in Manitoba holds among the most comprehensive set of administrative databases in the country, which includes individual-level linkable, de-identified records of all Manitoba residents using their Personal Health Information Number (PHIN). Unique in North America, the Repository at MCHP has capabilities to link individuals across several generations with their social, health, education, and justice data [36]. To receive health services, most individuals in Manitoba have a Personal Health Information Number (PHIN) that can be encrypted and used to link deterministically across datasets for research purposes. Permission must be obtained by researchers to access and use the databases, which are owned and governed by various provincial government departments. There are strict measures in place to protect the data. The data in the databases are collected for administrative reasons (e.g., physician remuneration); however, the utility and validity of the data has been previously documented and used extensively used for research purposes [37, 38]. Individuals’ data were linked across the data sources. Given that less than 5% of our sample was missing from our analyses, we did not attempt to adjust or impute the missing data. This small of amount of missing data is unlikely to bias the results.

First, we created a population-based cohort of children aged 0 to 17 years who were diagnosed with a special health care need (see Figure 1) and resided in Manitoba at any point during the 5-year study period, April 1, 2013, to March 31, 2018. The following databases were used to identify CYSHCN diagnoses (see Supplementary Table for CYSHCN Disability Definitions), receipt of Special Education funding and or publicly funded respite services:

The disability conditions and their corresponding International Classification of Diseases (ICD) codes were selected based on previous work conducted by researchers at the Manitoba Centre for Health Policy, as defined in the “Mental Health of Manitoba’s Children” report [39], and according to eligibility criteria for Manitoba Education Special Needs Funding Criteria Level 2 (moderate to severe impairment) and Level 3 (profound impairment) used in the “How are Manitoba’s Children Doing?” report [40]. The study cohort consisted of 14,759 children (Figure 1). Residence in Manitoba was based on six-digit postal codes in the Manitoba Health Insurance Registry (“Registry”), a population-based registry maintained by the provincial health department. Children who met at least one study criterion (physician diagnosis of a developmental disability, spina bifida, or cerebral palsy, received special education funding, or received publicly funded respite services from CDS/CLDS) were included in the study (see Figure 1). The CLDS dataset includes all CYSHCN who turned 18 years of age at some point during the fiscal year (April 1-March 31), therefore some CYSHCN in the CLDS database would still be 17 years of age during a given fiscal year. Children who met multiple criteria were only included once for cohort development.

Next, we created a population-based cohort of mothers by linking children’s birth records to their mothers using the Registry (mothers can be identified with a greater degree of certainty than fathers) [40]. Almost all CYSHCN (N = 14,530, 98.5%) were linked to their mother’s health card number; 14,127 mothers were identified. Given that less than 5% of our sample was missing from our analyses, we did not attempt to adjust or impute the missing data. This small among of missing data is unlikely to bias the results. Finally, we created a sibling cohort from birth records linking the mothers’ records to her other children; 33,719 siblings were identified. Biological children were identified through their mother’s PHIN, and stepchildren and children adopted at birth were identified through the family registration number, which is connected to the family head (usually the father). The unit of analysis was the family (N = 14,127), which is the number of mothers, as mothers were the head of the household and defined the unit of analysis.

The Charlson Comorbidity Index (CCI) is based on 17 medical comorbid conditions[43] identified from ICD-9-CM and ICD-10 coding algorithms in hospital abstracts and medical services data [44–46]. Each comorbidity has an associated weight from one to six based on mortality risk, with the sum of weights resulting in a singular comorbidity score [44, 47]. The CCI scores of zero to six reflect the range of comorbidities; higher scores predict higher mortality risk [48] and higher resource use [44, 47, 49]. The CCI has been validated [44, 47] and used in population-based research [50, 51], and has also been used with pediatric samples to predict illness severity, mortality, and the need for hospitalization for patients with COVID-19 [52–54]. While there are pediatric-specific comorbidity measures (e.g. pediatric comorbidity index) [55], the CCI was available to us through the MCHP at the time of analysis and is a stronger predictor of mortality in pediatric populations than physical health indicators.

Receipt of provincially funded respite services was defined as the number of respite services plans, recorded as ‘respite service records’ (e.g., Respite>Regular Hours/Service, or Self-Managed>Respite) approved for the CYSHCN cohort and their families and delivered to them during the study period. A respite service plan is based on a needs assessment conducted by CDS/CLDS program staff. Respite service plans vary in duration (e.g., 2 months to 12 months), after which needs are assessed and a new respite service plan is created and approved. The study variables are presented in Table 1.

Chi-square and independent t-tests were performed to test for statistically significant differences (p < 0.05) between those who received respite services and those who did not. Next, we used multivariable logistic regression to examine factors associated with receipt of respite services. Variables were selected a priori from the literature and by conceptual relevance. Additionally, variables were partly included based on the bivariate associations. Variables were sequentially entered in blocks (child, maternal, household) and the models were compared using the Bayesian Information Criterion (BIC). Odds ratios and 95% confidence intervals were calculated. Lastly, we calculated descriptive statistics to summarize receipt of funded respite services. Analyses were conducted using SAS v 9.4.

This study was approved by the Health Research Ethics Board at the University of Manitoba (#H2019:106) and the Health Information Privacy Committee of the Government of Manitoba (File No.2019/20-01).

In total, there were 27,308 respite service/program records (respite services plans) attributed to 3,489 children, meaning these CYSHCN received one or more respite services/programs during the study period (Table 4 Respite program/services and sub-services provided). Among the respite services/programs received, half (52.0%; 14,190) were for self-managed services (respite and summer gap-programming to support CYSHCN to maintain skills), while less than one quarter (22.3%; 6,090) were attributed to provincial department or regional health authority staff delivery (regular hours/service, nursing (funded)).

Of the 27,308 respite services records, 27,093 (99.2%) were linked to a mother. Overall (out of 27,093 records), 67.2% of the records were associated with a mother who had a diagnosed mental disorder and/or physical health condition, although the percentage varied by sub-service (as shown in Table 4).

To assist in identifying inequities in availability, accessibility, and affordability, we explored public (government) spending on respite services in Manitoba (Table 5). Approximately half of the total spending was allocated to self-managed (respite, summer gap) care, which is independently arranged and managed by families. The costliest sub-services were support staff for afterschool care, self-managed respite, and overnight respite.

To the best of our knowledge, this is the first population-based cohort study using linked administrative data to examine characteristics of families of CYSHCN relative to receipt of publicly funded respite care. A key finding is that less than one-quarter of Manitoba CYSHCN (24.2%) received at least one respite service. Furthermore, we demonstrated that there are marked economic differences in receipt of respite in this province as non-receipt of respite was more likely among those residing in lower-income neighbourhoods. Given the major role parents play in providing care for their CYSHCN, our finding that receipt of respite services was significantly associated with mothers’ age and marital status, and that approximately half of mothers in the total sample had a physician diagnosed mental health and physical health conditions is noteworthy. Finally, half of all respite service records were attributed to self-managed services, meaning that respite services were arranged and managed by families of CYSHCN.

Respite services can and do play a key role in the wellbeing of families of CYSHCN [6, 15] by “strengthening sibling, parent, and overall family functioning [and] improve[ing] family members’ ability to cope with and meet the needs of the person with a disability” [79, p. 481]. Based on our study findings, along with evidence from Canadian families and respite services providers [30, 34], it appears many Manitoba families may not be receiving publicly funded respite care, and may not be receiving any respite if they are not able to finance private or self-managed respite themselves. The call for a health system response to the respite needs of CYSHCN and their ‘all-too-often exhausted’ networks of parents...’ [80, p. 2] and ‘fragile’ families [81] has been made previously. This is especially important given the neighbourhood income disparities related to receipt of respite found in our study. Universal and targeted approaches [82] that engage multiple sectors (e.g., health, social services) and different levels of government are needed to address gaps in respite services to ensure equitable access and receipt of publicly funded respite for all CYSHCN in Manitoba, particularly for lower economic status families.

The physical and mental health of caregivers of CYSHCN is a priority given the complex care requirements for CYSHCN [3], which can be taxing for families [6, 7], particularly for mothers who are often the primary caregivers [83, 84]. The high rates of maternal mental and physical health diagnoses in our study sample, coupled with the Manitoba CDS trend toward self-managed respite care, points to the need for program and policy improvements that address the needs of the whole person, through family-centred [85], coordinated [86] and collaborative care across health and social care sectors.

The use of linked administrative data to assess receipt of respite among Manitoba families of CYSHCN enabled a broad assessment of family characteristics relative to respite services and provided new and valuable insights into potential gaps in respite services for this population. The benefits of this methodology are noteworthy, including objective measures (i.e., ones not based on recall), continuous data collection, and analysis of long-term outcomes [37, 87].

Study findings however are not without limitations. Regarding the administrative data, we only had access to records of respite provided by publicly funded services, thus data do not include private respite services received through out-of-pocket funding by families. Therefore, there may be families in the group that did not receive publicly funded respite but that did receive private respite services. We could also not account for the number of respite hours provided to families. The respite services data is not well-collected, documented, completed, or intuitive to interpret. Additionally, the datasets we used do not include racial, ethnic, or cultural identity; thus, we were not able to examine the impact of intersecting identities on receipt of respite services. Income quintiles may be less accurate for rural areas [82] as the MCHP uses different cut-offs to determine urban and rural income quintiles because of differences in income distribution and population density [88]. Health diagnoses only reflect those who sought treatment from physicians, as the MCHP administrative data does not capture therapy and counseling received from psychologists and other mental health professionals. We also did not account for the timing of mothers’ health diagnoses with respect to when respite was received.

While we strived for a whole-family approach, fathers were not included due to the uncertainty in linking children to their fathers [40]. If a CYSHCN could not be linked to their mother, their mother and siblings were not included; thus, the cohorts are incomplete. We also did not verify that the CYSHCN and their mother lived in the same household. This has implications for the geographic and income-related findings.

Finally, the use of the CCI may have influenced our findings, as it is not ideal for pediatric populations; pediatric-specific comorbidity measures (e.g. pediatric comorbidity index) may provide more appropriate assessments [55].

This study provides a basis for further research. This could include examining family socioeconomic and health characteristics by CYSHCN disability type and by the number of disabilities, which may help to clarify the burden of care experienced by families. In particular, the connection we found between non-receipt of respite and poor physical and mental health of mothers of CYSHCN warrants further attention. Additionally, examining the characteristics of families of CYSHCN based on the number of respite service records or plans by disability could provide insight on the level of respite services accessed by Manitoba families and inform future care planning and policy. Further, alternative analytic approaches (e.g. CART analysis) could be useful for identifying and visualizing the health and sociodemographic characteristics most strongly associated with not receiving respite services. Finally, an economic analysis could determine whether the costs of providing publicly funded respite services to families of CYSHCN are offset by savings in the Manitoba healthcare system, thereby providing a rationale for increasing funding for respite services.