Type 1 diabetes mellitus (T1D), a chronic endocrine disorder most often diagnosed in childhood or adolescence, is primarily managed at the individual, family, and community levels but has important implications for the hospital system [1–4]. In the absence of a known cure, therapeutic options include multiple daily injections of insulin or continuous subcutaneous insulin infusion using insulin pumps. Suboptimal management of T1D may increase the burden on acute hospitals due to diabetic ketoacidosis (DKA) and other complications, the risks of which may be exacerbated by socioeconomic conditions and healthcare funding resources [1, 2, 5]. Advances in diabetes-related technologies such as insulin infusion pumps can help improve health outcomes among children and adolescents living with T1D [3, 6–9], although they tend to incur significantly higher costs [10]. Disparities in access to insulin pumps persist within and across countries owing to differences in clinical practices as well as economic reasons, including inadequacy of reimbursement of these devices by public health systems and private insurance [3, 11]; such disparities have been found even in countries of universal medical coverage such as Canada, France, and New Zealand [12–14]. Yet, despite evidence that paediatric patients with T1D tend to stay in hospital longer than those without diabetes, the literature is limited on diabetes-specific hospitalization patterns among children [2]. Hospital length of stay (LOS) is considered a salient metric of health system costs and performance and may indicate different medical and non-medical factors, such as hospital characteristics, severity of complications, and continuity of care; however, research on social determinants of LOS is fragmented and incomplete [15–17].

Data limitations often preclude characterizing the broader social and policy context that may influence length of stay. The complexity of factors directly or indirectly related to LOS means much remains unknown [15]. Studies aiming to understand patient-level variations in LOS using routinely collected hospital administrative data tend to be limited to factors such as age, sex, diagnostic group, admission type, and source of insurance [18, 19]. Innovations in data linkages are facilitating population-based studies of the social correlates of cause-specific hospital stays [20]. In Canada, one study using linked data indicated that LOS among adult inpatients with chronic conditions was not associated with perceived unmet healthcare needs [21]. Other research found the risk of acute-care admissions was higher among children with diabetes of lower socioeconomic status [13]. We are unaware of any studies in Canada or elsewhere investigating associations of social environments, including public funding policies for insulin pumps, with LOS among paediatric inpatients with diabetes.

Echoing global epidemiological trends of rising prevalence of paediatric diabetes [4], the Canadian province of New Brunswick has seen its paediatric diabetes rate increasing over time: from 2.8 per thousand in 2000 to 4.2 per thousand in 2016 [22]. To address the escalating burden, the provincial government adopted a comprehensive diabetes strategy in 2011 which identified among its goals improving affordable access to insulin pumps and, ultimately, reducing hospital days by persons with diabetes [23]. However, the strategy document was silent on measures or information needs on the social determinants of diabetes and related health system outcomes [24]. To better inform policy and planning options aimed at minimizing time in hospital among the high-risk paediatric population, the objective of this research was to assess the associations of socioeconomic status and a public insulin pump funding policy with diabetes-related LOS among paediatric inpatients. A retrospective population-wide study design was used, drawing on linked administrative and socioenvironmental datasets covering all paediatric inpatients with diabetes in the province of New Brunswick, Canada.

The dataset linkages yielded an initial paediatric inpatient population of 392 diabetes-related hospitalizations among 279 individuals over the four-year period. Excluding those lacking a valid residential postal code history in New Brunswick or neighbourhood deprivation information (~1% of the records), the final analysis included 386 acute-care hospital stays (~0.5% entailing transfers). Based on provincial estimates of diabetes prevalence among children and adolescents [22], this translated to approximately 15.1 hospitalizations per 100 person-years in the paediatric population living with diabetes. The great majority (90.6%) of inpatient stays recorded T1D as the primary diagnosis, with the balance divided evenly between type 2 diabetes (4.7) and other or unspecified types (4.7%).

Among these paediatric inpatients, one-fifth (21%) had accessed the NB-IPP (Table 1). Drawing on nationally defined socioeconomic distribution indices, approximately one-fifth (20%) resided in the least materially deprived neighbourhoods (quintiles 1 and 2) and two-fifths (42%) resided in the most deprived neighbourhoods (quintile 5). There were somewhat more female inpatients (56%) than males (44%).

Hospital episodes due to paediatric diabetes averaged 3.87 days, with no discernible year-over-year pattern across the period of observation (Figure 1). The bivariate analysis revealed that the mean LOS was significantly lower among those having accessed the NB-IPP compared with those who had not: 3.25 days versus 4.03 days (p<0.05). Stays were somewhat shorter among those living in less deprived neighbourhoods (quintiles 1-4: 3.68 days) versus those in the most deprived neighbourhoods (quintile 5: 4.13 days), as well as among males (3.74 days) versus females (3.97 days), but the observed differences were not statistically significant.

According to results from the multiple regression (Table 2), paediatric inpatients who had accessed the NB-IPP experienced significantly shorter diabetes-related stays, on average 1.34 days less (95% CI: 0.63–2.05 days less), compared with those who did not access the social policy scheme, after controlling for other sociodemographic characteristics. Those in the oldest paediatric age group (15–18 years) experienced significantly shorter stays compared with the youngest (0–9 years). Some differences were observed by health region, with inpatients residing in Zone 2 (Saint John area) or Zone 3 (Fredericton area) having a shorter average LOS compared with those residing in Zone 1 (Moncton area). Neither the patient’s sex nor their residential neighbourhood deprivation index showed statistically significant differences in terms of time spent in hospital, all else being equal.

With regard to DKA-specific stays, which represented 40% of all diabetes-related stays, the LOS averaged 3.16 days. Patients who had accessed the NB-IPP averaged 1.51 fewer days in acute care (95% CI: 0.38–2.63 fewer days), compared with those who did not access the social policy scheme, adjusting for other socioenvironmental and demographic variables (Table 2, model 2).

Using population-representative data from multiple linked administrative and geospatial databases, this study represents the first exploration of social and policy correlates of hospital length of stay among paediatric inpatients with diabetes in the province of New Brunswick, Canada. We found paediatric inpatients who had accessed public resources made available through a new provincial insulin pump funding policy experienced significantly shorter diabetes-related stays compared with those who had not accessed the policy scheme, with a mean LOS of 1.34 fewer days (95% CI: 0.63–2.05 fewer days), after controlling for other demographic and socioenvironmental characteristics. A similar association held for DKA-specific hospitalizations, with the LOS averaging 1.51 days less (95% CI: 0.38–2.63 days less).

The findings suggest that, regardless of universal coverage for essential medical and hospital services, given the self-care skills and high out-of-pocket expenses required to manage diabetes, the health financing policy context influences health outcomes [1, 36]. Public resourcing may offer expanded opportunities for purchasing healthy foods, engaging in physical activities, and other lifestyle modifications which may shape health capabilities and, in turn, diabetes outcomes including preventable complications [36]. Enhanced health capabilities and, in particular, health literacy to address parental anxiety may help reduce hospital utilization among paediatric inpatients [37].

The results aligned with investigations of a social policy for funding of assistive devices in the Canadian province of Ontario, which suggested reasons to support public funding of paediatric insulin pumps based on favourable safety profiles and clinical perceptions of the need to reduce disparities in social support systems for children and youth living with T1D [13, 38].

Similarly to a study from New Zealand [14], significant (but unexplained) regional differences were found in population-based outcomes despite the public funding of insulin pumps. Further research is needed to better understand the clinical practices, referral patterns, support systems, or other patient characteristics that may drive such regional disparities. On the other hand, socioeconomic status, as measured through neighbourhood-level categorization of material deprivation, was not found here to be significantly associated with LOS. Only modest differences in LOS among low-income groups have been observed for other health interventions in the Canadian hospital system [20]. Moreover, a previous study from Toronto, Canada’s largest city, found a non-linear socioeconomic gradient in health outcomes among paediatric patients living with T1D [5]. It is possible that, given New Brunswick’s exclusively rural and small urban population, combined with the small number of paediatric hospitalizations (N = 386), there lacked statistical power in terms of population size and heterogeneity across deprivation categories to independently account for area-based disparities beyond health zones.

A key strength of this study was its population-wide design. Limitations included the lack of available information in the linkable data platform at the time of analysis on complete pharmaceutical records including insulin prescriptions, glycated hemoglobin levels upon hospital admission, diabetes duration among both NB-IPP participants and non-participants, actual pump utilization or discontinuation patterns, and other sources of financial support such as private insurance coverage. Also lacking from our analysis was information that could have offered a more thorough understanding of the facilitators and barriers to accessing the funding program at the patient and family level (such as parental education, ethnicity, and knowledge of official languages) and at the health system level (e.g., conformity with evidence-informed primary care management). It is possible that some children may have been diagnosed with diabetes upon hospitalization. The risk of selection bias is also possible, given that patients and their families eligible for program support must have already demonstrated regular follow-up with a diabetes healthcare team, which may thus affect the statistical results [39]. More generally, given the well-known lags between administrative data capture and data readiness for research use [40], compounded by the substantial resource demands of building linkable data platforms in smaller jurisdictions such as New Brunswick [28], this analysis was limited to the four-year period following NB-IPP implementation. As such, we were presently unable to track effects of the subsequent change in program age limit or of transition experiences from paediatric to adult care, when care disruptions may lead to increases in adverse health outcomes [41].

This novel characterization of social and policy correlates differentiating lengths of hospital stay among paediatric inpatients with diabetes highlighted that the health policy environment, specifically, access to a universal insulin pump funding scheme, was associated with significantly shorter stays. Results were suggestive, albeit not inferring causality, of positive returns to the hospital system of public policies for reducing the financial burden among families for diabetes-related technologies. Reliable estimates of the rate of insulin pump uptake among persons living with T1D are limited in Canada, as are evidences on whether government financing results in more equitable access to pumps [42]. More research is needed on cost-effectiveness of social policy initiatives to help inform decisions regarding equitable access and possible expansion of public funding for new assistive technologies for diabetes management among paediatric patients.