Reasons for hospitalisation in youth with type 1 diabetes, 2010–2019

To determine the incidence of hospitalisation for all diagnoses among Australian youth with type 1 diabetes.


| INTRODUCTION
Type 1 diabetes mellitus is an autoimmune endocrine disorder that can have devastating complications including for the cardiovascular, renal and neurological systems. 1 It often begins in childhood or adolescence, with incidence peaking between the ages of 10 and 14 years. 2A systematic review of 32 studies and over 114,000 youth with type 1 diabetes aged <15 years found that compared to those without diabetes, youth with type 1 diabetes were three times more likely to be hospitalised. 3As our understanding of diabetes has improved, studies have shown that the consequences of type 1 diabetes extend beyond the well-described vascular complications and include conditions such as respiratory infections and mental health disorders. 4However, most studies reporting hospitalisations among youth with type 1 diabetes have focused on diabetes-related admissions.To date, there has been no population-based study reporting the full range of reasons for hospitalisation among youth with type 1 diabetes.
There is also a paucity of evidence characterising diabetes-related admissions in detail.As the prevalence of type 1 diabetes continues to increase, 5 it is vital to understand the patterns of hospitalisation among youth with type 1 diabetes to inform management and resource allocation.As such, we sought to determine the incidence of hospitalisation for all diagnoses among Australian youth with type 1 diabetes compared to the general population.We also explored diabetes-related admissions in more detail to determine the specific conditions accounting for the bulk of these admissions, the temporal relationship of admissions to the date of diabetes diagnosis, and the characteristics of those who are readmitted.

| METHODS
The diabetes population for this study included registrants of the National Diabetes Services Scheme (NDSS), which was established to deliver subsidised self-management equipment to people with diabetes.The NDSS captures over 90% of Australians with diagnosed type 1 diabetes. 6,7or the first part of this study, which compared reasons for hospitalisation among youth with type 1 diabetes to the general population, we included NDSS registrants aged under 20 years with type 1 diabetes from the states of Victoria and Queensland (approximately 50% of the national NDSS population) registered on the NDSS as of 1 July 2010, and all new registrants until 30 June 2017.An algorithm involving the original diagnosis and clinical characteristics was used to distinguish between type 1 diabetes and type 2 diabetes (Data S1).This cohort was linked by the Australian Institute of Health and Welfare (AIHW) to hospital admission data sets and to the National Death Index (NDI).We excluded individuals with hospitalisations or death codes for cystic fibrosis (Data S1).The general population was the whole Australian population aged under 20 years derived from the AIHW National Morbidity Database. 8The two states included for the diabetes population in this analysis were selected as these were the only ones for which the NDSS linked data set contained both public and private hospitalisations, as does the general population data source.
The outcome of interest for the first part of the study was the primary diagnosis recorded for a hospitalisation at the International Classification of Diseases, Tenth Revision (ICD-10) three-digit diagnosis level.We excluded admissions with primary diagnoses in ICD-10 chapters that did not correspond to specified medical conditions or with insufficient data (Data S1).Any diagnoses with no hospitalisations in the type 1 diabetes population for any year in the study period were also excluded.Population size data for the general population were obtained from the Australian Bureau of Statistics (ABS) Estimated Resident Population 9 for the years 2010 to 2017.To compare the risk of hospitalisation for each ICD-10 chapter and three-digit diagnosis among those with type 1 diabetes to the general population, we calculated relative risks (RRs) with 95% confidence intervals using Poisson regression, adjusted for financial year.We ranked diagnoses according to relative risks of hospitalisation among those with type 1 diabetes

What's new?
What is already known?
• Youth with type 1 diabetes are more likely to be hospitalised compared to those without diabetes.
What has this study found?
• Youth aged under 20 years with type 1 diabetes had an increased risk of hospitalisation for almost all diagnoses compared to the general population.• Over 60% of hospitalisations in youth with type 1 diabetes were directly related to diabetes; almost half of these were for ketoacidosis.
What are the implications of the study?
• The broad range of complications experienced by youth with type 1 diabetes should be considered in clinical management.compared to the general population, excluding any codes that are for diagnostic testing (e.g.sleep studies).
For the second part of this study, which explored diabetes-related admissions among youth with type 1 diabetes, we included NDSS registrants aged under 20 years with type 1 diabetes from Victoria, Queensland, New South Wales (NSW) and the Australian Capital Territory (ACT) (approximately 80% of the national NDSS population) registered on the NDSS as of 1 July 2010, and all new registrants until 30 June 2019.This analysis population could be expanded compared to the other analysis to maximise the sample size as it did not require comparison to the general population.Other Australian states and territories (South Australia, Tasmania, Northern Territory, Western Australia) were not included because of restrictions on obtaining private hospital data from these jurisdictions.Application was sought for NSW and ACT as it was anticipated that private hospital data would be accessible; however, access to only public hospital data was granted due to lack of private hospital data completeness.The geographical scope was expanded as complete capture of private admissions was not essential, since over 90% of diabetes-related admissions occurred in public hospitals.The study period was expanded as two additional years of data where private hospital data were incomplete were available.Follow-up time for this analysis was therefore from an individual's date of registration on the NDSS until 30 June 2019.
We included all type 1 diabetes-related admissions (ICD-10 code E10), which were split at the four-digit diagnosis level.Admissions for type 1 diabetes with other specified complications (E106) were further split at the five-digit diagnosis level.For the most common diabetesrelated diagnoses (ketoacidosis E101, hyperglycaemia E1065, hypoglycaemia E1064), we determined the proportion and characteristics of individuals with readmissions for a particular diagnostic code within 12 months of the first admission for the same code.Follow-up time for this analysis was therefore 12 months from an individual's first admission for a diabetes-related code.These individuals were characterised according to Index of Relative Socioeconomic Disadvantage (IRSD) scores, which summarise information about socio-economic status of people in a geographic area. 10Low IRSD scores indicate greater disadvantage, while high scores indicate less disadvantage.IRSD scores were split into fifths.Pearson's chi-squared was used to compare sex proportions of those with readmissions for the most common diabetes-related diagnoses to those with single admissions for these diagnoses and to those with no admissions for any of these diagnoses.Kruskal-Wallis equality-of-populations rank test was used to compare those admitted for ketoacidosis, hyperglycaemia and hypoglycaemia, by IRSD and frequency of readmissions.One-way analysis of variance (ANOVA) was used to determine age differences between these three groups.Logistic regression was used to identify risk factors for admission and readmission for each of these diagnoses.
All analyses were performed in Stata, version 16.0.This study was approved by the Alfred Hospital Ethics Committee (Project No: 463/18) and the AIHW Ethics Committee (EO2018/5/501).

| Reasons for hospitalisation among youth with type 1 diabetes compared to the general population
The population size of youth with type 1 diabetes in Victoria and Queensland varied from n = 4794 in 2010-11 to n = 5207 in 2016-17 (Table S1).The general Australian population aged under 20 years varied from n = 5,658,746 in 2010-11 to n = 6,047,751 in 2016-17.Among youth with type 1 diabetes, 21,898 hospitalisations occurred between 2010-11 and 2016-17, 86% of which occurred in public hospitals.There were 4,308,143 admissions in the general population.The cohort of youth with type 1 diabetes was 52.6% male, had a median age of 14.4 years, a median age of diabetes diagnosis of 8.5 years, and a median duration of diabetes of 4.3 years (Table S2).
At a broad category level, endocrine disorders were the most common reason for hospitalisation among those with type 1 diabetes, accounting for over 65% of total admissions (Table 1).Compared to the general population, youth with type 1 diabetes were over three times more likely to be hospitalised for infectious diseases (RR 4.32, 95% CI 4.06-4.61),mental health disorders (RR 3.38, 95% CI 3.11-3.68),and skin diseases (RR 3.66, 95% CI 3.36-4.00).Conversely, hospitalisations for ear (RR 0.62, 95% CI 0.52-0.74)and respiratory diseases (RR 0.88, 95% CI 0.82-0.95)occurred less frequently among those with type 1 diabetes compared to the general population.
At a three-digit diagnosis level, the largest relative risks of hospitalisation for youth with type 1 diabetes compared to the general population were observed for type 1 diabetes admissions and for the diagnosis of poisoning and adverse effects of hormones (Table 2).This diagnosis (T38) consisted almost exclusively (98.9%) of adverse effects of insulin (ICD-10 four diagnosis T383).Aside from these diagnoses, some of the largest relative risks of hospitalisation were observed for infectious diseases, particularly anogenital herpesviral infections (RR 54.83, 95% CI 33.21-90.53).However, this diagnosis only accounted for 16 admissions among those with type 1 diabetes.The median age of those with type 1 diabetes hospitalised for this condition was 18 years and median length of stay was 3 days.Hospitalisations for mental health disorders, especially personality disorders (RR 9.70%-95% CI 8.02-11.72),were much more likely among youth with type 1 diabetes compared to the general population.Large relative risks of hospitalisations were also observed for skin conditions including pilonidal cyst and sinus (RR 6.87, 95% CI 5.66-8.33).
Hospitalisation for almost all conditions occurred more frequently among youth with type 1 diabetes compared to the general population.The only exceptions were nonsuppurative otitis media (RR 0.31, 95% CI 0.22-0.43),chronic diseases of tonsils and adenoids (RR 0.66, 95% CI 0.56-0.79),and dental caries (RR 0.80, 95% CI 0.66-0.97)(Table 3).Hospitalisation for several respiratory conditions was more common in the general population; however, these findings were not significant (Table S4).Diabetesrelated admissions, infectious diseases and mental health disorders were among the most common reasons for admission among youth with type 1 diabetes (Table S5).

| Diabetes-related admissions among youth with type 1 diabetes
Between 1 July 2010 and 30 June 2019, a total of 45,685 youth with type 1 diabetes were registered on the NDSS across Victoria, Queensland, NSW and the ACT.The cohort was 51.9% male, had a median age of 14.4 years, a median age of diabetes diagnosis of 8.4 years and a median duration of diabetes of 4.3 years (Table S3).In this cohort, 25,555 hospitalisations for type 1 diabetes occurred among 9946 individuals.Almost half (46.8%) of diabetes-related admissions were for ketoacidosis (Figure 1).Of all ketoacidosis admissions, 15.4% occurred within 3 months of diabetes diagnosis.One fifth of diabetes-related admissions were for hyperglycaemia, almost all of which (92.5%) occurred more than 3 months after diagnosis.Almost one quarter of diabetes-related admissions (23.5%) were for uncomplicated diabetes (E109), while 8.7% were for hypoglycaemia.
Of those who were admitted for ketoacidosis, one quarter (25.8%) were readmitted for ketoacidosis within 12 months.The proportion of those with readmissions within 12 months was significantly smaller for hyperglycaemia (15.9%) and hypoglycaemia (13.0%) (p < 0.001).Over 55% of those readmitted for ketoacidosis, hyperglycaemia or hypoglycaemia were female (Table S6).The proportion of females among those readmitted for any of these three diagnoses was significantly higher compared to those admitted for any of these diagnoses who were not readmitted (50.2% female) (p < 0.001) and compared to those who were not admitted for any of these diagnoses (46.3% female) (p < 0.001).Youth residing in most disadvantaged areas (first IRSD fifth) accounted for the greatest share of those readmitted for ketoacidosis (28.2%) and hyperglycaemia (26.8%), while those readmitted for hypoglycaemia were most likely to reside in areas within the third IRSD fifth (24.4%).Those admitted for ketoacidosis or hyperglycaemia were significantly more likely to reside in T A B L E 1 Incidence of hospitalisations in youth with type 1 diabetes compared to the general population by disease category.
The results of the logistic regression that explored risk factors for admission and readmission for the three most common diabetes-related diagnoses are displayed in Table 4. Female sex and residence in the first IRSD fifth were associated with increased risk of admission for all three diagnoses, and increased risk of readmission for ketoacidosis (p < 0.05).Age was associated with a slightly increased risk of admission and readmission for ketoacidosis, while being inversely associated with admission for hypoglycaemia (p < 0.05).Age at diabetes diagnosis showed a slight inverse association with admission for all three diagnoses, and with readmission for ketoacidosis (p < 0.05).

| DISCUSSION
In this population-based linkage study, we found that Australian youth with type 1 diabetes were at increased risk of hospitalisation for most conditions compared to the general population, particularly infectious diseases, mental health disorders and skin conditions.While the substantial risk of hospitalisation for infectious disease we found among youth with type 1 diabetes is consistent with the emerging literature, the specific infections responsible for the greatest relative risks among those with type 1 diabetes compared to the general population were somewhat surprising.The greatly increased risk of admissions for anogenital herpesviral infections among young adults with type 1 diabetes is a novel finding.Although there were only 16 admissions in our diabetes data set, T A B L E 3 Reasons for hospitalisation occurring more frequently in youth in the general population compared to those with type 1 diabetes.

Relative risk of hospitalisation (95% CI)
Nonsuppurative otitis media F I G U R E 1 Characteristics of hospital admissions for type 1 diabetes (E10) among youth on the National Diabetes Services Scheme.

| 7 of 9
TOMIC et most herpes simplex infections are uncomplicated and are managed in the community. 11These severe cases requiring hospitalisation are likely to represent a small proportion of total herpesviral infections among youth with type 1 diabetes.If youth with type 1 diabetes truly have an increased risk of these infections, this would have important implications for clinicians with regards to screening and treatment.The high risk of gastrointestinal infections among those with type 1 diabetes is also a novel finding; however, it is consistent with results from our recent analysis of Australians with type 2 diabetes where we found a substantial burden of excess hospitalisations for gastroenteritis. 12Chronic hyperglycaemia is thought to contribute to gastrointestinal infectious processes, 13 which may explain the increased risk of intestinal infections in those with diabetes.Additionally, children with type 1 diabetes who have gastroenteritis may be more likely to be admitted compared to children without diabetes so that glucose monitoring and their insulin and fluid requirements can be closely managed while they experience vomiting and diarrhoea.
The substantially increased risk of numerous skin conditions we found among youth with type 1 diabetes compared to the general population is consistent with the literature. 14We also reinforce the growing body of evidence linking type 1 diabetes to mental health disorders.Increased incidence of anxiety, 15 depression 16 and personality disorders 17 have all been previously reported in studies of young people with type 1 diabetes relative to the general population.The replication of our previous finding of increased risk of alcohol use disorders among adults with type 2 diabetes 12 in youth with type 1 diabetes attests to the need for holistic management of young people with diabetes to minimise the detrimental effects of alcohol on their physical and mental health.Some of the increased risks of hospitalisation among youth with type 1 diabetes may in part be explained by the fact that clinicians may have a lower threshold to admit those with type 1 diabetes compared to those without diabetes for similar medical conditions of similar severity.The finding of a greatly increased risk of acute pancreatitis among youth with type 1 diabetes is another novel finding that should be explored in other population data sets of youth with type 1 diabetes.As those with type 1 diabetes admitted for this condition were mostly older teenagers (median age 17.1 years), this finding may be related to gallstone disease, which is a major cause of pancreatitis. 180][21] It may be driven by the fact that youth with type 1 diabetes in our data set are more likely to come from less socio-disadvantaged households compared to the general population.Low socio-economic status has been linked to increased prevalence of childhood ear infections. 22Also, youth of higher socio-economic advantage have greater access to private community dental care to prevent tooth decay from deteriorating to the point where hospitalisation is required.Although youth with type 1 diabetes were at increased risk of hospitalisation for many respiratory conditions, hospitalisation for respiratory diseases at the broad category level was more common in the general population.This finding may in part be a statistical artefact of low numbers as there were no hospitalisations among those with type 1 diabetes in some years for many respiratory conditions.The slightly reduced risk for some common respiratory conditions among those with type 1 diabetes, albeit not statistically significant, would have also contributed towards this finding.One of these conditions was pneumonia, which was somewhat surprising given there is other evidence that type 1 diabetes is associated with increased risk of pneumoniarelated hospitalisation. 23Like for ear and tooth diseases, this may be in part explained by socio-economic factors, as children from low socio-economic status backgrounds are at increased risk of pneumonia. 24he substantial proportion of ketoacidosis admissions among youth with type 1 diabetes is notable, albeit unsurprising in the context of an AIHW report showing increasing ketoacidosis hospitalisation rates among Australian youth with type 1 diabetes between 2009-10 and 2014-15. 25We found that most ketoacidosis admissions occurred 3 months or more after diabetes diagnosis, suggesting many of these hospitalisations could be prevented by better community management of young people after diagnosis of type 1 diabetes.We found that one quarter of those admitted for ketoacidosis had a recurrent admission for ketoacidosis within a 12-month period, and that readmissions were most common in those residing in areas of highest socio-economic disadvantage, which is consistent with findings from a recent study of US youth with type 1 diabetes. 26Our findings suggest that clinicians and policymakers should consider targeted management and resource allocation towards areas of highest socio-economic disadvantage where youth with type 1 diabetes may be predisposed to ketoacidosis admission and readmission.Notably, the association between residence in areas of socio-economic disadvantage and hospitalisation was stronger for hyperglycaemia than for hypoglycaemia.This may be in part due to the greater risk for poor outcomes among those residing in areas of highest socio-economic disadvantage being partly offset by more aggressive glucose-lowering management among those residing in less disadvantaged areas.
The strengths of our study include the use of population-based data sets and the level of granularity at which hospitalisations were analysed.One of the limitations was the absence of private hospital data for some Australian states, meaning our comparison to the general population could only include Victoria and Queensland.As we were unable to exclude people with type 1 diabetes from the general population, the magnitude of all associations would have been slightly diminished in the first part of our study.The general population data source did not contain unit record data, so it was not possible to explore the individual burden of admissions for various diagnoses among those with type 1 diabetes compared to the general population.The results are only generalisable to other high-income countries.In conclusion, we have shown that Australian youth with type 1 diabetes were at increased risk of hospitalisation for almost all conditions compared to the general population.As our understanding of the diversity of diabetes complications grows, clinical strategy should be refined to prevent hospitalisation and other adverse outcomes.We found that around 30% of all hospitalisations in those with type 1 diabetes were for ketoacidosis, most of which occurred more than 3 months after diabetes diagnosis, underscoring the importance of education of people and carers.We reinforce the association between residence in areas of high socio-economic disadvantage and hospitalisation for ketoacidosis, attesting to the need for targeted resource allocation in these areas.
Top 25diagnoses with highest relative risks of hospitalisation in youth with type 1 diabetes compared to the general population at individual-diagnosis levels.
Note: Relative risks were calculated using Poisson regression, adjusted for financial year.Abbreviations: CI, confidence intervals; ICD-10, International Classification of Diseases, Tenth Revision.Note: Relative risks were calculated using Poisson regression, adjusted for financial year.Abbreviations: CI, confidence intervals; ICD-10, International Classification of Diseases, Tenth Revision.

ratio for admission (95% CI) Odds ratio for readmission within 12 months of index admission (95% CI)
Factors associated with admission and readmission within 12 months of the index admission for ketoacidosis, hyperglycaemia and hypoglycaemia among youth with type 1 diabetes.Odds T A B L E 4Abbreviations: CI, confidence intervals; IRSD, Index of Relative Socioeconomic Disadvantage.