Risk Factors for Diabetic Ketoacidosis in Children

Abstract: Background: Diabetic ketoacidosis (DKA) is a life-threatening acute complication of diabetes. It is preventable if timely administration of sufficient amounts of insulin is initiated. A large proportion of children are affected by DKA annually, especially at onset of type 1 diabetes (T1D). Mortality is relatively rare in developed countries but high in developing parts of the world and is mostly related to cerebral edema. Although severe, permanent cerebral injury is uncommon in the majority of children with DKA, there may be a risk of cognitive impairment, both in the mid- and long-term setting. Moreover, it has been shown that acute kidney injury is common during DKA. The long-term consequences of this are yet to be investigated. Acute renal impairment resolves uneventfully in most children after DKA, but there is concern that tubular injury in the acute may increase the risk of chronic kidney disease in children with diabetes. Since DKA is preventable, it is important to investigate factors that may be associated with increased risk of this condition. Aims: The main aims of this thesis were 1) to describe the extent of delayed referral among children admitted to hospital for new-onset T1D and DKA, and to analyze the effect of delayed treatment on the severity of DKA; 2) to investigate if continuous subcutaneous insulin infusion (abbreviated CSII, commonly known as insulin pumps) are associated with increased risk of DKA in children, compared with multiple daily injections (abbreviated MDI, commonly known as “insulin pens”); 3) to estimate the effect of an internationally standardized measurement of low economic standard on the risk of DKA at onset of T1D, and 4) to analyze the effect of DKA at onset of T1D on the risk of a first-recorded recurrent DKA episode. Secondary aims were to assess the effects of age at diagnosis, parental level of education and family income on the risk of another episode of DKA. Methods: Study I and II were prospective, national population-based studies. Questionnaires were designed to collect survey data from all patients admitted to hospital in Sweden for DKA, from February 1, 2015, to January 31, 2017. The pediatric part of the Swedish National Diabetes Register (NDR) was used to assess the extent of missing data compared with the register, and to complete data when possible. Medical records were checked by the children’s attending physicians in some cases. Two questionnaires were used for data collection and presented to participating caregivers and children at the time of hospital admission. One questionnaire was filled out by the caregivers (together with children if older than 15 years) and used for pre-admission data and assess for coherency with register data. Another questionnaire was filled out by attending pediatricians (and specialized nurses in some cases) primarily for in-hospital data and for coherency with the caregivers’ questionnaires and the register. Study III and IV were register-based, retrospective national population studies. Individual data from the pediatric part of the NDR and from Statistics Sweden (a public, national statistical agency) were merged. For socioeconomic variables, the longitudinal integrated database for health insurance and labor market studies (LISA) provided by Statistics Sweden was used and study-specific variables derived from LISA. For the main exposure variable in study III, relative low economic standard as defined by the European Union’s statistical agency and Statistics Sweden was used. By this definition, an individual is at persistent risk of poverty if disposable income, weighted for household composition, is below 60% of the population median. For study IV, Cox regression models were built to estimate the effect of DKA, age of the child, parental level of education and disposable income quartile at diagnosis of T1D, on the risk of a later episode of DKA during established T1D. Censoring was set to five years after diagnosis to limit the extent of uncontrolled confounding. Results: Delayed referral to hospital was common in Sweden during the study period in study I. It was estimated that delay occurred in 43% of the cases, and delay was associated with significantly more severe ketoacidosis at hospital admission. Parental suspicion of diabetes was associated with milder ketoacidosis. The most common misdiagnosis was gastroenteritis. Treatment with CSII was associated with significantly higher risk for DKA than MDI in study II. However, the increased risk with CSII was only seen in cases of mild DKA (pH 7.29 – pH 7.20) but not in children hospitalized for moderate to severe DKA (pH<7.20). Children with MDI had significantly higher mean HbA1c levels (102.7 mmol/mol, SD 23.3 mmol/mol) compared with children with CSII (73.9 mmol/mol, SD 18.1 mmol/mol). In both children with CSII and MDI, the use of sensor-based continuous glucose monitoring was lower than the national average, but it was higher among children with CSII (56%) than in children with MDI (28%). Persistent low economic standard was significantly associated with 42% higher risk of DKA at onset of T1D (RR 1.42, C.I. 1.13 - 1.79, p = 0.003). There was no evidence that DKA at onset of T1D in children was associated with another episode of DKA during established diabetes (hazard ratio 0.78, C.I. 0.39 - 1.6, p=0.47). There was a higher risk of DKA during established T1D for primary level of education compared to tertiary level (hazard ratio 5.8, p=0.001), and for secondary level of education compared to tertiary level (hazard ratio 1.8, p=0.031). There was also significant association between higher risk of DKA for children from the lowest household income quartile compared to the highest income quartile (hazard ratio 0.30, p=0.011). Conclusions: Delayed referral is common in children with new-onset T1D who present with DKA. Delayed referral leads to worse outcome measured as pH at hospital admission. CSII is associated with increased risk of mild DKA, but not with moderate or severe DKA. Persistent low economic standard is associated with increased rates of DKA at new-onset T1D in children, even in a high-income country with completely reimbursed public healthcare like Sweden. There is no evidence that DKA at new-onset T1D is associated with a later episode of DKA and the risk factors for DKA at onset of T1D and during established T1D may be partially different. However, parental education level and low household disposable income are risk factors of DKA, both at onset of T1D and during established T1D.