Prevalence and factors associated with diabetic ketoacidosis at diagnosis of type 1 diabetes: A report from a tertiary medical center in Central Pennsylvania

Abstract Objective To explore the rate and factors associated with diabetic ketoacidosis (DKA) at diagnosis of type 1 diabetes (T1D) in a single tertiary medical centre in Central Pennsylvania. Methods Retrospective chart review of all individuals ≤ 18 years of age who were diagnosed with T1D (N = 350) at the Penn State Hershey Pediatric Diabetes Clinic from January 2017 to December 2019. We report logistic regression models for DKA at diagnosis of T1D for age, gender, race/ethnicity, BMI percentile, health insurance, outcome of any healthcare encounter 30 days prior to T1D diagnosis, HbA1c level, altered mental status at diagnosis, and diagnosis of autism spectrum disorder and a multivariable logistic regression model including all aforementioned variables. Results Of the 350 newly diagnosed children with T1D from 2017 to 2019, 161/350 (46%) presented in DKA. Among patients with DKA, there were 45 (28%) in mild DKA and 116 (72%) in moderate/severe DKA, which represents 13% and 33% of all patients diagnosed with T1D, respectively. Variables associated with increased risk of DKA at presentation of T1D included age (<3 or 9‐13), BMI percentile (<3% or > 97%), no referral during preceding healthcare encounter, HbA1c level and altered mental status. In a multivariable model, age (<3 or 9‐13), no referral during preceding healthcare encounter, HbA1c level and altered mental status were associated with DKA at presentation, whereas gender, race/ethnicity, BMI percentile, health insurance and autism spectrum disorder diagnosis were not. Discussion Our study notes an overall higher rate of DKA at diagnosis (46%) compared to the SEARCH study (approximately 30%) but a lower rate compared to a recent study in Colorado children (58%).


| INTRODUC TI ON
Diabetic ketoacidosis (DKA) at the time of diagnosis in type 1 diabetes (T1D) is a preventable complication given the recognizable symptoms (polyuria, polydipsia, enuresis, weight loss and fatigue) of T1D. 1,2 Children diagnosed in DKA have increased morbidity, mortality, and incur higher medical costs and healthcare resource utilization including ICU level care. 3 Several previous studies have identified risk factors for children to present in DKA, including age (<5 or [10][11][12], 4 Hispanic and African American race, low socioeconomic status, misdiagnosis at an initial encounter and lack of private health insurance. [5][6][7][8] A recent study in Colorado children was the first to report similar rates of DKA among privately insured children and Medicaid aligning with the implementation of the Affordable Care Act (ACA) in 2010. 9 Established rates of DKA at diagnosis of T1D differ between different countries and states within the United States. 8,10 The aims of this study were to (a) determine the rate of DKA at presentation of T1D at a single tertiary medical centre in central Pennsylvania and (b) evaluate the factors associated with their presentation in DKA.

| Statistical analyses
First, differences in diagnoses were examined by each variable. Chisquared tests were used to test for differences in rates of DKA at diagnosis of T1D for the following variables: age at diagnosis, gender, race/ethnicity, BMI percentile, health insurance, outcome of any preceding healthcare encounter 30 days prior to diagnosis of T1D (immediate referral, no referral, emergency department [ED]), HbA 1c , altered mental status and diagnosis of autism spectrum disorder. A logistic regression model for DKA diagnosis was generated separately for each variable. The fitted logistic regression models were shown graphically using estimated probabilities. A multivariable logistic regression model was fit for all variables aforementioned.
Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were reported from the models. A total of 16 patients (5%) had a missing value for at least one of the variables included in the multivariable model and were subsequently excluded when fitting the model.

| Patient population
There were 350 patients aged < 1 to 18 included in the study. The

| Family history
The majority of children (216, 66%) reported no family history of T1D, 39 (12%) had a first degree relative, 71 (22%) had a second degree relative and 24 (7%) lacked information. Children with a first degree relative with T1D had the lowest rate of DKA at 28%, compared to children with a second degree relative (56%) or no family history (46%). There were more children in the immediate referral group 203 (58%) than the no referral group 47 (13%). Children in the no referral group had the highest rate of DKA (75%), compared to the rates of DKA for children in the immediate referral group (37%) or those that presented to the ED directly (51%).

| Clinical factors associated with DKA
In univariable analysis, we found very young patients (<3) and patients aged 9 to 13 had the highest risk of DKA at diagnosis (Figure 1).
Children at the extremes for BMI percentile (<15th or > 85th percentiles) had higher probability of being diagnosed with DKA. Children with a HbA 1 c ≥ 10% (86 mmol/mol) or altered mental status had a higher risk of DKA at diagnosis. There were no differences in DKA at diagnosis based on gender, ethnicity/race, autism spectrum disorder diagnosis or insurance coverage. Figure 1 shows the estimated probabilities of DKA from logistic regression models fit separately for each variable. In the multivariable model (Table 1), clinical factors associated with increased risk for presentation in DKA at diagnosis included age (<3 and 9-13), no referral during any healthcare encounter 30 days prior to diagnosis, HbA 1 c ≥ 10% (86 mmol/mol) and altered mental status.

| CON CLUS IONS
Our study was the first to report the rate of DKA at diagnosis of T1D individuals. The authors proposed a potential driver is increased enrolment in high-deductible private insurance plans that disincentive families from seeking timely care. 9 We report similar findings of no difference in the risk of DKA at diagnosis based on private insurance or Medicaid in a patient population after the implementation of the ACA in 2010; however, we still note an overall lower rate of DKA than their study. Previous reports before the implementation or the first years of the ACA found children with private insurance had lower rates of DKA compared to Medicaid. 7,11,12 An alarming proportion of children who had a healthcare encounter within 30 days of their diagnosis of T1D but who had not been referred for concern of T1D presented in DKA (75%), which was much higher than our average rate of DKA (46%), children in the immediate referral group (37%), or those that presented to the ED directly (51%). Our findings support previous reports that a misdiagnosis of a child presenting in T1D is associated with an increased rate of DKA. 7,12 One report found rates of misdiagnosis of T1D are highest among very young children (<5) and independent of type of healthcare insurance. 12  diagnostic screening for T1D in the primary care setting. Limitations of this study include the nature of a retrospective study, which is limited by the accuracy of documented information provided within the medical record and all scanned media, and the single-centre design.
In conclusion, our study was the first to report the rate of DKA at diagnosis of T1D within a tertiary medical centre in central Pennsylvania at 46% and the associated risk factors. Future interventions such as general public awareness campaigns, educating childhood contacts on the presenting symptoms of T1D or islet autoantibody screening for high-risk populations were proposed to reduce the rate of children presenting in DKA at diagnosis of T1D. 13,14

ACK N OWLED G EM ENTS
The authors would like to thank the dedicated staff of the Penn State Hershey Pediatric Diabetes Clinic.

CO N FLI C T O F I NTE R E S T
No authors report a conflict of interest.

AUTH O R CO NTR I B UTI O N
KTB and KBK performed the research. KTB, DH and KBK designed the research study. ES analysed the data. KTB, DH, ES and KBK wrote and edited the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.