Outcomes for continuous subcutaneous insulin infusion users in young adults from lower socioeconomic backgrounds

Abstract Objective Diabetes technology is available and its efficacy and safety have been demonstrated; however, there is little evidence as to how this technology is being utilized and its effectiveness in vulnerable populations. This study evaluated differences in outcomes for young adults in the United States (U.S.) from lower socioeconomic (SES) backgrounds with type 1 diabetes (T1D) managed on continuous subcutaneous insulin infusion (CSII) versus multiple daily injections (MDI) or fixed‐dose insulin (FDI). Research design, methods and participants Utilizing the Optum® de‐identified Electronic Health Record data set between 2008 and 2018 to perform a retrospective, cohort study, we identified 805 subjects with T1D aged 18–30 years with Medicaid. We evaluated median difference in HbA1c between CSII and MDI/FDI users for 24 months. Predictors of diabetic ketoacidosis (DKA)‐associated hospitalizations by CSII use were evaluated using logistic regression. Results CSII users showed statistically significant lower median HbA1c values at 24 months of follow‐up compared to individuals on MDI/FDI. Non‐white individuals were at lower odds of receiving treatment with CSII. Subjects on CSII were not more likely to be hospitalized for DKA compared to subjects treated with MDI/FDI. Older subjects were at lower odds of being hospitalized for DKA. Males and subjects followed by Endocrinologists were at higher odds of being hospitalized for DKA. Conclusions Young adults with T1D from lower SES backgrounds show improved glycaemic control when in CSII compared to MDI/FDI without increases in hospitalizations for DKA.

of the population. 2,3 Large database and registry data have demonstrated that many youth and young adults with T1D do not meet established haemoglobin A1c (HbA1c) goals. 4,5 In fact, according to data from the T1D Exchange Registry, between 2016 and 2018, only 17% of youth achieved the American Diabetes Association (ADA) HbA1c goal <7.5%. 6 While reasons for poor glycaemic control in this group are multifactorial, lack of diabetes technology utilization may play a role.
Diabetes technologies as stand-alone insulin pumps and CGMs in addition to systems in which the pump and sensor fully communicate are being increasingly utilized to treat T1D. After almost four decades since insulin pumps became commercially available starting with Medtronic's Minimed 502 in 1983, there are substantial data that individuals with T1D on CSII demonstrate better HbA1c outcomes compared to MDI. 7 Furthermore, subjects with T1D on CSII demonstrate improvements in microvascular outcomes including retinopathy and peripheral neuropathy compared to management with multiple daily injections (MDI). 8 CGM was next to make its way to the T1D community and has demonstrated numerous benefits including HbA1c reduction, reduction of glycaemic variability and less hypoglycaemia. 9,10 Moreover, individuals with CGM who are switched from MDI to CSII spend more time in range (TIR) defined as the glucose concentration 70-180 mg/dL compared to those on CGM and MDI. 11 Results seem to improve further the more automated the system, as seen in SAP therapy which has been shown to reduce HbA1c and time spent in hypoglycaemia. 12 Most recently, HCL systems are available and illustrate encouraging time in range data. 13 13,14 Additionally, it appears that there are clear racial and ethnic biases regarding who receives newer technologies even when controlling for SES. 3 The data regarding racial and ethnic biases are predominantly from the paediatric literature. In a study by Willi et al using data from the T1D Exchange Clinic Network, fewer Black compared to white children across all income strata were on insulin pumps. Higher HbA1c values were seen even in high-income Black families, perhaps because fewer of them were managed with insulin pump therapy. Black children with private insurance were less likely to be on insulin pumps compared with white children without private insurance. 3 In another study in a paediatric population of subjects with T1D by Lin et al, subjects of non-Hispanic white race and higher socioeconomic status were more likely to be placed on pumps in the first year following diagnosis. 19 In a recent study examining racial-ethnic inequity in young adults with T1D, Black young adults had the lowest insulin pump use, despite similar rates of public insurance as Hispanic young adults. 20 When diabetes technology companies reach out to insurance companies through their managed care teams, they lobby for cover-

| Data source
We performed a retrospective cohort study using the Optum® deidentified Electronic Health Record (EHR) data set. 21 The data set contained EHR data from 5 million adults (age 18 and older), nationally distributed across the United States. EHR data contains ICD-9 and ICD-10 codes, prescription medication orders, vital signs, laboratory results, procedure codes and demographic measures. The

| Study sample
The study sample was limited to T1D patients with an encounter between 2008 and 2018 and had activity for at least two years after first encounter date. Subjects aged 18 to 30 years, on Medicaid, and with a diagnosis of T1D were identified. Medicaid was used as a proxy for lower SES. 23 Subjects with a diagnosis code of pregnancy and those who received glucose-lowering medications for type 2 diabetes (T2) were excluded from the analysis.

| Measures
Prescription and diagnosis codes for all variables are listed under supplemental material.

| Exposure
Treatment modality was classified as CSII versus MDI or FDI.
Subjects were assigned to CSII or MDI or FDI based on encounter diagnosis codes (see Table S1 for a list of diagnosis codes). For subjects with a diagnosis code for CSII, the first record indicating CSII was considered the date of insulin pump initiation. Patients on MDI and FDI were combined as the non-CSII group and were defined using prescription codes, and the date of enrolment in the cohort was considered their treatment start date (see Table S2). All laboratory results and hospitalization data prior to this treatment start date were excluded from this analysis.

| Outcomes
The primary outcome was HbA1c (%) measurements obtained at 3-6, 7-12, 13-18 and 19-24 months following the assignment of treatment modality. HbA1c values were considered if occurred at least 3 months after CSII prescription. We used the last HbA1c for every patient with multiple HbA1c records per each 3-month interval. For graphical illustrations, laboratory data in the form of HbA1c were analysed at 6-month intervals.
The secondary outcome was hospitalization for DKA. Unique DKA encounters were defined as an inpatient diagnosis code for DKA corresponding to a distinct inpatient visit encounter ID. We calculated the total number of unique DKA episodes of DKA per subject over the 24-month follow-up period.

| Covariates
We controlled for the age, race, ethnicity, gender and income of the subject. Age was classified as 18-26 years and 27 or older at the time of first encounter. This was based on knowledge that parental healthcare coverage for dependent children ends after age 26.
Blacks, Asians and other races were grouped into one racial category and compared to whites in this analysis. Ethnicity was also re-categorized into Hispanics vs. other ethnicities. Subjects making less than $45,000 a year were considered low-income individuals.
Because adult Endocrinologists may differ in their approach to prescribing CSII to their patients when compared to physicians from other specialties, a composite variable was generated specifying if the subject had at least one encounter with an adult Endocrinologist and was adjusted for in the analysis.

| Statistical analyses
Characteristics and outcomes of the study sample overall and stratified by CSII use are presented as medians and interquartile ranges or frequencies and percentages, as appropriate. Bivariate comparisons for all variables by CSII use were examined using the Mann-Whitney U tests for continuous variables and Chi-square tests for categorical variables. Differences in HbA1c between CSII and MDI/FDI users were compared at baseline and at each determined interval for the 24-month follow-up period. Univariate and multivariable logistic regression analyses were used to identify factors associated with CSII prescription and DKA-related hospital admissions. A histogram was used to illustrate the HbA1c at 6-month intervals. Model fitness was tested using the likelihood test. All tests were two-sided, and the alpha level of significance was set at 0.05. All analyses were done using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).

| RE SULTS
Demographic characteristics of the 805 subjects that met inclusion criteria are shown in Table 1

| DISCUSS ION
In our sample of young adults with T1D on Medicaid, we found that median HbA1c levels were statistically significantly lower in individuals managed with CSII compared to MDI/FDI at 24 months of follow-up. These results are consistent with an analysis of randomized clinical trials comparing CSII with MDI in subjects with T1D in which CSII was shown to lead to statistically lower HbA1c values regardless of whether regular human insulin or rapid-acting analogue insulin was utilized. 7 In addition to being statistically significant, the difference between median HbA1c values in the CSII group versus the MDI/FDI group is clinically relevant. According to landmark data from the Diabetes Complications and Control Trial (DCCT) and the

Epidemiology of Diabetes Interventions and Complications (EDIC)
follow-up, intensive glucose control reduces the risk of microvascular complications. 24 In fact, the risk reduction is non-linear, and thus, even greater risk reduction is seen in those individuals with initial higher HbA1c levels. This is particularly meaningful in the lower SES population that we describe, who if faced with further disabilities due to microvascular changes including blindness, dialysis and amputation, the further down on the socioeconomic ladder he or she will fall. Ultimately, this will create an economic burden and result in increased healthcare utilization.
Risk for DKA was not greater in the CSII group; a significant finding given the reduced utilization of CSII in the Non-white population. It is possible that there is inherent bias by practitioners to avoid prescribing CSII to Non-white individuals and low SES groups due to a fear that those populations may not be able to successfully operate the technology. The lack of increased risk of DKA for persons on CSII may provide more assurance to providers that they can be prescribed to patients for whom technology has been historically withheld out of fear of adverse outcomes. Predictors of DKA included following with an adult Endocrinologist which may be explained by the fact that typically more complex individuals with T1D follow with specialists. Male subjects were also at increased odds for hospitalization for DKA which warrants further explanation.
Older individuals were at less risk of DKA perhaps owing to more   should be more inclusive of young adults with T1D of lower SES.

ACK N OWLED G EM ENTS
This study was made possible by the Clinical Research Scholarship Program (CRSP) and the Saint Louis University Department of Health and Clinical Research Outcomes.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest relevant to this article to disclose.

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