We used the 2007 MarketScan Commercial Claims and Encounters (CCE) Database (MarketScan Database; Thompson Medstat, Ann Arbor, MI). This database, which has been used extensively in studies of health care costs, including studies of diabetes costs (12,14–16), contains fully adjudicated and paid claims for several million people with employer-sponsored health plans, including employees, spouses, and dependents (17). The CCE database includes patient-level data on inpatient, outpatient, and drug claims that can be linked using encrypted enrollee information. Types of health plans from which enrollees obtain services can be divided into fee-for-service (FFS) plans and fully or partially capitated plans. FFS plans include preferred provider organization (PPO) plans, exclusive provider organization plans, point-of-service (POS) plans, consumer-directed health plans, and indemnity plans. Fully or partially capitated plans include HMO plans and capitated POS plans (18). Cost estimation for those enrolled in FFS health plans is straightforward, because claims represent the payments for each service. For those enrolled in capitated plans, payment records often reflect encounter status and not the actual payment (18). Therefore, we limited our analysis to youth in FFS plans. All expenses are in 2007 U.S. dollars.

Our study population consisted of U.S. youth who had health care service and prescription drug coverage through an FFS plan between 1 January and 30 December 2007.

Youth with DKA and severe hypoglycemia were identified in three steps. First, following criteria used in a previous study (19), enrollees were identified as having diabetes if during calendar year 2007 they had at least one claim for diabetes-related drugs and supplies and at least two outpatient encounters ≥30 days apart in which diabetes was a primary or secondary diagnosis or at least one inpatient admission with diabetes as a primary or secondary diagnosis. The requirement of at least two outpatient encounters excludes those who were misdiagnosed at the first encounter but later determined not to have diabetes (16). We used the ICD-9-CM codes 250.00–250.93, 357.2, 362.0–362.02, and 366.41 (18) as diabetes indicators in the outpatient and inpatient claims. In the second step, youth with DM were identified to have ITDM if they had at least one drug claim for insulin as indicated by therapeutic class code 172. Last, youth with ITDM were identified to have DKA if they had any claims coded with ICD-9-CM code 250.1 (2) and with severe hypoglycemia if they had any claims coded with ICD-9-CM codes 251.0, 251.1, 251.2, or 250.8 (12).

To determine the extent to which the number of episodes of DKA or severe hypoglycemia was associated with excess expenditures, we dichotomized youth with these complications into those having one episode and those having more than one episode. We considered an individual to have had one episode of DKA if they received one service with a DKA code during the study period and to have had more than one episode if they had received at least two such services at least 30 days apart. We used the same approach to determine the number of severe hypoglycemia episodes.

Complete records were available for 7,724 youth with ITDM. We excluded 39 with missing census region or urbanity of residence (urban versus rural). We also excluded 129 people with uncommon chronic conditions (congenital heart failure, hemiplegia, lymphoma, Down's syndrome, autism, leukemia, liver diseases, and congenital heart defects). We could not consider these conditions because the number of youth experiencing each condition was very small (e.g., ≤10 among DKA or severe hypoglycemia youth). Thus, our final analytic database had 7,556 youth. Because of its relative high prevalence and possible interaction with diabetes expenditures, we did not exclude 309 (4.1%) youth with asthma.

We used Stata version 10.1 (Stata, College Station, TX) for all analyses. We estimated four expenditure models (outpatient, inpatient, drug, and total). The presence or absence of DKA or severe hypoglycemia were modeled using two indicator variables in each of the four models. Covariates included age, sex, census region (midwest, south, west, and northeast), residence urbanity, health benefit plan type (PPO versus non-PPO), and presence of asthma. We also estimated models including the DKA/severe hypoglycemia interaction term.

We compared the unadjusted means of the characteristics of youth by their DKA or severe hypoglycemia status with Student t tests. We used a generalized linear model with log link and gamma distribution to model total, outpatient, and drug expenditures because all youth had positive expenditures and expenditures were highly skewed and right tailed (20). We used a two-part model to model inpatient expenditures to account for those with no inpatient expenditures (21,22). In the first part of the two-part model, we used logistic regression to estimate the probability that someone would have had any inpatient expenditures, and, in the second part, we used a generalized linear model with log link and gamma distribution to estimate the inpatient expenditures for those who had such expenditures.

We calculated model-based predicted marginal medical expenditures by DKA status. In this approach, predictions were made for all observations assuming no DKA status for all observations and then again assuming DKA. All other variables remained at their original values. The predicted excess expenditure associated with DKA was calculated as the difference between the predicted expenditures for youth with and without DKA. We used the same approach to estimate mean medical expenditures for youth by severe hypoglycemia status and predicted excess expenditure associated with severe hypoglycemia. We used 100 nonparametric bootstrap replications to calculate the estimates' standard errors. We considered results significant if P < 0.05.

Unadjusted total mean medical expenditures were $6,191 more for youth with at least one DKA episode than for those with none and $5,151 more for those with at least one severe hypoglycemia episode than for those with none (P < 0.05) (Table 2). Except for drug expenditures by severe hypoglycemia status, the unadjusted means for all expenditure categories were also significantly greater among youth with DKA or severe hypoglycemia than among those without (P < 0.05). A total of 83% of the youth in our sample had zero inpatient expenditures (Table 2).

Total medical expenditures were significantly less among youth residing in the midwest and south than in those residing in the northeast (P < 0.05) (Table 3). Expenditures were significantly greater among youth with asthma, DKA, or severe hypoglycemia compared with their counterparts (P < 0.05). Outpatient expenditure was greater among girls than among boys (P < 0.05). Those in the south and west had significantly lower outpatient expenditures than those in the northeast (P < 0.05). Outpatient expenditures were significantly greater among youth with asthma, DKA, or severe hypoglycemia than among their counterparts (P < 0.05). The probability of having inpatient expenditures decreased with age; however, the expenditure increased with age (P < 0.05). Those residing in the midwest and south were significantly less likely to be admitted to the hospital than those in the northeast (P < 0.05). Both probability of having inpatient expenditure and expenditures were greater among youth with asthma than those without and greater among those with at least one severe hypoglycemia episode than among those with none (P < 0.05). Those with DKA were significantly more likely than those without to be admitted to the hospital (P < 0.05). Prescription drug expenditures decreased with age and were greater among boys than among girls (P < 0.05). Residents of other regions had significantly less prescription drug expenditures than northeastern residents (P < 0.05). Prescription drugs expenditures were greater in urban areas than in rural areas, greater among youth with asthma than those without, and greater among those with at least one severe hypoglycemia episode than among those with none but significantly lower among those with DKA than those without DKA (P < 0.05). The terms for the DKA/severe hypoglycemia interaction were not significant and therefore excluded in all the analyses.

The predicted mean total medical expenditure was $8,398 for youth with no DKA episodes and $5,837 more for those who had at least one (P < 0.05). Mean outpatient and inpatient expenditures were significantly greater among youth who had at least one DKA episode (P < 0.05) (Table 4). Inpatient expenditures accounted for slightly >40% of the total medical expenditures incurred by youth who had DKA, but excess inpatient expenditures were >90% of the excess expenditure attributable to DKA. In contrast, mean prescription drug costs were lower among youth who had DKA than those who did not (P < 0.05). The mean excess total medical expenditure for youth who had one episode of DKA was $3,554, which was about half than that of youth who had more than one ($8,455). Excess inpatient expenditures accounted for >90% of the excess medical expenditures associated with both one and more than one episode of DKA. The mean annual total medical expenditure for those who had DKA was 1.7 times greater than for those without and was even greater for those with multiple DKA episodes.

The predicted mean total medical expenditure was $8,970 for ITDM youth who had no severe hypoglycemia and $3,880 more for those who had at least one severe hypoglycemia (P < 0.05) (Table 4). Compared with youth who had no severe hypoglycemia episodes, the mean excess total medical expenditure was $2,888 among those who had one episode and $5,929 among those who had more than one episode. Excess outpatient expenditures accounted for 46.7% of the total excess expenditures associated with severe hypoglycemia. The mean annual total medical expenditure for youth who experienced at least one severe hypoglycemia episode was 1.4 times greater than for those without severe hypoglycemia episodes.