We analyzed data from the MarketScan Commercial Claims and Encounters Database (CCE) (MarketScan Database; Thompson Medstat, Ann Arbor, MI). Widely used to estimate the health expenditures of various illnesses—including diabetes among adults (6,12,13)—the CCE database annually compiles fully adjudicated and paid health insurance claims from more than 100 large employers and health plans. The claims come from several million individuals, including employees, their spouses, and dependents, who are covered by employer-sponsored private health insurance (14). The CCE includes patient-level data on inpatient, outpatient, and drug claims. All three can be linked through encrypted and unique enrollee identifiers (15).

Enrollee health plans are 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 plans, consumer-directed health plans, and indemnity plans. Fully or partially capitated plans include health maintenance organizations and point-of-service plans with capitation (15).

We restricted our analyses to 3,366,791 youth aged ≤19 years who were continuously enrolled in FFS health plans with prescription drug coverage from 1 January through 30 December 2007 (Fig. 1). We limited our analysis to individuals enrolled in FFS plans because their insurance claims represent actual payment for the services, whereas payment records in capitated plans often only reflect encounters (15). We excluded 0.5% of the youth because of missing records for census region (northeast, midwest, south, and west) or residence (rural vs. urban).

We identified an individual as having diabetes if the record showed 1) at least two outpatient encounters that were at least 30 days apart, coded for diabetes as a primary or secondary diagnosis (16), or 2) at least one inpatient admission coded for diabetes as a primary or secondary diagnosis, and 3) at least one pharmaceutical drug claim filled for insulin or oral hypoglycemic agents, or both. The requirement for at least two outpatient encounters protected against inclusion of youth who were misdiagnosed at the first encounter and were later determined not to have diabetes (17). To indicate diabetes, we used the ICD-9-CM codes 250.00–250.93, 357.2, 362.0–362.02, and 366.41 in outpatient and inpatient claims (12,18) and the Therapeutic Class Codes of 172, 173, and 174 in drug claims. We identified 8,226 youth who had diabetes.

Because the presence of other chronic conditions could substantially influence health care expenditures, the model should account for these conditions. Among the 8,226 youth with diabetes, 4.6% had asthma, the most commonly reported chronic disease in youth, and we included data on these individuals. A total of 159 youth (<2%) had one or more of the following uncommon chronic conditions: congenital heart failure, hemiplegia, lymphoma, Down syndrome, autism, leukemia, congenital heart defects, and liver diseases. Because the number of observations for each of the uncommon chronic conditions was too small to produce a reliable estimated coefficient for that condition, we excluded data on these youth. Remaining in our sample were 8,226 youth with diabetes.

We used bootstrapping to estimate the SEs of predicted and excess mean expenditures. Bootstrapping the entire database of diabetes and without diabetic youth exceeded our computation capacity, so we used a case-control design to reduce the sample size. Each youth with diabetes (case subject) was directly matched one-to-one with five youth without diabetes (control subject), on the multiple variables: age (±2 years), sex, U.S. census region, and residence. For each case, control subjects were selected from a randomized subset of combination of matching variables without replacements. Each case-control matched set was identified as a unique cluster. Our analytic sample comprised 49,356 youth.

Data limitations precluded us from conclusively distinguishing type 1 and type 2 diabetes. We distinguished youth with diabetes by treatment mode (ITDM vs. NITDM). Those who had at least one prescription filled for insulin (Therapeutic Class Code of 172) were identified as having ITDM. By use of this method, we identified 7,556 diabetic youth who had ITDM. In contrast, remaining in our sample were 670 individuals with diabetes but without insulin prescriptions. These youth were considered to have NITDM.

We used t tests to assess differences in the means of sample characteristics and compared youth with diabetes and those without diabetes and youth with ITDM and those with NITDM.

Of the 49,356 youth, 15.2% had no medical expenditures, 18.0% had no outpatient expenditures, 95.9% had no inpatient expenditures, and 35.1% had no prescription drug expenses. Of those identified with diabetes, 91.9% were taking insulin, alone or combined with oral hypoglycemic agents, and 84.0% had not been hospitalized during the year.

The characteristics of the study population by diabetes status and treatment mode are summarized in Table 1. Among youth with diabetes, those with ITDM were significantly younger (12.7 vs. 14.8 years; P < 0.05) and less likely to be girls (47.7% vs. 67.5%; P < 0.05). Compared with youth with NITDM, those with ITDM were more likely to be from the west (13.3% vs. 16.7%; P < 0.05) and less likely to be from the south (52.1% vs. 42.8%; P < 0.05). The percentage of youth with asthma was significantly greater among those with NITDM than among those with ITDM (9.7% vs. 4.1%; P < 0.05).

The predicted mean annual total medical expenditure was $9,061 for youth with diabetes and $1,468 for youth without diabetes, an excess of $7,593 among diabetic youth (P < 0.05; Table 2). The outpatient, inpatient, and drug expenditures were all significantly greater for diabetic youth than those without (P < 0.05). Regardless of diabetes status, outpatient expenditures (43 and 66% for those with or without diabetes, respectively) accounted for the largest share of the total expenditures, followed by prescription drugs (39 and 19%, respectively) and inpatient expenditures (18 and 15%, respectively). However, excess expenditures on prescription drugs accounted for the largest proportion of the excess total expenditures ($3,228; 43%), followed by excess expenditures on outpatient care ($2,970; 39%) and inpatient care ($1,406; 19%). For youth with diabetes, the predicted mean expenditures were $811 on diabetes supplies and $1,688 on medications (data not shown).

The predicted mean total medical expenditure was $9,333 for youth with ITDM and $5,683 for those with NITDM, an excess of $3,650 (P < 0.05) for youth with ITDM (Table 2). The three component expenditures were all significantly greater for youth with ITDM than for youth with NITDM (P < 0.05). Regardless of treatment mode, the outpatient expenditures comprised the largest share of total expenditure (43 and 58%, respectively, for ITDM and NITDM youth), followed by prescription drugs (39 and 26%, respectively) and inpatient expenditures (18 and 17%, respectively). However, of the predicted excess total medical expenditure for youth with ITDM, predicted excess expenditures on prescription drugs accounted for the largest proportion ($2,137; 59%), followed by expenditures on inpatient ($712; 20%) and outpatient care ($697; 19%). The expenditure on diabetes supplies was $863 for ITDM and $82 for NITDM youth, and the expenditure for diabetes medication was $1,826 for ITDM and $228 for NITDM youth (data not shown).