Medical Care Expenditures for Individuals with Prediabetes: The Potential Cost Savings in Reducing the Risk of Developing Diabetes

The National DPP is a partnership of public and private organizations providing intensive training in diet, physical activity, and behavior modification. Increased physical activity and weight loss achieved from participation in the program lower the risk of diabetes by improving the body's ability to use insulin and process glucose.¹⁷ These programs are based on evidence from a randomized clinical trial that showed that people with prediabetes who took part in a structured lifestyle change program cut their risk of developing type 2 diabetes by 58%.¹⁰ This risk reduction was a result of a 5% to 7% body weight loss through healthier eating and 150 minutes of physical activity a week. These group-based programs are offered both in person and virtually and require a 1-year time commitment with 16 weekly sessions followed by 6 additional monthly sessions.^17–19 The 5% weight loss is a minimum target specified in the national standards of the program and further scientific evidence may alter the minimum. For example, a CMS report found that, on average, attending more sessions led to greater weight loss, such that those who completed 16 or more sessions experienced weight loss of 6.2%.^14,17,20 Staying in the program for the full year is essential to help establish new habits and avoid slipping back into old habits.²¹ The impact of the program can last for years. Even after 10 years, people who completed diabetes prevention lifestyle change programs were one third less likely to develop type 2 diabetes.²²

This study utilized individual-level data from the Truven Health MarketScan^® Lab Database, which is a 4.4-million person subsample of the Truven Health MarketScan Treatment Pathways. This de-identified data set allowed for a specific focus on the commercially insured population. The database contained integrated claims, lab test results, and enrollment information submitted to Truven under business agreements with more than 150 employers and 200 commercial insurance carriers. Data from 2008–2012 were analyzed for continuously enrolled adults between the ages of 18–64 years in 2008 with an order to screen hemoglobin A1c (HbA1c) in 2009 and a valid result reported within 6 months of the initial laboratory screening. They were selected from patients who met the American Diabetes Associations (ADA) recommendation for HbA1c levels.²³ Patients excluded were those diagnosed with diabetes or other conditions associated with diabetes (including gestational diabetes), and who were pregnant or taking metformin for at least 6 months prior to the screening.

In the base year, prediabetes was defined as an initial HbA1c test result between 5.7% and 6.4%.^23,24 Other laboratory screens such as a fasting plasma glucose and oral glucose tolerance were excluded from this analysis because patient reports of fasting are often unreliable. The final sample of 8229 adults with prediabetes in the base year was identified after applying the inclusion criteria in Figure 1. Claims for the prediabetes cohort were tracked based on International Classification of Diseases, Ninth Revision codes for primary and secondary diagnosis of diabetes mellitus (all codes with prefix of 250 and 249) and other conditions associated with diabetes (357.2, all codes with prefix of 362, 366.41, all codes with prefix of 648).

Total medical expenditures included payments made by both insurance providers and individuals and are reported in nominal US dollars (not adjusted for inflation). A subset of expenditures was studied for primary care physician (PCP) office visits, specialty care office visits, nonphysician office visits, inpatient admissions, other outpatient care, emergency room visits, pharmaceutical prescription drugs, and laboratory tests.²⁵ Following the initial HbA1c screen in 2009 (the base year), medical expenditures and diagnoses of diabetes were tracked for 3 years (2010 to 2012). The 3-year time horizon from detection of prediabetes was selected for assessing the net savings and ROI from lifestyle intervention programs because it generally aligns with that of health plans and employers who see employee turnover or change health care plans every few years.²⁶

Two tailed z tests were used to compare proportions and 2-tailed t tests were used for comparisons of means. Statistical significance was defined as P < 0.05. All analyses were conducted using STATA 13 (StataCorp LP, College Station, TX).

To help inform the private payers' decision whether or not to include coverage for the National DPP, results from research on medical expenditures was combined with findings from previous studies¹⁰ to develop an algorithm to estimate net cost savings and the ROI from enrollment and participation in a National DPP for eligible employees.²⁷ First, among the adult population with lab results, approximately 37% were expected to have prediabetes.^1,28 Actual prevalence rates may vary by demographics and the lab indicators used, and there is limited research for prevalence rates specific to a 3-year cohort that aligns with the screening criteria used in this study. Next, employers may anticipate a 10%-50% enrollment rate in National DPPs, which is a broad range around a 30% midpoint enrollment rate.²⁹ Enrollment is defined as eligible participants attending at least 1 National DPP session. Furthermore, in order to see the health benefits of enrollment in a National DPP, participants must complete the programs. The CDC defines completion rate as the number of enrollees who completed 12 or more sessions (9 during the first 6 months and 3 during the second 6 months) divided by the number of eligible enrollees. Alternate definitions of benchmark completion include attending 9 or more sessions of the core 16, attending all sessions among those who attend the initial core sessions, or by using attrition rates.^18,29,30 Therefore, the available evidence suggests the program completion rate is expected to be 40%-70%.^18,29,30 Finally, the 3-year cumulative incidence of diabetes among the prediabetes population with and without the lifestyle intervention is used to measure how many individuals completing the lifestyle change program would prevent the onset of diabetes.¹⁰ It is assumed that the lifestyle change program implementation would be based on programs adopted from National DPPs³¹ and that achieved milestones for enrollment and completion rates are similar to those from prior research.

Baseline descriptive statistics for selected demographic and clinical characteristics for the 8229 commercially insured adults (39% of the sample) identified with prediabetes based on lab results in the base year are shown in Table 1. These screening results aligned well with national estimates.¹ Among adults between the ages of 18–64 years, a majority (61%) was between the ages of 45–59 years with the remaining individuals equally distributed between the younger and older age brackets. Additionally, the study cohort was almost evenly distributed by sex. The cohort was more heavily concentrated in the south and west regions of the country; however, sampling weights were not available to adjust for location. A breakdown of other health conditions among this prediabetes cohort is shown in Table 1.

The onset of diabetes among this cohort of individuals with prediabetes was tracked for 3 years. Further descriptive details for the individuals with no onset of diabetes and those with an onset of diabetes are shown in Table 1. Over 3 years, 29% (n = 2414) of the cohort of adults with prediabetes developed diabetes. Among those who were diagnosed, 53% were diagnosed in the first year, 28% in the second year, and 19% in the third year. Interestingly, this 3-year prevalence rate matches the 29% rate found in the original DPP trial.¹⁰

The results from the univariate chi-square tests show that among demographic characteristics, only age was associated with the onset of diabetes (P < 0.001). Compared to individuals with no diagnoses of diabetes 3 years from the base year, those diagnosed with diabetes were less likely to be between the ages of 35–44 or 45–54 years (P < 0.001), and more likely to be between the ages of 60–64 or 65+ years (P < 0.001). Chronic conditions also were found to be significantly associated with the progression from prediabetes to a diagnosis of diabetes. Hypertension was more prevalent among individuals who developed diabetes than those who did not (P < 0.001) (Table 1). There were significant, but only slight differences between the 2 groups in the prevalence of chronic obstructive pulmonary disease (P < 0.001) and congestive heart failure (P < 0.001). Depression was no more likely among individuals with diabetes than among individuals with prediabetes (P = 0.626).

The next part of this study tracked medical expenditures over time and compared the differences in spending between those who developed diabetes and those who did not in the at-risk population identified. Average annual medical expenditures for the prediabetes cohort were $6837 (Fig. 2), and were higher among those diagnosed with diabetes in subsequent years compared to those who were not diagnosed. Overall, medical expenditures rose with time and the initial spike in the first year costs for those newly diagnosed with diabetes most likely occurred from poorer health status or other utilization of medical care that led to the discovery of diabetes.^32,33 To measure the differential in spending by prevalence of costly chronic diseases between the 2 groups, propensity score matching was used to identify individuals with similar characteristics to the diabetes group in the non-diabetes group based on demographic and medical claims variables (results not shown). The difference in medical spending in the base year was found to be almost negligible after matching, suggesting that preventing the onset of diabetes can indeed result in medical cost savings (P > 0.05). It is important to note that the results do not suggest that the added costs were simply from diagnosing diabetes, but a combination of health factors among these individuals. The annual differential in average medical expenditures was approximately 35%, 44%, and 30% greater among individuals with diabetes in the first year (2010), second year (2011), and third year (2012) following the HbA1c screening, respectively (P < 0.05) (Fig. 2).

Figure 3 shows the average medical expenditures for individuals diagnosed with diabetes and individuals who were not diagnosed in the third year after the initial screen by selected medical care categories where the differences in average expenditures were statistically significant (P < 0.05) between the 2 groups. For those who incurred a medical expense, the average per capita medical care expenditures for individuals with diabetes were significantly higher for pharmaceuticals (P < 0.05), laboratory tests (P < 0.05), PCP office visits (P < 0.05), and specialty care (P < 0.05). Similarly, Figure 4 shows a 3-year total for cumulative average expenditures for individuals with diagnosed diabetes compared to individuals not diagnosed. The differences in expenditures between the 2 groups of patients was 40% for PCP office visits (about 2 additional visits per year), 23% for lab tests, 22% for prescription drugs, and 10% for specialty care (all P < 0.05). It is important to note that these were average expenditures specific to each sector and do not represent the total average medical spending trends displayed in Figure 2.

For simplicity and ease of interpretation, these findings and assumptions were applied to an algorithm based on existing evidence to a sample of 1000 individuals to simulate experience for commercially insured beneficiaries with coverage and access to a CDC-recognized DPP. It is anticipated that 370 individuals (37%) would have a lab screening result in the prediabetes range.^1,28 Of the eligible individuals, 37–185 individuals (10%-50%) would enroll in a National DPP,²⁹ and 15–130 individuals (40%-70%) would successfully complete the program.^18,29,30 Based on results from this analysis and findings in previous studies, among those who complete, 4–37 individuals (29%) would be diagnosed with diabetes within 3 years if they had not participated in the intervention.¹⁰ However, with the intervention, only 2–19 (14%) individuals are expected to develop diabetes.¹⁰ Therefore, the cumulative potential cases of diabetes prevented by the lifestyle program in 3 years would be the differences between the 2 groups, consisting of 2–19 individuals. It is important to note that the core conversion rate of diabetes among those with prediabetes was based on a 3-year cumulative rate. Further, the reduction in incidence rates was based on enrollment and completion of the National DPP research study.

It is estimated that individuals participating in a CDC-recognized National DPP translated to community or primary care settings with explicit adoption of National DPP training materials would be charged $400-$500 per person depending on program costs for group-based settings.³⁴ For an employer with 1000 beneficiaries with similar risk and severity factors, using these estimates for the 37–185 individuals with prediabetes enrolled in a National DPP, the estimated total program costs would range from $16,650 to $83,250 ($450 multiplied by the estimated 37 to 185 individuals).

An employer's investment in a lifestyle change program based on these default prevalence, enrollment, completion, and incidence rates would generate potential net savings over 3 years ranging between −$3131 to $35,037 and a corresponding 3-year ROI ranging from −19% to 42%. Assuming the individuals with prediabetes were able to maintain their health status and prevent the progression to diabetes, net savings in the long run would be higher because of lower medical expenditures over time.¹³ In addition, these values are raw estimates that do not include medical insurance components on both ends including premiums, co-pays, and deductibles as well as all payments made by the insurance carrier. Further, it is important to note that this net savings estimate for participation in the National DPP interventions does not include the savings from avoiding lost wages and earnings resulting from increased absenteeism and reduced productivity of the employed population^2,35 or the benefits from decreasing the loss of quality of life. Overall, just as with the Medicare population, preventing the onset of diabetes among the commercially insured population with prediabetes also can result in improved health outcomes and reduced medical spending.