Measuring the Value of Better Diabetes Management
Published Online: April 12, 2013
Darius N. Lakdawalla, PhD; Michael R. Eber, BSE; Felicia M. Forma, BSc; Jeffrey Sullivan, MS; Pierre-Carl Michaud, PhD; Lily A. Bradley, MBA; and Dana P. Goldman, PhD
The growing burden of type 2 diabetes mellitus has outpaced the modest progress in the efficacy of diabetes medications. However, it is unclear whether we are using our existing medications optimally. This article quantifies the value of addressing underuse of existing diabetes medications in the United States. Interventions—new technologies, public policies, or clinical approaches—that double the rate of initiation of insulin generate a value on average of more than $15,000 over the lifetime of a patient developing diabetes between ages 51 and 60 years, or $12.6 billion in the aggregate. Interventions that improve adherence would generate a value of more than $13,000 on average for the same patients, or $10.7 billion in the aggregate. The value of such interventions is on par with highly optimistic projections of technological progress in medication efficacy.
Type 2 diabetes mellitus (T2DM) continues to grow at a remarkable rate. Diabetes incidence in the United States rose by more than 45% from 2002 to 2010, and its substantial costs make this trend particularly troubling. Diabetes decreases quality of life, increases morbidity, reduces life expectancy by approximately 3 years in near-elderly cohorts, and imposes additional lifetime healthcare costs of roughly $200,000.1
Prevention is a natural response to the growing epidemic, and its social benefits appear to be quite large.1 Yet the policy question of how to improve and extend the lives of the tens of millions of patients who already have or will soon develop the disease remains. Technological progress in treatment is one possibility. Indeed, the discovery of 8 new drug classes in the past 15 years has expanded the arsenal of medications for management of T2DM. Most of the benefits have accrued to patients who do not respond adequately or durably to older therapies (eg, metformin, sulfonylureas). As David Nathan observes, drug classes “that have been developed recently are generally no more potent, and often less effective in lowering glycemia, than the 3 oldest classes.”2 In other words, the newer drugs have benefited patients who are inappropriate for or unresponsive to established treatments by preventing longer-term macrovascular and microvascular complications.3
Since a substantial segment of patients have not enjoyed major benefits from new treatments, it is worth asking whether we are using existing therapies to the greatest benefit. In fact, evidence suggests underuse along at least 2 dimensions: 1) imperfect medication adherence4,5 and 2) delays in or lack of insulin initiation.6,7
First, it is widely understood that patients with chronic illness do not always adhere to their providers’ instructions. Patient adherence to oral blood glucose– lowering medication and insulin regimens generally ranges from 60 to 85 percent, with even lower adherence reported for certain populations, such as those covered by Medicaid.5 A number of earlier researchers have documented the impact of poor adherence on diabetes-related outcomes. 8-10
Second, concerns arise about underuse of diabetes treatments. Many center on insulin because the injection regimen imposes inconvenience and other personal costs.7 This is unfortunate because compared with prevailing standards, earlier use of insulin may be warranted, particularly in patients with poor glycemic control (ie, glycated hemoglobin [A1C] >9).11,12 Clinicians may be reluctant to initiate patients on insulin because it is taken as a sign of “defeat.”13
The healthcare delivery system also can delay insulin initiation because of the need for referral to an endocrinologist in complex cases.
In light of these issues, we study whether and how poor adherence to and lower uptake of insulin therapy quantitatively contribute to the costs of diabetes on patients and the healthcare system. Both the expansion of insulin use and the reduction of poor adherence are likely to require the investment of resources to develop new policies, clinical approaches, and technologies. We are interested in the potential return on such investments.
To investigate these questions, we calculate and compare the lifetime benefits to patients of scenarios that improve adherence and increase insulin use. We then compare these benefits with the value of a scenario generating further improvements in the efficacy of diabetes medicines. In particular, we calculate the future technological progress that would be required in this scenario to match the value of expanding insulin therapy.
For all such scenarios, we simulate outcomes for a cohort of newly diagnosed US diabetes patients who are aged 51 years in either 2003 or 2004 and who acquire diabetes between ages 51 and 60 years. We then examine the costs and benefits of each of the above scenarios, compared with current diabetes management.
Overview of the Model
We use a well-established dynamic microsimulation model, the Future Elderly Model (FEM),14-17 to model the lifetime consequences of expanded insulin use, improved patient adherence, and future progress in the efficacy of diabetes medicines. The model simulates health status, health spending, and mortality experience of the US population with diabetes over the age of 50 years. The FEM is based on data from the Health and Retirement Study (HRS), a biennial survey of Americans 51 years and older that has been ongoing since 1992. We supplement the HRS with medical spending data from the Medicare Current Beneficiary Survey and Medical Expenditure Panel Study for persons who would not be eligible for Medicare. The FEM is well suited for this study, because it allows the simulation of alternative lifetimes for patients aged 51+ years in different treatment environments. Further detail on the mechanics and implementation of the model are provided in the eAppendix (available at www.ajmc.com).
Figure 1 gives a schematic view of how patients progress in the model. We model 3 steps in the management of diabetes. First, the model assigns an initial A1C level. Second, it assigns a first-line oral drug treatment regimen. Finally, the model predicts whether insulin therapy is initiated and whether oral drug therapy is modified.
Oral medications are categorized in 3 drug classes: metformin (biguanides), sulfonylureas, and thiazolidinediones (TZDs). They represent more than 90% of the oral diabetes medications taken by HRS respondents.
We allow for addition of insulin therapy as a second method of treatment starting in the second year of the simulation, because few T2DM patients are treated with insulin as a first-line agent. Upon initiating insulin, patients face a one-time decision whether to keep taking their oral medication or to depend solely on insulin.
For effects of drug treatment on health transitions, we focused on A1C, weight, and mortality, end points that were consistently reported in the literature and could be mapped to the health conditions in the model.
Table 1 presents estimates for clinical effects of oral diabetes medications used in the model by drug class and outcome. Since evidence on the effects of combination therapy is not universally available, we conservatively assumed that the effect of a combination therapy equals that of its most efficacious component therapy, relative to placebo.
For insulin therapy, clinical effects used in the model are summarized in Table 2. Because we focus on the effect of adding insulin to existing treatment regimens, we combined evidence of the effect of adding insulin treatment on A1C reduction with epidemiological evidence linking A1C to mortality, since the evidence on the direct effect of adding insulin is sparse.
In both cases, evidence of the effect of glycemic control from treatment on mortality is taken from the recent 10-year follow-up study of the United Kingdom Prospective Diabetes Study trial.18
Because poor adherence to diabetes medications decreases the effectiveness of treatment,5 we include adherence measures in the model. We map adherence to effectiveness, using estimates from the literature (see eAppendix).
Monetizing Lifetime Consequences
Finally, it is necessary to monetize effects on life expectancy, in order to compare them with the financial costs of increased treatment. The monetary value of longevity extensions is controversial. Viscusi and Aldy estimated based on a review of the literature that the value of a statistical life ranges primarily from $5 million to $12 million.19 Assuming a 3% real rate of interest and a constant flow of value implies figures between $150,000 and $360,000 for each statistical life-year. We chose a value of statistical life-year of $200,000 for our calculations, which is inside but toward the lower end of this range. Note that we conservatively assume that the value of morbidity reductions, independent of mortality, is zero. Future benefits and costs were discounted at 3% to compare scenario outcomes.
PDF is available on the last page.