We and others have shown that timely detection and treatment of retinopathy in Americans with insulin-dependent (IDDM) and non–insulin-dependent diabetes mellitus (NIDDM) result in considerable savings for the federal budget and, in some cases, for health payer organizations [8-13]. Preventive eye care programs for diabetic persons have been proposed by the National Eye Institute [14], the Centers for Disease Control and Prevention [15], and the American Academy of Ophthalmology [16]. The importance of ophthalmologic screening for diabetic persons is highlighted by its inclusion in the Health Plan Employer Data and Information Set (HEDIS II; National Committee for Quality Assurance, Washington, D.C.) adopted throughout the managed care industry. Increasingly, a cost-effectiveness analysis based on dollars spent per quality-adjusted life-year (QALY) saved has been advocated as a strategy for societal health care allocation decisions. In this report, we evaluate the potential cost-effectiveness of screening, recruitment, and treatment programs for diabetic eye disease. Unlike our previous analyses, which dealt specifically with cost savings to the federal budget, this analysis is expressed in terms of cost-effectiveness from the perspective of health insurers.

Methods

Top Methods Results Discussion Author & Article Info References

Model Structure

We used the PROPHET (Prospective Population Health Event Tabulation) Modeling System detailed in previous articles [8-10]. The PROPHET model is an epidemiology-based net-work simulation program designed for modeling the progression of a chronic irreversible disease. The model is based on Monte Carlo simulation and analyzes events and costs incurred during the course of an irreversible chronic disease; it considers each patient individually. Monte Carlo simulation allows a simple, probability-based solution, using random number generation, of complex disease progression processes over time that would otherwise need to be expressed as multivariate differential equations. For instance, if there is a 1% likelihood of a person in the model progressing to proliferative retinopathy over the next time interval, the simulation software draws a random number between 0 and 1. If that number is 0.01 or less, the person in question is recorded as having developed proliferative retinopathy. If the random number is greater than 0.01, that person remains in the baseline state. The simulation begins with a theoretical cohort representing all Americans in a given age group who develop diabetes in a single year. A complete set of clinical assumptions used in the model is shown in the Appendix.

Data on incidence of diabetes, treatment with insulin, progression of diabetic eye disease, and vision loss are drawn from cross-sectional and longitudinal studies [17-24]. Outcomes of laser photocoagulation for proliferative retinopathy and macular edema are based on published clinical trial results. In the Diabetic Retinopathy Study, panretinal photocoagulation was initially reported to reduce the likelihood of severe vision loss by 60% as compared with no treatment. However, many patients enrolled in this study had advanced disease and, hence, a worse outcome than patients identified for care today. When the published results of panretinal photocoagulation in the Early Treatment Diabetic Retinopathy Study are compared with the "no treatment" group of the Diabetic Retinopathy Study, an 84% reduction in progression to severe vision loss is found (1.48% as compared with 8.7%). We believe this to be a conservative finding because reanalysis of the primary data suggests that the treatment effect may be over 90% [25].

In the case of treatment for macular edema, published data have used the clinically relevant end point of "doubling the visual angle" rather than the end point of legal blindness, which determines eligibility for federally funded insurance and entitlement programs. However, additional data are available from the Early Treatment Diabetic Retinopathy Study, which used a visual acuity end point (that is, the proportion of eyes with vision worse than 20/100) corresponding more exactly to current definitions of legal blindness [26]. The relative benefit of treatment as compared with no treatment is assumed to be permanent, based on two 15-year follow-up studies of patients treated with photocoagulation. In applying these data, we do not assume that successfully treated persons will not lose vision but rather that the reduction in risk for vision loss associated with treatment is constant over time [27, 28].

In our model, each person is assessed for disease progression and mortality by age [29], disease duration, and disease severity [30-33]. Persons being screened, according to the preferred practice pattern for diabetic retinopathy of the American Academy of Ophthalmology [34], are assessed for the presence of proliferative retinopathy and macular edema using published sensitivities for ophthalmic examination [35, 36]. Those persons predicted to have proliferative retinopathy or clinically significant macular edema detected by screening are next assessed for treatment outcome. Treatment failures are determined, net costs and benefits during the cycle are tabulated, and the cycle repeats itself at 2-month intervals throughout the lifetime of the cohort. Monte Carlo simulation is used for determining outcome of all events in each cycle, including disease progression, disease detection, treatment outcome, and mortality [37]. The remaining assumptions about disease progression and outcome of treatment are unchanged from our previous reports.

Incident Cases

The total number of incident cases of diabetes mellitus in the U.S. population is derived from previously published age-specific incident rates [17]. When applied to age-specific 1988 U.S. population figures [38], as shown in Table 1, these rates predict that 576 136 persons will develop NIDDM in the United States annually. This estimate is consistent with previous predictions [39, 40].

Expenditures

Screening and treatment costs are derived from the average Medicare charges in 1990 [41]. Although costs of care may vary among insurers, the Medicare fee schedule is increasingly becoming the basis for computing medical care reimbursement for many indemnity and managed care insurance plans. To the extent that actual costs of care are less than the charges used as inputs in this model, ophthalmologic screening for diabetic persons will be more cost-effective than predicted in our analysis.

All costs are expressed in 1990 U.S. dollars using a discount rate of 5%, based on published techniques [42-48]. Some debate remains among health economists about whether investments in health care should be subjected to present value analysis (that is, discounted) just like any other financial investment. The purpose of discounting is to account for the time-value of money or universal preference for funds now rather than funds later. This value is most appropriately measured by common borrowing costs over and above inflation. Public investments are often discounted at 4% to 5% because it is the rate over inflation at which governments typically borrow funds (that is, sell bonds). The justification for discounting an investment such as screening for diabetic eye disease is that the investment value will be realized only over a period of years, whereas alternative investments in health care—for example, prenatal care or childhood vaccination—may be realized over a shorter term. For this reason, we have discounted person-years of sight and QALYs by 5% (except in Table 3 where comparability to other published cost-effectiveness figures is needed). As can be seen from Table 2 and Table 3, applying discounted values is a more conservative approach because discounting reduces the predicted cost-effectiveness of the proposed intervention.

Screening Implementation

The estimated present implementation of appropriate screening in patients with NIDDM is derived from analysis of the studies by Klein and coworkers [49] and Witkin and Klein [50], who found that 40% of these patients in Wisconsin were not meeting the suggested guidelines for ophthalmic care. Analysis of the 1989 National Health Interview Survey suggests that 50% of persons with self-reported diabetes reported no dilated eye examination in the past year [51]. Consequently, we have chosen 60% as an estimate for the prevailing implementation of appropriate eye care in patients with NIDDM. We assumed that screening under an ideal program is done according to the preferred practice pattern for diabetic retinopathy published by the American Academy of Ophthalmology [34], which suggests using dilated ophthalmoscopy annually for patients with no retinopathy and examination every 6 months for those with retinopathy. This pattern is consistent with recommendations published by the American College of Physicians and the American Diabetes Association [52].

Assignment of Quality-Adjusted Life-Years to Outcomes

In 1989, 74.4% of 12 500 legally blind persons were successfully rehabilitated, and 25.6% were considered "poorly adjusted" after rehabilitation efforts (Avery C, U.S. Department of Education. Personal communication). Drummond [53] estimated that a year of blindness for "well-adjusted" and "poorly adjusted" persons corresponds to 0.48 and 0.36 QALYs, respectively. If these values are accurate, then the total QALYs saved can be determined from person-years of sight. As part of a national study of cataract outcomes, patient utilities for various states of vision were elicited using a standard rating scale method. Utilities associated with severe vision loss were consistent with those reported by Drummond [53] (Bass E. Personal communication).

Results

Top Methods Results Discussion Author & Article Info References

Previous reports have shown that over 413 200 person-years of sight are currently saved and over 710 800 person-years of sight could be salvaged if all diabetic patients had appropriate ophthalmic screening and treatment [8]. These data permit the cost per person-years of sight saved to be calculated as shown in Table 2. The costs of providing currently suggested screening and treatment of diabetic retinopathy is $1757 per person-year of sight saved.

The cost per QALY associated with detecting and treating diabetic eye disease ranges from $1996 per QALY for those with IDDM to $3530 per QALY for those with NIDDM who do not require insulin. Overall, the cost-effectiveness of detecting and treating eye disease in Americans with diabetes is $3190 per QALY (Table 2).

The cost per QALY was reported for various medical interventions by Detsky [54] in 1989. These values were expressed in 1986 U.S. dollars and are shown in Table 3. In Detsky's analysis, QALYs were not discounted. Therefore, the cost–utility ratios associated with treating diabetic retinopathy shown in Table 3 have been adjusted to 1986 U.S. dollars without discounting QALYs so as to compare directly with Detsky's report. As can be seen from Table 3, screening for and treating diabetic eye disease was more cost-effective than any medical interventions studied by Detsky. This remains true, even if one compares the discounted values for ophthalmologic screening and treatment Table 2 with the undiscounted figures for nonophthalmologic interventions shown in Table 3.

Quality-of-life issues are notoriously subjective and difficult to quantify precisely. Figure 1 presents a sensitivity analysis showing the relation of QALYs per year of blindness and the cost–utility results. Note that a year of blindness may increase to 0.65 to 0.75 QALYs while the cost–utility of screening and treatment for diabetic retinopathy still remains less costly than the least expensive intervention reported by Detsky [54].

View larger version (31K): [in this window] [in a new window]   Figure 1. Sensitivity analysis for the cost-effectiveness of ophthalmologic screening and treatment of retinopathy in insulin-dependent (type I) and non–insulin-dependent diabetes mellitus (type II). Assumptions used in the text (74.4% "well-adjusted" patients equals 0.48 QALYs; 25.6% "poorly adjusted" patients equals 0.36 QALYs) represent an x-axis value of 0.45. All costs are expressed in 1990 U.S. dollars.

Discussion

Top Methods Results Discussion Author & Article Info References

Detection and treatment of diabetic eye disease has previously been shown to result in cost savings to the federal budget. In this report, we show that appropriate care for diabetic eye disease is highly cost-effective, using the widely applied QALY method. Prevention of this secondary complication of diabetes, on a cost-effectiveness basis, ranks far above medical interventions commonly provided in the United States. Laupacis and coworkers [55] have suggested that health interventions costing less than $20 000 per QALY are readily adopted by society. Certainly, by these criteria, detection and treatment of diabetic eye disease meets any standard of cost-effectiveness.

Despite its high level of efficacy, clinical effectiveness, and cost-effectiveness, screening and treatment for diabetic eye disease remains a highly underutilized technology. Recent data show that half of those with diabetes have not had an eye examination in the last year. Moreover, half of those identified with high-risk retinopathy in a population-based study had not been treated and at the time of identification had no plans to be treated. In this model, cost-effectiveness of screening and treatment is based only on those who receive eye care because there are no costs included for major public health initiatives to increase screening participation. If such programs existed, then screening for eye disease in diabetic persons would appear increasingly cost-effective as program participation rates increased. In the absence of such public health programs, failure to participate in screening programs contributes neither cost nor savings in years of sight to the model.

Annual eye examination of diabetic persons has now been incorporated into the HEDIS II quality guidelines adopted throughout the managed care industry. These data underscore the wisdom of including diabetic eye disease in those guidelines and the benefit to diabetic persons and society that might result from their implementation.

Dr. Aiello: One Joslin Place, Beetham Eye Institute, Joslin Diabetes Center, Boston, MA 02215.