BACKGROUND: Strong evidence supports the efficacy of laser treatment for diabetic retinopathy; however, compliance with screening has been disappointing, and the optimum frequency of screening remains controversial. This cost-utility analysis compares cost, duration of blindness, and quality-adjusted life years (QALYs) for several intervals of screening.
POPULATION STUDIED: A hypothetical cohort of people 40 years and older with diabetes was modeled on the Third National Health and Nutrition Examination Survey. Estimates for retinopathy development and progression were taken from the United Kingdom Prospective Diabetes Study. Thus, the study population is probably similar to that seen by the typical family physician, though caution should be exercised in generalizing results to practices with higher proportions of Hispanics, African Americans, or other ethnic populations for whom the prevalence and rate of progression of retinopathy may be different from the reference populations.
STUDY DESIGN AND VALIDITY: This cost-utility study used a Markhov model to compare the impact of annual versus less frequent screening on the progression of diabetic retinopathy and macular edema. The perspective was that of a payer; costs included an ophthalmology visit, laser photocoagulation, and fluorescein angiogram. Age and hemoglobin (Hb) A1C levels were used to define patient groups at high, medium, and low risk. Predictions for life expectancy were adjusted for time spent blind on the basis of a utility of 0.69. A 3% discount was applied to all costs and years of life. The methodology of this study was strong. The authors defined alternatives explicitly and calculated incremental analysis of cost. Their model addressed accuracy of diagnosis, and they performed thorough univariate and multivariate sensitivity analyses. Weaknesses included the lacks of evidence for effectiveness of population screening for diabetic retinopathy, attention to the rate of poor follow-up, and recognition of the impact of visual impairment without blindness.
OUTCOMES MEASURED: The primary outcomes measured were costs, days of blindness, and QALYs. Costs directly due to blindness, the impact of labeling, and patient satisfaction were not addressed.
RESULTS: The effectiveness of screening is dramatically greater for patients at a higher risk of retinopathy; increasing screening frequency modestly decreases time spent blind. For example, for a low-risk patient (eg, 75 years old with a Hb A1C of 7%) screening every third year prevents 1 day of blindness during the patient’s lifetime, and annual screening prevents an additional 2 days of blindness (compared with no screening). For a high-risk patient (eg, 45 years old with a Hb A1C of 11%), screening every third year prevents 188 days of blindness, and annual screening saves an additional 21 days of blindness. Within any given risk group, the cost of screening increases dramatically with increasing frequency. For example, even for the high-risk patient for whom screening is most cost-effective, the marginal cost-effectiveness of increasing screening from every other year to every year was $181,850 per QALY gained. Extrapolating the results to the whole population and using $50,000 per QALY as a criterion of cost-effectiveness, screening every second year is cost-effective ($49,760/QALY), while annual screening is not ($107,510/QALY). Sensitivity analysis did not change the results substantially, except that lowering the utility score for blindness increased the cost-effectiveness of all screening.
This study provides good evidence that regular screening for retinopathy in high-risk patients with diabetes is cost-effective, but suggests that insurance companies and the developers of the Health Plan Employer Data and Information Set reconsider their insistence on annual screening for all patients as a measure of the quality of care. Clinicians should be cautious about extending this analysis to individual patients, but these findings underscore the clinical importance of the risk of retinopathy and give permission to clinicians to tailor recommendations to individuals on the basis of their risk. The findings are substantially limited by the lack of evidence that population screening for diabetic retinopathy is effective in usual practice.