The carefully conducted and clear cost–utility analysis by Javitt and Aiello [1] provides useful information on the economics of diabetic retinal screening for different populations from the perspective of a health insurance agency. There are, however, two major limitations to the application of the study findings to decision making.

The valuation of sight considers two levels of outcome, depending on whether the affected person is well or poorly adjusted to his or her disability. The proportion of patients reaching these end points is estimated from national figures on the outcome of rehabilitation. This estimate assumes that blind diabetic patients have an outcome from rehabilitation similar to that of other blind persons. This assumption is probably untrue because diabetic persons often have other disabilities when severe eye disease develops. Loss of light, discriminatory touch, and joint position sense is particularly common. It is uncertain how this difference in outcome affects valuation of sight; the complications of the condition mean that these patients are likely to start with a worse quality of life but tend to value their sight more. This makes interpreting comparisons of the costs per quality-adjusted year of life with those of other interventions particularly difficult.

In determining the costs of the intervention, however, it appears from the assumptions of Javitt and Aiello's model that the averted cost of the state of a blind person was subtracted from the cost of delivering the intervention. This confuses the picture because this realm no longer belongs just to the health insurer. Much of the cost of blindness will result from making social rather than health care provisions.