Objective: To provide scientifically based screening rules for the primary care setting designed to identify, through evaluation of a prescribed and limited portion of the posterior fundus, those patients with diabetes who have retinopathy severe enough to need referral to eye care specialists.

Design: Retrospective analysis of the Early Treatment Diabetic Retinopathy Study (ETDRS) photographic data base.

Participants: The fundus photographic grading data from 3711 patients with diabetes enrolled in the ETDRS.

Methods: Multivariate regression techniques were used to identify retinopathy lesions in photographic fields 1, 2, 3, or a combination thereof that predict proliferative diabetic retinopathy (PDR) or clinically significant macular edema (CSME) within the seven standard fields. These were used to construct a family of screening rules with optimal combined sensitivity and specificity on which to base referrals to eye care specialists.

Main outcome measures: Presence of moderate to severe nonproliferative diabetic retinopathy (NPDR), PDR, or CSME in graded fundus photographs.

Results: Hemorrhages and microaneurysms (h/ma) temporal to the macula (photographic field 3), as severe as or more severe than ETDRS standard photograph 1 (h/ma 3 > or = 3), identified 87% to 89% of eyes with PDR and 92% to 93% of eyes with moderately severe to severe NPDR, which are at high risk for developing PDR. Extrapolating the results using retinopathy prevalence data from epidemiologic studies for the general older onset diabetic population, the calculated sensitivity for detecting PDR on a single examination is 87%, the specificity 80%; for moderate NPDR or worse, the sensitivity is 81 %, specificity 93%. Applying the presence of h/ma 3 > or = 3 as a screening rule to the older onset population, 26.5% of patients would be referred and 73.5% would not be referred. Any hard exudate within one disc diameter of the macular center detects CSME with sensitivity 94%, specificity 54%. Hard exudate of moderate or worse severity anywhere in the macular region (field 2) predicts CSME with sensitivity 89%, specificity 58%.

Conclusions: Screening protocols based on assessing retinopathy lesion severity in the posterior fundus have the potential to identify most diabetic patients with vision-threatening retinopathy. If the protocols can be implemented effectively in a primary care setting, patients requiring referral for specialty care could be reliably identified, and the total number of patients needing specialty referral could be substantially reduced from current guidelines.