Methods: The study was divided in two phases. In the first phase, the four participants were instructed in the interpretation of retinographies. The second phase involved the evaluation of 1000 images of 200 patients, 100 without retinopathy and 100 with signs of diabetic retinopathy. The four participants had to decide if the images did or did not show evidence of diabetic retinopathy. Kappa index was used to assess the extent of agreement. A percentage disagreement of 15% with a precision of 5% (+/-5%) with a confidence level of 95% was considered adequate.
Results: The percentage of coincident diagnoses among ophthalmologists and primary care physicians was between 89 and 97.5%. With respect to the assessment of the agreement, the kappa index was between 80 and 95%. In all cases the confidence interval was at least 85%.
Conclusions: After an adequate training process, the reliability of evaluation of non-mydriatic retinographies of diabetic patients by primary care physicians was very high. This could allow the establishment of screening for diabetic retinopathy at the primary care level. Advantages of this system include a greater involvement of primary care physicians in the global management of diabetic patients and a lower demand for ophthalmic attention.
Andonegui J., Berástegui L., Serrano L., Eguzkiza A., Gaminde I., Aliseda D. Agreement among ophthalmologists and primary care physicians in the evaluation of retinographies of diabetic patients. Arch Soc Esp Oftalmol. 2008;83(9):527-31.