A single observer reviewed 842 of the 917 known diabetic patients registered with 40 GPs in the Poole area. A midstream urine specimen was tested for proteinuria using Albustix (Ames) and cultured to detect bacterial infection. After the first 3 months of the survey, the aliquot of this specimen was frozen for later determination of the random albumin/creatinine ratio (R-Alb/Creat). Patients were requested to submit a timed overnight urine collection for estimation of urinary albumin excretion rate (AER). Of the 842 patients reviewed, 493 (59%) submitted timed overnight urine collections; 43 were excluded because of urinary infection and/or proteinuria. One hundred and thirty-three (30%) of 450 diabetic patients were found to have microalbuminuria, although only 31 (7%) had an AER greater than 30 micrograms/min. Six hundred and seven urine samples were collected for R-Alb/Creat but 68 were excluded because of infection and/or proteinuria; in 10 further samples urinary creatinine was not measured. Two hundred and four (38%) of 532 diabetic patients were found to have an elevated R-Alb/Creat. There was a significant correlation between AER and R-Alb/Creat (r = 0.32, p less than 0.001) but a considerable number of patients showed either a normal AER and high R-Alb/Creat or the reverse. The value of R-Alb/Creat or an overnight urinary albumin concentration, or an overnight urinary albumin/creatinine ratio (ON-Alb/Creat) as screening tests to predict AER greater than 30 micrograms/min was assessed. An ON-Alb/Creat greater than 2.0 mg/mmol was the optimal screening test (sensitivity 96% and specificity 99.7%).(ABSTRACT TRUNCATED AT 250 WORDS)