Combining three different sources as a valid tool to identify known diabetic patients: use in a prevalence study in...

This study was carried out to validate the prevalence of known diabetes estimated from three different simultaneous sources (medical records from out-patient departments, registers and pharmaceutical prescriptions) as against the diagnosis of family doctors who had received ad hoc training before the study. To this effect, a selected Local Health Unit (Unità Sanitaria Locale or USL) in Bari, southern Italy, was examined. 2917 diabetic patients were found, with a 2.82% prevalence. The prevalence of known diabetes estimated from separate sources was 1.41% from medical records, 1.52% from the registers and 1.59% from prescriptions. With reference to the family doctor's conclusions, sensitivity, positive predictive value and efficiency of medical records were 49.8%, 92.5% and 57.4% of USL registers 53.8%, 95.6% and 61.8% of prescriptions 56.3%, 71.2% and 47.9%. 46% of the diabetic subjects were only identified from one source. In conclusion, by combining several sources of information together, a higher number of diabetic patients are identified than would be done by using separate sources; also, active cooperation from the family practitioners seems to be needed to correct the large number of false positive patients (n = 798) mostly identified from prescriptions.