RESUMO
A computerized population health problem targeting program for diabetes mellitus and other common non-communicable disease was tested at St Luke's Hospital, Malta. The program utilizes a patient questionnaire by dialogue with a computer lasting about 20 min. The system evaluates and targets health problems that are followed up through the database. During a period of 10 months, two groups of people were screened using the system, consisting of 128 recently diagnosed diabetic patients presenting consecutively and 320 people from the general population. The age distributions of the two groups were similar. The system indicated 27% of the control group as being within the Diabetes Risk Group; 5% of this control group were confirmed to have diabetes. Eighty-two percent of newly referred diabetic patients were detected, giving false positive results in 23% and a false negative result in 18% of cases. The classification rules in the system were tuned to the population of Moscow, Russia. We have since adjusted these to better suit the Maltese population. The adjusted rules now indicate 20% of the general population as being within the risk group and have correctly detected 84% of the diabetic people, giving a false positive result in 15% and a false negative result in 16% of cases.
