Objective: to estimate the prevalence of diabetes in a Lazio province, Italy, using a record linkage offour patient lists and applying capture-recapture techniques.
Design and setting: we selected records of individuals living in Viterbo province (Lazio region, Italy) from four lists obtained from three different health administrative registries in 2004: the hospital discharge registry of Lazio region (ICD9-CM code 250, 2488 subjects); the registry of outpatient care prescriptions (subjects exempted from medical charge for health care related to diabetes, 3582 subjects); subjects with prescriptions of glicate haemoglobin (7169 subjects); the registry of exemptions, from medical charge for health care of the local health unit of Viterbo province, related to diabetes (8529 subjects). A deterministic record linkage using 5 different keys of linkage obtained from demographic information, was applied The estimates were derived using two capture-recapture techniques: log-linear models and sample-coverage approach.
Results: Record linkage results using the 5 different keys of linkage were similar. Prevalence estimates varied between 63% and 7.8% when all four lists were considered and they varied between 5.7% and 6.1% when the glicate haemoglobin list was excluded.
Conclusions: Our estimates were slightly higher compared to those recently reported in the literature; estimates were particularly influenced by the inclusion/exclusion of the haemoglobin glicate prescriptions list. Estimates were not influenced either by the key of linkage, or by the capture-recapture technique. Health administrative registries can be used to estimate the prevalence of diabetes through record linkage and capture-recapture techniques. This approach leads to a reduction of time and costs compared with traditional health surveys.