BACKGROUND: Validation of diagnosis codes is a highly recommended step in order to offer reliable epidemiological measures in research performed in automated healthcare databases.
OBJECTIVES: To describe the validity of recorded diabetic retinopathy (DR) and maculopathy (DMP) diagnoses, including macular oedema (DMO) in The Health Improvement Network (THIN) UK primary care database.
METHODS: In two independent computer searches, we identified 20,838 diabetics aged 1–84 years with a first DR computer Read code entry in 2000–2008, and 4,064 with a first DMP entry. A two-step strategy was used to validate both outcomes: (1) for all DMP patients and a random sample of 500 DR patients a manual review of computerized patient profiles was conducted. Profiles included free- text comments from primary care practitioners' (PCPs) with referral information and test results. We classified subjects into probable case, possible case, and non-case according to the plausibility of the diagnosis. (2) for a random sample of 200 subjects with DR and 200 subjects with DMP (including 36 DMO) questionnaires and additional medical records information were requested to PCPs and reviewed. Confirmation of diagnosis by PCP was considered the gold standard.
RESULTS: After revision of the random sampled patient profiles with free-text comments, we categorized 418 DR cases and 3,676 DMP cases as probable/possible (including 711 DMO). After review of the information received from PCPs, probable/possible cases of DR and DMP were respectively confirmed in 87.3% and 87.2% of instances. Confirmation rate for DMO was 90.3%. The confirmation rates, once applied to the whole population of automatically computer-detected patients, translated into a weighted confirmation rate of 78.0% for DR codes, 79.0% for DMP codes, and 86% for DMO codes.
CONCLUSIONS: Read codes for DR, DMP, and DMO showed a moderate accuracy in identifying incident cases of these ophthalmologic complications. The validity further improved when incorporating PCPs' text comments to the patient's profile.
