In reply
We thank de Melo Neto et al for their insightful comments about our article.1 We agree that using ICD-9 codes can potentially introduce a bias in any analysis using such data. We have previously validated hypertension ICD-9 codes against structured medical chart review as the gold standard.2 Among HIV-infected persons enrolled in the 3-site Veterans Aging Cohort Study, the prevalence of hypertension by ICD-9 codes and medical chart reviews was 28% and 32% respectively, with an agreement of 85.2% and a κ value of 0.65. Hence, there may have been a slight underreporting of hypertension in the HIV-infected group. We are currently evaluating whether diagnosis of hypertension differs by HIV status using actual blood pressure measurement. For diabetes, we used a combination of laboratory measurements, use of insulin and oral hypoglycemic medications, and ICD-9 codes, as used in several previous studies.3,4