Diabetes and Tuberculosis: harmonization of population-based datasets

BACKGROUND Leveraging existing datasets is an efficient and effective mean of advancing health research. However, finding and pooling these datasets for certain comorbid conditions, such as diabetes and tuberculosis, can be challenging because most open-access datasets may focus on only one of these conditions. AIMS To develop a data infrastructure that can be utilized to study diabetes and tuberculosis in diverse populations. METHODS We cleaned, harmonized, and pooled three population-based datasets and one with electronic medical records: i) Hispanic Community Health Study/Study of Latinos (HCHS/SOL); ii) Coronary Artery Risk Development in Young Adults (CARDIA); iii) National Health and Nutrition Examination Survey (NHANES); and iv) All of Us (AoU) Research Program. Diabetes was defined as a combination of self-reported diagnosis, blood tests, and ICD-10 codes; tuberculosis was defined using self-reported diagnosis and ICD-10 codes. We provided user guides and analytical code for researchers to develop analytical datasets to study diabetes and tuberculosis. RESULTS There were 116 unique individuals who met criteria for both diabetes and tuberculosis: 20-year-round CARDIA (n = 19); baseline evaluation of SOL (n = 89); and two NHANES waves (n = 8). For AoU, we identified 253 unique individuals who had a diabetes diagnosis prior to the tuberculosis diagnosis, 18 who had both diagnoses recorded simultaneously, and 213 whose tuberculosis diagnosis was registered before the diabetes diagnosis. CONCLUSIONS Leveraging available data resources, which are predominantly utilized for research on noncommunicable diseases, is feasible to expedite the field and integrate both noncommunicable and infectious diseases.