Abstract
In epidemiology studies, identification of diabetes type (type 1 vs. type 2) among study participants with diabetes is important; however, conventional diabetes type identification approaches that include age at diabetes diagnosis as an initial criterion introduces biases. Using data from the National Health and Nutrition Examination Survey, we have developed a novel algorithm which does not include age at diagnosis to identify participants with self-reported diagnosed diabetes as having type 1 vs. type 2 diabetes. METHODS: A total of 5457 National Health and Nutrition Examination Survey participants between cycles 1999-2000 and 2015-2016 reported that a health professional had diagnosed them as having diabetes at a time other than during pregnancy and had complete information on diabetes-related questions. After developing an algorithm based on information regarding the treatment(s) they received, we classified these participants as having type 1 or type 2 diabetes. RESULTS: The treatment-based algorithm yielded a 6-94% split for type 1 and type 2 diabetes, which is consistent with reports from the Centers for Disease Control and other resources. Moreover, the demographics and clinical characteristics of the assigned type 1 and type 2 cases were consistent with contemporary epidemiologic findings. CONCLUSION: Applying diabetes treatment information from the National Health and Nutrition Examination Survey, as formulated in our treatment-based algorithm, may better identify type 1 and type 2 diabetes cases and thus prevent the specific biases imposed by conventional approaches which include the age of diabetes diagnosis as an initial criterion for diabetes type classification.