Abstract
OBJECTIVE: An accurate estimate of the prevalence, demographic characteristics, and geographic distribution of thalassemia in the United States is needed to plan for the health care needs of people with this disease. We developed and evaluated the predictive value of administrative case definitions for correctly identifying people living with thalassemia. METHODS: We conducted a retrospective study of the diagnostic accuracy of 3 claims-based case definitions to identify people with thalassemia in Medicaid administrative data from 2012 through 2019. Case definition 1 was ≥5 encounters with a code for thalassemia; case definition 2 was ≥1 encounter with a code for thalassemia and ≥6 encounters with a transfusion code; and case definition 3 was ≥2 encounters with a code for thalassemia and a transfusion code occurring on the same encounters. We validated our findings by using confirmatory laboratory assessment and expert review by clinicians at thalassemia treatment centers in Georgia (Children's Healthcare of Atlanta) and California (University of California San Francisco) as the gold standard. RESULTS: Of the 327 people identified, thalassemia was confirmed in 173 (52.9%), excluded in 68 (20.7%), and found indeterminate in 86 (26.2%) people. Case definition 1 had the lowest positive predictive value (PPV) (range, 55%-77%). For case definition 2, the PPV range was 80% to 86%. For case definition 3, the PPV range was the highest (82%-96%) but also captured more indeterminate cases. CONCLUSIONS: Accurately identifying patients with thalassemia using a case definition based on administrative claims data is feasible. Extending our method to other health care databases beyond Medicaid may allow for an estimate of the national prevalence of transfusion-dependent thalassemia. However, cases of nontransfusion-dependent thalassemia were difficult to define with sufficient precision.