Validity of Primary Immunodeficiency Disease Diagnoses in United States Medicaid Data

Purpose Primary immunodeficiency diseases (PID) are a rare group of disorders with a wide array of clinical presentations. The absence of validated methods to identify these diseases in electronic databases has limited understanding of their epidemiology and the impact of drug therapies on outcomes. We measured the positive predictive values (PPVs) of ICD-9 diagnoses for identifying PID within US Medicaid. Methods We identified Medicaid patients from California, Florida, New York, Ohio, and Pennsylvania with PID ICD-9 diagnoses (common variable immunodeficiency [279.06], X-linked agammaglobulinemia [279.04], hyper-immunoglobulin M syndrome [279.05], Wiskott Aldrich Syndrome [279.12]) recorded at least twice from 1999 to 2007. Outpatient records were reviewed by a clinical immunologist to adjudicate diagnoses. PPVs with 95 % confidence intervals (CIs) for confirmed outcomes were determined for individual ICD-9 diagnoses and combinations of diagnoses and Current Procedural Terminology codes for a quantitative immunoglobulin test (82784) or immunoglobulin infusion (96365). Results Among 83 patients with PID ICD-9 diagnoses, 16 were adjudicated as having the condition (PPV, 19.3 %; 95 % CI, 11.4-29.4 %). Individual ICD-9 diagnoses had low PPVs (range, 16.7-33.3 %). Requiring procedural codes for quantitative immunoglobulins or intravenous immunoglobulin did not increase PPVs of these diagnoses (range, 11.1-41.7 %). An X-linked agammaglobulinemia diagnosis plus intravenous immunoglobulin had the highest PPV among the algorithms evaluated (PPV, 41.7 %; 95 % CI, 15.1-72.3 %). Conclusions Algorithms comprising PID ICD-9 diagnoses and procedures for quantitative immunoglobulin tests and immunoglobulin infusion had low PPVs for adjudicated diagnoses in Medicaid. Alternative data sources should be evaluated to study the epidemiology of these diseases.