DEVELOPMENT AND VALIDATION OF AN ALGORITHM FOR IDENTIFYING PATIENTS WITH HEMOPHILIA A IN AN ADMINISTRATIVE CLAIMS DATABASE

• Published in: Value in health Vol. 20; no. 5; p. A1
• Main Authors: Lyons, J, Desai, VC, Jemison, J, Xu, Y, Ridgeway, G, Finkle, W, Solari, PG, Sullivan, SD, Lanes, S
• Format: Journal Article
• Language: English
• Published: Lawrenceville Elsevier Science Ltd 01.05.2017
• Subjects: Algorithms; Clotting; Diagnosis; Emergency medical care; Hemophilia; Inpatient care; Medical diagnosis; Medical records; Medical screening; Patients; Prediction models; Regression analysis; Validity; United States > US
• ISSN: 1098-3015; 1524-4733
• DOI: 10.1016/j.jval.2017.05.005

OBJECTIVES: Develop and validate an algorithm to identify patients with hemophilia A in an administrative claims database. METHODS: We first created a screening algorithm using diagnosis and treatment codes to identify potential hemophilia A patients from administrative claims data in the US HealthCore Integrated Research Database between 01/01/06 and 04/30/15. Medical records for a randomly selected subset of patients were reviewed to confirm case status. In this validation sample, we used lasso logistic regression with cross-validation to develop a predictive model using covariates in claims data to estimate the probability of being a confirmed hemophilia A case. RESULTS: Using the screening algorithm, we identified an initial cohort of 2,252 patients with potential hemophilia A. Of 400 medical records reviewed, 248 (62%) patients were classified as hemophilia A cases, 131 (33%) were false positives, and 21 (5%) were of indeterminate status. The lasso regression model evaluated 36 potential covariates and identified several strong predictors of hemophilia A that were not included in the screening algorithm, including: >1 inpatient, outpatient or emergency room visit for hemophilia A; diagnosis after clotting factor level tests; diagnosis made by a hematologist and >1 hemophilia A diagnosis over 3 months. A probability threshold of >0.6 resulted in a PPV of 94.7% (95%CI: 92.0-97.5), sensitivity of 94.4% (95%CI: 91.5-97.2), and specificity of 90.1% (95%CI: 85.0-95.2) in the validation sample. We applied this model to the initial cohort to identify a refined cohort of 1,507 patients. The refined cohort was more likely to be male, be under the care of a hematologist, and have fewer comorbidities. CONCLUSIONS: We developed and validated an algorithm to identify hemophilia A cases in an administrative claims database with high PPV, sensitivity and specificity. This algorithm uses widely available variables that can be applied in other claims databases.