Preferred provider organization claims showed high predictive value but missed substantial proportion of adults with high-risk conditions

Background and Objective: We assessed the validity and utility of a claims-based ICD-9-CM algorithm for identifying preferred provider organization (PPO) enrollees ages 18-64 years at high risk for influenza complications. Methods: PPO enrollees with >= 2 encounters in an ambulatory setting or >= 1 encounters in an inpatient or emergency room setting with ICD-9-CM diagnosis codes for the high-risk conditions were considered algorithm positive. Stratified random sampling was used to select 1,001 algorithm-positive and 330 algorithm-negative enrollees for medical chart abstractions. Results: The prevalence of high-risk conditions using claims data was 2.5% compared to 18.2% according to medical records. The algorithm had a sensitivity of 12% and a specificity of 99%. Positive and negative predictive values were 87 and 84%, respectively. Sensitivity was twofold higher among adults aged 50-64 years than among younger adults (17 vs. 9%). Applying an algorithm definition of >= 1 encounters in any setting resulted in an increased sensitivity, but captured a higher proportion of false positives. Conclusion: A claims-positive record was highly indicative of the presence of high-fisk conditions, but such claims missed a large proportion of PPO enrollees with high-risk conditions. It is important to assess the validity of administrative data in different age groups. (c) 2005 Elsevier Inc. All rights reserved.