Authentication of Algorithm to Detect Metastases in Men with Prostate Cancer Using ICD-9 Codes

BACKGROUND: Metastasis is a crucial endpoint for patients with prostate cancer (PCa), but currently lacks a validated claims-based algorithm for detection. OBJECTIVE: To develop an algorithm using ICD-9 codes to facilitate accurate reporting of PCa metastases. METHODS: Medical records from 300 men hospitalized at Robert Wood Johnson University Hospital for PCa were reviewed. Using the presence of metastatic PCa on chart review as the gold standard, two algorithms to detect metastases were compared. Algorithm A used ICD-9 codes 198.5 (bone metastases), 197.0 (lung metastases), 197.7 (liver metastases), or 198.3 (brain and spinal cord metastases) to detect metastases, while algorithm B used only 198.5. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for the two algorithms were determined. Kappa statistics were used to measure agreement rates between claim data and chart review. RESULTS: Algorithm A demonstrated a sensitivity, specificity, PPV, and NPV of 95%, 100%, 100%, and 98.7%, respectively. Corresponding numbers for algorithm B were 90%, 100%, 100%, and 97.5%, respectively. The agreement rate is 96.8% for algorithm A and 93.5% for algorithm B. CONCLUSIONS: Using ICD-9 codes 198.5, 197.0, 197.7, or 198.3 in detecting the presence of PCa metastases offers a high sensitivity, specificity, PPV, and NPV value.