Concordance analysis of two databases to search for potential drug interactions in onco-hematologic patients

Abstract

Introduction: Potential drug interactions exert a significant impact on patient safety, especially within intricate onco-hematological treatments, potentially resulting in toxicity or treatment failures. Despite the availability of databases for potential drug interaction investigation, persistent heterogeneity in concordance rates and classifications exists. The additional variability in database agreement poses further complexity, notably in critical contexts like onco-hematology. Aim: To analyze the concordance of two databases for researching potential drug interaction in prescriptions for hematological patients at a University Hospital in the Midwest region of Brazil. Method: Cross-sectional study developed in a Brazilian hospital. The search for potential drug interaction was conducted in Micromedex® and UpToDate®. The variables were: the presence of potential drug interaction, severity, mechanism, management, and documentation. Data was analyzed in terms of frequency (absolute and relative), Cohen's kappa, and Fleiss kappa. Results: The presence of potential drug interaction, showed a lack of concordance between the databases (k = −0.115 [95% CI: 0.361–0.532], p = 0.003). Regarding the mechanism, a strong agreement was observed (k = 0.805, p < 0.001 [95% CI: 0.550–0.941]). The management concordance showed a fair agreement, 46.8% (k = 0.22, p < 0.001 [95% CI: 0.099–0.341]). Stratifying the categories, significant concordance was observed in “Adjustment of dose + Monitoring” (k = 0.302, p = 0.018) and “Monitoring” (k = 0.417, p = 0.001), while other categories did not reach statistical significance. Conclusion: Our study emphasizes the variability in potential drug interaction research, revealing disparities in severity classification, management recommendations, and documentation practices across databases.