Developing Indicators of Inpatient Adverse Drug Events Through Nonlinear Analysis Using Administrative Data

• DOI: 10.1097/mlr.0b013e3180616c2c
• Publication Date: 2009-03-04T23:50:38Z
• Authors: Jonathan R. Nebeker; Paul R. Yarnold; Robert C. Soltysik; Brian C. Sauer; Shannon A. Sims; Matthew H. Samore; Randall W. Rupper; Kathleen M. Swanson; Lucy A. Savitz; Judith Shinogle; Wu Xu
• Keywords: Male; Risk Management; Data Collection; Incidence; Delirium; Hemorrhage; Blood Coagulation Disorders; Middle Aged; Hospital Records; Psychoses, Substance-Induced; 3. Good health; 03 medical and health sciences; 0302 clinical medicine; Nonlinear Dynamics; ROC Curve; International Classification of Diseases; Predictive Value of Tests; Utah; Adverse Drug Reaction Reporting Systems; Humans; Female; Aged; Retrospective Studies

Background: Because of uniform availability, hospital administrative data are appealing for surveillance adverse drug events (ADEs). Expert-generated rules that rely on the presence International Classification Diseases, 9th Revision Clinical Modification (ICD-9-CM) codes have limited accuracy. Rules based nonlinear associations among all types available may be more accurate. Objectives: By applying hierarchically optimal classification tree analysis (HOCTA) to data, derive and validate bleeding/anticoagulation problems delirium/psychosis. Research Design: Retrospective cohort design. Subjects: A random sample 3987 admissions drawn from 41 Utah acute-care hospitals in 2001 2003. Measures: Professional nurse reviewers identified ADEs using implicit chart review. Pharmacists assigned Medical Dictionary Regulatory Activities ADE descriptions identification clinical groups events. Hospitals provided patient demographic, admission, ICD9-CM data. Results: Incidence proportions were 0.8% drug-induced 1.0% The model bleeding had very good discrimination sensitivity at 0.87 86% fair positive predictive value (PPV) 12%. delirium excellent 94%, 0.83, but low PPV 3%. Poisoning event designed targeted sensitivities and, when forced in, degraded Conclusions: Hierarchically is a promising method rapidly developing clinically meaningful resultant anticoagulation useful retrospective screening rate estimation.