Comparison Of Centralized And Traditional Methods Of Heart Failure Hospitalization Endpoint Capture

Introduction Pragmatic clinical trials (PCTs) employ innovative approaches to improve clinical trial efficiency and reduce trial-related costs. Leveraging electronic medical records (EMR) has the potential to lower trial burden and cost related to patient follow-up and data acquisition. Hypothesis Centralized (EMR-based) and traditional (trial-based) methods of heart failure (HF) hospitalization endpoint capture are comparable. Methods In the Veteran Affairs (VA) national integrated EMR, we used four distinct algorithms: 1) any (primary or secondary) discharge diagnosis code of HF; 2) any discharge diagnosis code of HF and IV diuretic use during associated hospitalization; 3) principal discharge diagnosis code of HF; 4) principal discharge diagnosis code of HF and IV diuretic use during associated hospitalization, to identify HF hospitalizations for 405 Veteran participants of the Influenza Vaccine to Effectively Stop Cardio Thoracic Events and Decompensated heart failure (INVESTED) PCT. HF hospitalizations identified by EMR were compared to INVESTED study events adjudicated according to FDA definitions by a blinded clinical events committee. Sensitivity and specificity of the EMR algorithms were estimated with the adjudication by the clinical events committee as the gold standard. Results Sensitivity and specificity for each algorithm are shown in Table 1. The algorithm comprising any discharge diagnosis code of HF had the highest sensitivity, while the principal discharge diagnosis code of HF and IV diuretic use during associated hospitalization algorithm had the highest specificity. The algorithm containing any discharge diagnosis code for HF with IV diuretic use during associated hospitalization had the optimal balance between correctly capturing and classifying HF hospitalizations and distinguishing non-HF hospitalizations. Conclusions Centralized endpoint capture of HF hospitalizations is sensitive and highly specific when compared to traditional methods for all algorithms examined. Any discharge diagnosis code of HF combined with IV diuretic use during associated hospitalization had the best combination of sensitivity and specificity. Pragmatic clinical trials (PCTs) employ innovative approaches to improve clinical trial efficiency and reduce trial-related costs. Leveraging electronic medical records (EMR) has the potential to lower trial burden and cost related to patient follow-up and data acquisition. Centralized (EMR-based) and traditional (trial-based) methods of heart failure (HF) hospitalization endpoint capture are comparable. In the Veteran Affairs (VA) national integrated EMR, we used four distinct algorithms: 1) any (primary or secondary) discharge diagnosis code of HF; 2) any discharge diagnosis code of HF and IV diuretic use during associated hospitalization; 3) principal discharge diagnosis code of HF; 4) principal discharge diagnosis code of HF and IV diuretic use during associated hospitalization, to identify HF hospitalizations for 405 Veteran participants of the Influenza Vaccine to Effectively Stop Cardio Thoracic Events and Decompensated heart failure (INVESTED) PCT. HF hospitalizations identified by EMR were compared to INVESTED study events adjudicated according to FDA definitions by a blinded clinical events committee. Sensitivity and specificity of the EMR algorithms were estimated with the adjudication by the clinical events committee as the gold standard. Sensitivity and specificity for each algorithm are shown in Table 1. The algorithm comprising any discharge diagnosis code of HF had the highest sensitivity, while the principal discharge diagnosis code of HF and IV diuretic use during associated hospitalization algorithm had the highest specificity. The algorithm containing any discharge diagnosis code for HF with IV diuretic use during associated hospitalization had the optimal balance between correctly capturing and classifying HF hospitalizations and distinguishing non-HF hospitalizations. Centralized endpoint capture of HF hospitalizations is sensitive and highly specific when compared to traditional methods for all algorithms examined. Any discharge diagnosis code of HF combined with IV diuretic use during associated hospitalization had the best combination of sensitivity and specificity.