Abstract 17226: Enhancing the Prediction of 30 Day Readmission after PCI Using Unstructured Data Extracted by Semi-Automated Querying of the Electronic Medical Record
Background: Models predicting readmission have poor discrimination. We examined if information from the electronic medical record (EMR), abstracted with semi-automated methods, adds incremental predictive information for 30-day readmission after percutaneous coronary intervention (PCI).
Methods: We matched readmitted patients to controls 1:2 on a validated readmission risk model, and used a search-assisted chart review to extract data from the EMR, including past need for medical interpretation, albumin level, medical adherence, previous visits to the emergency department, atrial fibrillation and atrial flutter, syncope and presyncope, end-stage liver disease, malignancy, and anxiety. We assessed differences in the rates of these conditions between cases and controls, and estimated their independent association using logistic regression conditioned on matched groups.
Results: The 893 cases were matched to 1776 controls. Cases and controls did not differ on predicted risk of readmission (16.3% vs. 16.2%, p = 0.748). The readmission rate was 9.8%. In univariate analysis, cases and controls were significantly different with respect to requiring a medical interpreter (7.9% for cases vs. 5.3% for controls, p = 0.009), number of emergency department visits in the past year (1.12 vs. 0.77 s, p < 0.001), homelessness (3.2% vs. 1.6%, p = 0.007), anticoagulation (33.9% vs. 22.1%, p < 0.001), atrial fibrillation/atrial flutter (32.7% vs. 28.9%, p = 0.045), presyncope or syncope (27.8% vs. 21.3%, p < 0.001), and anxiety (69.4% vs. 62.4% for controls, p < 0.001). After multivariable logistic regression, anticoagulation (OR: 1.70 [1.41-2.06]), number of ED visits (OR: 1.07 [1.02-1.13]), and anxiety (OR: 1.26 [1.04-1.53]) were independently associated with readmission.
Conclusions: Characteristics found by semi-automated queries of the EMR can be used to refine risk prediction for early hospital readmission following PCI.
Author Disclosures: J.H. Wasfy: Employment; Significant; Massachusetts General Physicians Organization. G. Singal: None. C. O’Brien: None. D.M. Blumenthal: Consultant/Advisory Board; Significant; Exercise.com. K.F. Kennedy: None. J.A. Spertus: Research Grant; Significant; American College of Cardiology Foundation. Ownership Interest; Modest; Health Outcomes Sciences. R.W. Yeh: Research Grant; Significant; National Heart Lung and Blood Institute, American Heart Association, Massachusetts General Hospital. Other Research Support; Significant; Harvard Clinical Research Institute. Consultant/Advisory Board; Modest; Gilead Sciences, Abbott Vascular.
This research has received full or partial funding support from the American Heart Association
- © 2014 by American Heart Association, Inc.