SHM Converge Scientific Abstract ePoster Gallery-Predicting 90-Day Copd Readmission Risk: Chart Review vs. Administrative Data

Pdf Summary

This study compared two ways of building a model to predict 90-day readmission after hospitalization for acute exacerbation of COPD (AE-COPD): one using manual chart review as the “gold standard,” and one using administrative data alone.<br /><br />Researchers reviewed charts for all eligible COPD admissions at a single hospital between January 2008 and October 2018. Patients were categorized by likelihood of true AE-COPD, and those rated “probably” or “definitely” were included. A random forest model was trained on demographic data, comorbidities, discharge labs, and vital signs. Its performance was then compared with a previously developed administrative-data model using the same chart-reviewed test set.<br /><br />A total of 452 COPD admissions were identified through chart review; 174 resulted in readmission within 90 days. Readmitted and non-readmitted patients had similar age, sex, race, ethnicity, and length of stay, but readmitted patients had a higher comorbidity burden by the Elixhauser Index.<br /><br />The chart review-based model had an AUC of 0.62 (95% CI 0.52–0.72). The administrative-data model had an AUC of 0.67 (95% CI 0.58–0.77). The difference was not statistically significant (DeLong p=0.4).<br /><br />The main conclusion was that a model built using manual chart review did not perform significantly better than one built using administrative data. Because chart review is labor-intensive and difficult to scale, the findings support continued development of COPD readmission prediction tools using administrative data. However, the authors note that larger, multi-site studies are needed to confirm these results.

Asset Subtitle

Sukarn Chokkara

Meta Tag

Author List Juan Rojas, Kyle Carey, Laura R. Venable, Matt Bonomo, Matthew Churpek, Maximilian Hemmrich, Michael Hermsen, Sukarn Chokkara, Valerie G. Press

Category Research

Concept COPD Exacerbation

Concept Readmission Risk

Concept COPD Readmission Prediction

Concept Chart Review

Concept Administrative Data

Distinguished Non-Finalist

Presenter Organization University of Chicago Medicine

Presenting Author Sukarn Chokkara

Track Technology in Hospital Medicine

Keywords

COPD

acute exacerbation

readmission prediction

90-day readmission

chart review

administrative data

random forest model

AUC

Elixhauser Index

hospitalization

COPD Exacerbation

Readmission Risk

COPD Readmission Prediction

Chart Review

Administrative Data