Identification of postoperative complications using electronic health record data and machine learning - 18/10/19

Doi : 10.1016/j.amjsurg.2019.10.009

Michael Bronsert ^a,^b , Abhinav B. Singh ^b , William G. Henderson ^a,^b,^c , Karl Hammermeister ^a,^b,^d , Robert A. Meguid ^a,^b , Kathryn L. Colborn ^c,^⁎
^a University of Colorado Anschutz Medical Campus, Adult and Child Consortium for Health Outcomes Research and Delivery Science, Aurora, CO, USA
^b Surgical Outcomes and Applied Research Program, Department of Surgery, University of Colorado School of Medicine, Aurora, CO, USA
^c University of Colorado Anschutz Medical Campus, Colorado School of Public Health, Department of Biostatistics and Informatics, Aurora, CO, USA
^d University of Colorado Anschutz Medical Campus, School of Medicine, Department of Cardiology, Aurora, CO, USA

^∗Corresponding author. University of Colorado Anschutz Medical Campus, 13001 E. 17th Place, Room W4126 Aurora, CO, 80045, USAUniversity of Colorado Anschutz Medical Campus13001 E. 17th PlaceRoom W4126 AuroraCO80045USA

Sous presse. Épreuves corrigées par l'auteur. Disponible en ligne depuis le Friday 18 October 2019

Abstract

Background

Methods

We used an elastic-net model to estimate regression coefficients and carry out variable selection. International classification of disease codes (ICD-9), common procedural terminology (CPT) codes, medications, and CPT-specific complication event rate were included as predictors.

Results

Of 6840 patients, 922 (13.5%) had at least one of the 18 complications tracked by NSQIP. The model achieved 88% specificity, 83% sensitivity, 97% negative predictive value, 52% positive predictive value, and an area under the curve of 0.93.

Conclusions

Using machine learning on EHR postoperative data linked to NSQIP outcomes data, a model with 163 predictors from the EHR identified complications well at our institution.

Le texte complet de cet article est disponible en PDF.

Highlights

•	A model for identifying patients with postoperative complications was developed.
•	The model had 83% sensitivity, 88% specificity, and AUC of 0.93.
•	This model could be used for electronic surveillance of postoperative complications.

Le texte complet de cet article est disponible en PDF.

Résumé

Using the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) complication status of patients who underwent an operation at the University of Colorado Hospital, we developed a machine learning algorithm for identifying patients with one or more complications using data from the electronic health record (EHR). The model achieved 88% specificity, 83% sensitivity, 97% negative predictive value, 52% positive predictive value, and an area under the curve of 0.93. The model developed could be used for electronic postoperative complication surveillance to supplement manual chart review.

Le texte complet de cet article est disponible en PDF.

Keywords : NSQIP, Postoperative complications, Elastic-net, Machine learning