Predicting Incident Adenocarcinoma of the Esophagus or Gastric Cardia Using Machine Learning of Electronic Health Records

Abstract

The Kettles Esophageal and Cardia Adenocarcinoma predictioN (K-ECAN) Tool was developed and validated using a form of artificial intelligence of routine electronic medical records to predict which patients would develop particular types of cancer of the esophagus or stomach years in advance.

Publication
Gastroenterology

Background and Context: Tools that can automatically predict incident esophageal adenocarcinoma (EAC) and gastric cardia adenocarcinoma (GCA) using electronic health records (EHR) to guide screening decisions are needed.

New Findings: In this retrospective case-control analysis using machine learning, the Kettles Esophageal and Cardia Adenocarcinoma predictioN (K-ECAN) Tool was well calibrated and more accurate than available alternatives. While GERD was associated with EAC/GCA, other factors added more information to the model.

Limitations: K-ECAN was developed and validated within the Veterans Health Administration (VHA) population. Non-Veterans may differ in terms of the frequency of medical encounters and laboratory blood draws, and non-VHA settings may differ in their practice of diagnostic coding.

Clinical Research Relevance: The Kettles Esophageal and Cardia Adenocarcinoma predictioN Tool is an automated, valid tool predicting incident esophageal