AI screening identifies patients’ risky preoperative alcohol use

Editor's Note:

A recent study suggests artificial intelligence (AI) can be valuable for identifying patients who consumed risky amounts of of alcohol prior to surgery. Findings appeared in the journal Alcohol, Clinical and Experimental Research on January 8.

For the study, researchers extracted 3 years of text-based clinical records from more than 53,000 patients enrolled in the Michigan Genomic Initiative who were scheduled to receive surgery. Data included outpatient and inpatient admission, progress, procedure and discharge notes, and other communications. Using Natural Language Processing (NLP, a variety of machine learning), they correctly identified 87% of at-risk patients, compared to 29% for International Classification of Diseases (ICD) diagnostic codes.

Defined as more than 2 drinks/day prior to surgery, risky perioperative alcohol use is one of the most common surgical risk factors and is on the rise, researchers note. Additionally, screening for risky alcohol is often absent or happens too close to surgery to allow for a meaningful intervention. Patients who drink beyond healthy limits are prone to surgical complications as well as increased infections, wound complications, longer hospital stays, and increased hospital readmissions.