A recent analysis published in the journal Alcohol: Clinical & Experimental Research suggests that artificial intelligence (AI) can play a vital role in recognizing risky alcohol use in patients undergoing surgical procedures. The study utilized a natural language processing model to examine the medical records of over 50,000 patients who had surgeries between 2012 and 2019.

Patient records typically contain diagnostic codes, but they also contain valuable information such as notes, test outcomes, and billing data that may indicate potential alcohol misuse. To identify contextual cues, researchers programmed a natural language processing model to detect both diagnostic codes and other markers of risky alcohol consumption, such as a history of medical complications related to alcohol abuse or exceeding recommended drinking limits.

It has been established that consuming alcohol excessively before surgery can lead to higher infection rates, prolonged hospital stays, and other complications. Among the patients in the study, 4.8 percent of their records had a diagnosis code related to alcohol use. However, with the assistance of contextual clues, the AI model successfully classified an additional 9.7 percent of patients as being at risk, bringing the total to 14.5 percent.

Remarkably, the AI model's accuracy closely resembled that of a panel of human experts in alcohol use, matching their classifications around 87 percent of the time when examining a subset of records.

Lead author V.G. Vinod Vydiswaran, an associate professor of learning health sciences at the University of Michigan Medical School, highlighted that this analysis sets the foundation for identifying other potential risks in primary care and beyond, with appropriate validation. By using AI to extract relevant information from providers' notes, it can save time for healthcare professionals who no longer need to review the entire record to find crucial details.

Although the researchers plan to eventually make the AI model publicly available, they note that it will have to be trained using medical records specific to each facility. This ensures that the model is based on data that accurately reflects the patient population and healthcare practices of individual institutions.

With the ongoing advancements in AI technology and its ability to analyze vast amounts of medical data, the integration of AI into healthcare settings shows promising potential for improving patient safety and enhancing healthcare decision-making processes.