On a busy day in a primary care clinic, doctors often ask patients about their alcohol consumption, often using brief surveys completed on paper or smartphones. Many people consider this a formality. However, recent research reveals that critical warning indicators of risky drinking are commonly overlooked – and that artificial intelligence may help bridge this gap. 

The study, published in the journal Drug and Alcohol Dependence, found that only around 10% of persons who match the criteria for alcohol use disorder, or AUD, receive any type of assistance in a given year. Many people are unaware that their drinking has reached dangerous levels, despite the fact that it can increase the risk of injury, sleep problems, mental health issues, and cancer. 

Researchers from the University of Michigan Addiction Centre examined anonymised electronic health records from over 133,000 primary care patients at U-M Health. They employed a type of artificial intelligence known as natural language processing, or NLP. Simply said, this technology examines and evaluates written language, similar to how spellcheck reads words, to identify patterns that humans may overlook. 

Using only formal diagnosis codes and regular alcohol questionnaires, doctors identified 820 patients with risky drinking, also known as AUD. The AI tool identified roughly the same group. However, when it examined the free-text notes left by doctors and nurses—remarks about a patient's lifestyle, stress, or symptoms—it identified more than 47,500 additional patients who were drinking dangerously. 

According to Anne Fernandez, PhD, an addiction psychologist who assisted in leading the study, this shows that most people who drink at unsafe levels are currently being neglected. "Doctors can't read every clinical note from every provider and appointment in a patient's chart, but automated tools can do this quickly and easily," she told me. "Our study shows that these notes contain useful information about alcohol use that we hope can improve clinical care in the long term." 

To assess accuracy, the researchers contacted 170 patients identified by routine screening or by the AI technology. The AI methodology identified 17 additional cases of alcohol use disorder and 23 additional cases of unsafe drinking that traditional methods had overlooked. Those detected by routine screening were more likely to have depression or anxiety and had already sought treatment. 

Risky drinking is defined as alcohol consumption that raises the risk of harm, even if a person does not feel "addicted". Because patients may underreport drinking on questionnaires, reviewing daily clinical records may provide a more complete picture. 

Fernandez emphasised that AI cannot replace doctors. Alcohol use varies over time, data can be incomplete, and AI can make mistakes. However, she stated that it may enable earlier conversations and timely assistance. 

Many patients, including some physicians, are unaware that prescription AUD medications are available and covered by Medicaid, Medicare, and most insurance plans. Another tried-and-true strategy, known as SBIRT (screening, short intervention, and referral to treatment), is frequently discussed. 

The study, which was funded by the National Institute on Alcohol Abuse and Alcoholism in the United States, points to a future in which technology assists clinicians in identifying risky drinking earlier and providing aid before major harm happens.