Universities have long grappled with alcohol among the many pressures facing young people, and a newly developed scale out of Turkiye offers a fresh, evidence based way to flag problematic alcohol use before it spirals into something harder to manage. Researchers at Çanakkale Onsekiz Mart University’s Center for Combating Addiction Application and Research surveyed 599 students and used an AI assisted process to build a fifteen item screening tool they have named the Student Alcohol Risk Assessment Scale 15, or SARAS 15 for short.

Published in the International Journal of Methods in Psychiatric Research, the study set out to address a gap that has followed university alcohol screening for years. Existing instruments tend to focus narrowly on drinking behaviour itself, while the wider ripple effects, missed lectures, mounting debt, strained mental health, often go unmeasured, leaving many cases of risky alcohol use undetected until problems have already taken hold.

Why campuses need a sharper lens on risky alcohol use

Alcohol use remains the most common risky behaviour reported among university students, and problematic alcohol use has repeatedly been linked in earlier research to disrupted academic performance, strained finances and long term health difficulties. In Turkiye specifically, earlier reviews have put lifetime alcohol use among students at around 46 percent, with drinking in the previous month reported by roughly a third of the student population.

The new study’s own sample reflected a similar pattern. Just under half of participants, 45 percent, said they had consumed alcohol in the past month, and 9.8 percent reported drinking on three or more days a week during that period. Binge drinking, defined in the study as five or more standard drinks in a sitting for men and four or more for women, was reported by 6.7 percent of students overall, rising to 14.1 percent among men.

Existing tools such as AUDIT, CAGE and RAPS4 QF remain useful and were used throughout the study to check the new scale’s accuracy. But the research team argued that a brief, student focused measure covering financial strain, academic disruption, emotional wellbeing and awareness of the need to cut back could sit alongside these established instruments rather than replace them.

Building a scale to detect problematic alcohol use

The development process began with 72 candidate questions drawn from a wider 81 item survey covering family background, drinking patterns and psychosocial consequences. After removing overlapping or unsuitable items, 59 candidates were scored using the ChatGPT 4.5 model against six criteria, including relevance to alcohol related risk, clarity, brevity and coverage of academic or financial impact.

The twenty highest scoring items were then reviewed independently by four experts, who kept only those items reaching at least 80 percent agreement on suitability. Fifteen questions survived this process, covering matters such as whether a student has felt they needed help because of drinking, whether alcohol has affected attendance at classes or work, whether they have found themselves in dangerous situations after drinking, and how often they have felt guilt or regret afterwards.

Each answer is converted into a numerical score, giving a possible total of between zero and twenty seven. Students scoring zero to five are classed as low risk, six to twelve as medium risk, and thirteen or above as high risk.

How well the new scale performs

Researchers tested several machine learning approaches to see how accurately the fifteen item scale could sort students into these risk bands. A logistic regression model performed best, correctly classifying risk group membership with 93.5 percent accuracy, alongside a sensitivity of 0.826 and specificity of 0.959. A separate random forest model reached 87 percent accuracy and was especially strong at identifying the high risk group, achieving a perfect F1 score of 1.00 in that category.

Reliability testing also held up well. Internal consistency, measured using Cronbach’s alpha, came to 0.811, described in the paper as acceptable for a scale still in its early stages of validation. When checked against the established AUDIT questionnaire, SARAS 15 scores correlated strongly, with a coefficient of 0.861, and showed similarly strong agreement with RAPS4 QF at 0.793, supporting its use as a genuine measure of problematic alcohol use rather than a standalone curiosity.

Among the fifteen items, feelings of guilt or regret after drinking carried the heaviest weight in the model’s predictions, followed closely by financial difficulties, average drinking quantity and physical health problems. Gender, by contrast, contributed almost nothing to the model’s predictive power, suggesting the scale captures risk in a way that applies broadly across the student population rather than skewing towards one group.

What the risk categories revealed

Applying SARAS 15 to the full sample of 599 students placed 74.6 percent in the low risk category, 22.9 percent in the medium risk band and 2.5 percent in the high risk group. Men scored higher on average than women, at 4.37 compared with 3.27, and were more likely to fall into the medium risk category.

The gap between risk groups was stark on several measures. Among students classed as high risk, 80 percent reported binge drinking in the previous month, compared with just 4.9 percent of those in the low risk group. Students in the medium and high risk bands also reported more psychological complaints, more smoking and more relatives who drink, pointing to the layered social and emotional context that often surrounds problematic alcohol use on campus.

Financial strain linked to drinking was reported by 10.5 percent of the full sample, while 8.5 percent said alcohol had affected their attendance at classes or work, and 15.5 percent felt their alcohol use was connected to symptoms of depression or anxiety. These figures echo the very consequences the SARAS 15 questionnaire was built to capture, reinforcing why a broader, student specific measure may add real value alongside traditional screening tools.

A tool for early identification, not diagnosis

The research team were careful to frame SARAS 15 as a complement to established instruments rather than a replacement, and stressed it is not designed to diagnose alcohol use disorder. Instead, they see its value in flagging students who might benefit from further assessment, counselling or preventive support before difficulties become entrenched.

The authors also acknowledged limits to the current findings. The sample was drawn from a single university using a cross sectional design, meaning the results cannot be read as national prevalence figures. The risk category thresholds were set from the observed data and reviewed by experts for plausibility, but have not yet been tested against a clinical diagnostic interview. Artificial intelligence played a supporting role in ranking candidate items only, with all final decisions made by the human expert panel.

Further validation across different universities and cultural settings is planned, the authors say, but the early results suggest that a brief, multi domain scale built around the everyday realities of student life, missed classes, financial pressure, disrupted sleep, guilt after a heavy night, can offer a genuinely useful early warning signal for identifying risky alcohol use before it becomes a bigger problem.

Source: dbrecoveryresources