Frequency and machine learning predictors of depressive, anxiety, obsessive-compulsive symptoms, and suicidal ideation among university students

IntroductionProspective studies on mental health of students showed that young adults enroled in university are affected by poorer mental health than other working peers or adults, and this condition is responsible for a large proportion of disability-adjusted life-years MethodsWe enrolled 1388 students at the baseline (and 557 completed follow-up after six months) who reported their demographic information and completed self-report questionnaires on depressive, anxiety and obsessive-compulsive symptoms. We applied multiple regression modelling and supervised machine learning to evince associations and predict the risk factors of poorer mental health at baseline and follow-up ResultsApproximately one out of five students reported severe depressive symptoms and/or suicidal ideation. An association of economic worry with depression was evidenced at the beginning of the study (when there was a high frequency of worry OR = 3.11 [1.88 - 5.15]) and during follow-up. Supervised machine learning exhibited high accuracy in predicting the students who maintained well-being (balanced accuracy = 0.85) or absence of suicidal ideation, but its performance was almost null in identifying those whose symptoms worsened. ConclusionsStudents severe mental health problems are reaching worrying percentages, and few demographic factors can be leveraged to predict poor mental health outcomes. Further research including people with lived experience is crucial to assess students needs and improve the prediction of those at risk of developing worse symptoms.