Accurately Estimating PHQ-9 Scores from Clinical Notes using a Machine Learning Model

A new machine learning model accurately predicts depression levels from both complete and incomplete clinical notes, according to a poster presented at the American Society of Clinical Psychopharmacology annual meeting. Researchers aimed to use an innovative data science method to provide accurate estimated scores on the Patient Health Questionnaire-9 (PHQ-9) from unorganised and partially organised clinical notes. The study drew data from more than 490,000 individuals with major depressive disorder, whose PHQ-9 scores and clinical notes were available for analysis. After applying the model, PHQ-9 scores were generated for 2.2 million patient encounters – 2.7 times the number of recorded PHQ-9 scores – and encounters for 208,692 patients, or 1.2 times the original individuals with scores.