Background: Ever greater importance is given to psychological distress in oncological settings. Distress could have a severe adverse effect on adherence and compliance to therapies, medical treatments, and quality of life (NCCN, 2015). Moreover, psychological research is progressively increasing in oncological settings and psychological distress has rapidly gained popularity – leading to the development of scales aimed at its evaluation. In this regard, the revised version of the Psychological Distress Inventory – the PDI-R (Rossi et al., 2022) – has demonstrated excellent psychometric properties but its screening properties were limited to self-report questionnaires. Consequently, to fill this gap, this study aimed to evaluate the accuracy of the PDI-R in identifying the levels of distress of oncological outpatients compared to a diagnostic interview – using a machine learning approach. Methods: Oncological outpatients (n = 603; mean age = 68.38, SD = 9.31; 316 males) were enrolled at the Presidio Ospedaliero of Saronno, ASST Valle Olona, Italy. Patients were tested with the PDI-R which is composed of 8 items that evaluate internal, external, and general distress. In this study, the PDI-R shows good internal consistency: McDonald’s omega = .922. Then, patients were interviewed by a psycho-oncologist who assessed distress levels of each patient. In accordance with the methodological procedures for the application of machine learning algorithms, patients were randomly divided into two different databases. The first database (n = 400) was used to train the machine learning algorithm and the second one (n = 203) was used to test the accuracy of its predictions. The k-nearest neighbors (k-NN) algorithm was applied. Results: According to the clinical interview, the training database (mean age = 67.78, SD = 9.10; 212 males) shows that 76% of patients were without distress (or mild distress) and 24% had moderate or severe distress. Moreover, according to the clinical interview, the testing database (mean age = 69.58, SD = 9.62; 104 males) shows that 71% of patients were without distress (or mild distress) and 29% had moderate or severe distress. Using the 8 items of the PDI-R, the k-NN algorithm, correctly identified 193 subjects (144 true negatives and 49 true positives). At the same time, only 10 subjects (10 false negatives) were incorrectly classified – thereby showing a machine learning-based accuracy of 95%. Conclusions: These findings suggest that the PDI-R is a reliable self-report questionnaire with strong psychometric properties and high accuracy in identifying psychological distress. Its use in research and clinical practice is therefore recommended to increase the quality of both assessment and treatment of psychological distress in patients with oncological difficulties.