Background: Remote monitoring of cancer patients is known to improve survival by allowing early reporting and management of adverse events (AEs). Moreover, there is growing interest in connected, digital tools to monitor AEs by collecting patient-reported outcomes (PROs). However, countless PRO data can be overwhelming for daily clinical practice. To be simple and efficient, PRO tools should provide clear and actionable information to practitioners. In this study, we sought to validate an algorithm, Cureety TechCare, that proposes a simple clinical classification derived from AEs reported by patients. The primary objective was to assess the diagnostic performance of the algorithm using real-life data collected by the Cureety digital platform. Methods: Cancer patients using the Cureety Tech Care medical device were prompted to complete, at least once a week on their phone or computer, a PRO questionnaire personalized to their treatment and disease. An algorithm then computed, from the reported AEs, a simple "clinical classification" indicative of the patient's health state, with just four possible levels: red, orange, yellow, green (from most to less at-risk). These classifications allow the medical team to efficiently prioritize patients most at risk, by considering the patient either "flagged" as needing additional phone or in-person follow-up (red or orange), or "not flagged" (yellow or green). For this observational study, 400 questionnaires were randomly selected from patients that used the Cureety platform between October 1st, 2019, and September 30th, 2022. The questionnaire data were then independently assessed by 2 oncologists who were blinded to the Cureety TechCare classification. These expert classifications provided the references to calculate the sensitivity (ratio of true positive in the "flagged" group) and specificity (ratio of true negative in the "not flagged" group) of the algorithm. Results: Out of the 400 patients, 60 were in the "flagged" group based on the expert evaluations (with 47 true positives from the algorithm) and 340 in the "not flagged" group (with 321 true negatives). The sensitivity of Cureety TechCare was 78.3% (95% CI: 67.9%-88.8%), and its specificity was 94.4% (95% CI: 92.0%-96.9%). In addition, the algorithm correctly marked as "flagged" 95.8% (23/24) of the most at-risk patients, identified as "red" by the experts. Conclusions: The study demonstrated satisfactory accuracy of Cureety TechCare against a clinician-driven assessment. Its use to complement standard of care is thus clinically relevant, and it can be trusted as a useful supplementary indicator of the patient’s health state. The simplicity of the output makes it particularly useful as a tool to prioritize care, with a clear four-level, color-coded clinical classification to represent the combined AEs reported by the patient. Clinical trial information: NCT05653609.