OBJECTIVES: The objective of this proof of concept study was to evaluate alerts generated by a patient-reported outcome measure (PROM)-based algorithm for monitoring patients with rheumatoid arthritis (RA). METHODS: The algorithm was constructed using an example PROM score of an equally weighted mean of visual analogue scale (VAS) general health, VAS disease activity and VAS pain. Based on the PROM score, red flags are generated in 2 instances: the target level of disease activity is not met; change in disease activity surpasses an early alert threshold. To reduce false alarms, 3 consecutive red flags are needed to trigger an alert to the physician. Time series data from patients included consecutively in the practice-based Nijmegen Early RA cohort were analysed to select an appropriate autoregressive integrated moving average (ARIMA) model. This allowed for advanced interpolation of PROM scores and weekly data evaluation. Alerts were evaluated against disease-modifying antirheumatic drug (DMARD)/biologic medication intensification registered in the cohort. RESULTS: Data of 165 patients followed in their second year postdiagnosis were analysed. In 89.8% of 716 visits, the algorithm did not generate an alert and medication was not escalated. Positive predictive value, sensitivity and specificity were 24.6%, 55.6% and 69.7%, respectively. Comparable performance was found when analyses were stratified for baseline Disease Activity Score 28-joint count (DAS28) level. CONCLUSIONS: When using the algorithm to screen scheduled visits, the overall chance of missing patients in need of medication intensification is low. These findings provide evidence that an off-site monitoring system could aid in optimising the number and timing of face-to-face consultations of patients with their rheumatologists.