Hand function impairment is among the most common symptoms of several Rheumatic and Musculoskeletal Diseases (RMD), making its timely assessment essential for diagnosis and disease management. However, standard clinical evaluations are often subjective, irregular, and inconclusive. In this study, we explore the potential of keyboard-derived features, such as inter-key flight and hold times, as digital biomarkers of fine-motor skill impairment in individuals with RMDs. Data from 59 participants (31 patients, 28 controls) was retrieved from the COTIDIANA dataset. Each participant completed a transcription task on a smartphone, from which keyboard typing dynamics were retrieved. We defined a binary classification task to distinguish (i) RMD patients with and (ii) healthy controls without hand joint pain. We applied multiple Machine Learning (ML) pipelines using keyboard features, functional tests, and Patient-Reported Outcome Measures (PROM), comparing their predictive performance via balanced accuracy, F1-score, and AUROC. Our best keyboard-based ML predictive model achieved a test AUROC of 0.87 and F1-score of 0.83, outperforming most traditional PROM-based assessments such as EQ-5D-5L, HADS, and functional tests (MPUT). Feature importances revealed that RMD patients exhibited longer typing latency than controls, which is consistent with previous Parkinson’s and Multiple Sclerosis research. Findings suggest fine-motor impairments can be effectively captured through typing dynamics, supporting the generalization of this approach across motor-related conditions. Integrating digital assessment tools into rheumatology care could promote more objective, data-driven strategies to improve patient identification and referral. This would enable the fast-screening of individuals exhibiting early fine-motor signs of decline, fostering efficient access to clinical care.