A Machine Learning Framework for Efficient Triage of Long-COVID Patients to Specialists

Long-COVID poses a significant burden on healthcare systems, necessitating efficient triaging strategies to optimise patient care. This study presents a machine learning (ML) framework for prioritising long-COVID patients for specialist consultations in cardiology, pulmonology, and psychiatry. Utilising a dataset of 175 patients, a Support Vector Machine model was developed, and subsequently enhanced with Synthetic Minority Oversampling Technique for class imbalance handling. The model yielded an accuracy of 0.67 and an Area-Under-the-Curve of 0.84, outperforming alternative models such as Random Forest, k-nearest neighbours, XGBoost, and Decision Trees. Notably, the SF-36 General Health score emerged as the most critical factor in patient classification, followed by bodily pain and anxiety levels. These findings demonstrate the potential of ML-based approaches to streamline healthcare resource allocation and improve patient outcomes. Future work will assess the clinical impact of this approach in prospective studies.Clinical Relevance- This study proposes a machine-learning-driven method to enhance specialist triaging for long-COVID patients, offering potential for improved healthcare resource allocation and timely patient care.