Severe Arrhythmias During Six-Minute Walk Tests in Early Outpatient Cardiac Rehabilitation: A Two-Year Multilevel Analysis and Machine Learning-Enhanced Prediction Model in a Thai Provincial Hospital

Objective: To determine the prevalence and associated factors of severe arrhythmias during the Six-Minute Walk Test (6MWT) in early outpatient cardiac rehabilitation, and to develop a prediction model using multilevel analysis and machine learning techniques.Material and Methods: This retrospective analytical study spanned a period of 2 years; from January 2022 to December 2023. It included 357 patients undergoing the 6MWT in the early outpatient cardiac rehabilitation at Uttaradit Hospital, in Northern Thailand. Multilevel structural equation modeling and machine learning-enhanced logistic regression were used to analyze individual and institutional factors associated with severe arrhythmias. Research instruments included: a patient data recording form (45-item structured checklist, Content Validity Index (CVI))=0.92, Cohen’s kappa=0.87) and an institutional factors questionnaire (30-item Likert scale, CVI=0.89, Intraclass Correlation Coefficient (ICC)=0.91). Results: The prevalence of severe arrhythmias during the 6MWT was 4.2% (95% confidence interval (CI)): 2.38-6.84%). Significant individual-level risk factors included age (OR=1.56, p-vaue=0.002), male gender (OR=2.31, p-vaue=0.019) and heart failure (odds ratio (OR))=3.42, p-vaue<0.001). Institutional factors; such as staff experience >10 years (OR=0.51, p-vaue=0.037) and emergency equipment readiness (OR=0.45, p-vaue=0.020), were protective factors. The prediction model demonstrated high accuracy in identifying high-risk patients, with a sensitivity of 80.0% and specificity of 76.1% at the optimal cut-off point. The model showed excellent discriminative ability having an area under the Receiver Operating Characteristic (ROC) curve of 0.843 (95% CI: 0.781-0.905).Conclusion: Severe arrhythmias during the 6MWT in early cardiac rehabilitation are relatively uncommon, but clinically significant. The developed prediction model, incorporating both individual and institutional factors, shows promising performance and could enhance risk stratification and safety protocols in cardiac rehabilitation programs. These findings highlight the importance of considering both patient-specific and institutional factors in managing arrhythmic risk during cardiac rehabilitation.