Objective:

To externally validate a personalized machine learning-based web application designed for outcome assessment using the five-repetition sit-to-stand (5R-STS) test.

Approach:

• Study Cohort: Participants were healthy individuals from various countries, recruited to evaluate the model's calibration and classification accuracy.
• Measurement and Data Collection: The 5R-STS test was conducted following a standardized protocol, with sociodemographic data collected through questionnaires.
• Model Description: The model uses quantile regression and Lasso penalty to predict personalized upper limits of normal (ULN) based on demographic factors.
• External Validation: The model was applied to an independent dataset without modifications to assess its predictive capabilities.

Key Findings:

• The machine learning model demonstrated improved classification accuracy for objective functional impairment (OFI) compared to traditional methods.
• External validation confirmed the model's adaptability across diverse demographics and testing conditions.

Limitations:

• The study focused solely on healthy participants, which may limit the applicability of findings to clinical populations.
• The sample size was determined by the availability of eligible participants, which may introduce bias.

Conclusion:

The external validation supports the use of the machine learning-driven web tool for assessing functional impairment via the 5R-STS test.

Sources:

• Machine Learning Model Web Application