A method to validate scoring systems based on logistic regression models to predict binary outcomes via a mobile...

Area of Science:

• Biostatistics
• Medical Informatics
• Health Services Research

Background:

• Points systems require rigorous validation for accurate binary event prediction.
• Discrimination (Area Under the Curve) and calibration (smoothed calibration plots) are key validation metrics.
• Performing these statistical validations can be complex and time-consuming.

Purpose of the Study:

• To develop a mobile application for validating points systems.
• To simplify the calculation of discrimination and calibration using bootstrapping.
• To provide an accessible tool for researchers and clinicians.

Main Methods:

• Developed an Android application using native programming languages.
• Integrated statistical methods including logistic regression, bootstrapping, Area Under the Curve (AUC), and smoothed calibration plots.
• Validated the application using simulated data for Intensive Care Unit (ICU) mortality prediction, comparing results with the R statistical package.

Main Results:

• The mobile application produced results consistent with the R statistical package.
• No significant differences were observed between the application and R for the applied statistical techniques.
• The application successfully performed complex statistical validations on a mobile platform.

Conclusions:

• The developed methodology and mobile application effectively validate points systems for binary event prediction.
• The tool can be applied to various point systems and other predictive models.
• This offers a more accessible approach to validating predictive models in healthcare.