Functional data analysis with application to periodically stimulated foetal heart rate data. II: functional logistic...

Area of Science:

• Biostatistics
• Medical Statistics
• Longitudinal Data Analysis

Background:

• Modeling binary outcomes with functional covariates presents unique statistical challenges, particularly in singular longitudinal data analysis where measurements exceed the number of subjects.
• Existing methods may not adequately handle the complexity of functional data, especially when dealing with time-series measurements like physiological signals.

Purpose of the Study:

• To develop and present a robust statistical basis solution for modeling binary response data incorporating a functional covariate and scalar covariates.
• To extend this methodology for analyzing functional data with repeated stimuli, specifically applied to predicting high-risk birth outcomes using fetal heart rate tracings.

Main Methods:

• Utilized a basis expansion technique for parameter estimation in the presence of functional covariates.
• Employed a modified Fisher scoring algorithm to find maximum likelihood parameter estimates.
• Applied the developed technique to analyze periodically stimulated fetal heart rate tracings.

Main Results:

• The proposed basis solution effectively models binary responses with functional and scalar covariates.
• The extended technique successfully predicted the probability of high-risk birth outcomes using stimulated fetal heart rate tracings.
• Specifically, the stimulated fetal heart rate tracings achieved a prediction accuracy of 94.1% for high-risk pregnancies.

Conclusions:

• The developed statistical model provides an effective approach for analyzing complex functional data in binary outcome prediction.
• Periodically stimulated fetal heart rate tracings are highly predictive of high-risk pregnancies, outperforming unstimulated measurements.
• This methodology holds significant potential for improving prenatal risk assessment and clinical decision-making.