Exploring flexible polynomial regression as a method to align routine clinical outcomes with daily data capture...

Area of Science:

• Biostatistics
• Medical Informatics
• Pediatric Pulmonology

Background:

• Clinical outcome data capture frequency often lags behind high-volume remote technology data.
• A data disparity exists between clinical outcomes and remote monitoring, hindering comprehensive patient analysis.
• Flexible polynomial regression offers a method to reconcile these data volume differences.

Purpose of the Study:

• To investigate flexible polynomial regression for estimating personalized trends in continuous clinical outcomes over time.
• To align infrequent clinical outcome data with frequent remote monitoring data.
• To assess the utility of flexible polynomials in pediatric cystic fibrosis patient data.

Main Methods:

• Electronic health records were used to calculate flexible polynomial regression models (1st-4th order).
• Models predicted forced expiratory volume in 1 second (FEV1) over time in children with cystic fibrosis.
• Individualized best-fit models were selected using the lowest Akaike Information Criterion (AIC).

Main Results:

• 8,549 FEV1 measurements from 267 children were analyzed.
• Polynomial predictions were accurate for individuals with over 15 measurements.
• Model performance for those with fewer than 15 measurements was conditional on data quantity and timing.

Conclusions:

• Flexible polynomials effectively extrapolate clinical outcomes to match frequent remote monitoring data intervals.
• This approach aids in aligning disparate data sources for a more cohesive patient health overview.
• The method was successfully validated using Body Mass Index (BMI) data in the same pediatric population.