A better performing algorithm for identification of implausible growth data from longitudinal pediatric medical...

Area of Science:

• Pediatric growth monitoring
• Biostatistics
• Public health

Background:

• Tracking pediatric body size trajectories, like body mass index (BMI), is vital for assessing chronic disease risk.
• Inaccurate height and weight measurements in electronic medical records can lead to biologically implausible data, complicating growth trajectory analysis.
• Existing methods for identifying erroneous pediatric growth data may not be optimal.

Purpose of the Study:

• To develop and validate open-source algorithms for detecting and removing biologically implausible pediatric height and weight measurements.
• To improve the accuracy of body mass index (BMI) trajectory modeling in children.
• To enhance the analysis of associations between exposures, BMI trajectories, and subsequent health conditions.

Main Methods:

• Development of novel open-source algorithms for identifying implausible values in pediatric height and weight data.
• Comparison of the developed algorithms against three existing published algorithms using Monte Carlo simulations.
• Evaluation of algorithm performance based on sensitivity, specificity, and speed, using simulation inputs from longitudinal epidemiological cohorts.

Main Results:

• The newly developed algorithms demonstrated higher specificity compared to three previously published algorithms.
• The sensitivity and speed of the new algorithms were comparable to the existing methods.
• The findings indicate superior performance in cleaning longitudinal pediatric growth data.

Conclusions:

• The developed open-source algorithms are effective in detecting and removing biologically implausible pediatric growth data.
• Adoption of these algorithms can improve the quality and reliability of longitudinal pediatric growth datasets.
• Enhanced data quality facilitates more accurate modeling of BMI trajectories and their health implications.