Predicting Low Cognitive Ability at Age 5-Feature Selection Using Machine Learning Methods and Birth Cohort Data.

Area of Science:

• Pediatric cognitive development
• Machine learning in healthcare
• Biostatistics

Background:

• Early identification of low cognitive ability is crucial for timely intervention.
• Birth-cohort studies offer rich data for predictive modeling.
• Machine learning algorithms can analyze complex datasets to identify risk factors.

Purpose of the Study:

• To develop and validate a machine learning model for predicting low cognitive ability at age 5.
• To identify key maternal, infant, and sociodemographic predictors of cognitive development.
• To assess the predictive performance of a Random Forest model.

Main Methods:

• Utilized the Irish population-based BASELINE cohort (n=1,070).
• Trained a Random Forest model to predict intelligence quotient (IQ) ≤90 at age 5.
• Performed internal validation using 10-fold cross-validation (5 repetitions).

Main Results:

• The top five predictors were maternal schooling, infant Apgar score, socioeconomic index, maternal BMI, and first-trimester alcohol consumption.
• A parsimonious model with 11 features achieved 95% classification accuracy.
• The model demonstrated excellent predictive ability on internal validation.

Conclusions:

• Machine learning effectively identifies important features for predicting cognitive ability.
• The developed model shows promise for risk prediction in pediatric populations.
• External validation in diverse cohorts is necessary to confirm generalizability.