Development of a prediction model for infant hospitalization and death using clinical features assessed by community health workers during routine postnatal home visits in Dhaka, Bangladesh
Journal:
medRxiv
Published Date:
Jan 23, 2026
Abstract
Introduction: To improve upon the World Health Organization (WHO) 8 danger signs used to identify young infants (<2 months) requiring referral during community health worker (CHW) home visits, aggregative features (e.g., cumulative visits with fever) rather than visit-specific features (e.g., fever at a single visit), and a machine learning random forest model, may enhance predictive performance. Applying these approaches, we aimed to develop a prediction model for infant hospitalization and/or death using CHW-assessed clinical features during home visits in Dhaka, Bangladesh. Methods: We analyzed data from generally healthy infants prospectively enrolled at birth and assessed at 11 scheduled CHW visits from 3-60 days of age. To predict first hospitalization or death, we developed two models -- time-varying Cox regression and random forest -- using the same set of candidate predictors (45 clinical features of which 8 were WHO danger signs, and 12 additional covariates) with aggregative features incorporated. We evaluated discrimination (C-statistic) and calibration (calibration plots). Performance was compared to a time-varying Cox model using only WHO danger signs. Results: Among 1906 infants, 176 (9.2%) had an event (173 hospitalizations, 3 deaths). The best-performing Cox model (C-statistic=0.71; 95% CI 0.68-0.75) consisting of three baseline covariates (any perinatal/delivery complication, umbilical cord care, gestational age) and four visit-specific clinical features (nasal congestion, cough, jaundice, skin rash), and a Cox model with these four features plus WHO danger signs (C-statistic=0.70; 95% CI 0.67-0.74), demonstrated higher discrimination than WHO danger signs alone (C-statistic=0.56; 95% CI 0.54-0.60), with similar calibration. A random forest model (42 predictors) was well-calibrated with comparable discrimination (C-statistic=0.69; 95% CI 0.64-0.73). Conclusion: Aggregative features and random forest did not outperform a time-varying Cox model using baseline covariates and visit-specific features. Adding four features to WHO danger signs may improve predictive performance by capturing a broader spectrum of infant illnesses requiring hospitalization.
