Predicting in-hospital length of stay for very-low-birth-weight preterm infants using machine learning techniques
Wei-Ting Lin1, Tsung-Yu Wu1, Yen-Ju Chen1
1Department of Pediatrics, National Cheng Kung University Hospital, College of Medicine, National Cheng-Kung University, Tainan, Taiwan.
Insights
Machine learning models can predict prolonged hospital stays for very-low-birth-weight (VLBW) infants. This study benchmarks length of stay (LOS) in Taiwan, showing ML improves prediction accuracy for VLBW infant care.
Area of Science:
- Neonatalogy
- Medical Informatics
- Biostatistics
Background:
- In-hospital length of stay (LOS) for very-low-birth-weight (VLBW) infants (BW < 1500 g) is a key indicator of care quality and impacts resource allocation.
- Understanding and predicting LOS is crucial for managing VLBW infant care and healthcare resources.
Purpose of the Study:
- To analyze the length of stay (LOS) for very-low-birth-weight (VLBW) infants in Taiwan.
- To develop and compare the performance of machine learning (ML) models for predicting LOS in VLBW infants.
Main Methods:
- Retrospective analysis of LOS data from 3519 VLBW infants born between 2016-2018 in the Taiwan Neonatal Network.
- Development of prediction models using 21 early neonatal variables and six ML algorithms, assessing continuous (LOS, PMA) and categorical (discharge group) outcomes.
- Model performance evaluated using R² for continuous variables and AUC for categorical variables.
Main Results:
- Mortality was high (59%) within the first 7 days for VLBW infants.
- Median LOS differed significantly between surviving (62 days) and deceased (5 days) infants.
- Machine learning models showed limited accuracy (R² < 0.6) for predicting overall LOS or postmenstrual age (PMA), but logistic regression (AUC: 0.724) and random forest (AUC: 0.712) demonstrated better performance for predicting prolonged LOS.
Conclusions:
- This study establishes a benchmark for VLBW infant LOS across gestational age groups in Taiwan.
- Machine learning techniques show promise in enhancing the accuracy of prolonged LOS prediction models for VLBW infants.
Background/Purpose:
The in-hospital length of stay (LOS) among very-low-birth-weight (VLBW, BW < 1500 g) infants is an index for care quality and affects medical resource allocation. We aimed to analyze the LOS among VLBW infants in Taiwan, and to develop and compare the performance of different LOS prediction models using machine learning (ML) techniques.
Methods:
This retrospective study illustrated LOS data from VLBW infants born between 2016 and 2018 registered in the Taiwan Neonatal Network. Among infants discharged alive, continuous variables (LOS or postmenstrual age, PMA) and categorical variables (late and non-late discharge group) were used as outcome variables to build prediction models. We used 21 early neonatal variables and six algorithms. The performance was compared using the coefficient of determination (R2) for continuous variables and area under the curve (AUC) for categorical variables.
Results:
A total of 3519 VLBW infants were included to illustrate the profile of LOS. We found 59% of mortalities occurred within the first 7 days after birth. The median of LOS among surviving and deceased infants was 62 days and 5 days. For the ML prediction models, 2940 infants were enrolled. Prediction of LOS or PMA had R2 values less than 0.6. Among the prediction models for prolonged LOS, the logistic regression (ROC: 0.724) and random forest (ROC: 0.712) approach had better performance.
Conclusion:
We provide a benchmark of LOS among VLBW infants in each gestational age group in Taiwan. ML technique can improve the accuracy of the prediction model of prolonged LOS of VLBW.
