Prediction of Nursing Need Proxies Using Vital Signs and Biomarkers Data: Application of Deep Learning Models

Area of Science:

• Healthcare Informatics
• Artificial Intelligence in Medicine
• Nursing Administration

Background:

• Deep learning applications in nursing administration for workload management are limited.
• Accurate prediction of patient nursing needs is crucial for effective staffing and care delivery.

Purpose of the Study:

• To develop and compare deep learning models (RNN, LSTM) against traditional regression for predicting nursing need proxies in hospitalized patients.
• To assess the efficacy of predictive models in enhancing nursing resource allocation.

Main Methods:

• Cross-sectional secondary data analysis of 20,855 adult patients' electronic health records.
• Utilized patient data (vital signs, biomarkers, demographics) from the preceding 48 hours.
• Employed Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) algorithms for sequential data processing.

Main Results:

• Both RNN and LSTM models demonstrated superior predictive performance for nursing need proxies compared to traditional regression.
• A reduction in prediction accuracy was observed during periods of rapid patient condition changes.

Conclusions:

• Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) models offer enhanced predictive capabilities for nursing needs.
• Proactive application of these deep learning models can aid in timely detection of patient demands, supporting effective and safe nursing services.