Using nursing data for machine learning-based prediction modeling in intensive care units: A scoping review.

Area of Science:

• Nursing Informatics
• Clinical Data Science
• Machine Learning in Healthcare

Background:

• Nursing data is crucial for early detection of patient deterioration and outcome prediction.
• Advancements in machine learning necessitate robust clinical prognosis prediction models for intensive care unit (ICU) patients.
• A comprehensive understanding of nursing data types used in predictive models is lacking.

Purpose of the Study:

• To conduct a scoping review of machine learning models utilizing nursing data for predicting ICU patient health outcomes.
• To identify and categorize the types of nursing data incorporated into these predictive models.

Main Methods:

• Systematic literature search across seven databases until December 2023.
• Inclusion of studies using machine learning with nursing data for adult ICU patient prognosis prediction.
• Data extraction and organization focusing on study, model, and nursing data characteristics.

Main Results:

• 151 studies published between 2004-2023, with a notable increase since 2018.
• Supervised learning was the predominant machine learning method, with regression being most common.
• Nursing scales were the most frequently utilized nursing data type (n=150), followed by assessment records, notes, and activity records. Clinical outcomes (87.0%) were predicted more often than nursing-sensitive outcomes (13.0%).

Conclusions:

• Nursing scales are currently the most utilized nursing data for predicting critically ill patient prognoses.
• The potential of diverse nursing data types (e.g., nursing notes) in prediction models warrants further investigation.
• Findings offer insights for healthcare providers and researchers developing clinical prognosis prediction models in the ICU.