Using Machine Learning to Study Factors Affecting Discharge Destination in Recovery Units.

Area of Science:

• Medical informatics
• Machine learning in healthcare
• Stroke rehabilitation research

Background:

• Machine learning is increasingly used in medicine to analyze complex patient data.
• Multidimensional datasets are essential for advanced medical analysis and factor identification.

Purpose of the Study:

• To develop a predictive model for stroke patient discharge destination (home vs. facility).
• To utilize supervised machine learning, specifically a random forest algorithm, for this prediction.
• To construct a comprehensive dataset of 50 items for analysis.

Main Methods:

• Analysis of 30 cerebrovascular disease patients' discharge data.
• Dataset included patient characteristics, physical/cognitive function, Functional Independence Measure (FIM), blood data, and social factors.
• Random forest algorithm employed for classification, with accuracy assessed via five-fold cross-validation.

Main Results:

• Functional Independence Measure (FIM) and cognitive function (including memory) were identified as critical predictors.
• The random forest model achieved an 87.1% accuracy in predicting discharge destination.
• Mean decrease Gini was used to quantify the importance of each factor in the classification.

Conclusions:

• Activities of daily living (ADL) and cognitive function are significant factors influencing discharge decisions for stroke patients.
• This study highlights the potential of machine learning in optimizing stroke patient care pathways.
• Accurate prediction of discharge destination can aid in resource allocation and patient management.