Retrieving cases for treatment advice in nursing using text representation and structured text retrieval

Area of Science:

• Health Informatics
• Medical Data Analysis
• Nursing Research

Background:

• Nursing databases typically store patient details and treatments in structured formats.
• Extracting actionable insights from this data often requires advanced retrieval methods.
• Existing methods like rule-based systems and standard text retrieval have limitations in complex case analysis.

Purpose of the Study:

• To transform structured nursing database records into a text-based case collection.
• To develop and evaluate a case-structured retrieval method for patient data.
• To compare the performance of this new method against standard text retrieval and rule-based systems for treatment prediction.

Main Methods:

• Patient case days with history were created from a nursing database.
• The cosine similarity measure was adapted to compute whole case similarity.
• A linear regression model was used to learn optimal weights for combining case component similarities.
• Two treatment prediction tasks were evaluated on 1355 records.

Main Results:

• Standard text retrieval outperformed the rule-based system on one prediction task.
• The proposed case-structured retrieval method demonstrated at least an 18% improvement on both prediction tasks.
• This indicates the effectiveness of incorporating case structure into retrieval functions.

Conclusions:

• Case-structured retrieval offers a significant advancement over traditional methods for nursing data analysis.
• The developed method shows promise for improving treatment prediction accuracy in clinical settings.
• Further research into advanced retrieval techniques for medical records is warranted.