Measuring agreement between healthcare survey instruments using mutual information.

Area of Science:

• Health Services Research
• Information Theory
• Biostatistics

Background:

• Healthcare researchers frequently employ multiple survey instruments to assess patient symptoms.
• Comparing outcomes from different instruments raises questions about agreement and consistency.
• Existing methods for measuring agreement may have limitations.

Purpose of the Study:

• To introduce a novel procedure using mutual information for quantitatively measuring agreement between multiple healthcare survey instruments.
• To demonstrate the application of mutual information in comparing the outcomes of different assessment tools.
• To provide a framework for understanding information shared across instruments.

Main Methods:

• Review of existing methods for measuring agreement/disagreement between instruments.
• Development and application of a procedure utilizing mutual information.
• Quantitative measurement of shared information between instrument outcomes.
• Worked examples to illustrate the proposed approach.

Main Results:

• A case study analyzing instruments for detecting delirium superimposed on dementia (DSD) in older adults.
• Comparison of mutual information with traditional measures like odds ratio and Cohen's kappa.
• Demonstration of mutual information's utility in assessing instrument agreement.

Conclusions:

• Mutual information analysis effectively explains agreement and disagreement among multiple instruments.
• The proposed approach offers new insights into the application of healthcare survey instruments.
• This method has the potential to improve the reliability and validity of multi-instrument studies.