A new look at patient satisfaction: learning from self-organizing maps

Area of Science:

• Nursing Science
• Data Analysis
• Artificial Intelligence

Background:

• Patient satisfaction surveys necessitate a variety of data processing techniques for comprehensive understanding.
• Integrating diverse analytical methods is crucial for accurately interpreting survey results.

Purpose of the Study:

• Introduce the self-organizing map (SOM), an artificial neural network, to nursing science.
• Demonstrate the application of SOM for clustering and analyzing patient satisfaction data.

Main Methods:

• Secondary analysis of patient satisfaction data from 5,283 adult patients across four Finnish hospitals (2008, 2010).
• Utilized self-organizing maps (SOM) for data clustering (respondents and questionnaire items) and as a preprocessor for multinomial logistic regression.
• Conducted missing data analysis to enhance data interpretation.

Main Results:

• Combined SOM and logistic regression analyses identified links between satisfaction levels, satisfaction components, and item nonresponse.
• The perceived positive association between patient age and satisfaction may be influenced by age-related item nonresponse patterns.

Conclusions:

• Self-organizing maps are effective for clustering questionnaire data, even with low dimensionality.
• Incorporating nonresponse data into analyses can help identify potentially misleading non-causative relationships in patient satisfaction studies.