Contrasting temporal trend discovery for large healthcare databases

Area of Science:

• Health Informatics
• Data Mining
• Healthcare Analytics

Background:

• Increasing adoption of electronic health records (EHRs) drives demand for advanced data analysis techniques.
• Traditional methods struggle to identify complex temporal trends within diverse patient populations.

Purpose of the Study:

• To introduce a novel data mining approach for exploring and comparing temporal trends in in-patient subgroups.
• To leverage visual analytics and associated rule mining for enhanced big data exploration.

Main Methods:

• Utilized the Apriori algorithm and linear model-based recursive partitioning for associated rule mining.
• Employed visual analytics, including regression trees with scatter plots and trend lines, for big data trend discovery.
• Evaluated the approach using the Nationwide Inpatient Sample (NIS) database.

Main Results:

• Demonstrated the discovery of opposite temporal trends in age and sex-based patient subgroups.
• Highlighted the inability of traditional trend-tracking techniques to identify these nuanced trends.
• Showcased the effectiveness of the proposed visual analytics approach for big data trend analysis.

Conclusions:

• The novel approach effectively uncovers complex temporal trends in healthcare data, particularly those missed by conventional methods.
• This technique offers valuable decision support for policymakers and hospital management in identifying and responding to critical health trends.
• Visual analytics integrated with data mining provides a powerful tool for understanding patient subgroup dynamics over time.