Building Data-Driven Pathways From Routinely Collected Hospital Data: A Case Study on Prostate Cancer.

Area of Science:

• Health Informatics
• Big Data Analytics in Healthcare
• Clinical Pathway Analysis

Background:

• Routinely collected hospital data is complex, heterogeneous, and fragmented across multiple Hospital Information Systems (HIS).
• This big data holds potential for disease understanding, pattern discovery, and service improvement but its quality is often inconsistent.
• Existing algorithms struggle with unstructured data and lack clinically meaningful visualizations, necessitating advanced support systems.

Purpose of the Study:

• To explore the development of data-driven patient pathways from routine hospital data.
• To propose a framework for constructing, assessing quality, and visualizing these pathways for research and decision support.
• To apply the framework to a prostate cancer case study.

Main Methods:

• Extracted and validated prostate cancer patient data from eight HIS and a cancer registry.
• Constructed data-driven pathways for 1904 patients.
• Utilized an expert knowledge base for pathway quality assessment and developed software for pathway visualization.

Main Results:

• Developed a framework and pathway formalism for summarizing, visualizing, and querying complex patient information.
• Enabled computation of quality indicators and dimensions for the constructed pathways.
• Introduced a novel graphical representation for synthesizing pathway information.

Conclusions:

• Data-driven clinical pathways from routine hospital data uncover previously unavailable insights into patients and diseases.
• This approach unifies heterogeneous data onto a single model and facilitates data quality assessment.
• The developed methods support further research in prostate cancer, biomarker analysis, and the mining/visualization of routine health data.