PEPRVision: a visualization framework for Pharmacoepidemiology research

Area of Science:

• Pharmacoepidemiology
• Health Informatics
• Data Visualization

Background:

• Medical chart review is standard in epidemiology but structured healthcare data, like administrative claims, are difficult to interpret.
• Existing data formats hinder efficient patient profile review in pharmacoepidemiologic research.

Purpose of the Study:

• Introduce PEPRVision, a novel visualization framework to translate structured longitudinal healthcare data into intuitive graphics.
• Enhance patient profile review and pattern recognition in pharmacoepidemiology.

Main Methods:

• Developed PEPRVision based on preattentive visual processing principles.
• Utilized visual attributes (color, position, shape, length) for rapid pattern recognition.
• Demonstrated the framework using US administrative claims data across four case studies.

Main Results:

• PEPRVision effectively supports evaluation of algorithms for health outcomes (e.g., congenital hearing loss).
• The framework aids in identifying biases like reverse causation in drug safety studies.
• Facilitates assessment of potential teratogenicity signals and identifies exposure mechanisms for risk mitigation.

Conclusions:

• PEPRVision enhances the validity and clinical relevance of pharmacoepidemiologic research by enabling review of patient-level variation.
• Complements population-based causal inference studies by providing intuitive data interpretation.
• Offers a valuable tool for improving study design and risk management strategies in drug safety research.