Developing an expert panel process to refine health outcome definitions in observational data.

Area of Science:

• Pharmacovigilance and Drug Safety
• Observational Data Analysis
• Health Informatics

Background:

• Accurate measurement of health outcomes of interest (HOI) is crucial for drug safety surveillance using observational data.
• Existing methods for HOI identification in claims databases can be improved for validity and efficiency.

Purpose of the Study:

• To develop and evaluate a novel method for reviewing HOI definitions in claims databases.
• The method integrates web-based digital tools, a systematic expert panel review, and a data collection process.

Main Methods:

• An interactive web-based dashboard was developed to present de-identified patient data for HOI assessment.
• A modified Delphi method was employed, with six panelist pairs independently reviewing ~200 cases for three HOIs (acute liver injury, kidney injury, myocardial infarction).
• Panelists assessed HOI presence, severity, and contributing data, with discrepancies resolved through consensus.

Main Results:

• The dashboard facilitated efficient and consistent case review, with panelists reporting ease of use and reliability.
• Iterative development refined the dashboard and review process, demonstrating its adaptability.
• The expert review process, supported by the dashboard, proved effective for HOI classification.

Conclusions:

• The developed web-based dashboard and expert review process effectively support accurate HOI measurement in observational data.
• This methodology is applicable to future case identification and classification in pharmacovigilance.
• Refinement of the dashboard and review process can be achieved by applying these methods to additional health outcomes.