Medicine utilization studies in Australian individual-level dispensing data: A blinded, multi-center replicated...

Area of Science:

• Health Services Research
• Pharmacoepidemiology
• Data Science

Background:

• Medication dispensing data require extensive preparation for research.
• Analytical decisions during data preparation can lead to non-replicable results.
• Standardizing these decisions is crucial for reliable research outcomes.

Purpose of the Study:

• To examine the impact of analytical decisions on research findings.
• To assess reproducibility in studies analyzing medication dispensing data.
• To evaluate a harmonized protocol for enhancing reproducibility in pharmacoepidemiology.

Main Methods:

• Four Australian sites independently applied the HARmonized Protocol template to Enhance Reproducibility (HARPER).
• Analyses were conducted on the Australian Pharmaceutical Benefits Scheme 10% sample dataset.
• Key outcomes included cohort size, demographics, treatment discontinuation, switching, and intensification, with concordance assessed across sites.

Main Results:

• Good agreement was observed for cohort size, gender, and age group.
• However, poor agreement was found for treatment discontinuation, switching, and intensification rates due to differing methods for estimating days supply and operational definitions.
• Time-to-event estimates and hazard ratios also showed poor agreement.

Conclusions:

• Variations in analytical decisions when processing dispensing data significantly affect research replicability.
• Detailed analytical protocols, such as HARPER, are essential for ensuring transparency in operational definitions and interpretations.
• Standardized protocols are critical for improving the reliability and reproducibility of studies using medication dispensing data.