A systematic review of how missing data are handled and reported in multi-database pharmacoepidemiologic studies

Area of Science:

• Pharmacoepidemiology
• Health Informatics
• Biostatistics

Background:

• Multi-database studies (MDBS) are crucial for evaluating drug safety and effectiveness.
• Missing data in MDBS can lead to biased results and reduced precision.
• Standardized reporting and handling of missing data in MDBS are needed.

Purpose of the Study:

• To assess the reporting and handling of missing data in pharmacoepidemiologic MDBS.
• To identify common methods used to address missing data in these studies.

Main Methods:

• Systematic literature search of PubMed for pharmacoepidemiologic MDBS published 2018-2019.
• Inclusion criteria: studies using ≥2 distinct databases for the same drug outcome.
• Extracted data included MDBS strategies, missing data reporting (type, bias), and methods for addressing missing data.

Main Results:

• 62 studies were included, with most data from North America and Europe.
• 56% of studies reported missing data; 35% of those acknowledged potential bias.
• 19 studies reported methods to address missing data, with complete case analysis being most common (68%).

Conclusions:

• Just over half of recent pharmacoepidemiologic MDBS reported missing data.
• Two-thirds of studies reporting missing data also described how they addressed it.
• Increased vigilance in reporting and addressing missing data is essential to mitigate bias in MDBS.