Beware of registries for their biases

Area of Science:

• Epidemiology
• Biostatistics
• Health Services Research

Background:

• Patient registries are widely utilized for data collection.
• However, observational and retrospective data from registries and administrative databases are susceptible to various biases.
• Control group selection is a critical challenge in observational studies.

Purpose of the Study:

• To highlight common and emerging biases in patient registries and observational studies.
• To emphasize the critical role of control group selection in data integrity.
• To introduce and explain the concept of "mortality bias".

Main Methods:

• Review of established biases in observational research.
• Identification and definition of specific biases including "immortal time bias" and "mortality bias".
• Discussion of the impact of control group selection on study outcomes.

Main Results:

• Observational studies and registries are prone to selection biases.
• Common biases include confounding by disease severity, channeling bias, and immortal time bias.
• A novel bias, "mortality bias," is described where deaths in a source population reduce observed mortality in a registry.

Conclusions:

• Careful consideration of biases is essential for valid scientific conclusions from patient registries.
• Proper control group selection is paramount to avoid misleading results.
• Awareness of "mortality bias" is crucial to prevent misinterpretation of registry data.