Best (but often forgotten) Statistical Practices: Measuring Real-World Intervention Effectiveness using Electronic...

Area of Science:

• Health Services Research
• Clinical Epidemiology
• Data Science

Background:

• Randomized controlled trials (RCTs) form the primary evidence base for interventions.
• Clinical practice often differs from RCT conditions, impacting intervention effectiveness.
• Electronic health data offers opportunities to study real-world intervention effectiveness.

Purpose of the Study:

• To identify key barriers in real-world effectiveness studies using electronic health data.
• To propose statistical best practices for overcoming these challenges.
• To improve the quality of evidence generated from real-world data.

Main Methods:

• Review of challenges in real-world effectiveness studies.
• Discussion of statistical methodologies for addressing data quality.
• Exploration of techniques to mitigate selection bias and confounding by indication.

Main Results:

• Key barriers include data quality issues, selection bias, confounding by indication, and limited generalizability.
• Specific statistical methods can help address these challenges.
• High-quality real-world evidence is achievable with appropriate analytical approaches.

Conclusions:

• Real-world effectiveness studies using electronic health data are crucial but challenging.
• Adopting robust statistical best practices is essential for reliable findings.
• This work provides a framework for improving the validity of real-world evidence.