Methodological approaches in analysing observational data: A practical example on how to address clustering and...

Area of Science:

• Biostatistics
• Epidemiology
• Health Services Research

Background:

• Randomized controlled trials (RCTs) cannot answer all effectiveness questions.
• Observational studies require methods to minimize bias, particularly selection bias and clustering.
• Advanced statistical techniques are essential for accurate effect estimates in real-world data.

Purpose of the Study:

• Introduce propensity score matching and mixed models for real-world data analysis.
• Implement and demonstrate these methods using R statistical software for reproducibility.
• Address bias and clustering in observational health research.

Main Methods:

• Employed a two-level analytic strategy.
• Utilized generalized models for binary data analysis, accounting for dependencies.
• Applied genetic matching and covariate adjustment to mitigate selection bias.
• Analyzed data from two population samples: one matched, one full sample.

Main Results:

• Different analytical methods yielded varying results but consistent directional findings.
• The probability of receiving a case conference was higher in the treatment group.
• Both genetic matching and covariate adjustment demonstrated limitations but offered complementary insights.

Conclusions:

• Statistical approaches effectively reduced bias in observational studies.
• Methodological choices are constrained by the specific sample characteristics.
• A careful evaluation of the pros and cons of each method is crucial for individual studies.