Comparison of predictor approaches for longitudinal binary outcomes: application to anesthesiology data

Area of Science:

• Medical Statistics
• Anesthesiology Research
• Clinical Data Analysis

Background:

• Longitudinal binary data analysis is crucial in clinical studies.
• Traditional statistical methods are insufficient for complex clinical hypotheses.
• Advanced techniques like GEE and GLMM offer improved analytical capabilities.

Purpose of the Study:

• To compare Generalized Estimating Equations (GEE) and Generalized Linear Mixed Models (GLMM) for modeling repeated binary responses.
• To evaluate the suitability of GEE and GLMM in an anesthesiology context.
• To analyze factors associated with hypotension during spinal anesthesia.

Main Methods:

• Comparative analysis of GEE and GLMM statistical approaches.
• Utilized a dataset of 375 patients undergoing spinal anesthesia.
• Modeled the relationship between hypotension and clinical variables (age, gender, surgical factors, vital signs, anesthetic agents).

Main Results:

• Parameter estimates from GLMM were generally larger than GEE, except for time-after, Marcain-Heavy, and Fentanyl.
• Standard errors were larger in the GLMM compared to GEE.
• GLMM demonstrated greater suitability for analyzing hypotension in this dataset.

Conclusions:

• Generalized Linear Mixed Models (GLMM) are recommended over Generalized Estimating Equations (GEE) for analyzing hypotension during spinal anesthesia.
• The choice of statistical model significantly impacts the interpretation of clinical findings in longitudinal binary data.
• Accurate modeling is essential for understanding and managing patient outcomes in anesthesiology.