Logistic regression: a brief primer

Area of Science:

• Medical Statistics
• Biostatistics
• Epidemiology

Background:

• Regression techniques are vital in medical research for measuring associations, predicting outcomes, and controlling confounding variables.
• Logistic regression offers an efficient method to analyze the impact of independent variables on binary outcomes.

Purpose of the Study:

• To elucidate the application and considerations of logistic regression in medical research.
• To detail variable selection, assumption checking, and model building strategies for logistic regression.

Main Methods:

• Logistic regression utilizes components of linear regression on the logit scale to identify key predictors.
• Key considerations include independent variable selection, assumption verification (e.g., linearity, multicollinearity), and model-building strategies (direct, sequential, stepwise).
• Model validity (internal and external) and goodness-of-fit are assessed using diagnostic statistics and measures.

Main Results:

• Independent variable contributions are quantified, and results are typically reported as odds ratios (ORs) with 95% confidence intervals (CIs).
• Adequate events per variable (10-20) are recommended to prevent model overfitting.
• Model fit is evaluated by comparing observed and predicted values.

Conclusions:

• Logistic regression is a versatile tool for medical research, providing insights into variable effects on binary outcomes.
• Proper application requires careful attention to variable selection, assumption adherence, and model validation for reliable results.