Prediction versus aetiology: common pitfalls and how to avoid them.

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Prediction and etiological research are distinct fields within epidemiology.
• Both utilize multivariable modeling but differ in aims and interpretation.
• Confusion between these research types leads to low-quality publications.

Purpose of the Study:

• To differentiate prediction research from etiological research.
• To highlight the distinct aims and interpretation of results for each study type.
• To address common pitfalls arising from the confusion between prediction and etiology.

Main Methods:

• The study differentiates prediction and etiological research conceptually.
• It contrasts their underlying research aims and interpretation of multivariable models.
• Examples of common pitfalls in interpreting prediction models are provided.

Main Results:

• Etiological research seeks causal effects, adjusting for known confounders.
• Prediction research focuses on accurate outcome risk prediction using collective predictors.
• Prediction models rely on statistical associations, not necessarily causal ones.

Conclusions:

• Clear separation of prediction and etiological research is crucial.
• Misattributing causal meaning to individual predictors in prediction models is a significant issue.
• Understanding these differences improves research quality and applicability.