A Guide for the Application of Statistics in Biomedical Studies Concerning Machine Learning and Artificial...

Area of Science:

• Orthopaedic research
• Machine learning applications
• Artificial intelligence in medicine

Background:

• The orthopaedic literature shows a surge in machine learning (ML) analyses.
• Existing literature often focuses on ML definitions and basic concepts, lacking depth.
• Traditional statistical methods require explicit variable selection, potentially introducing bias and limiting performance optimization.

Purpose of the Study:

• To critically evaluate the current state of machine learning in orthopaedic research.
• To advocate for the application of scientific method principles for interpreting ML study results.
• To guide readers in drawing independent conclusions from ML-based orthopaedic studies.

Main Methods:

• Review and critical analysis of existing machine learning literature in orthopaedics.
• Comparison of machine learning approaches with traditional statistical methodologies.
• Emphasis on the principles of scientific method for evidence appraisal.

Main Results:

• Current ML studies in orthopaedics often provide repetitive explanations of ML concepts.
• The clinical utility of published ML algorithms is frequently limited and requires external validation.
• A gap exists in guiding readers to critically interpret ML study findings.

Conclusions:

• Readers need guidance to apply scientific principles for evaluating ML research in orthopaedics.
• Critical appraisal is essential to understand the true potential and limitations of ML algorithms.
• Further validation and a robust interpretative framework are necessary before widespread adoption of ML in clinical practice.