Machine Learning-Driven Probability Scoring Enhances Diagnostic Certainty and Reduces Costs in Suspected...

Area of Science:

• Orthopedic Surgery
• Medical Diagnostics
• Artificial Intelligence in Medicine

Background:

• Accurate periprosthetic joint infection (PJI) diagnosis is difficult, especially in culture-negative cases, leading to diagnostic uncertainty.
• Current diagnostic methods often result in high levels of indecision among physicians.
• SynTuition, an AI-based probability score, was developed to aid PJI diagnosis using preoperative biomarkers.

Purpose of the Study:

• To compare the diagnostic performance of SynTuition against standard physician practice for suspected PJI.
• To evaluate the economic impact of SynTuition in diagnosing PJI.
• To assess SynTuition's ability to reduce diagnostic uncertainty in challenging PJI cases.

Main Methods:

• 12 physicians diagnosed 274 clinical vignettes of suspected PJI.
• SynTuition probabilities were converted to binary classifications using a validated threshold.
• Diagnostic accuracy, agreement, indecision rates, decision curve analysis, and misdiagnosis costs were evaluated.

Main Results:

• SynTuition achieved 96.0% agreement versus expert adjudication, outperforming physicians (90.8%).
• SynTuition provided definitive diagnoses in ambiguous cases where physicians showed high indecision (38-48%).
• SynTuition demonstrated higher net benefit, reducing projected unnecessary revisions by up to 5.8% and saving an estimated $4000 per suspected case.

Conclusions:

• SynTuition exhibits superior diagnostic accuracy and lower uncertainty compared to routine physician practice for PJI.
• The AI tool offers significant clinical and economic advantages, particularly in diagnostically ambiguous PJI cases.
• Integration of SynTuition into clinical decision-making is supported for suspected PJI diagnosis.