Common pre-diagnostic features in individuals with different rare diseases represent a key for diagnostic support...

Area of Science:

• Medical Diagnostics
• Machine Learning in Healthcare
• Rare Diseases

Background:

• Rare diseases (RD) present diverse clinical symptoms, posing significant diagnostic challenges.
• A hypothesis suggests shared phenomena exist among RD patients before diagnosis.
• Identifying these commonalities can improve diagnostic pathways.

Purpose of the Study:

• To identify commonalities in pre-diagnostic experiences across various rare diseases.
• To develop a machine learning-based diagnostic support tool for rare diseases.

Main Methods:

• Qualitative analysis of 20 interviews with rare disease patients.
• Development of a 53-questionnaire based on pre-diagnostic experiences.
• Application of machine learning algorithms (4 single methods + fusion) to analyze questionnaire data from RD, non-rare, chronic, and psychosomatic groups.

Main Results:

• 1763 questionnaires were collected and analyzed.
• Machine learning models achieved high sensitivity: 88.9% for RD, 86.6% for NRO, 87.7% for CD, and 84.2% for PSY.
• Distinct answer patterns were identified for rare disease patients.

Conclusions:

• Rare disease patients exhibit shared pre-diagnosis experiences despite diverse presentations.
• The developed questionnaire and data-mining approach successfully identified unique patterns for diagnostic support.
• These findings suggest a potential for improved rare disease diagnosis through pattern recognition.