Patient-Oriented Questionnaires and Machine Learning for Rare Disease Diagnosis: A Systematic Review

Area of Science:

• Medical Informatics
• Computational Biology
• Rare Diseases Research

Background:

• Delayed diagnosis of rare diseases (RDs) is a significant challenge due to nonspecific symptoms and limited clinician experience.
• Patient-oriented questionnaires (POQs) offer a patient-centric data source, capturing daily experiences beyond clinical signs.
• Machine learning (ML) techniques can potentially analyze POQ data for improved diagnostic pathways.

Purpose of the Study:

• To systematically review the current research on using POQs and ML for disease diagnosis.
• To identify the potentials and limitations of applying ML to POQ data for rare and common diseases.
• To assess the state of research regarding predictive indicators derived from POQs.

Main Methods:

• Systematic literature search adhering to PRISMA guidelines across PubMed, Semantic Scholar, and Google Scholar (until June 2023).
• Inclusion criteria focused on studies using POQs for diagnosis and applying ML to the resulting data, including prediction generation and indicator identification.
• Data extraction included publication year, questionnaire details, ML algorithms, input data, performance metrics, and questionnaire development.

Main Results:

• The search yielded 421 results, with 26 studies selected for further consideration.
• Sixteen studies focused on disease analysis using ML algorithms, while ten provided supporting research.
• Potentials and limitations of the POQ-ML approach were discussed.

Conclusions:

• ML algorithms demonstrate promising results when applied to POQ data for disease diagnosis.
• The full potential of this methodology remains underexploited, warranting further investigation.
• Clinical application and real-world medical practice integration of ML-driven POQ analysis require further study.