Diagnostic Support for Selected Paediatric Pulmonary Diseases Using Answer-Pattern Recognition in Questionnaires...

Area of Science:

• Pediatric Pulmonology
• Medical Informatics
• Data Mining Applications

Background:

• Pediatric pulmonary disease symptoms are often non-specific, leading to missed diagnoses of rare conditions like primary ciliary dyskinesia (PCD), cystic fibrosis (CF), and protracted bacterial bronchitis (PBB) by general practitioners (GPs).
• Early and accurate diagnosis is crucial for effective management and improved patient outcomes.

Purpose of the Study:

• To develop and validate a novel diagnostic support tool for pediatric pulmonary diseases.
• The tool integrates parental observations via a questionnaire with data mining techniques for enhanced diagnostic accuracy.

Main Methods:

• Parental interviews informed the creation of a disease-specific questionnaire.
• A database was established from questionnaires completed by parents of children with various pulmonary conditions.
• Multiple machine learning classifiers and an ensemble model were trained and evaluated using ten-fold cross-validation.

Main Results:

• The diagnostic tool achieved high accuracy across multiple conditions, including perfect diagnosis for cystic fibrosis (CF), asthma (AS), primary ciliary dyskinesia (PCD), and acute bronchitis (AB).
• High diagnostic rates were observed for pneumonia (90.5%) and protracted bacterial bronchitis (94.4%).
• The ensemble classifier demonstrated an overall sensitivity of 98.8%, with ROC analyses confirming its diagnostic accuracy.

Conclusions:

• A questionnaire-based diagnostic support tool leveraging data mining shows significant promise for assisting GPs in identifying pediatric pulmonary diseases.
• This tool can enhance physician awareness of rare conditions such as PCD and CF, facilitating timely intervention.
• The system's applicability in daily general practice was highlighted through case studies.