A COVID-19 Risk Assessment Decision Support System for General Practitioners: Design and Development Study

Area of Science:

• Public Health
• Health Informatics
• Epidemiology

Background:

• COVID-19 presents a global health crisis requiring robust community engagement and effective healthcare delivery by general practitioners (GPs).
• GPs face significant challenges in managing infectious disease outbreaks, with a lack of suitable mobile systems for dynamic patient assessment and follow-up.
• Existing healthcare systems struggle to provide GPs with tools for real-time data collection, risk stratification, and patient management during pandemics.

Purpose of the Study:

• To design, develop, and deploy DDC19, a mobile decision support system tailored for GPs during the COVID-19 pandemic.
• To enhance GPs' capabilities in collecting patient data, dynamically assessing COVID-19 risk, and managing patient follow-up.
• To provide an effective tool for triaging patients and supporting clinical decision-making in a pandemic setting.

Main Methods:

• Developed a mobile-based decision support system (DDC19) comprising patient-end and GP-end applications and a central database.
• Constructed a COVID-19 dynamic risk stratification model using a multiclass logistic regression algorithm, prioritizing high recall and clinical interpretability.
• Utilized a 10-fold cross-validation on 2243 clinical cases with up to 36 features for model training and quantitative evaluation of risk stratification.

Main Results:

• The DDC19 system integrates mobile apps and a database for wireless data transmission and management.
• The dynamic risk stratification model demonstrated good predictive ability across various scenarios, with AUC values above 0.71.
• Model performance was evaluated using datasets with increasing feature dimensions, including demographic, symptom, contact history, blood test, and CT imaging data.

Conclusions:

• DDC19 is a functional mobile decision support system designed to aid GPs in assessing COVID-19 patient risk.
• The system effectively supports GPs in providing dynamic risk assessments and managing patients during the COVID-19 outbreak.
• The developed risk model shows a robust ability to predict patient risk levels across different data availability scenarios.