A predictive internet-based model for COVID-19 hospitalization census.

Area of Science:

• Epidemiology
• Health Informatics
• Time Series Analysis

Background:

• The COVID-19 pandemic significantly impacted hospital resources, highlighting the need for accurate patient demand forecasting.
• Existing research explored correlations between Google Trends data and COVID-19 case counts.
• Effective resource allocation and staffing require robust predictive models for healthcare systems.

Purpose of the Study:

• To develop and evaluate a Vector Error Correction Model (VECM) for forecasting COVID-19 patient census within a healthcare system.
• To integrate internet search trends and healthcare chatbot data as predictors in the VECM.
• To assess the VECM's predictive accuracy and potential utility in public health surveillance.

Main Methods:

• A Vector Error Correction Model (VECM) was employed to forecast the number of COVID-19 patients (Census).
• The model incorporated data from Google search trends and healthcare chatbot interactions.
• Statistical hypothesis testing and mean absolute percentage prediction error were used to evaluate model performance.

Main Results:

• The VECM demonstrated a good fit for the COVID-19 patient census data.
• The model exhibited very good forecasting performance, validated through rigorous statistical analysis.
• The integration of internet search activity and chatbot scores proved effective in enhancing predictions.

Conclusions:

• The developed VECM shows potential as a valuable tool for COVID-19 surveillance within healthcare systems.
• The study highlights the utility of incorporating real-time internet data for public health forecasting.
• Despite limitations, the rigorous methodology supports the VECM's application in managing healthcare demands during health crises.