Forecasts of cardiac and respiratory mortality in Tehran, Iran, using ARIMAX and CNN-LSTM models

Area of Science:

• Public Health
• Biostatistics
• Epidemiology
• Artificial Intelligence in Healthcare

Background:

• Cardiovascular diseases (CVDs) are a primary cause of mortality and disability globally, with Iran facing a projected doubling of disease burden by 2025.
• Accurate forecasting models are crucial for predicting CVD progression, mortality rates, and identifying risk factors to inform public health interventions.

Purpose of the Study:

• To compare the predictive accuracy of two time series models: Autoregressive Integrated Moving Average with Exogenous variable (ARIMAX) and Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM).
• To evaluate the effectiveness of these models in forecasting cardiovascular disease trends in Iran.

Main Methods:

• Time series analysis was conducted using ARIMAX (6,1,6) and CNN-LSTM models.
• Model performance was evaluated based on Mean Squared Error (MSE).
• The influence of Nitrogen dioxide (NO2) as an exogenous variable was assessed in the ARIMAX model.

Main Results:

• The ARIMAX model achieved an MSE of 0.655, with significant associations noted at lags 4 and 6, and NO2 significant at lag 6.
• The CNN-LSTM model demonstrated superior performance with a significantly lower MSE of 0.21.
• Deep learning models, specifically CNN-LSTM, provided more accurate forecasts compared to the classical ARIMAX method.

Conclusions:

• Deep learning approaches, exemplified by CNN-LSTM, significantly outperform traditional statistical methods like ARIMAX for cardiovascular disease time series forecasting.
• The findings highlight the potential of advanced AI models to improve the accuracy of epidemiological predictions and support public health strategies for managing cardiovascular diseases.