Forecasting cardiovascular disease mortality using artificial neural networks in Sindh, Pakistan

Area of Science:

• Public Health
• Epidemiology
• Biostatistics

Background:

• Cardiovascular disease (CVD) is a major global health concern with increasing prevalence.
• Accurate modeling and forecasting of CVD mortality are essential for public health planning and intervention evaluation.
• The Sindh province of Pakistan faces a growing burden of cardiovascular disease.

Purpose of the Study:

• To model and forecast cardiovascular disease (CVD) mortality in Pakistan's Sindh province.
• To compare the predictive accuracy of classical time series models with a machine learning approach.
• To identify the most effective model for CVD mortality forecasting in the region.

Main Methods:

• Utilized a time series dataset of CVD mortality cases from 1999-2021 from a civil hospital in Nawabshah, Sindh.
• Applied classical time series models: Naïve, Holt-Winters, and Simple Exponential Smoothing (SES).
• Compared classical models against the Artificial Neural Network Auto-Regressive (ANNAR) machine learning model.

Main Results:

• The Artificial Neural Network Auto-Regressive (ANNAR) model demonstrated superior performance in forecasting CVD mortality.
• ANNAR significantly outperformed Naïve, Holt-Winters, and SES models based on error metrics (RMSE, MAE, MAPE).
• The study identified ANNAR as the optimal model for predicting future CVD mortality trends in Sindh.

Conclusions:

• The ANNAR model is the most effective tool for forecasting cardiovascular disease mortality in Pakistan's Sindh province.
• Accurate CVD mortality forecasts can inform health policy, resource allocation, and intervention strategies.
• This predictive capability aids in mitigating the future disease burden of cardiovascular disease.