Predicting Multimorbidity Using Saudi Health Indicators (Sharik) Nationwide Data: Statistical and Machine Learning...

Area of Science:

• Public Health
• Epidemiology
• Health Informatics

Background:

• The Saudi population exhibits a high prevalence of multimorbidity.
• Identifying modifiable behavioral risk factors is crucial for mitigating multimorbidity.
• Understanding predictors of multimorbidity is essential for targeted public health interventions.

Purpose of the Study:

• To identify and predict factors associated with multimorbidity in the Saudi population.
• To compare the efficacy of statistical and machine learning models in predicting multimorbidity.
• To provide evidence for public health policy development in Saudi Arabia.

Main Methods:

• Utilized data from 23,098 Saudi residents from the "Sharik" Health Indicators Surveillance System 2021.
• Employed binary logistic regression models to identify multimorbidity predictors.
• Applied a backpropagation neural network model, validated with training, validation, and testing data.

Main Results:

• Females and smokers demonstrated the highest likelihood of multimorbidity.
• Age and reduced fruit consumption were significant predictors.
• The backpropagation neural network achieved higher accuracy (80.7%) compared to logistic regression (77%).

Conclusions:

• Machine learning models, specifically backpropagation neural networks, are effective in predicting multimorbidity among Saudi adults.
• Identified risk factors provide actionable insights for public health policy and intervention.
• This study highlights the potential of machine learning for epidemiological research in the Middle East.