An Effective Random Generalised Linear Model to Predict COPD

Chronic obstructive pulmonary disease (COPD) is a type of chronic lung illness that worsens with time and leads to a restriction in the outflow of air from the lungs. According to the World Health Organisation, The World Health Organization ranks COPD as the third leading cause of death. Clinically, the diagnosis of this disease is relatively difficult; therefore, early identification of individuals at risk of developing COPD is vital for implementing preventative strategies. This research work has developed a generalised linear model (GLM) to predict the COPD status of the patients. A dataset of 1262 patients (688 COPD cases and 574 controls) was used. Exploratory data analysis (EDA) was utilised to observe how potential covariates were related to the response variable (COPD status). By employing rigorous model selection techniques (forward selection and backwards elimination) according to (AIC) which stand from Akaike information criterion and (BIC) which stand from Bayesian information criterion (BIC), a consensus was reached that the most suitable model is a binomial logistic regression model which includes the smoking history, gender, and age. The model was validated using an independent test set with an accuracy of 73%. Such a model, once fully validated, has the ability for predicting the risk of developing COPD in patients with existing lung conditions, including but not limited to, asthma.