Modelling binary logistic regression on natural toxin poisoned patients in Malaysia

Unintentional poisoning is on the rise globally and thus a statistical analysis has been conducted. This study focuses on implementing a binary logistic regression analysis to study the factors that lead significantly to unintentional and intentional poisoning in Malaysia. The data consists of information from 592 patients with 7 predictor variables and a binary response variable. Preliminary analysis and chi-square test are conducted before performing the analysis. To obtain a good logistic model, there are three main steps: model building, diagnostic checking and drawing inferences from the parameters. The model provides parameter inferences such as odds ratio and estimation of predicted probabilities. Based on the results, unintentional poisoning is significantly influenced by physical form of poison, type of natural toxin and exposure route. Odds ratio estimation indicates that patients exposed to poisoning through bites or stings are more likely to have unintentional poisoning compared to other exposure routes. Generally, estimation of predicted probability are high for unintentional poisoned patients when they are bitten by animals.