Multinomial logistic regression modeling of effect of some selected demographic factors on patients’ blood sugar level

Abstract

In the world today, a good number of people are affected by blood sugar irregularities, which is often characterized by imbalance level of glucose (sugar) in the blood, resulting from the body’s insulin issues. In this work, data on risk factors of blood sugar irregularities, which include age, gender, body mass index (BMI), blood group, and genotype, were collected on 464 patients from a hospital in Abaji, Abuja. Then the probability of the patients’ blood sugar level falling in either low, normal or high category based on the effect of these risk factors was modelled using multinomial logistic regression (MNLR) technique. The significance of the effects of each of the risk factors on the modelled probability of the patients’ blood sugar level was assessed. Pearson and deviance techniques for MNLR were used to assess the fitness of the model. Overall significance of the predictors was then assessed using Likelihood Ratio Test (LRT) technique. Lastly, the MNLR predicted probabilities for certain levels of the risk factors were estimated and plotted against Age and BMI.  The results revealed an odds ratio of 1.022 for the AGE factor, indicating a 2.2% higher odds of low sugar level than normal sugar level for every unit increase in the patient’s age. For the BMI factor, the odds ratio was 1.055, implying a 5.5% higher odds of low blood sugar level than the normal level for every unit change in the patient’s BMI. The same result is true for other demographic factors in this category. The graphs show that the odds of having normal blood sugar level falls slowly with each decade of life and sharply as the BMI increases. It was also observed that the probability of maintaining a low blood sugar level decreases as age increases, while it increases as the BMI increases.