BINOMIAL LOGISTIC REGRESSION OF INCIDENCE OF CARDIOVASCULAR DISEASE AMONG STAFF IN HIGHER INSTITUTION

Abstract

Cardiovascular disease is a cluster of diseases and injuries that affect the heart and blood vessels of people across the globe. This study evaluates the incidence of the disease among 200 randomly sampled staff of the Federal Polytechnic, Ile-Oluji, Nigeria from existing records of Health Services Unit of the Institution. Variables include Age, Gender, Weight, Pulse Rate, Heart Disease Condition and Maximal Aerobic Capacity (V02 Max). The logit transformation of continuous dependent variables (Age, Weight and V02 Max) reveals that they are linearly related to the dependent variable (Heart Disease Status). Evaluating the observed and predicted classification with respect to staff heart disease condition, the logit transformation: correctly classifies 84.0% of the overall cases with the addition of the independent variables; correctly predicted 97.6% of staff not having the disease; correctly predicted 66.7% cases with the observed characteristic compared to the total number of cases predicted as having the disease; and correctly predicted 85.1% cases without the observed characteristic compared to the total number of cases predicted as not having the disease. Chi-square test of independence reveals that age and gender are dependent on the heart disease status. Wald test reveals that Age (P-value = 0.001) and Weight (P-value =0.030) added significantly to the model, but Gender (P-value = 0.069) and VO2 Max (P-value = 0.235) are not significant. Hence, prediction is made such that the probability that a staff will have heart disease on a one-unit change is 2.602 times greater for males compared to females.