Using Binary Logistic Regression for Categorical Predictors to Assess the Likelihood of Counterfeit Medicine in Nigeria

• Activity Number: 287 - Contributed Poster Presentations: Government Statistics Section
• Type: Contributed
• Date/Time: Tuesday, August 9, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Government Statistics Section
• Author(s): Oluwole Adegoke Nuga* and Akintunde Azeez
• Companies: Bells University of Technology and NAFDAC
• Keywords: Binary Logistic Regression; NAFDAC; TruScan; odd ratio ; Lemeshow test

Abstract

This work attempts to evaluate the likelihood of counterfeit medicines in Nigeria by using a Binary Logistic Regression (BLR) with three categorical predictors. Central Laboratory (CL) test results of 4263 medicines comprising of antimalaria, antibiotics, antidiabetic and others (antihypertensive, antidepressant), were collected from the National Agency for Food and Drug Administration and Control (NAFDAC) with the corresponding TruScan results as well as zones where medicines were sourced. Chi-square was used to test for multicollinearity prior to fitting the BLR model. The estimated parameters' statistical significance was determined, and the Lemeshow test was used to assess model prediction capability. No multicollinearity was present and the odd ratio for TruScan was 76.6, indicating a high likelihood of the device returning an accurate CL result. The results also showed higher odd of antimalaria being counterfeited among drug types as well as a higher possibility of fake medicines from southwestern region of the country. Lemeshow value (9.91) with a p-value of 0.19 showed insignificance at 5% denoting that the fitted model has high prediction capability.