PEMODELAN M-ADAPTIVE GENERALIZED POISSON REGRESSION SPLINE PADA KASUS MDR-TB DI KALIMANTAN BARAT

Abstract

Tuberculosis is a disease caused by the Mycobacterium tuberculosis. Multi-Drug Resistant Tuberculosis (MDR-TB) is the term used to describe Mycobacterium tuberculosis that is resistant to one or more Anti-TB drugs. This study aims to determine the factors that affect the number of patients recovering from MDR-TB, by modeling the number of MDR-TB cured patients using Multivariate Adaptive Generalized Poisson Regression Spline (MAGPRS) method. The predictor variables are the average age (X1), percentage of patients who fail category 2 treatment (X2), percentage of patients who fail category 1 treatment (X3), percentage of patients relapsed (X4), percentage of patients neglecting treatment (X5), and percentage history of close contact with other patients (X6). A combination of BF (Basis function), MI (Maximum interaction), and MO (Minimum observation), the BF value is two to four times of predictor variables, MI has value of 1,2, and 3, and MO has value of 0,1,2, and 3. From the result, the best model was obtained from the combination of BF=24, MI=3, and MO=1, with GCV values of 0,3504 and R2 of 88,3%, and there are 14 BF that affect the response variable . The most influential predictors variables in a row, are X6, X3, X5, and X2.