Survival modelling of tuberculosis treatment: Insights from Kaplan-Meier and Cox proportional hazard models

The study explores the crucial role of statistics in data science and analytics, particularly in statistical concepts and methods from the basis of data analysis, machine learning, and decision-making. The importance of statistical thinking in extracting insight from data and making informed decisions is also highlighted. Real-life survival data consisting of 203 tuberculosis patients with six variables, including age, gender, marital status, treatment (drug), censored (outcome), and time in days, is used to validate the study. Different analyses are carried out using R and Python software to depict summary statistics, plotting raw data, histograms, and percentages in pie plots, survival times for four different treatments include: Rifampicin, Isoniazid, Levofloxacin, and Bedaquiline, CoxPH model is fitted, and the Kaplan Meier survival curve for the treatment. The results reveal that Bedaquiline is given to 3 patients with no right-censored patient recorded, Levofloxacin was administered to 4 patients, only 1 is right censored, Isoniazid is administered to 36 patients, 8 are right censored, and Rifampicin was applied to 160 patients, and 8 are right censored. However, Rifampicin exhibits the highest survival rate, followed by Isoniazid, while Levofloxacin and Bedaquiline indicate greater uncertainty through their curves and shaded regions in the survival curve.