Bootstrapping of compliance to anti-Tuberculosis medication: South Africa, 2014

Abstract

Background: Although, Tuberculosis (TB) is curable if the treatment is adhered to and completed it is still a major cause of death globally including South Africa. The success rate for TB treatment was 77.2% in 2014, of which more than 37 000 lives were lost because of it in South Africa. Several studies have been carried out on this subject, but the difference between the present study and the previous work done is the methodology proposed to establish the determinants of anti-TB medication compliance. Understanding the determinant of anti-TB medication compliance will help the policymakers on the appropriate decision to reduce the menace of the disease.

Methods: In this study, we proposed logistic regression to a sample of individuals taken from the National Income Dynamics Survey data that self-reported to have been TB diagnosed. A comparison of the classical logistic regression and parametric bootstrap estimation methods was done for this data to determine the model that best describes the data.

Results: The results obtained from the two methods were similar and identified gender, language, alcohol, English literacy, belief in religion and household SES as the determinants of TB patients on medication. The standard errors for the bootstrap logistic model were bigger than the standard errors of the classical model.

Conclusion: We conclude that the classical model is better and for this scenario, there was no need to resample. The outcome of this study supports the existing findings that controlling the social and economic determinants of health will help eradicate TB.

Keywords: Anti-TB medication compliance; Bootstrapping; Logistic regression; South Africa; Tuberculosis.