AUTHOR=Kiplimo Richard , Kosgei Mathew , Mwangi Ann , Onyango Elizabeth , Ogero Morris , Koske Joseph 

TITLE=Longitudinal-Survival Models for Case-Based Tuberculosis Progression

JOURNAL=Frontiers in Public Health

VOLUME=Volume 9 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2021.543750

DOI=10.3389/fpubh.2021.543750

ISSN=2296-2565

ABSTRACT=Introduction: Tuberculosis (TB) continue to be responsible for a high global disease burden. An important feature presented by this disease is that individual patients are followed over a period of time and thus would present both repeated measures and survival times in addition to other baseline covariate information.
Separate analysis has been the norm when dealing with such data despite it coming from the same patient, a pointer that there could be association between them. This study aimed to investigate the association between sputum smear results and the risk of experiencing unfavorable outcome and predict survival probabilities beyond treatment duration.
Method: A joint model for longitudinal and time-to-event data was obtained by relating longitudinally measured smear results with time to experiencing unfavorable outcome. The study considered a generalized linear mixed-effects model for the longitudinal process and the Cox proportional hazards model for the survival process with baseline hazard specified using B-Splines. Inference of the parameters were obtained based on Bayesian method. The longitudinal and event submodels were assumed to be related via current value association structure. The obtained joint model was then used to predict the subject’s future risk of survival based on longitudinal outcome trajectories. Data was sourced from routinely collected TB data in Kenya between 2014 and 2016.
Results: The average baseline age was 35.10 (SD:14.82). Female constituted 34%. Patients with previous history of treatment were 8.42% while HIV co-infection was at 24.50%. Model diagnostics using MCMC chains showed longitudinal and survival submodels convergence. The association parameter was found to be 1.63. Adjusting for HIV status (HR:1.37 CI[0.79,2.39]) and history of TB treatment (HR:1.65 CI[0.75,3.62]), sex (HR:0.58 CI[0.36,0.99] was shown to be statistically significant. 
Conclusion: The obtained joint model showed strong underlying association between the sputum smear and survival data. Men are at higher risk of unfavorable outcomes.