PO 8417 RISING TRENDS IN TB MORTALITY AMID DECLINE IN CASES NOTIFIED IN A RURAL COUNTY IN KENYA: COHORT STUDY

Abstract

BackgroundDespite introduction of rapid and accurate diagnostic tools and aggressive treatment for tuberculosis (TB), it is still a global health problem. In 2016, globally, 1.7 million people died of TB, 95% from resource-poor countries. This study aimed to estimate changing trends in all-cause mortality rate and identify features associated with mortality among suspected TB patients on treatment.MethodsA cohort study of patients registered in a TB surveillance system from 2012 to 2016 and followed up for six months during TB treatment. The outcome was all-cause mortality within six months of TB treatment. The exposures examined were demographic and clinical features at the time of starting TB treatment.ResultsA total of 10,717 participants, median (IQR) age 33 (24–45) years, of which 3163 (30%) were HIV-infected were included in the analyses. During follow-up of 5175.5 person-years (PY), 585/10,717 (5.5%) participants died; mortality rate was 12.2 (95% CI 11.3,13.3) deaths per 100PY. The yearly mortality rate increased from 7.79 (95% CI 6.35, 9.54) in 2012 to 17.73 (95% CI 14.93, 21.06) in 2016 per 100PY (Ptrend <0.001) but the number of suspected-TB notifications declined from 2610 (24%) in 2012 to 1689 (16%) in 2016 (Ptrend=0.02). 77% of all deaths occurred by month three. Mortality among HIV-infected participants was higher (325/3163; 10.3%) than among HIV-non-infected participants (251/7413; 3.4%; p<0.001). Old age, being a female, type of TB diagnosis used, body mass index (BMI) <18.5, HIV status and year of diagnosis were associated with mortality in the multivariate regression model.ConclusionThis large population level TB study identifies an alarming trend of patients dying within months of starting treatment. These early deaths could be due to late diagnosis and multidrug-resistance. The study warrants further investigation to go beyond already established indicators which remained constant (including HIV co-infection), to explore host, disease or health system related factors that may explain the observed trend.