Predictors of unfavorable tuberculosis treatment outcomes: prospective cohort study among notified cases in Vihiga County, Kenya

Abstract

Each year, tuberculosis (TB) causes an estimated 10 million illnesses and 1.4 deaths globally. Kenya, one of 30 high-TB burden countries, has made progress in identifying more TB cases, but achieving the global treatment success rate target of >90% has remained a challenge. With TB mortality rate of 13% and treatment interruption rate of >5%, Vihiga is one of the counties with highest rates of unfavorable TB treatment outcomes in Kenya. Understanding their predictors is critical for improving TB epidemiology and advancing the global zero TB epidemic targets. A prospective cohort study was conducted among 291 notified TB patients from 20 health facilities in Vihiga County to: describe distribution of TB disease and treatment outcomes by patients’ characteristics; identify determinants of treatment interruption; and investigate factors associated with survival distribution and the occurrence of mortality. Baseline and follow-up data were gathered using questionnaires, while qualitative data was obtained using treatment interruption tracing form and the mortality audit tool. Patients’ demographic, socioeconomic, behavioral, and clinical characteristics were summarized descriptively, while probabilities of treatment completion, survival, and event-time intervals were estimated using Kaplan-Meier estimator. Log-rank test was used to quantify statistical differences in survival probability based on univariable patients’ characteristics. Cox proportional hazard model was fitted to identify determinants of TB treatment interruption and factors associated with the occurrence of all-cause mortality through the calculation of hazard ratios (HR) at 95% Confidence Intervals (CI) and p ≤ 0.05. Qualitatively, reasons for treatment interruption were identified thematically while causes of mortality, and associated circumstances were categorized by HIV status. Of the 291 patients, 72% were male, and nearly half were aged 25-34 years (23.4%) and 35-44 years (23.4%). During follow-up, 32 (11%) patients interrupted treatment while 45 (15%) died. Higher incidences of treatment interruption (59%, p <0.001) and mortality (78%, p<0.001) occurred during the intensive phase of treatment. Alcohol consumption (HR = 9.2, 95% CI; 2.6–32.5, p< 0.001); being female (HR = 5.01, 95% CI; 1.68–15.0, p = 0.004) and; having primary or lower of education level (HR = 3.09, 95% CI; 1.13–8.49, p < 0.029) increased risk for treatment interruption, while having a treatment supporter (HR = 0.33, 95% CI; 0.14–0.76, p = 0.009) was protective. The main reasons for treatment interruption were feeling better soon after beginning treatment and alcohol use. Severe illness (HR = 5.06, 95% CI; 1.59–16.1, p = 0.006); HIV coinfection (HR = 2.56, 95% CI; 1.28–5.12, p = 0.008); comorbidities (HR = 2.72, 95% CI; 1.36–5.44, p = 0.005); and smoking (HR = 2.79, 95% CI; 1.01–7.75, p = 0.049) were associated with increased risk of occurrence of mortality. Mortality among HIV-negative patients was ascribed to lung complications while advanced HIV disease was the leading cause of mortality among the HIV-positive. This study's findings indicate that interruption of treatment and mortality from TB remain problematic and occur early in the treatment. The study also acknowledges the important role person's characteristics play in the distribution of TB disease and predicting unfavorable treatment outcomes. Comprehensive public health interventions using multisectoral approach may help reduce rates of unfavorable TB treatment outcomes.