Abstract Body

Background: While the individual-level benefits of antiretroviral therapy (ART) are well established, much remains to be understood about the population-level impacts of ART scale-up in Kenya and other Sub-Saharan African countries. Most knowledge of mortality levels and determinants comes from relatively small or unrepresentative study sites that may not accurately reflect full national programs. In this study we analyze data from a large sample of Kenyan ART facilities, representing the general population receiving ART.

Methods: We randomly extracted between 50 and 250 adult ART charts at a nationally-representative sample of Kenyan facilities. The charts were paired with a survey of facility resources and practices. A competing risk analysis was used to calculate crude cumulative mortality and loss to follow up (LTFU) after 6, 12 and 24 months on treatment. Adjusted mortality was also calculated for the same time points, which included a correction to account for death among LTFU patients. The correction was estimated at the facility-level, based on a meta-analysis of prior literature demonstrating the relationship between total percent LTFU and the percent of those that had died. Additional analyses examined cumulative mortality stratified by sex, and initial CD4. Regression analysis was done at the facility-level to assess predictors of adjusted mortality rates.

Results: We extracted charts from 16,015 adult ART patients across 63 facilities. The crude cumulative mortality rate was 7%, 9% and 10% at 6, 12 and 24 months respectively. Cumulative LTFU for the same points was 22%, 27%, and 34%. After applying the LTFU correction, adjusted cumulative mortality was 14%, 17% and 18% at 6, 12 and 24 months. Public and rural facilities had significantly higher mortality rates, after controlling for the demographic and clinical composition of their patient populations. Other program characteristics, including nurse-led treatment, were not significant predictors.

Conclusions: The adjusted mortality rates contribute important information about outcomes for a large, unbiased sample of Kenyan patients. Differences in mortality by location and ownership suggest the presence of inequities affecting rural and public clinics. The lack of significant differences across other dimensions indicates that comparable care may be possible with cost-saving changes such as task-shifting.