Publications

INTRODUCTION

Current estimates of retention among HIV-infected patients on antiretroviral therapy (ART) in Africa consider patients who are lost to follow-up (LTF) as well as those who die shortly after their last clinic visit to be no longer in care and to represent limitations in access to care. Yet many lost patients may have "silently" transferred and deaths shortly after the last clinic visit more likely represent limitations in clinical care rather than access to care after initial linkage.

METHODS

We evaluated HIV-infected adults initiating ART from 1/1/2004 to 9/30/2007 at a clinic in rural Uganda. A representative sample of lost patients was tracked in the community to obtain updated information about care at other ART sites. Updated outcomes were incorporated with probability weights to obtain "corrected" estimates of retention for the entire clinic population. We used the competing risks approach to estimate "connection to care"--the percentage of patients accessing care over time (including those who died while in care).

RESULTS

Among 3,628 patients, 829 became lost, 128 were tracked and in 111, updated information was obtained. Of 111, 79 (71%) were alive and 35/48 (73%) of patients interviewed in person were in care and on ART. Patient retention for the clinic population assuming lost patients were not in care was 82.3%, 68.9%, and 60.1% at 1, 2 and 3 years. Incorporating updated care information from the sample of lost patients increased estimates of patient retention to 85.8% to 90.9%, 78.9% to 86.2% and 75.8% to 84.7% at the same time points.

CONCLUSIONS

Accounting for "silent transfers" and early deaths increased estimates of patient retention and connection to care substantially. Deaths soon after the last clinic visit (potentially reflecting limitations in clinical effectiveness) and disconnection from care among patient who were alive each accounted for approximately half of failures of retention.

David Dowdy and colleagues discuss the complexities of costing new TB diagnostic tests, including GeneXpert, and argue that flexible analytic tools are needed for decision-makers to adapt large-sample cost-effectiveness data to local conditions.