Background: Loss to follow-up (LTF) from HIV programs can present a barrier to effective evaluation of patient outcomes, such as death, and may be misinterpreted as an indicator of engagement in care. We used data from defaulter tracing for a random sample of patients LTF in one PEPFAR-supported HIV clinic to correct death estimates and estimate true disengagement from care for the larger population from which the sample was derived.
Methodology: Patients of the HIV clinic at Gatundu District Hospital in Kenya with at least 1 clinic visit between 2008-12 were assessed for LTF status using electronic data, with LTF defined as no visit in the past 3 (ART) or 6 (pre-ART) months and not dead or transferred-out. Of these, a random sample, stratified by pre-ART and ART status, was selected for study tracing. Tracers tracked patients and completed a questionnaire for patients or contacts found, collecting information including patient vital status (alive/dead) and engagement in care (in care/disengaged). Rates and percent dead and in care were presented using the initial clinic data and data updated with outcomes from defaulter tracing, weighted to represent all patients LTF. Patients reported by a contact as alive were classified as in care (optimistic scenario) or disengaged (pessimistic scenario) in separate analyses.
Results: 413 (21%) of the 1,974 clinic patients were LTF, of which 66 (16%, 40 pre-ART, 26 ART) were sampled. Questionnaires were completed for 65 (98%) patients (46 patients, 19 contacts). Seven (18%) pre-ART and 6 (23%) ART patients reported being disengaged from HIV care. Nine (23%) pre- ART and 5 (19%) ART patients were reported to have died. Updating the initial data with sample outcomes increased overall retention in care for pre-ART patients from 61% to 69% (pessimistic) or 75% (optimistic) and for ART patients from 84% to 88%; deaths increased from 10% to 18% for pre-ART patients and from 8% to 10% for ART patients. Incidence rates for death were lower and for LTF/disengagement substantially higher in the initial data than in the updated data, for both pre-ART and ART patients (see table).
Conclusions: This suggests that only a minority of patients classified as LTF by one large HIV clinic in Kenya are actually disengaged from HIV care. Also, death rates were underestimated. While our data are from a single facility, these findings suggest that high levels of LTF observed in routinely-collected data may be a poor proxy for disengagement from care and obscure higher death rates.