Loss to follow-up (LTFU) in Zambia’s national treatment program threatens progress toward achieving HIV epidemic control. With CDC/PEPFAR funding, the Centre for Infectious Disease Research in Zambia (CIDRZ) supports the national program in Lusaka Province where a drop in 12-month retention was noted from 76% in FY17 to 65% in FY18. The recent CIDRZ BetterInfo study used a multi-stage sampling approach to generate revised, regionally representative estimates for mortality and LTFU in the national program. We adapted the study methodology to create a field ready “BetterInfo lite” sampling approach to ascertain true program status for patients apparently “LTFU” and drivers of retention decline.
Using routine data, we selected 10 high-volume facilities in Lusaka for “BetterInfo Lite” based on the largest net drop in program retention between Q3 and Q4 FY18. We randomly selected 15% of newly “LTFU” clients between Q3 and Q4 in each facility to be traced by phone or in person and to complete a brief vital status and retention questionnaire. Leveraging existing CIDRZ platforms, trained peer educators contacted patients or their contacts to determine patient mortality and true program status (i.e. alive in care, transferred, disengaged from care, dead, or LTFU).
1,023 of 6,968 LTFU patients (15%) were randomly selected for tracing. Tracing was attempted in 1,001 (98%), with 604 of these (80%) reached by phone or in person. 397 (40%) could not be reached. Of the 604 contacted, 491 (81%) were found alive and in care at their original clinic or a neighboring satellite health post, 80 (13%) were “silent” transfers, 17 (3%) disengaged from care and 16 (3%) were deceased (figure 1).
It is feasible to adapt a rigorous sampling-based strategy used in a research context for routine program use to revise HIV treatment program retention estimates. We observed a high proportion of patients alive in care at the facility where they were flagged as “LTFU”, suggesting data quality issues, likely due to increasingly decentralized ART distribution and data collection. We also observed numerous “silent” transfers after ART initiation and identified 16 missed deaths, resulting in a mortality underestimation and LTFU overestimation. Scale up of electronic data systems to decentralized ART dispensation points, use of unique identifier tools such as biometrics, and enhanced early patient support and follow-up are needed to improve, and better monitor, program retention.