Intra-cluster correlation estimates for HIV-related outcomes from care and treatment clinics in Dar es Salaam, Tanzania

Abstract

Introduction Researchers planning cluster-randomized controlled trials (cRCTs) require estimates of the intra-cluster correlation coefficient (ICC) from previous studies for sample size calculations. This paper fills a persistent gap in the literature by providing estimates of ICCs for many key HIV-related clinical outcomes. Methods Data from HIV-positive patients from 47 HIV care and treatment clinics in Dar es Salaam, Tanzania were used to calculate ICCs by site of enrollment or site of ART initiation for various clinical outcomes using cross-sectional and longitudinal data. ICCs were estimated using linear mixed models where either clinic of enrollment or clinic of ART initiation served as the random effect. Results ICCs ranged from 0 to 0.0706 (95% CI: 0.0447, 0.1098). For most outcomes, the ICCs were large enough to meaningfully affect sample size calculations. For binary outcomes, the ICCs for event prevalence at baseline tended to be larger than the ICCs for later cumulative incidences. For continuous outcomes, the ICCs for baseline values tended to be larger than the ICCs for the change in values from baseline. Conclusion The ICCs for HIV-related outcomes cannot be ignored when calculating sample sizes for future cluster-randomized trials. The differences between ICCs calculated from baseline data alone and ICCs calculated using longitudinal data demonstrate the importance of selecting an ICC that reflects a study’s intended design and duration for sample size calculations. While not generalizable to all contexts, these estimates provide guidance for future researchers seeking to design adequately powered cRCTs in Sub-Saharan African HIV treatment and care clinics.

The impact of continuous quality improvement on coverage of antenatal HIV care tests in rural South Africa: Results of a stepped-wedge cluster-randomised controlled implementation trial

“…We further assumed that half of all pregnant women would be HIV positive and that pregnant women would make 3 ANC visits. We assumed an intracluster correlation coefficient (ICC) of 0.10, which is a conservative assumption compared to other ICCs measured in similar settings [43]. We assumed missing data from 15% of enrolled women.…”

Community delivery of antiretroviral drugs: A non-inferiority cluster-randomized pragmatic trial in Dar es Salaam, Tanzania

“…We assumed an intra-cluster correlation coefficient (ICC) of 0.03. This assumption was based on a study in Dar es Salaam [ 18 ], which found the healthcare facility ICC value for the 6-month cumulative incidence of non-adherence to ARVs (defined as a 50% drop in CD4 cell count from its peak value and return to pre-ART CD4 cell count or lower after 168 days on ART, or a VL greater than 10,000 copies/ml after 168 days on ART) to be 0.016. We took the upper bound of the 95% CI of this estimate (which was 0.03) as a conservative estimate of the ICC for our primary endpoint.…”

The impact of community health worker-led home delivery of antiretroviral therapy on virological suppression: a non-inferiority cluster-randomized health systems trial in Dar es Salaam, Tanzania

“…We used a range of intra-cluster correlation coefficients (ICCs) from 0.005 to 0.020. Barnhart et al calculated ICC values for CD4-cell count measures in Dar es Salaam [ 34 ]. The 6-months cumulative incidence for non-adherence to ART (defined as a 50% drop in CD4-count from its peak value and return to pre-ART CD4-count or lower after 168 days on ART or a viral load greater than 10,000 after 168 days on ART) had an ICC value of 0.016 (95% confidence interval: 0.009 to 0.029).…”