1. Describe the study design you will employ in order to determine if your intervention has had an effect on the outcome variable of interest.
Though I considered a cluster RCT design, I ultimately decided that I do not have access to a large enough sample size to accomodate this design. So, I have decided to randomize at the individual level instead.
2. Define the unit-of-analysis for your main outcome evaluation, the minimum meaningful effect size, and the sample size necessary to detect this effect size.
Here are the sample size calculations I put together for the study protocol:
Sample size calculations were generated based on preliminary data from the Mfangano Island Healthy Network Impact Study (MIHNIS). The MIHNIS study was a quasi-experimental study with the intervention delivered to one community and neighboring communities acting as control groups.
During the six months prior to delivery of the intervention in the MIHNIS study, 151 of 407 (37%) patients on ART at the Sena Health Center missed a clinic visit by ≥28 days. Of these, 133 (88%) joined the study. Among the control group in the MIHNIS study, 23 of 86 (27%) participants who missed a clinic visit by 28 days during the 6 months enrollment period had a 90-day clinic absence within the next 12 months. Among patients eligible for the intervention and who missed a clinic visit by ≥28 days during the enrollment period, 31 of 52 (60%) actually participated in the microclinic intervention. In the MIHNIS study participants in the intervention community had a more than 50% lower rate of 90-day disengagement than those in the intervention community, regardless of intervention uptake.
Thus, we assume an event rate of 0.27 in the control group and 0.135 in the intervention group. Setting alpha to 0.05 and beta to 0.2 and assuming an average of 2 participants per group with a coefficient of variation of 0.25, we will need 142 patients per study arm. This sample size accounts for non-independence of outcomes introduced by the clustered nature of the intervention. Allowing for 10% participant attrition, we plan to sample 160 participants per study arm. The MIHNIS study ascertained outcomes for 98.6% of study participants, thus the allowance for 10% attrition in the proposed study is conservative.
The accessible population for this study includes all patients in HIV care and treatment, regardless of ART status, at the Sena (excluding patients residing in the East sub-location), Yokia, Ugina, Nyakweri, Wakula, Soklo, Remba and Ringiti Health Centers. Excluding patients residing in the Mfangano East sublocation from the total in Table 1, results in approximately 1500 current adult patients in these clinics. If we assume that 40% will be eligible for our study and that 60% will be willing to participate in the study, we will have an accessible population of 360 or 180 per arm. Thus, because the accessible population is larger than the estimated sample size, we will extend our sample size to include up to 180 participants in each arm to allow for errors in these sample size calculations.