Background:  Optimizing ART clinic treatment capacity in resource-constrained settings is essential to successful HIV treatment scale-up. Planning tools that forecast the need for medical staff, supplies and sites at local to national levels can guide policymakers as they implement scale-up. We developed an interactive decision-support tool for use in planning ART clinic design.

Methods:  We modeled a hypothetical ART clinic to determine maximal enrollment and visit capacity at steady-state operation under various resource constraints. Model components include random and scheduled patient arrival, routing to different “stations” (e.g., clerk, doctor), processing times, and re-visit intervals. Model parameters were derived from field work in Tanzania. Our clinic saw a mix of 2 patient types:  10 to 15 enrollees/day with monthly physician-based care and 20 to 25 enrollees/day with monthly nurse-based care and a physician visit every 3 months. The 10-room baseline clinic included 2 clerks, 8 counselors, and 6 each of nurses, doctors, and phlebotomists (5, 20, 6, 6, and 6 minutes/patient, respectively). We ran 100 6-month trials using the ARENA simulation package with VBA interfaces to determine clinic capacity and mean time to first turned away patient (TTFTA).

Results:  The baseline clinic model accommodated 87 patients/day at steady-state, with resource utilization ranging from a high of 81% (rooms) to a low of 30% (physicians) and a TTFTA of 59 days. Changing clinic rooms—the limiting resource—had dramatic non-linear effects on outputs (see the figure). Eliminating 3 rooms (30% reduction) decreased treatment capacity to 64 patients/day (26% reduction) and decreased TTFTA to day 30 (49% earlier). Adding 3 rooms increased treatment capacity to 107 patients/day (23% increase) and increased TTFTA to day 70 (19% later) Adding 1 staff member to all stations increased capacity by 5% to 91 patients/day but reduced the TTFTA by 1 day due to earlier over-enrollment. In contrast, varying physician staff alone by ±50% did not appreciably change outputs.

Conclusions:  We created a stochastic simulation model of an ART clinic for use in planning HIV treatment scale-up. The model predicts the dynamics of clinic staffing and patient load, and can help planners anticipate expanding clinical service needs. A disease progression module will be added to predict clinical outcomes in light of various resource use strategies.