Evaluating Alternative Designs of a Multilevel HIV Intervention in Maharashtra, India: The Impact of Stakeholder Constraints

Abstract

Background. Multilevel interventions combine individual component interventions, and their design can be informed by decision analysis. Our objective was to identify the optimal combination of interventions for alcohol-using HIV+ individuals on antiretroviral drug therapy in Maharashtra, India, explicitly considering stakeholder constraints. Methods. Using an HIV simulation, we evaluated the expected net monetary benefit (ENMB), the probability of lying on the efficiency frontier (PEF), and annual program costs of 5,836 unique combinations of 15 single-focused HIV risk-reduction interventions. We evaluated scenarios of 1) no constraints (i.e., maximize expected value), 2) short-term budget constraints (limits on annual programmatic costs of US$200,000 and $400,000), and 3) a constraint stemming from risk aversion (requiring that the strategy has >50% PEF). Results. With no constraints, the combination including long individual alcohol counseling, text-message adherence support, long group counseling for sex-risk, and long individual counseling for sex-risk (annual cost = $428,886; PEF ∼27%) maximized ENMB and would be the optimal design. With a cost constraint of $400,000, the combination including long individual alcohol counseling, text-message adherence support, brief group counseling for sex-risk, and long individual counseling for sex-risk (annual cost = $374,745; PEF ∼4%) maximized ENMB. With a cost constraint of $200,000, the combination including long individual alcohol counseling, text-message adherence support, and brief group counseling for sex-risk (annual cost = $187,335; PEF ∼54%) maximized ENMB. With the risk aversion constraint, the same configuration (long individual alcohol counseling, text-message support, and brief group counseling for sex-risk) maximized health benefit. Conclusion. Evaluating the costs, risks, and projected benefits of alternatives supports informed decision making prior to initiating study; however, stakeholder constraints should be explicitly included and discussed when using decision analyses to guide study design.

Keywords: HIV; constrained optimization; decision analytic modeling; economic evaluation; study design.

Conflict of interest statement

The following author is employed by the sponsor: Kendall Bryant. The other authors have not conflicts of interest to disclose.

Figures

Figure 1

Simulation structure and dynamics. In order to simulate interventions that had individual patient benefits and population benefits, this analysis involved an individual-level microsimulation combined with a compartmental dynamic transmission model. Further details of the simulation calculations and logic can be found in the technical appendix.

Figure 2

Optimal options considering programmatic budget constraints. (A) The optimal options with a 1-year program cost below $200,000. If a funder perceived uncertainty in future health care budgets, they may impose a restriction on what to study based on annual program costs. (B) The optimal options with 1-year program cost below $400,000. A constraint on annual spending is a manifestation of an implicitly high discount rate. It may be driven in part by uncertainty in future sources of financing. (C) Under no cost constraint, all five options are considered. These five configurations were the most efficient, but each had different programmatic costs and probability of being most efficient. Decision analysis eliminated 5,815 of the 5,836 options, leaving 22 choices to consider.

Figure 3

Optimal options considering a risk-averse decision maker. (A) The optimal options with a constraint of having at least a 50% chance of being on the efficiency frontier would leave two of the five options. (B) Decision makers may have alternative levels of risk requirements, so presenting the risk along with the expected value can make the tradeoff between more extensive intervention packages and risk of inefficiency relative to less intense intervention packages more explicit.