Incentives versus Defaults: Cost-Effectiveness of Behavioral Approaches for HIV Screening

Many HIV positive individuals are still unaware of their infected status, which has led health systems to try many alternative approaches to expand HIV testing.  The CDC, US Preventive Service Task Force, and others have encouraged the use of opt-out default screening regimens for patients, while some systems have implemented small incentives to increase testing uptake.  We conducted a two-year randomized trial in the emergency room of a large safety net hospital to directly test the effectiveness of alternative default choices against the effectiveness of small incentives, as well as to test the degree to which defaults and incentives were complements or substitutes.  In separately reported results, we found that defaults and incentives each had substantial effects on raising HIV testing rates.  For example, with zero incentives, changing from opt-in to active choice raised testing rates from 38% to 51%, and changing to opt-out achieved a 66% rate.  Small incentives by themselves could raise testing rates by as much as 20%, depending on the risk-group and default scenario.  In the current paper we build on this analysis to quantify the cost-effectiveness of different combinations of defaults and incentives.  The main benefit metric is the incremental number of new HIV diagnoses, estimated from a mathematical screening model exploiting survey information on the risk level of each patient.  We also modeled the relative effectiveness in terms of future HIV infections averted. We find the highest incremental cost-effectiveness ratios (ICERs) when moving from opt-in to active choice or opt-out.  When starting from a base case of opt-in with no incentives, changing the defaults dominates adding small incentives in terms of cost per new diagnosis.  However, the effectiveness of the opt-out regimen is further enhanced by additional small incentives, though at a greater marginal cost per diagnosis. These results hold across a wide range of alternative parameter values.  We suggest a research agenda to similarly compare competing incentive versus choice architecture effects for a wide range of behavioral change efforts in other domains.