A behavioral design approach to improving a Chagas disease vector control campaign in Peru

Individual behavior change is a critical ingredient in efforts to improve global health. While improving access to high quality health services and strengthening health systems capacity have been important foci of global health and development initiatives, in recent decades the “behavioral revolution” has focused attention on how individual behavior and decision-making drive the effectiveness of most global health programs [1, 2]. Effective, evidence-based interventions for infectious disease prevention—e.g., child vaccination, bed nets, HIV treatment regimens—often fail at the last mile due to individual and household decisions, habits, and behaviors [1, 2]. As countries undergo rapid demographic and nutritional transitions [3, 4], the role of behaviors in driving chronic disease risk is further amplified [5].

Central to the focus on behavior has been a growing understanding of how the human brain makes decisions, from motivations and mindsets to unconscious biases and cognitive shortcuts that we all use to simplify a complex world that demands attention and focus [6, 7]. Recent work in the field of behavioral economics and related fields has contributed a rich menu of insights, principles, empirical regularities and strategies that can be engineered into global health programs to attempt to increase impact and reach [8].

For example, present bias, the tendency to place more weight on costs and benefits realized today and less weight on those realized in the future, makes us prioritize immediate pleasure over actions that are in our long-term interest [9, 10]. Correctly-designed financial incentives can leverage present bias by offering immediate, tangible rewards for current health-related behaviors that may be tedious, unpleasant, or otherwise costly. Loss aversion, the tendency to value losses more highly than gains of the same face value, makes us hold onto assets that we might not be willing to pay for again [11]. An incentive structure that endows participants with up-front resources that are forfeited if behavioral targets are not met leverages loss aversion to increase impact.

In parallel with the behavioral revolution, other important trends are transforming global health research and practice. Interest is growing among the global policy community in the power (and limitations) of randomized controlled trials (RCTs) to evaluate the impact of health and development interventions—the so-called “evaluation revolution” [12]. In the past decade, many RCTs of innovative interventions informed by behavioral insights have been funded and fielded, generating policy-relevant evidence to inform future investments in global health programs. Finally, the “design for development” movement has expanded the global health practitioner’s toolkit with new ways to think about innovation, idea generation, user-defined needs, and user experience [13‐15].

Datta & Mullainathan [12] and others [16] have identified a systematic “behavioral design” approach to leveraging the potential of behavioral economics insights into health and development programs. Their framework calls for a four-step process that begins well before the design and evaluation of behaviorally-informed interventions. In theory, a rigorous approach to intervention design should improve the effectiveness of public and private-sector health programs. However, the feasibility of behavioral design as a method for global health research remains unknown. We present here our application of behavioral design to a door-to-door Chagas disease vector control campaign in Arequipa, Peru. We begin with an explanation of the campaign context and the motivation to intervene to improve household participation. Next, we describe the design phases of the study. We report both our original intervention designs as they emerged from the design process, as well as modifications that had to be made to the designs during the Test phase (a pragmatic field trial). Lastly, we discuss lessons learned for future research projects and programs using behavioral design.

Datta & Mullainathan’s [12] behavioral design approach comprises a four-step process that begins well before the design and evaluation of behaviorally-informed interventions (see Fig. 1 in Datta & Mullainathan). Similar to a product design or engineering process, behavioral design starts with problem definition: what are you trying to fix, and why? Next, a diagnosis phase identifies “actionable bottlenecks” or possible targets of intervention where psychological factors are driving behavior. Only then do the investigators proceed to intervention design, matching behavioral insights to actionable bottlenecks. In the final step, these interventions are tested, ideally in a randomized controlled counterfactual evaluation complemented by a factual evaluation of mechanisms, implementation, and participant experience [29].

Our research team is focused on a specific health-related behavior: participation in a government-run insecticide spray campaign to interrupt the vectorial transmission of Chagas disease, a vector-borne parasitic disease. While we were interested in applying principles from behavioral economics to address participation challenges, we were reluctant to show up with a behavioral economics hammer in search of some intervention nails. The behavioral design framework discussed here [12] provided a systematic methodology for defining the behaviors of interest, diagnosing reasons for household reluctance or refusal to participate, designing interventions to address actionable bottlenecks, and then testing those intervention in a rigorous counterfactual context. Behavioral design offered us a broader range of strategies and approaches than pure price changes and community education and health promotion.

Of course, we have to note that the results from our cluster randomized trial of almost 5000 households showed no significant increases in participation for our behaviorally-designed interventions compared to the control campaign [41]. We discuss possible reasons for this finding elsewhere [41], but are reluctant to infer from this one trial that the behavioral design approach is ineffective. The pragmatic, real-world setting of our study made the behavioral design approach challenging, productive, and focused. Behavioral design challenged us to do more—and more different—types of formative work than we had previously conducted in Arequipa around the Chagas campaign. The approach also demanded a new level of collaboration and integration with our partners in the Ministry of Health in order to rethink both the vector control campaign design and implementation processes. We argue that behavioral design was ultimately productive in uncovering previously unstudied explanations for household reluctance and refusal to participate in the campaign, while also keeping the intervention design focused on addressing key actionable bottlenecks rather than letting fashionable or conceptually interesting behavioral insights drive that process. We found that the complexity of an ongoing public health campaign, the interdisciplinary composition of our team, and a long candidate list of compelling and potentially relevant behavioral insights to apply in this setting required a framework that could organize and focus our work in an efficient and transdisciplinary way.

We are not the first team to employ a behavioral design approach in empirical studies. The Datta and Mullainathan framework has been applied to interventions in the agricultural [42], financial services [43‐48], WASH (water, sanitation and hygiene) [49, 50] and global health [16, 51] sectors. In addition, numerous governments (although few in developing countries) have instituted “nudge units” or “behavioral insights teams” to bring a behavioral perspective to policy design and implementation [52‐54]. It will be important, but challenging, to evaluate the marginal impact of behavioral design on program effectiveness and implementation. One potential evaluation strategy would be random assignment of behavioral design approaches across multiple organizations or sites that are developing and rolling out similar interventions.

An important question going forward is the impact of employing a behavioral design approach on internal and external validity of global health research. In our case behavioral design led to more complex interventions that addressed multiple actionable bottlenecks and employed multiple behavioral insights. Increasingly complex interventions can make interpretation and inference challenging and can make treatment fidelity hard to assess in the field. The implicit pragmatic focus of behavioral design also means that interventions developed through this process are often being deployed in real-world settings where tight control of intervention and sampling is not possible, thereby also undermining internal validity.

Behavioral design can also hinder external validity by tailoring interventions so precisely to the local context through identified actionable bottlenecks that results have limited generalizability to other settings. In our case it is unlikely that the interventions we designed, even if shown to dramatically improve participation in Arequipa’s vector control campaign, would be useful in the same form in another South American Chagas campaign or, for example, in a bed bug elimination campaign in a major US city. This tension is of course not new to the design and evaluation of global health programs and is a key theme in the evaluation revolution. Behavioral design’s contribution to generalizability may therefore lie in the process itself: The define-diagnose-design-test steps can be implemented in virtually any setting for any behaviors of interest, and, if applied rigorously, can expand the use of relevant behavioral insights to solve persistent health challenges.

A related question is how behavioral design approaches may impact scalability of interventions. Scalability refers to the ability to expand successful interventions to a broader population or larger scale [55]. Key dimensions of scalability include cost, workforce to carry out the intervention with fidelity, and interaction of the intervention with contextual factors. Our experience using behavioral design to “supercharge” an existing government-run vector control campaign has convinced us that the approach is a powerful facilitator of scalability, by encouraging strategic tailoring of large-scale interventions to local context. The definition and diagnoses phases of behavioral design can also help to identify the lowest common denominator or minimum effective dose of an intervention prior to scale.

We applied behavioral design, a four-phase approach to intervention development and evaluation, to the problem of declining household participation in a Chagas disease vector-control campaign in Arequipa, Peru. The approach helped us link problem definition and formative work on diagnosis of actionable bottlenecks to the design and testing of interventions informed by behavioral insights. While we did not follow the prescription to the letter, the framework provided useful guidance. Careful attention to how behavioral design may affect internal and external validity of evaluations, and scalability of interventions, is needed going forward. We recommend behavioral design as a useful complement to other intervention design and evaluation approaches in global health programs.