UC Berkeley

The epidemiological transition has resulted in an increased burden of non-communicable diseases throughout lower- and middle-income countries (LMICs). As countries adapt to this new environment, they must find cost-effective ways of screening for and treating these diseases. With the proliferation of mobile technology as well as the increased evidence for the effectiveness of cash transfers, a randomized-controlled trial in rural Tanzania was conducted to examine the usage of an SMS-based approach of informing individuals of the dangers of non-communicable diseases and offering financial incentives for hypertension screening. The study originally included 961 participants randomized into two treatment arms and a control arm across three villages. The treatment arms received text messages notifying them of the importance and availability of screening at their local dispensary. One treatment arm also received two thousand Tanzanian shillings if they came in for a visit. The study was rolled out to the three villages over the course of seven weeks. Beyond estimating the overall impact of this program and the utility of negative prices, the study design enabled the measure of spillover to individuals who were not initially enrolled in the experiment. Further, information on social networks was leveraged to estimate the impact of not being treated but knowing someone who was. Without accounting for spillover when analyzing the impact of the intervention, impact is significantly underestimated. The results that lack any control for network or spillover effects show a six to eight percentage point change in screening behavior, whereas accounting for them shows a thirty percentage point impact. This effect is largest for those in the cash arm who knew other people in the same treatment arm. There is also a negative effect on screening for individuals in the cash arm who knew people in the information arm. Therefore, the study suggests both the potential viability of using SMS-based approaches with cash transfers to increase non-communicable disease screening as well as a need to better understand the network effects associated with such a program. Importantly those designing randomized controlled trials should account for the possibility of network interactions when selecting their samples.