Causal Outcomes: 25% Reductions in Emergency Room Visits Implementing Behavioral Economics and Predictive Analytics in a Targeted Medicaid Population
Can overuse of emergency departments (EDs) be significantly reduced in Medicaid populations without intensive and costly programs, such as individual counseling of members by physicians or nurses, to change the behaviors of members having a history of avoidable uses? Can overuse of emergency departments (EDs) be significantly reduced in Medicaid populations without intensive and costly programs, such as individual counseling of members by physicians or nurses, to change the behaviors of members having a history of avoidable uses?
This paper describes the results of deploying a low-cost, highly scalable software platform, NextNudgeTM, which uses machine-learning techniques to target “nudges” – carefully crafted messages that apply principles of behavioral economics to elicit desired changes in behavior – to those members who are predicted to be most likely to respond to them. The platform was used to identify members to nudge, such as relatively healthy members with no recent wellness visits but with a history of recent ED use, and to track the results of nudging them. Nudges consisted of a combination of live outbound calls and follow-up direct mail offering information about a free Nurse Advisory Line (NAL) and the availability of nearby primary care physicians (PCPs) with convenient locations and hours; a refrigerator magnet with the NAL free telephone number was also included in the direct mail. The difference in subsequent ED visit rates (ED visits per member per quarter) between nudged members and similar non- nudged members were used to determine the causal impact of the nudges on ED visits.
Within six months of deploying the NextNudgeTM platform, the average ER visit rate and costs had dropped by roughly 26% and 39% respectively among nudged compared to historical averages among the target populations – a highly statistically and economically significant reduction. These preliminary results suggest that relatively inexpensive, information-based nudges can be very effective in changing member behaviors when they are targeted to members who are most likely to respond.
