Patient Nudges: Can Predictive Analytics Alter Patient Decisions?

Indirect suggestions like nudges have been shown to change employee behavior in a variety of employer benefit settings, ranging from selecting retirement savings plans to choosing healthcare providers based on price differences. This study takes advantage of a natural experiment to evaluate the effects of a tool that uses predictive analytics to nudge high-risk employees towards active health management.

I analyze six years of healthcare claims data on roughly 40,000 employees from a large company who manages their healthcare expenditures. In the sixth year of this dataset, the company launched healthcare nudge campaigns targeting conditions that can be actively managed to mitigate downstream costs in acute care settings. The campaigns were targeted to individual employees based upon their actual healthcare claims histories, and focused on pain management (e.g., lower back, hip, and knee), primary care services, emergency room mitigation, weight management (e.g., encouraging weight appraisals), diabetes management, encouraging colonoscopies, and cardiac care management.

The natural experiment emerged when an unexpected re-organization halted the launch of the nudge programs for about one quarter of the company. This shock was exogenous to the individual employee traits on which these programs were triggered and created control groups (comprised of individuals who would have otherwise been targeted for each of the campaigns) necessary to evaluate the effect of these nudges on employee decisions about preventative health services.

Preliminary results using an event study and differences-in-differences design and controlling for employee risk scores, various fixed effects, and group-specific time trends suggest that the program significantly increased some basic ambulatory services. These services include lower back pain related physician visits, physical therapy sessions, chiropractic care, and drug prescriptions, as well as weight appraisals and visits for diabetes management. However, these nudge campaigns do not appear to have reduced the short-run occurrences or costs of related downstream acute care services such as back injections, back surgeries, heart attacks, heart surgeries, diabetes-related hospital visits, or diabetes surgeries.