Abstract
We document that interactions with manipulated Clinical Decision Support (CDS) systems can induce not only short-term but also long-term changes in physicians' opioid prescribing behavior. Physicians in our sample adopted electronic health record software from a list of federally certified vendors in 2011. Between 2016 and Spring 2019, one vendor secretly embedded a biased CDS function designed to promote extended-release opioid sales. Affected physicians not only increased opioid claims relative to the control group during the treatment window but also maintained a higher propensity to prescribe opioids even after the biased function was removed. This long-term behavioral change persisted even after affected physicians moved to new locations, changed their affiliations, or faced stricter state-level opioid regulations. Increasing physician awareness helped mitigate this impact. Using machine learning algorithms, we estimate that decision-making distortion accounts for approximately 54\% of the treatment effects in a physician decision model with dynamic learning.