Vigilant Algorithms and the Contagion of Risk in Automated Opioid Risk Assessment.

Amid the ongoing opioid crisis, U.S. health systems have extended prescription monitoring interfaces with proprietary reports that provide clinicians with algorithmic risk scores to guide opioid decisions. However, little is known about how such tools shape clinical encounters, professional authority and accountability in everyday care. This article examines the case of NarxCare, a widely adopted risk-scoring system, through a multi-sited ethnographic study combining clinician interviews, interface walkthroughs, documentary analysis and clinic observation. We find that NarxCare operates through multiple opacities, presenting scores as authoritative yet unexaminable. It produces an amalgamated risk by collapsing diverse concerns-overdose, liability and societal harm-into a single number. Through its logic of contagion, physicians themselves become risk subjects, their reputations and careers tethered to the scores of patients they treat. These dynamics illustrate how this algorithmic automation recasts prescribers as governable agents of the opioid crisis; however, by unsettling established epistemic cultures of safety, it fails to completely enrol them as actuarial enforcers. In practice, clinicians respond by refusing, anticipating, or domesticating the scores, but cannot evade them fully. NarxCare thus reconfigures rather than resolves opioid risk, a dynamic with significant implications as algorithmic infrastructures expand across healthcare and scoring enters new domains.