UCSF

High expectations for new technologies coexist with wide variability in the actual adoption and impact of information technology (IT) projects in clinical settings, and the frequent failure to incorporate otherwise "successful" projects into routine practice. This paper draws on actor-network theory to present an in-depth, sociotechnical analysis of one such project--a computer kiosk designed to diagnose and expedite treatment of urinary tract infections (UTI) in adult women. Research at a hospital urgent care clinic demonstrated the kiosk program's effectiveness at diagnosing UTI and reducing patient wait times, and the kiosk was subsequently adopted by the clinic for routine patient care. However, a study promoting the adoption of the device at emergency departments (ED) was characterized by persistent staff resistance and lower-than-expected patient eligibility for kiosk-assisted care. The device was ultimately abandoned at all but one of the new sites. Observations and interviews with ED staff and the design/research team were conducted at four California EDs between April and July 2011 and point to conflicting understandings of evidence for the device's usefulness and reasons for its (non)adoption. The kiosk program's designers had attempted to "rationalize" medical work by embedding a formal representation of triage practices in the kiosk's software. However, the kiosk's "network" failed to stabilize as it encountered different patient populations, institutional politics, and the complex, pragmatic aspects of ED work at each site. The results of this evaluation challenge the persistent myth that a priori qualities and meanings inhere in technology regardless of context. The design and deployment of new IT projects in complex medical settings would benefit from empirically informed understandings of, and responses to, the contingent properties of human-technology relations.