Paper short abstract:

This paper investigates an emergent medical and technical innovation we call "Predictive Care." Drawing on pilot data from a computational ethnography we examine how the imagined potential of Predictive Care "makes live" new social arrangements and forms of "care" in medicine.

Paper long abstract:

This paper investigates an emergent medical and technical innovation we call "Predictive Care." Predictive Care is an approach to healthcare driven by advances in Artificial Intelligence (e.g. Machine Learning or ML) that combines Big Data (on whole populations) and Small Data (data on any single person) to identify patterns and predict outcomes or behaviors. While few Predictive Care tools are currently in use, in psychiatry there is high demand for a model designed to anticipate the likelihood of inpatient violence or aggression. The impetus behind these efforts is a desire to generate more accurate assessments of risk and facilitate proactive and targeted interventions (e.g., seclusion), while also preventing unnecessary ones (eg, restraints). At present, nurses assess patients every 2-3 hours for seven different signs of aggression (eg, irritability) and generate a daily predictive score (0-7). However, high scores (4+) may result in longer Emergency Department (ED) stays and increase the likelihood of being restrained and/or secluded. Further, internal data suggests that Black men are more frequently brought to the ED by police and restrained. Drawing on pilot data from a hybrid approach to shared datasets (computational ethnography) we examine how the imagined potential of Predictive Care 'makes live' new social arrangements and forms of 'care' in psychiatry. We show how these emergent tools reconfigure everyday workflows for clinicians who must alter their actions in the present to produce desirable future outcomes.