e-HAIL Research in Progress: Improving Clinical Prediction of Adverse Events by Integrating Features of the Care Environment

Acute clinical deterioration—an unexpected and sudden worsening of a patient’s health—happens to one in fifteen hospitalized adults. Early warning systems based on real-time clinical prediction models can alert providers to impending deterioration; yet, these models often perform inconsistently across patients, clinical settings, and hospitals. A potential limitation of early warning systems in widespread use is that they ignore features of the care environment (e.g., patient occupancy, clinical workload, and clinician staffing) when making predictions, despite growing awareness that these factors play a causal role in patient outcomes. We hypothesize that incorporating features of the care environment into clinical prediction models will improve their predictive accuracy and consistency while informing us about how patients interact with their hospital environment while recovering from acute illnesses.