VALIDATION AND INTEGRATION OF A PREDICTIVE WARNING SYSTEM COMBINING PHYSIOLOGICAL MEASURES AND BEDSIDE EVALUATIONS

Purpose:  To detail the validation and systems integration of a clinical early warning system for adult inpatients in non-intensive care settings as a means for risk stratification by acuity level with the goal to provide clinical decision support for acute clinical deterioration.

Method: We developed an analytical framework to quantify clinical deterioration by integrating a published Early Warning Score (EWS) and a new Nurse Screening Assessment (NSA) tool focusing on fundamental health domains and patient safety assessments not traditionally incorporated into EWS. The new clinical EWS created a measure of patient acuity with improved abilities in detecting adverse events (defined as medical emergency team activation, escalation of care, and in-hospital mortality). Data were collected between December 2015 and March 2016 including over 1,000,000 clinical observations and administrative data for patients in 26 units at two acute care hospitals. Measures of association were evaluated to determine the independence among NSA and physiological components. Survival Analysis models and Monte Carlo methods were used for validation. Cost Analysis was utilized to examine the impact on cost, resource utilization, and potential harm reduction.

Result: There were no significant pairwise dependency structures among NSA and the physiological parameters (p < 0.05). Further analysis exposed significant differences between surgical and medical populations with respect to medical emergency team activations (p < 0.01). The relative risk of experiencing an adverse event within 24 hours was shown to increase exponentially with increases in the EWS. Results demonstrated improved discrimination with the inclusion of additional parameters based on the acute nature of deterioration onset and time elapsed in a critical state. Initial cost analysis indicated opportunities for cost avoidance related to a reduction in patients with an escalation of care with consideration of a potential increase in resource utilization.

Conclusion: Early detection of clinical deterioration can lead to a reduction in adverse outcomes. Our clinical EWS highlights the opportunities for assisting providers in identifying and appropriately responding to clinical deterioration up to 24 hours prior to an adverse patient outcome by integrating both physiological indicators and bedside evaluations. Inclusion of time-based metrics allowed for enhanced detection based on rapidity and duration of clinical deterioration. Iterative implementation processes highlighted the role of education and the importance of staged pilot testing when re-engineering complex high-reliability systems.