A Real-Time Automated Machine Learning Algorithm for Predicting Mortality in Trauma Patients: Survey Says it's Ready for Prime-Time.

BACKGROUND: Though artificial intelligence ("AI") has been increasingly applied to patient care, many of these predictive models are retrospective and not readily available for real-time decision-making. This survey-based study aims to evaluate implementation of a new, validated mortality risk calculator (Parkland Trauma Index of Mortality, "PTIM") embedded in our electronic healthrecord ("EHR") that calculates hourly predictions of mortality with high sensitivity and specificity. METHODS: This is a prospective, survey-based study performed at a level 1 trauma center. An anonymous survey was sent to surgical providers and regarding PTIM implementation. The PTIM score evaluates 23 variables including Glasgow Coma Score (GCS), vital signs, and laboratory data. RESULTS: Of the 40 completed surveys, 35 reported using PTIM in decision-making. Prior to reviewing PTIM, providers identified perceived top 3 predictors of mortality, including GCS (22/38, 58%), age (18/35, 47%), and maximum heart rate (17/35, 45%). Most providers reported the PTIM assisted their treatment decisions (27/35, 77%) and timing of operative intervention (23/35, 66%). Many providers agreed that PTIM integrated into rounds and patient assessment (22/36, 61%) and that it improved efficiency in assessing patients' potential mortality (21/36, 58%). CONCLUSIONS: Artificial intelligence algorithms are mostly retrospective and lag in real-time prediction of mortality. To our knowledge, this is the first real-time, automated algorithm predicting mortality in trauma patients. In this small survey-based study, we found PTIM assists in decision-making, timing of intervention, and improves accuracy in assessing mortality. Next steps include evaluating the short- and long-term impact on patient outcomes.

Clinical Diagnosis of Infection in Surgical Intensive Care Unit: You're Not as Good as You Think!

Background: Because of the everincreasing costs and the complexity of institutional medical reimbursement policies, the necessity for extensive laboratory work-up of potentially infected patients has come into question. We hypothesized that intensivists are able to differentiate between infected and non-infected patients clinically, without the need to pan-culture, and are able to identify the location of the infection clinically in order to administer timely and appropriate treatment. Methods: Data collected prospectively on critically ill patients suspected of having an infection in the surgical intensive care unit (SICU) was obtained over a six-month period in a single tertiary academic medical center. Objective evidence of infection derived from laboratory or imaging data was compared with the subjective answers of the three most senior physicians' clinical diagnoses. Results: Thirty-nine critically ill surgical patients received 52 work-ups for suspected infections on the basis of signs and symptoms (e.g., fever, altered mental status). Thirty patients were found to be infected. Clinical diagnosis differentiated infected and non-infected patients with only 61.5% accuracy (sensitivity 60.3%; specificity 64.4%; p = 0.0049). Concordance between physicians was poor (κ = 0.33). Providers were able to predict the infectious source correctly only 60% of the time. Utilization of culture/objective data and SICU antibiotic protocols led to overall 78% appropriate initiation of antibiotics compared with 48% when treatment was based on clinical evaluation alone. Conclusion: Clinical diagnosis of infection is difficult, inaccurate, and unreliable in the absence of culture and sensitivity data. Infection suspected on the basis of signs and symptoms should be confirmed via objective and thorough work-up.