Hypotension During Vasopressor Infusion Occurs in Predictable Clusters: A Multicenter Analysis.

Background: Published evidence indicates that mean arterial pressure (MAP) below a goal range (hypotension) is associated with worse outcomes, though MAP management failures are common. We sought to characterize hypotension occurrences in ICUs and consider the implications for MAP management. Methods: Retrospective analysis of 3 hospitals' cohorts of adult ICU patients during continuous vasopressor infusion. Two cohorts were general, mixed ICU patients and one was exclusively acute spinal cord injury patients. "Hypotension-clusters" were defined where there were ≥10 min of cumulative hypotension over a 60-min period and "constant hypotension" was ≥10 continuous minutes. Trend analysis was performed (predicting future MAP using 14 min of preceding MAP data) to understand which hypotension-clusters could likely have been predicted by clinician awareness of MAP trends. Results: In cohorts of 155, 66, and 16 ICU stays, respectively, the majority of hypotension occurred within the hypotension-clusters. Failures to keep MAP above the hypotension threshold were notable in the bottom quartiles of each cohort, with hypotension durations of 436, 167, and 468 min, respectively, occurring within hypotension-clusters per day. Mean arterial pressure trend analysis identified most hypotension-clusters before any constant hypotension occurred (81.2%-93.6% sensitivity, range). The positive predictive value of hypotension predictions ranged from 51.4% to 72.9%. Conclusions: Across 3 cohorts, most hypotension occurred in temporal clusters of hypotension that were usually predictable from extrapolation of MAP trends.

Diagnostic suspicion bias and machine learning: Breaking the awareness deadlock for sepsis detection.

Many early warning algorithms are downstream of clinical evaluation and diagnostic testing, which means that they may not be useful when clinicians fail to suspect illness and fail to order appropriate tests. Depending on how such algorithms handle missing data, they could even indicate "low risk" simply because the testing data were never ordered. We considered predictive methodologies to identify sepsis at triage, before diagnostic tests are ordered, in a busy Emergency Department (ED). One algorithm used "bland clinical data" (data available at triage for nearly every patient). The second algorithm added three yes/no questions to be answered after the triage interview. Retrospectively, we studied adult patients from a single ED between 2014-16, separated into training (70%) and testing (30%) cohorts, and a final validation cohort of patients from four EDs between 2016-2018. Sepsis was defined per the Rhee criteria. Investigational predictors were demographics and triage vital signs (downloaded from the hospital EMR); past medical history; and the auxiliary queries (answered by chart reviewers who were blinded to all data except the triage note and initial HPI). We developed L2-regularized logistic regression models using a greedy forward feature selection. There were 1164, 499, and 784 patients in the training, testing, and validation cohorts, respectively. The bland clinical data model yielded ROC AUC's 0.78 (0.76-0.81) and 0.77 (0.73-0.81), for training and testing, respectively, and ranged from 0.74-0.79 in four hospital validation. The second model which included auxiliary queries yielded 0.84 (0.82-0.87) and 0.83 (0.79-0.86), and ranged from 0.78-0.83 in four hospital validation. The first algorithm did not require clinician input but yielded middling performance. The second showed a trend towards superior performance, though required additional user effort. These methods are alternatives to predictive algorithms downstream of clinical evaluation and diagnostic testing. For hospital early warning algorithms, consideration should be given to bias and usability of various methods.

End-user evaluation of an interface for clinical decision support using predictive algorithms.

There have been decades of interest in advanced computational algorithms with potential for clinical decision support systems (CDSS), yet these have not been widely implemented in clinical practice. One major barrier to dissemination may be a user-friendly interface that integrates into clinical workflows. Complicated or non-intuitive displays may confuse users and may even increase patient management errors. We recently developed a graphical user interface (GUI) intended to integrate a predictive hemodynamic model into the workflow of nurses caring for patients on vasopressors in the intensive care unit (ICU). Here, we evaluated user perceptions of the usability of this system. The software was installed in the room of an ICU patient, running for at least 4 hours with the display hidden. Afterward, we showed nurses a video recording of the session and surveyed their perceptions about the software's potential safety and usefulness. We collected data for nine patients. Overall, nurses expressed reasonable enthusiasm that the software would be useful and without serious safety concerns. However, there was a wide diversity of opinions about what specific aspects of the software would be useful and what aspects were confusing. In several instances, the same elements of the GUI were cited as most useful by some nurses and most confusing by others. Our findings validate that it is possible to develop GUIs for CDSS that are perceived as potentially useful and without substantial risk but also reinforce the diversity of user perceptions about novel CDSS technology. Clinical Relevance- This end-user evaluation of a novel CDSS highlights the importance of end-user experience in the workflow integration of advanced computational algorithms for bedside decision support during critical care.

Validation of a 'Usual Care' Model for Vasopressor Initiation in a Cohort of Emergency Department Patients with Sepsis.

Usual care regarding vasopressor (VP) initiation is ill-defined. We aimed to further validate a quantitative model for usual care in the Emergency Department (ED) regarding the timing of VP initiation in sepsis. We retrospectively studied a cohort of adult critically-ill ED patients who also received antibiotics in the ED. We applied a multivariable model previously developed from another patient cohort which distinguishes between time points at which patients were or were not subsequently started on a continuous VP infusion. The model has six independently significant predictors (respiratory rate, Glasgow Coma Scale score, systolic blood pressure, SpO2, administered intravenous fluids, and elapsed time). The outcome was initiation of VP infusion, either within the ED or within 6 hours after leaving the ED. We applied the model to all time points, beginning when all model input parameters were first available for a given patient, and ending when either VP were first started, or the patient left the ED. Out of 55,963 adult ED patients during the two-year study interval, we identified 1,629 who met our inclusion criteria. The area under the receiver operating characteristic curve was 0.81 for all patients, and 0.72 for the subset with at least one hypotensive blood pressure measurement. At a model threshold with sensitivity and specificity 0.74 and 0.74, respectively, the median advance detection time was 170.5 minutes (IQR 53 - 363).