SESSION TITLE: Patient Safety
SESSION TYPE: Original Investigation Slide
PRESENTED ON: Tuesday, October 29, 2013 at 02:45 PM - 04:15 PM
PURPOSE: To develop a severe sepsis early warning system using data from the electronic medical record (EMR) to notify clinical surveillance personnel of the presence of severe sepsis in all hospital inpatients, especially those not in ICUs. We hypothesized that screening of alerts by surveillance personnel would provide more actionable information to bedside clinicians.
METHODS: This is a descriptive study of CareVeillance, an electronic surveillance tool developed jointly by CSC Consulting and University of Kansas Hospital. Software for collecting sepsis-specific information from the EMR (Epic) was developed using standard criteria for diagnosis of severe sepsis and an iterative process involving clinical and quality improvement personnel with software engineers. A pilot was performed in 2/2012. Clinical personnel (one critical care surveillance nurse, backed up by one intensivist) evaluated all patients detected by CareVeillance to determine the tool's sensitivity and specificity by reviewing each patient's EMR and, when necessary, by conversation with the patient's physician and/or bedside nurse.
RESULTS: The pilot study ran 11 days. CareVeillance identified 655 potential sepsis episodes involving 477 patients. Of these, numerous episodes were removed on evaluation by clinical personnel for the following reasons: No infection - 297, organ dysfunction non-acute - 177, patients on palliative care - 4, no SIRS on review - 15, patient already in ICU at detection - 66, flagged for reassessment - 58, beyond window for early goal-directed therapy (EGDT) - 19. In all, 19 patients were deemed by clinical personnel to have severe sepsis and eligible for EGDT. Bedside clinical personnel were notified only for these 19 patients. On average, the review process required 1.9 +/- 1.6 minutes per episode (mean +/- SD).
CONCLUSIONS: CareVeillance appears to be highly sensitive for the identification of severe sepsis, but lacks specificity. Assessment of the alerts by clinical surveillance personnel can allow for transmission of specific, actionable alerts to bedside clinicians.
CLINICAL IMPLICATIONS: Previous work with best practice alerts in the EMR had demonstrated that they were often ignored, and the lack of specificity in the current data may indicate why a bedside nurse would find such alerts untrustworthy. Use of a clinical surveillance nurse and/or physician along with the software may allow for improved specificity at the bedside.
DISCLOSURE: The following authors have nothing to disclose: Bristol Brandt, Amanda Gartner, Michael Moncure, Chad Cannon, Liz Carlton, Steven Simpson
No Product/Research Disclosure Information