View Abstract

A4504 - Attitudes and Perceptions of Medical Trainees Towards an Electronic Medical Alert System for Sepsis
Author Block: K. Swenson1, J. Ferguson1, L. Shieh1, A. Rogers2; 1Internal Medicine, Stanford University, Stanford, CA, United States, 2Pulmonary and Critical Care Medicine, Stanford University, Stanford, CA, United States.
Rationale: Electronic sepsis alerts are key tools for early sepsis recognition. To optimize case detection, false positive rates among sepsis alerts are often high, which can cause alarm fatigue. Outcome studies of sepsis alerts often lack assessment of clinician perception regarding the utility of sepsis alerts, especially among medical trainees. Methods: We conducted a survey of 121 Internal Medicine trainees at a large academic medical center in 2016 on perceptions and responses to positive sepsis alerts among ward patients, following the adoption of an electronic sepsis alert based on vital sign and laboratory abnormalities based on 2012 Surviving Sepsis Campaign definitions for severe sepsis. An anonymous, cross-sectional questionnaire was distributed electronically to all trainees. Perceptions of alert accuracy were compared to known test characteristics of our institution’s electronic sepsis alert.1 Results: 107 of 121 providers completed the questionnaire (response rate of 88%). On average, 10% of positive alerts were felt to have identified a new case of sepsis. By comparison, the positive predictive value of our institution’s electronic alert is 50.3% based on retrospective chart review.1 54% of alerts were thought to identify patients already diagnosed by the provider with sepsis, and a further 36% were thought to identify patients with sepsis-mimicking diagnoses. The average respondent felt that it was appropriate to receive 5 electronic alerts in order to identify one new sepsis diagnosis (acceptable positive-predictive value of 20%). Respondents felt that 85% of alerts did not alter clinical management; after such alerts, most respondents felt that they were less likely to respond to a positive alert in the future, either from the same or a different patient. Among open-ended comments, themes included high alert frequency for known sepsis cases, and undue pressure on trainees to perform additional diagnostics and therapeutics following a positive alert. Conclusions: Medical trainees understand the importance of early sepsis recognition and will accept a low positive predictive value for identifying new cases of sepsis. However, trainees feel that the vast majority of true positive alerts identify patients already recognized as septic, and most do not alter management decisions (85%), perceptions not captured by retrospective evaluation of electronic alert test characteristics. Alarm fatigue, including desentization to alerts from other patients, could decrease the efficacy of electronic sepsis alerts. References: 1. Rolnick et al. ""Validation of test performance and clinical time zero for an electronic health record embedded severe sepsis alert."" Applied clinical informatics 7.2 (2016): 560.