Piloting an Electronic Sepsis Alerting System to Reduce Inpatient Sepsis-related Serious Safety Events.

Abstract

Background: Electronic surveillance systems that enable continuous patient screening for emerging signs of sepsis may improve sepsis recognition and patient outcomes. Objectives: An electronic sepsis alerting system was piloted on Pediatric Medicine units at a tertiary care hospital. The pilot served to refine the system and accompanying workflow before hospital-wide implementation, aiming to reduce inpatient serious safety events related to delayed sepsis recognition. Methods: The multitiered intervention included (1) an electronic sepsis screening and alerting system; (2) an interdisciplinary huddle workflow; (3) risk stratification nomenclature (continue routine care, sepsis watch, code sepsis) to reconcile alerts; and (4) streamlined documentation and ordering tools. Iterative changes were made to the tools and workflow using plan, do, study, act cycle methodology. Measures of clinician adherence to the workflow were tracked monthly during the pilot (% alerts reconciled, % alerts with documented huddle note). Alert to huddle time and huddle to reconciliation time (target: 10 minutes) served as outcome measures for the pilot. The frequency of consults to the hospital’s critical care response team for the indication of query sepsis was monitored as a balancing measure. Alert frequency, rate of corresponding clinical concern for sepsis, and missed sepsis cases were tracked as measures of screening tool utility. Results: During the 3-month pilot, 86% of alerts were reconciled and 67% had a documented huddle note. The median alert to huddle time decreased from 9 to 3 minutes over the course of the pilot. The median huddle to reconciliation time decreased from 14 to 8 minutes. The frequency of CCRT consults during the pilot was 1 per month, which was increased slightly from the baseline frequency of 0.67 consults per month over the previous year. An alert was triggered for 4.1% of opportunities (average 2.2 alerts per day). Of the alerts reconciled, 29% indicated a concern for sepsis. There were four instances of clinical concern for sepsis where no alert was triggered. The sepsis huddle workflow was followed in three of these instances. Conclusions/implications: Pilot results suggest clinician adherence to a new streamlined approach to care for inpatients at risk for sepsis. Involving clinicians in system design and embedding electronic tools within existing care processes supported workflow uptake. Future efforts will focus on optimizing the alerting system’s clinical accuracy and examining its impact on clinical outcomes.