Background:
In-hospital mortality attributable to sepsis is higher than overall population mortality (Gaieski DF et al, 2013; Dombrovskiy VY, 2007). Furthermore, the rates of severe sepsis are increasing annually (Dombrovskiy VY,2007). Early detection and early intervention have been shown effective at reducing mortality among in-patients (Dombrovskiy VY, 2007).
Internal data suggests that the identification of sepsis that occurs mid-hospitlization in inpatients is delayed in our facility (Sankey C, unpublished data). With identification delayed, ability to provide potentially life-saving therapies is delayed, and with each hour delay in antibiotics, in-hospital mortality has been shown to increase 7.6% (Gaieski DF et al, 2010).
Purpose:
Primary: Improved recognition of sepsis-related clinical deterioration in medical inpatients via real-time notification of an attending hospitalist when a patient newly meets modified SIRS criteria.
Secondary: Reduced delays in the initiation of appropriate sepsis-related therapies, when indicated, in medical inpatients with new onset sepsis via rapid attending level identification of at-risk patients.
Description:
A fully automated, electronic health record (EHR)-based monitoring and alert system monitors all medical inpatients for modified SIRS criteria (abnormalities in pulse, respiration rate, temperature, white blood cell count, drop is systolic blood pressure, increase in creatinine). When a patient meets these criteria, the EHR generates a “Sepsis Alert” page to the attending physician on the Rapid Response Team (RRT). The RRT physician performs a chart review, and then intervenes as needed. This protocol is novel in that is uses no new resources, and is designed for sustainability, ease of implementation, and rapid scale-up.
To date, our EHR tool has been active for 11,894 adult inpatient hospitalizations. 82 Sepsis Alerts have fired. Preliminary data is largely focused on the validity of the Sepsis Alert, which is novel but adapted from prior studies to improve sustainability. Patients with an alert fire have a longer non-ICU length of stay (164 hours vs 80.8 hours, p<0.001), are more likely to have a sepsis diagnosis (0.229 vs 0.066, p<0.0001), and have a higher mortality rate (p<0.0001). Furthermore, there is a trend towards longer ICU length of stay in patients with alert fire (87.6 hours vs 51.6 hours, p=0.055). Regarding sustainability and ease of implementation, implementation is ahead of schedule and under budget. In less than one year, this protocol progressed from idea to fully-implemented on our medical service. On preliminary review, 14/14 sepsis alerts have resulted in attending-level chart review and documentation following the alert fire.
Conclusions:
Adoption of this protocol has been rapid and done with minimal resources. Preliminary data suggests our alert successfully identifies high acuity inpatients, provides a mechanism for rapid assessment by attending hospitalist staff, and physician adherence to the protocol is high. Furthermore, implementation has been rapid and well ahead of schedule, suggesting that our protocol is portable and easy to implement.