Sepsis is a common, costly, and mortal clinical syndrome. Many delays in sepsis recognition and intervention are due to “data latency,” the period of time between data suggestive of sepsis being entered in the EHR to identification and interpretation by a care provider. The magnitude of this delay can be significant, as the diagnostic components of sepsis are derived from multiple sources (ancillary staff, laboratory analyzers, nurses) and not listed together in the EHR. Efforts to study the impact of continuous monitoring and electronic provider notification specifically for sepsis are also few in number. 


We sought to improve recognition of sepsis-related clinical deterioration in medical inpatients. This was achieved via real-time notification of an attending Hospitalist when a patient met pre-specified sepsis criteria, with the intent to reduce delays in the initiation of appropriate sepsis-related therapies. The project was designed with a focus on ease of implementation, low resource requirement and sustainability. 


We performed a before and after study of a low-resource, electronic health record-based protocol for sepsis monitoring, alert, and intervention. Using our HER (Epic), a best practice alert (BPA) was designed to monitor for patients who newly met modified SIRS criteria. When criteria were met, a “Sepsis Alert” was sent via Epic to Yale-New Haven Hospital (YNHH) rapid response team (RRT) Hospitalist attending physician via pager in HIPAA-compliant manner. The attending physician then reviewed the EHR, determined the intervention, and documented a Sepsis Alert Note.

Our sample included 15,554 adults discharged from YNHH Medicine Service from April 2015 through March 2016 from a specific set of medical units. There were 128 deaths in this group (raw mortality rate 0.82%). In a convenience sample immediately prior to our study population, the unadjusted mortality rate was 0.29%, which did not reach statistical significance (3 deaths in 1027 discharges, OR 2.85, p=0.07). The odds of ICU admission are 1.24 times greater in the period with the program fully implemented compared to the period with no program implemented (p=0.04). The odds of ICU admission are 2.91 times higher in the period with a partially implemented program compared to the period with no program implemented (p<.0001). The partially implemented program refers to the rapid-cycle deployment of the intervention with units being brought live in a stepwise fashion. The sepsis BPA trigger and completed assessment had an OR of death 15.76 (8.63,28.77; p<0.0001) which approximated that of an activated RRT assessment (OR 15.38, [10.25,23.09; p<0.0001]). Furthermore, the BPA trigger was associated with increased risk of ICU admission during the hospitalization.


Given the low cost and rapid implementation, this trigger tool may be a useful and easily implemented instrument to assist hospitals in identifying sepsis patients at higher risk of death or ICU admission during their admission.