Preventing sepsis alert fatigue


If they’re too infrequent, alerts can delay sepsis identification and treatment. If they’re too abundant, the alerts can overwhelm providers. Finding the sweet spot for sepsis alerts, QI leaders say, can require time, technology, patience – and sometimes trial and error.

University Hospital in Salt Lake City wanted to broaden its sepsis recognition system to ensure that decompensating patients were seen and resuscitated quickly, regardless of the cause. Another hospital offered a lesson in what not to do when a staff member cautioned that a sepsis alert system based on SIRS alone had been a “total disaster” and left providers fuming. One report suggested that nearly half of all ward patients meet SIRS criteria at some point during their hospitalization, and that using the criteria for sepsis screening in hospital wards is both “time consuming and impractical.”1

Instead, University Hospital tweaked its MEWS or Modified Early Warning System, based on consultations with hospitalists, ICU physicians, and other providers about the appropriate thresholds for vital signs. “It’s kind of like asking someone, ‘Well, when are you really scared of the heart rate and when are you sort of scared and when are you not scared at all?’ ” said project coleader Devin J. Horton, MD, an academic hospitalist.

The team also analyzed the number of alerts per week per unit and their sensitivity and specificity in detecting sepsis. As junior faculty members, Dr. Horton and his collaborator, academic hospitalist Kencee K. Graves, MD, were mindful to avoid angering other doctors over being alerted too often. For their MEWS scoring system, they sacrificed a bit of sensitivity to ensure that the number of alerts remained manageable.

Before going live with its own new alert system, Middlesex Hospital in Middletown, Conn., had a subgroup spend several weeks testing the system in silent mode and tweaking different parameters such as respiratory rate and heart rate to reduce the potential for too many alerts. “If you look at each and every alert, then you can identify how to make your adjustment so that it’s not overly sensitive,” said Terri Savino, MSN, RN, CPHQ, the hospital’s manager of patient experience and service excellence.

A sepsis task force also shared data showing the hospital’s significant reductions in sepsis mortality, total hospital mortality, and sepsis length of stay. “Medical staff were willing to accept the frequency and high sensitivity of the alert because the data demonstrated that it was making a difference in the lives of our patients,” said David M. Cosentino, MD, the hospital’s chief medical information officer.

Other alert systems’ mixed performances have yielded important lessons. At the University of Pennsylvania, Philadelphia, one prototype detected clinically deteriorating patients and sent an alert to the nurse, physician, and a rapid response team. Alerted providers converged on the patient’s bedside within 30 minutes and decided whether to elevate the level of care. Craig Umscheid, MD, MSCE, of the department of epidemiology and vice chair for quality and safety in the department of medicine, said the system was associated with a suggestion of reduced mortality.2 But it was noisy and less helpful than it could have been, he said, because it didn’t separate out declining patients already known to the team from those who were still unrecognized.

Tools to predict which patients may develop severe sepsis or septic shock have worked even less well, he said. One triggered an alarm before patients showed signs of clinical deterioration. “The team didn’t know what to do with that prediction,” Dr. Umscheid said. As a result, the alert didn’t improve mortality or discharges to home. “If you’re making this prediction too early and providers don’t know what to do with the information, it’s not going to change care or affect patient outcomes,” he said. “It’s just going to frustrate providers.”


1. Churpek MM et al. Incidence and prognostic value of the systemic inflammatory response syndrome and organ dysfunctions in ward patients. Am J Respir Crit Care Med. 2015 Oct 15;192(8):958-64.

2. Umscheid CA et al. Development, implementation, and impact of an automated early warning and response system for sepsis. J Hosp Med. 2015 Jan;10: 26-31.

Recommended Reading

Doctors’ and nurses’ predictions of ICU outcomes have variable accuracy
The Hospitalist
Improving the risk assessment of patients with acute pulmonary embolism
The Hospitalist
Intense urine output monitoring beneficial in ICU
The Hospitalist
Alarm reductions don’t improve ICU response times
The Hospitalist
Inpatient antiviral treatment reduces ICU admissions among influenza patients
The Hospitalist
When should I transfuse a patient who has anemia?
The Hospitalist
FDA approves angiotensin II for shock patients
The Hospitalist
FVC deterioration signals increasing risk in rib fracture patients
The Hospitalist
Charting a new course in sepsis management
The Hospitalist
Elderly trauma patients at high risk for post-discharge mortality
The Hospitalist
   Comments ()