Investigators at Children’s Hospital of Philadelphia have developed a two-stage process to better identify children with sepsis while also minimizing alert fatigue. The approach includes an electronic alert tool that flags children with abnormal vital signs, but includes screening questions that enable clinicians to eliminate patients with no sign of infection. This approach is paired with a sepsis huddle to bring clinician judgment into the equation.

  • The first stage of the alert fires for abnormal vital signs, so either a high heart rate based on age, or for low blood pressure based on age. This alert goes off in about 10% of the ED patient population.
  • A series of three screening questions eliminates some patients from sepsis consideration, reducing the number of patients with potential sepsis to 1% of the ED population.
  • If a patient with an abnormal vital sign exhibits an abnormal mental status, a high-risk condition, or an abnormality in perfusion, that prompts the second stage of the alert to go off, triggering a sepsis huddle bedside.
  • The two-stage process has brought the number of young patients with missed sepsis diagnoses in the ED from 17% to 18% before the process was implemented down to the 4-5% range.

Identifying sepsis is always complex, but there are additional challenges involved with recognizing the life-threatening condition in children. “What is normal in terms of vital signs changes by age, and so you have to think about a child as he or she relates to other children their age,” explains Fran Balamuth, MD, PhD, MSCE, an emergency physician at Children’s Hospital of Philadelphia (CHOP). “Then, secondly, children are more likely than adults to present in compensated shock, meaning that their low blood pressure or hypotension often happens late in the episode of sepsis.”

Consequently, clinicians often struggle early on to find signs of compensated shock such as tachycardia, altered mental status, or restlessness, but many of these signs also are common in children who do not carry sepsis, Balamuth offers.

“Healthy children with fevers who do not have septic shock often present with tachycardia, so finding the ones among all those patients with compensated shock is a big challenge,” she says.

When CHOP’s sepsis quality improvement program first put a sepsis protocol in place, it purposefully did not include any sort of vital sign-based electronic alert because investigators were concerned it would misfire too often, resulting in alert fatigue.

“We treated patients on the protocol when we were clinically concerned about them, but there was no sort of screening tool,” Balamuth explains.

However, during the first few years of using that method, investigators observed that they were missing cases of sepsis. “There were some patients ending up in the ICU with severe sepsis and septic shock that we had not identified in the ED,” Balamuth notes. “We decided that maybe it was time to reconsider whether an electronic tool would help us, but we were still concerned about the alert fatigue issue, so we decided to look at it in an evidence-based way.”

Investigators decided to examine a potential alert retrospectively in 18 months of patients who presented with fever in the ED to see if the alert would have helped to define the patients with sepsis whom clinicians had missed in real time. “The electronic alert did help us find those patients — almost all of them, so based on that data we decided to implement a prospective alert,” Balamuth explains.

Guard Against Alert Fatigue

To prevent the alert from triggering the sepsis protocol on too many well patients, there is a two-stage process involved.

“The first stage of the alert fires for abnormal vital signs, so either a high heart rate based on age, or for low blood pressure based on age, and this alert goes off in about 10% of the ED patient population,” Balamuth says. For patients identified as exhibiting one or both of these abnormal vital signs, the nurse then has to answer whether the patient’s abnormal vital sign is due to possible infection.

“If the child’s heart rate is high because he or she has a broken leg or some other non-infectious reason, then the nurse can just click the alert away and does not have to answer any additional questions,” Balamuth says. “However, if the heart rate is high because of a potential infection, meaning that the patient either has a fever or there is some other reason to be concerned about infection, then this prompts the nurse to answer two additional screening questions.”

The first question asks whether the patient has any underlying conditions that may put him or her at risk for sepsis such as a history of cancer or a central line. The second question asks about the patient’s capillary refill time to assess perfusion, Balamuth explains. In addition to these questions, the alert factors in the patient’s mental status, which is assessed on all patients at triage.

“If the patient has an abnormal mental status, a high-risk condition, or an abnormality in perfusion, then that prompts the second stage of the alert to go off,” Balamuth notes. “And that leads to what we call a sepsis huddle, where the senior team is brought to the bedside to [make a decision on] whether this patient needs to be treated for sepsis or not.”

The sepsis huddle is triggered in 10% of patients who pass through the first stage of the alert, or roughly 1% of the patients who present to the ED, Balamuth explains.

“We are pretty happy that we don’t have to screen every patient who walks through the door, and that we can make an initial cut based on vital signs,” she says.

Further, while the approach isn’t perfect, it works well. “Before we put the alert in place [in May 2014], we were missing 17% to 18% of patients with sepsis,” Balamuth says, explaining that investigators define a missed patient as any patient who requires ICU care with sepsis or septic shock within 24 hours of an ED visit in which the patient was not treated with the hospital’s sepsis protocol. “We have been able to reduce that number to less than 4%, and we have been able to largely sustain that. So, over the past three years we have been in the 4% to 5% range [of missed sepsis diagnoses.].”1

Involve All Stakeholders

Balamuth acknowledges that changing behavior in a big, busy clinical setting is always challenging, but she notes that implementation of the alert tool was aided by the fact that the approach was developed by a multidisciplinary team.

“It included physicians, nurses, and nurse practitioners, and we all came up with this [approach] together, so the team-based approach really helped us on the implementation side,” she says. “We each had our stakeholder groups that we could get engaged with the process.”

Effective implementation required some learning time. For instance, at first, many of the nurses would immediately say no to the first question in the alert regarding whether they were worried about infection in the patient because they interpreted it to mean whether they were worried about sepsis rather than infection, Balamuth observes.

“We had to do some targeted educational interventions to make them understand that there is a low bar to answer the three questions,” she explains. “The combination of repeated educational interventions and having a big team of people that were respected in the ED and that worked well together — these are probably the biggest things that helped with implementation.”

Mary Kate Abbadessa, MSN, RN, RN-BC, CPEN, a clinical nurse specialist fellow with the sepsis program at CHOP, agrees that gathering everyone at the beginning of the process helped mitigate barriers.

“Our nursing informatics champion was able to really look through all of the little caveats of documentation and nursing workflow,” she says. “Each person [on the multidisciplinary team] could really speak to how to best implement the tool, and what would be their own potential barriers.”

When key stakeholders are missing, decisions often are made without a full appreciation of the implications for another group, Abbadessa observes. “Also, if you are not at the table, then you are less likely to want to buy into a change if it is negatively impacting your workflow.”

With new personnel constantly rotating in and out of the ED, the multidisciplinary sepsis group focuses its ongoing educational efforts on what Abbadessa terms “the constants.”

“We do have rotating residents, so our constants are our nursing staff, our tech staff, our attendings, and our nurse practitioner group,” she says. “Really having them champion the tool is quite effective, and it helps to ensure that each one of those groups understands their own role so that they can coach the others.”

Abbadessa adds that close and regular follow-up on all the personnel involved with implementing the alert tool is essential to ensuring sustained improvement. “Especially in large institutions, there can be misunderstandings or miscommunications on how to use a tool,” she says. “Providing one-to-one feedback to make sure all the staff know how this can positively impact patient care and ease their workload can help to increase their use of the tool. Monitoring compliance and giving quick, direct feedback will help to get it running more quickly.”

A key benefit of the two-stage sepsis alert tool is that it helps to get patients to their final destination safely and in the most efficient manner, Abbadessa offers.

“We have gotten really great at identifying these patients [with sepsis], getting them straight back to a room, and having all the providers there at once at the bedside, prioritizing their care,” she says. “For our sickest patients, that means identifying them, getting their fluids and antibiotics in quickly, getting their bed request in, and getting that report to the next team that will care for them as quickly as possible.”

Monitor Compliance

Fortunately, CHOP put solid data infrastructure in place so that key metrics can be monitored continually. “There is an application called ‘quick view’ that is updated every day, and helps us track our performance across a number of different QI [quality improvement] projects,” Balamuth explains.

For sepsis, the team tracks how quickly patients are getting antibiotics and fluids, and the number of patients with sepsis that were missed. “We have a QI team that meets monthly in the ED to see how things are going and talk about how we can improve,” Balamuth shares. “If we have a bad month or a bad week, we will do targeted education to the entire ED at our divisional meetings, and we send individual feedback to providers if they have missed the mark.”

If a diagnosis of sepsis was missed or a patient had to wait too long to receive antibiotics, the team will send the provider a survey to try to identify structural steps that would help improve care for the future, Balamuth adds. The hospital is eager to share its findings with colleagues. To that end, CHOP is part of a large sepsis collaborative that aims to improve sepsis outcomes across more than 40 pediatric hospitals. In addition, investigators are in the process of implementing the approach in a community hospital to see how it works in that setting.

“We have a needle-in-the-haystack problem at CHOP. We see lots of kids and most of them are OK, and we are trying to find the ones that aren’t,” Balamuth explains. “But that haystack is even bigger when you are talking about a community site that is not a referral center for kids with complicated problems, so sepsis is even more rare in healthy kids than it is in sick kids. I think there are probably some unique challenges at community-based sites.”

However, at any type of hospital or ED, Balamuth stresses that the electronic alert will only get one part of the way there in identifying sepsis.

“It still requires that bedside judgment to find some of those patients that the alert will miss,” she says.


  1. Balamuth F, Alpern ER, Abbadessa MK, et al. Improving recognition of pediatric severe sepsis in the emergency department: Contributions of a vital sign-based electronic alert and bedside clinician identification. Ann Emerg Med 2017 Jun 2. pii: S0196-0644(17)30315-3. doi: 10.1016/j.annemergmed.2017.03.019. [Epub ahead of print].


  • Mary Kate Abbadessa, MSN, RN, RN-BC, CPEN, Clinical Nurse Specialist Fellow, Sepsis Program, Children’s Hospital of Philadelphia. Phone: (800) 879-2467.
  • Fran Balamuth, MD, PhD, MSCE, Emergency Physician, Children’s Hospital of Philadelphia. Email: balamuthf@email.chop.edu.