Deteriorating child: pediatric warning scores

It is three in the morning in the pediatric ward. A six-year-old girl, admitted for pneumonia, is breathing a little faster than she was at the end of the afternoon. Her heart rate has gone up, she is quieter, and her mother thinks her daughter is just tired. The nursing technician adds up the points on the vital signs sheet and the number has passed the agreed threshold. She calls the resident.

This moment sums up the goal of pediatric alert scores. A hospitalized child's deterioration usually gives signs hours before a cardiac arrest or an urgent transfer to intensive care, but these signs are subtle and spread among several professionals and several shifts. The scores try to gather these clues into a number that triggers an action. This text shows what studies say about the accuracy, the impact and the limits of these tools in hospitals in several countries.

Key points

• Pediatric alert scores add up changes in vital signs, behavior and respiratory support.
• Accuracy for predicting deterioration is moderate to good in validation studies.
• The large international trial on the topic changed mortality very little.
• Hospitals in several countries, including low-resource settings, have implemented scores with good process results.
• The number only works when there is a clear response pathway and a team that listens to each other.

Why the child who deteriorates is misleading

Children compensate well for illness for a long time. Blood pressure may stay normal while heart rate rises and peripheral perfusion falls, and when blood pressure finally drops, the reserve is already exhausted. That is why pediatric deterioration often seems sudden to those who look only at the latest record, although it was gradual for those who followed the curve.

Another factor is the normal variation of vital signs with age. A respiratory rate that is acceptable in an infant is high for an adolescent, and a heart rate that is normal in a baby would be tachycardia in an older child. This variation makes interpretation harder for professionals who see children only occasionally, and it is one of the reasons for using tables and scores.

Signs spread among people and shifts

Information about deterioration is rarely in one place. Nursing notices the change in behavior, physical therapy notes the increased breathing effort, the family says the child is different, and the on-call physician sees an isolated record. Each one has a piece of the story.

Garza and colleagues, studying the implementation of a score, highlighted that teamwork and early recognition of deterioration were central to success (Garza and colleagues, 2021). The score works as a common language, which lets one professional call another with an objective reason. The text on clinical deterioration in the ward discusses the same logic in adults.

The role of the family

Parents and caregivers spend far more time with the child than any professional. They notice subtle changes, such as a different cry, refusal to play, unusual drowsiness or a new way of breathing. Many hospitals have included family concern among the items that justify a reassessment, regardless of the total score.

This listening has a double effect. It brings information that the vital signs record leaves out and, at the same time, strengthens the family's trust in the team. When the mother says her daughter is different, the right response is to examine the child and record what was seen, and the sentence stops being just a venting of tiredness.

What a pediatric alert score is

Pediatric early warning scores, known by the acronym PEWS, assign points to changes in domains such as behavior, the cardiovascular system and the respiratory system. The total indicates the degree of concern, and each point range corresponds to an action, such as reassessing sooner, calling the physician or triggering a rapid response team.

A review by Simon Junior and colleagues gathered 105 articles from 12 countries on pediatric alert scores (Simon Junior and colleagues, 2022). The authors found a wide variety of versions, with differences in parameters, cutoffs and associated actions (Simon Junior and colleagues, 2022). This variety makes it hard to compare studies and choose a single tool.

An example of a cutoff

In a study of 761 hospitalized pediatric patients, Solevåg and colleagues used a score of 3 or more as the alert threshold for medical assessment (Solevåg and colleagues, 2013). The work showed that the score could be applied in ward routine and that higher values were associated with the need for more care (Solevåg and colleagues, 2013).

The cutoff is always a choice. A low threshold triggers more calls and catches more children who deteriorate, but tires the team with false alarms. A high threshold reduces calls, but lets some deteriorations slip through. The text on pretest probability helps explain why the same threshold works differently in wards with different profiles.

Frequency of measurements

A score is only as good as the data feeding it. In many wards, vital signs are measured at fixed intervals, and a deterioration that begins right after a measurement can go hours without being recorded. That is why many protocols tie the score to the frequency of the next measurement: the higher the score, the shorter the interval until reassessment.

The quality of the measurement also matters. A respiratory rate counted over a few seconds, oximetry with a poorly positioned sensor or blood pressure measured with a wrongly sized cuff produce misleading numbers. Training the team to measure well is part of implementing any score.

How accurate the scores are

A review by Brown and colleagues analyzed the accuracy of pediatric alert scores in predicting clinical deterioration (Brown and colleagues, 2018). The area under the curve ranged from 0.73 to 0.91 in the included studies (Brown and colleagues, 2018). The review found 8 studies carried out in low- and middle-income countries (Brown and colleagues, 2018).

These values indicate moderate to good discrimination. The score reasonably separates children who will deteriorate from those who will remain stable, but it is wrong in some cases in both directions. Accuracy also depends on who fills it in, on the frequency of measurements and on the ward population.

The problem of manual calculation

Adding up points from memory in the middle of a busy shift leaves room for error. Sefton and colleagues compared score calculation by an electronic system with manual calculation by the team (Sefton and colleagues, 2017). Calculation accuracy was 98.5 out of every 100 with the electronic system and 85.6 out of every 100 with manual calculation (Sefton and colleagues, 2017). The time to calculate fell from 98 to 68 seconds with electronic support (Sefton and colleagues, 2017).

The result is a reminder that a decision tool can fail even before it is interpreted. A score added up wrongly leads to wrong decisions, and a difference of a few points can separate the child who is reassessed from the one who waits until the next round. The text on organizing reasoning in the medical record discusses how structured recording reduces this type of failure.

What impact studies show

Getting the prediction right is different from improving the outcome. Chong and colleagues carried out a systematic review of 15 studies on the impact of pediatric alert scores (Chong and colleagues, 2022). In the pooled analysis, the absence of the score was associated with a higher risk of death, with a relative risk of 1.18, and of code events, with a relative risk of 1.73 (Chong and colleagues, 2022). Cardiac arrest, on its own, showed no clear association with use of the score (Chong and colleagues, 2022).

The largest study was the EPOCH trial, conducted in 21 hospitals in 7 countries (Chong et al., 2022). In this trial, implementing the score changed in-hospital mortality little (Chong et al., 2022). The result is a reminder that a score on its own has limited effect when the surrounding system stays the same.

Why the impact is smaller than accuracy suggests

Several reasons explain the gap between predicting and changing the outcome. Mortality in pediatric wards in high-income countries is already low, which makes it hard to show a difference. Comparison hospitals often already had rapid response teams and other improvements. And the score depends on a whole chain of actions, from measuring vital signs to the call and the doctor's response.

When any link in this chain fails, the score loses value. A high number that nobody reads, or a call that takes long to be answered, produces the same result as not having the score. The text on clinical decision-making reminds us that a decision is a process, and a good test does not save a bad process.

There is also an effect that is hard to measure. In many services, implementing the score came together with culture changes, such as explicit permission for nursing staff to call the doctor without fear of reprimand and systematic review of deterioration events. These gains show up little in mortality rates, but they can change the way the team talks about the child who worries them. To evaluate a score in your own hospital, it is also worth looking at process indicators, such as the time between the alert and the medical assessment, the proportion of transfers to intensive care that were preceded by an alert, and the frequency of complete vital sign measurements.

What a score leaves out

Scores add up what is easy to measure. Other important signs are left out or enter crudely, such as how the condition has evolved since admission, the underlying diagnosis, the use of medications that alter vital signs, and each child's clinical context. A child with congenital heart disease may have low baseline saturation and score high without getting worse, while another may be in early shock with a modest score.

That is why the score must be read together with the history. It organizes attention and gives an objective reason to call, but the final interpretation depends on whoever knows the case. The text on baseline rate in diagnosis reminds us that the same number means different things in different groups.

Study	Population	Main finding	Source
Impact review	15 studies	Relative risk of death of 1.18 without the score	(Chong et al., 2022)
EPOCH trial	21 hospitals in 7 countries	Little change in in-hospital mortality	(Chong et al., 2022)
Accuracy review	Validation studies	Area under the curve from 0.73 to 0.91	(Brown et al., 2018)
Electronic calculation	Comparison with manual calculation	Accuracy of 98.5 versus 85.6 per 100	(Sefton et al., 2017)
EVAT collaboration	36 hospitals in 13 countries	Implementation in pediatric oncology centers	(Agulnik et al., 2022)

Scores in hospitals with few resources

Much of the recent research on pediatric scores comes from pediatric oncology hospitals in Latin America and other regions. The collaboration known as EVAT brought together 36 hospitals from 13 countries to implement alert scores in children with cancer (Agulnik et al., 2022). In another study, Agulnik et al. followed 16 centers in 10 countries, with 553 deterioration events and mortality of 29 per 100 among the children who got worse (Agulnik et al., 2021).

These numbers show the size of the problem in these services. In a previous study, Agulnik and colleagues reviewed unplanned transfers to intensive care and ward arrests at a North American hospital dedicated to pediatric hematology-oncology, from August 2014 to July 2016 (Agulnik and colleagues, 2020). There were 220 deterioration events in 35,945 patient-days, and 107 of them were critical deteriorations, a rate of 2.98 per 1000 patient-days (Agulnik and colleagues, 2020). Tracking these rates lets each hospital see its own progress over time.

Barriers to implementation

Agulnik and colleagues studied implementation barriers in 5 centers in 4 countries (Agulnik and colleagues, 2022). Difficulties included staff shortages, team turnover, rigid hierarchy between nursing and medicine, and difficulty sustaining training over time (Agulnik and colleagues, 2022). The centers that made progress had engaged local leaders and institutional support (Agulnik and colleagues, 2022).

Graetz and colleagues interviewed 83 professionals, 41 in Guatemala and 42 in the United States, about use of the score (Graetz and colleagues, 2020). Participants in both countries described the score as a way to give nursing a voice and to organize communication with physicians (Graetz and colleagues, 2020).

Lessons for any hospital

The experiences of low-resource hospitals offer lessons that apply to any service. The score tends to work best when it comes with ongoing training, a defined response flow and tracking of deterioration events. Nursing and medical leaders working together appear as a common factor in the centers that succeeded.

Another lesson is local adaptation. A tool designed for a university hospital with intensive care on the same floor needs adjustments in a service where transfer takes hours. The cutoff, the actions tied to each band and the expected response time should reflect the reality of each place.

The emotional side of the score

Graetz and colleagues also studied team emotions when facing a child who is deteriorating (Graetz and colleagues, 2021). Professionals reported anxiety, fear of making a mistake and, in some cases, a false sense of relief when the score stayed high for a long time without obvious worsening (Graetz and colleagues, 2021). This accommodation to a persistently high number is a risk, because the child may be compensating until the moment they stop compensating.

The phenomenon recalls alarm fatigue in intensive care. When a signal sounds all the time, the team starts to ignore it. The text on affective bias shows how fear, fatigue and sympathy enter clinical decisions, often without the professional noticing.

What changes for the clinician

The studies allow building a reasoning script for the hospitalized child who worries the team. The sequence below summarizes the most consistent points in the reviewed literature.

• Look at the trend of vital signs over the hours, not just the latest record.
• Check the score sum, because manual calculation is often wrong (Sefton and colleagues, 2017).
• Treat a high and persistent score as a warning sign, not as the new normal (Graetz and colleagues, 2021).
• Value the concern of nursing and the family, even when the score is below the threshold.
• Know the service escalation pathway, with names and response times.

The score supports judgment, and bedside assessment remains central. A child with a low score and a poor appearance needs assessment, and a child with a high score and a benign explanation also needs scheduled reassessment. The text on diagnostic uncertainty discusses how to plan that reassessment.

Where reasoning often fails

Three failures appear often. The first is attributing tachycardia to fever, pain or crying without reassessing after those causes have been treated. The second is normalizing the high score of a child who has been like this for hours. The third is taking too long to call for help for fear of bothering a more experienced colleague.

Sepsis is a common example of deterioration that gives warning signs beforehand. The text on sepsis in nursing triage shows how to recognize these signs early. Discussing deterioration cases in morbidity and mortality sessions helps the team find the weak links in the response chain.

How to communicate deterioration

When the score crosses the threshold, the way it is communicated influences the response. Handover structures, such as the situation, background, assessment and recommendation sequence, help nursing present the case objectively and ask for a concrete action. Stating the score value, the trend over the last hours and what is worrying at the bedside usually leads to a faster response than a vague account.

On the medical side, the response also needs to be clear. Agreeing when the child will be reassessed, what to do if the score rises further and who to call if the response is slow closes the loop. Recording this agreement in the medical record keeps the information from getting lost at shift change.

Shift change, in fact, is one of the highest-risk moments. The child who was being watched closely by the previous shift can become just another name on the next shift's list. Mentioning at handover which children had a high score, what the response was and what was agreed keeps attention on them and keeps the worsening trend from getting lost between one team and another.

Training with cases

Reading the vital signs curve and deciding to escalate are skills that can be trained. Fictional cases with vital signs in sequence allow practice of the central question: is this child improving, stable or worsening? The text with clinical reasoning exercises for pediatrics offers scenarios for this training, and differential diagnosis with clues helps in thinking about the causes of deterioration.

How NEXO helps train this reasoning

In NEXO, the case of the day is fictional, with a real clinical presentation and up to six progressive clues, and you propose the diagnosis by looking for the ICD-10 code.

The limits on using the game are described on the NEXO Economy page. If you research clinical reasoning teaching, the page for Researcher explains how to get in touch. Professionals who write or review cases find information on Curator and on Creators, and institutions can learn about the proposal on the partnerships page. Training with differential diagnosis and clues and dealing with diagnostic uncertainty complete this training.

Frequently asked questions

What is a pediatric early warning score?

It is a sum of points assigned to changes in behavior, circulation and breathing, and a review found 105 articles from 12 countries on these tools (Simon Junior et al., 2022).

Do the scores get the prediction of deterioration right?

The area under the curve ranged from 0.73 to 0.91 in the validation studies reviewed by Brown et al. (Brown et al., 2018).

Does the score reduce mortality?

The EPOCH trial, in 21 hospitals in 7 countries, showed little change in mortality (Chong et al., 2022), and the effect depends on the team's response.

Does it work in low-resource hospitals?

It works as a process tool, and the EVAT collaboration implemented scores in 36 hospitals in 13 countries (Agulnik et al., 2022).

Is it worth using electronic calculation?

It is. In one study, calculation accuracy was 98.5 per 100 with an electronic system and 85.6 per 100 with manual calculation (Sefton et al., 2017).

Take-home messages

• Deterioration in a hospitalized child usually gives signs hours in advance.
• Early warning scores gather scattered clues into a number that triggers an action.
• Accuracy is moderate to good, and the impact depends on the team's response.
• A high and persistent score is a warning sign, not a new normal.
• The concern of nursing and of the family deserves a response.

References

1. Agulnik A, Cárdenas A, Carrillo AK, et al. Clinical and organizational risk factors for mortality during deterioration events among pediatric oncology patients in Latin America: A multicenter prospective cohort. Cancer. 2021. PMID 33524166.
2. Agulnik A, Ferrara G, Puerto-Torres M, et al. Assessment of Barriers and Enablers to Implementation of a Pediatric Early Warning System in Resource-Limited Settings. JAMA Netw Open. 2022. PMID 35262714.
3. Agulnik A, Gonzalez Ruiz A, Muniz-Talavera H, et al. Model for regional collaboration: Successful strategy to implement a pediatric early warning system in 36 pediatric oncology centers in Latin America. Cancer. 2022. PMID 36161436.
4. Agulnik A, Gossett J, Carrillo AK, et al. Abnormal Vital Signs Predict Critical Deterioration in Hospitalized Pediatric Hematology-Oncology and Post-hematopoietic Cell Transplant Patients. Front Oncol. 2020. PMID 32266139.
5. Brown SR, Martinez Garcia D, Agulnik A. Scoping Review of Pediatric Early Warning Systems (PEWS) in Resource-Limited and Humanitarian Settings. Front Pediatr. 2018. PMID 30671424.
6. Chong SL, Goh MSL, Ong GY, et al. Do paediatric early warning systems reduce mortality and critical deterioration events among children? A systematic review and meta-analysis. Resusc Plus. 2022. PMID 35801231.
7. Garza M, Graetz DE, Kaye EC, et al. Impact of PEWS on Perceived Quality of Care During Deterioration in Children With Cancer Hospitalized in Different Resource-Settings. Front Oncol. 2021. PMID 34249696.
8. Graetz D, Kaye EC, Garza M, et al. Qualitative Study of Pediatric Early Warning Systems' Impact on Interdisciplinary Communication in Two Pediatric Oncology Hospitals With Varying Resources. JCO Glob Oncol. 2020. PMID 32673079.
9. Graetz DE, Giannars E, Kaye EC, et al. Clinician Emotions Surrounding Pediatric Oncology Patient Deterioration. Front Oncol. 2021. PMID 33718195.
10. Sefton G, Lane S, Killen R, et al. Accuracy and Efficiency of Recording Pediatric Early Warning Scores Using an Electronic Physiological Surveillance System Compared With Traditional Paper-Based Documentation. Comput Inform Nurs. 2017. PMID 27832032.
11. Simon Junior H, Schvartsman C, Sukys GA, et al. Pediatric emergency triage systems. Rev Paul Pediatr. 2022. PMID 35858040.
12. Solevåg AL, Eggen EH, Schröder J, et al. Use of a modified pediatric early warning score in a department of pediatric and adolescent medicine. PLoS One. 2013. PMID 23991121.