Paediatric Early Warning Scores (PEWS)are used in hospitalised patients to detect physiological deterioration and is being used increasingly throughout healthcare systems with a limited evidence based. There are two versions in general use that can lead to a clinical response, either by triggering an action or by reaching a ‘threshold’ when graduated responses may occur depending on the value of the score. Most evidence has come from research based on paediatric inpatients in specialist children’s hospitals, although the range of research is expanding, taking into account other clinical areas such as paediatric intensive care unit, emergency department and the prehospital setting. Currrently, it is uncertain whether a unified system does deliver benefits in terms of outcomes or financial savings, but it may inform and improve patient communication. PEWS may be an additional tool in context of a patient’s specific condition, and future work will include its validation for different conditions, different clinical settings, patient populations and organisational structure. The incorporation of PEWS within the electronic health records may form a keystone of the safe system framework and allow the development of consistent PEWS system to standardise practice.
- clinical procedures
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Clinical deterioration in hospitalised children is often preceded by periods of physiological instability that can be detected through the monitoring of vital sign observations and other clinical signs at the bedside. When recognised, this offers the opportunity for intervention with the potential to prevent adverse events such as cardiac arrest and death. However, these ‘early warning’ signs can be overlooked or inadequately treated by healthcare staff.1
Paediatric Early Warning Scores (PEWS) have been advocated as a mechanism to aid healthcare professional’s recognition of children ‘at risk’, prompting increased monitoring and escalation to staff with the appropriate emergency and critical care skills.2 Although there is no universally agreed definition, PEWS are generally a composite of a number of vital signs and other clinical indicators indicating organ dysfunction. Two main types exist. Score-based PEWS assess a number of clinical indicators against a scoring matrix, which quantifies the level of deviance from the accepted normal reference range. Excessively low or high values are considered a cause for concern. The summation of the individual clinical indicators score is commonly referred to as the early warning score and is considered a proxy for indicating the overall well-being of the child. Positive scores indicate children at risk of deterioration, with higher scores associated with increasing risk.3 These score-based PEWS are often accompanied by an escalation algorithm that indicates the action to be taken by the healthcare professional at each score. While score-based PEWS quantify abnormalities across a number of clinical parameters, trigger-based PEWS rely on the breeching of a single predetermined threshold. They are generally considered easier to use as there is no scoring matrix nor is summation of individual indicators required. The threshold for ‘abnormality’ is generally higher than that seen in score-based PEWS in order to prevent excessive false alarms. Both score-based and trigger-based PEWS rely on regular universal surveillance of patients to ‘track’ their progress and ‘trigger’ a sequence of events when predetermined criteria are met.4 Some indicators, such as heart and respiratory rate, may require age-appropriate criteria.
Early warning scores for deteriorating adults were first described in the late 1990s; however, it was not until 2005 that the first PEWS were published. The Brighton PEWS5 were derived from an adult tool with a 3×3 scoring matrix that assessed the child’s behaviour, cardiovascular and respiratory status. Additional score weighting for continuous nebulisers or persistent postoperative vomiting were added in response to the ‘failure’ to identify three children who the staff felt ‘should have scored higher’. No formal validation of the Brighton PEWS was undertaken. This unstructured method of development, based on expert opinion, sentinel events and ‘knee jerk’ response to individual cases of failed identification characterised the development of PEWS over the subsequent years and contributed to the proliferation of unvalidated PEWS, which are now in existence.
In 2008, a landmark confidential enquiry into childhood death conducted a multiprofessional case note review of 89 children who died in hospital1 Avoidable factors were present in 21% and potentially avoidable factors in 49% of deaths reviewed by the panel. Deficiencies in the recognition of serious illness in children were highlighted as a particular area of concern, and hospitals were recommended to adopt a standardised rational monitoring system with an embedded PEWS. This report was a key driver in the adoption of PEWS in the UK.
By 2013, the majority of hospitals in the UK reported having PEWS in place.6 Only a third were based on a previously published PEWS, with the remainder using a mix of systems adapted from other hospitals and those purposely designed for the individual unit. In all, 47 differing component parameters were identified, but respiratory and heart rate remained the most frequently used.
Where are they used and how do they perform?
Much of the PEWS literature has concentrated on specialist children’s hospitals.4 Two recent systematic reviews on PEWS and similar ‘track and trigger’ systems to identify deterioration in hospitalised children have been undertaken. The first identified 33 distinct PEWS from 55 papers.4 The second identified 38 primary studies on original and/or modified PEWS from 90 papers.2 Both noted considerable variety in the number and type of component parameters and scoring strategies. The heavy reliance of vital signs parameters was a feature of all the PEWS identified. Despite the wide range of PEWS, there remains no consensus on what type of PEWS should be implemented or on the constituent parameters. However, a head-to-head comparison of 18 PEWS did highlight significant differences in performance in a single-centre retrospective study.7
Despite the relatively large number of publications on PEWS, the evidence to support their use remains limited, and it is still not clear whether they improve patient outcomes.2 4 Implemented as a stand-alone intervention, the evidence to support the effect of PEWS on important patient outcomes such as death, unplanned transfer to paediatric intensive care unit (PICU) and other adverse events such as cardiac and respiratory arrest is very limited. When implemented as part of a package of intervention, such as education, access to teams of specialist personnel to ‘rapidly respond’ to children with developing critical illness, the picture is slightly different. There is moderate evidence that PEWS may decrease mortality and reduce the incidence of cardiorespiratory arrests. The level of evidence was weakened as it was all derived from observational studies. The majority of studies were conducted in specialist children’s hospitals with few studies examining PEWS in the non-acute setting.8 9
This year marks the publication of the first prospective multicentre randomised controlled trial of PEWS.10 The Bedside PEWS are notable for the rigour of development and prior validation in differing populations.11 12 Advocates of PEWS hoped that this landmark study would shine a light on their benefits. Disappointingly, the findings did not support the use of PEWS to decrease all-cause hospital mortality. Neither did PEWS have a positive effect on 21 other predefined outcomes, including cardiac arrest, death in PICU and emergency calls. There was, however, one positive finding. A composite measure indicating late PICU admission did show a significant reduction in the PEWS group. The overall rate of mortality was significantly lower than anticipated, and this may mean that the study was underpowered despite the enrolment of 144 539 patients.
A national approach to PEWS has been adopted in Scotland and Ireland,13 14 with the aim of standardising practice, improving communication and maximising resources. There have been calls for corresponding national PEWS for the UK, although it is uncertain whether a unified system delivers benefits in terms of outcomes or financial savings.
Emergency department (ED)
Other clinical settings have adopted and adapted PEWS. A number of studies have examined the ability of PEWS identify children in the ED who need to be hospitalised, and key studies are summarised in table 1. Instituting PEWS in the ED may have benefits in terms of continuity of assessment from first contact through to discharge. Performance appears to be variable, but it is unclear whether this is due to differences in patient populations, PEWS characteristics or other factors. A large prospective cohort study examined the performance of 10 differing PEWS in a single ED in the Netherlands.3 Six PEWS were classified as score based with the remaining four noted to be trigger based. Overall, the area under the receiver operator characteristic curve (AUROC) for predicting hospitalisation was found to be poor to moderate (range: 0.56 (95% CI 0.55 to 0.58) to 0.68 (95% CI 0.66 to 0.69)), and none of the PEWS had a high sensitivity as well as a high specificity. No statistical difference was detected between the differing systems; however, when the PEWS were placed in rank order, score-based systems outperformed trigger-based systems.
Two differing PEWS underwent a head-to-head comparison to see if they could act when used on an undifferentiated population.15 Both Brighton and Children’s Observation and Severity Tool PEWS scores performed similarly. A score of ≥3 had good specificity but poor sensitivity for predicting hospital admission and significant illness. Therefore, high PEWS should be taken seriously, but a low score is poor at ruling out the requirement for admission or serious underlying illness. Overall, PEWS were better at detecting significant medical illness compared with detecting the need for admission. They performed poorly in detecting significant surgical illness, but findings indicated they may be particularly useful in evaluating respiratory illness in the paediatric ED.
Studies have also explored whether PEWS can be used to detect children presenting to the ED who are in need of a PICU admission. A large US prospective study of 12 306 patients found that an elevated PEWS at initial assessment in the ED was associated with the need for admission directly to the PICU (AUROC 0.79).16 A smaller retrospective case-controlled study examined the ability of an ED-specific PEWS to differentiate between children requiring admission to the ward versus admission to intensive care.17 At a score of 3, the PAWS was able to identify children requiring PICU admission with a sensitivity of 70% and a specificity of 90% with an AUROC of 0.86. Comparative performance of the 10 PEWS3 revealed higher levels of performance in predicting PICU admission (AUROC range 0.60–0.82) and score-based systems continued to outrank trigger-based systems.
PEWS have been used for populations considered to be particularly high risk for adverse events. Two studies have examined the ability of PEWS to prevent readmission to the general18 and the cardiac PICU.19 In both studies, higher PEWS were associated with an increased risk of readmission; however, implementing a pretransfer PEWS assessment did not significantly reduce readmissions to the cardiac PICU.19 For the general PICU, analysis of the retrospective data indicated that implementation would overwhelm the resources available due to delayed discharges and inappropriate readmission.18 Both studies indicate that PEWS alone is unlikely to be an effective tool to identify and prevent readmissions to PICU and assessment needs to include other factors such as chronic disease variables.
PEWS have also been implemented in the haematology/oncology and bone marrow/stem cell transplants populations.20–22 One study reported an increase in days between ward cardiac arrests, although it is not clear if this effect was sustained or significant.21 The Bedside PEWS also performed very well with an AUROC of 0.9 in a retrospective analysis,22 indicating a high level of accuracy in this specialist population.
Some centres have developed condition-specific PEWS, largely as a consequence of perceived poor performance of existing PEWS in these specific patient groups. The Burn Centre PEWS23 and the modified PEWS for cardiac and renal populations24 were both developed using expert opinion and relevant literature rather than data. Both report being well received by staff; however, neither has undergone any form of formal evaluation of validity or utility so it is unclear whether these specialist PEWS outperform those that have undergone multicentre evaluation in differing populations, such as the Bedside PEWS.8 11 12 PEWS developed for children with cardiac conditions have demonstrated superior performance to generic PEWS (AUROC 0.917 vs 0.785) in a retrospective validation,25 and modification of PEWS to identify children with hand, foot and mouth disease demonstrated improved performance to the hospital-wide PEWS previously in use.26 However, both were retrospective, single-centre studies, and it is unclear whether implementation means these benefits are translated into improved clinical care.
There is a small body of evidence of PEWS utilisation in resource-limited countries for the general27 28 and oncology population.29 Implementation of modified PEWS decreased the number of unplanned PICU transfers and improved PICU utilisation in Guatemala and assisted in identifying children with cancer who would benefit from more intensive monitoring in the intermediate care environment.30 A prospective nested case-controlled study in Malawi indicated that a locally developed PEWS could identify children at high risk of death.31 However, further analysis revealed that children who died were much more likely to have no PEWS or vital signs recorded at all. Implementation of PEWS in these environments is likely to be more complex and needs to take account of the resources consumed in recording PEWS and balance this against diagnostic accuracy.
Are PEWS alone enough?
The adoption of PEWS has been widespread, and many were implemented in the absence of clear evidence of efficacy. It is now accepted that implementation of PEWS in isolation is unlikely to change outcome. Rather PEWS need to be embedded into a system-wide approach that addresses the multidisciplinary and complex nature of healthcare and improves individual and team situational awareness.32 33 Recent recommendations from NHS Improvement and The Royal College of Paediatrics and Child Health highlight six core elements that underpin a ‘safe system’ framework for hospitalised children (table 2).34
A recent study explored the experiences of nurses using PEWS and factors they considered to facilitate and inhibit their application.35 Lack of interdisciplinary awareness, particularly from members of the medical team, and the failure to integrate clinical judgement were noted to impede PEWS in clinical practice. Meaningful integration of clinical judgement into PEWS has been previously noted in a qualitative study of both nurses and physicians.36
There is now increasing interest in incorporating PEWS into the electric health record (EHR). This may improve accuracy and efficiency when compared with paper-based systems and offers the potential to automate the escalation of children with elevated scores.37 Using PEWS within an EHR also offers the opportunity to collect large sets of data to further develop and validate the evidence base for PEWS.38
Data from an EHR has been used to adapt a well-validated adult risk index to children using continuously age-adjusted measures.39 This may be particularly useful to hospitals seeking a single risk model applicable across both adults and paediatrics. Two recent studies also used an existing EHR to develop and evaluate a machine learning-based algorithms to predict the need for PICU for newly hospitalised children.40 41 Both performed significantly better than existing PEWS and have the advantage of evolving over time, eliminating the challenges associated with retrospective data.42 More sophisticated techniques that compare patients baseline data derived during periods of stability to contemporaneous postoperative data may offer the opportunity to analyse complex data in real time.43
Despite widespread use, the evidence base for PEWS remains limited. There remains no consensus on the characteristics of the most effective PEWS, and efforts to evaluate existing systems have been hampered by a lack of robust, valid and clinically meaningful outcomes.44 Calls for a national UK PEWS have not, as yet, been answered despite potential benefits in terms of standardisation, communication and research. Priorities for PEWS are presented in box 1.
Priorities for Paediatric Early Warning Scores (PEWS) development and research
Agree robust, valid and clinically meaningful outcomes to evaluate PEWS
Determine the predictive characteristics of PEWS in differing clinical settings, patient populations and organisational structures.
Evaluate the effect of the safe system framework on mortality and morbidity.
Investigate the barriers and facilitators to PEWS implementation.
Consider the development of a national PEWS to standardise practice.
Understand the role of ‘human factors’ in PEWS implementation and explore the role of technology in the identification and escalation of children at risk of deterioration.
Early identification of children at risk of clinical deterioration is an important step in preventing adverse events. Use of PEWS may assist recognition of the early signs of serious illness by identifying and quantifying deviations from the normal range of vital signs. Ongoing monitoring may facilitate healthcare professionals to ‘track’ a child’s progress and, if potential signs of deterioration are detected, ‘trigger’ key interventions, such as increased frequency of monitoring, senior staff review and transfer to a higher level of care. Their use may aid individual and team situation awareness of the children at risk of deterioration, particularly for junior staff or those new to paediatrics.32
There are many different PEWS currently in use. The majority are of unknown validity, and most studies have only been conducted in specialist children’s hospitals. Studying the predictive characteristics of all the existing PEWS may be helpful in determining which work best in differing clinical settings, patient populations and organisational structures. It is likely that differences may emerge, given that performances requirements for PEWS in a critical care area may be different to that used on a general ward with patients of lower acuity illness. Understanding and defining the properties that lead to high performance may support a more evidence-based approach to future PEWS development. Incorporation of additional clinical parameters may still be required to identify the early signs of deterioration in specific conditions such as congenital heart disease and biliary atresia.
Despite the natural appeal of PEWS, it is important to acknowledge that they will not identify all children at risk of deterioration, whether due to the speed or the mechanism of deterioration. Therefore, it is essential that all clinical staff are trained to recognise common patterns of deterioration and not to rely on PEWS as the sole mechanism to identify children at risk. However, they may be a valuable adjunct to clinical decision making, particularly for junior staff.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Commissioned; externally peer reviewed.