Critical illness in the patient and death can potentially be predicted and prevented. Deterioration of the clinical condition of hospitalised patients is often preceded by physiological changes up to 24 hours before death. Despite this several reports show that lack of identification and proper actions in patients developing acute and critical illness remains a problem.

The purpose of this study is to investigate if Paediatric early Warning Score (PEWS) optimises identification of acute and critically ill children and prevents life-threatening situations. The hypothesis is that implementation of PEWS supported by directions for action algorithms for intervention will have an impact on number of unplanned transfers to intensive care in already hospitalised children.

This study is a multi-centre randomised controlled intervention study designed within a Complex Intervention framework; the study sheds light on the problem, validation of the data collection instrument, testing of the intervention and evaluation. The study involves all paediatric departments and some acute departments in Central Denmark Region. The study is designed as a randomised controlled intervention study where children are randomised to one of two different Paediatric Early Waning Score models.

Development and implementation of PEWS is expected to contribute to reduce the number of children developing acute critical illness, number of admissions to intensive care. PEWS is also expected to contribute to increase professional skills and competences in health professionals. It is expected that this study will contribute towards working with a joint PEWS model in Denmark. Last but not least it must be expected that a PEWS model will contribute to reducing the costs for society as an intensive care hospital bed is more expensive than a hospital bed at a general paediatric department.

Introduction Evolving critical illness in the patient and death can potentially be predicted and prevented. Observation studies indicate that deterioration of the clinical condition of hospitalised patients is often preceded by physiological changes up to 24 hours before death. Focus in adult patients has for a number of years been on development of risk scoring system, also called "Early Warning Score" (EWS) systematically measuring the patients vital status. International experience shows that systematic observation by EWS accompanied by action algorithms contributes to prevent cardiac arrest, death and transfer to intensive care.

In many areas children are different from adults; there are e.g. physiological and anatomical differences and an EWS system developed for adult can thus not be used in children. Challenges in assessing if a child is critically ill are related to the child's symptoms of serious illness often being uncharacteristic. Children can seem relatively unaffected until a short time before circulatory insufficiency and cardiac arrest.

Thus, there is a need for developing and investigating if an EWS model for ill children has a significant influence on children's development of acute conditions.

Background In Denmark, more than 100,000 children below the age of 15 years are admitted annually. International studies show that 8.5 - 14.0% of cardiac arrest incidences in intensive care units involve paediatric patients. Survival is only between 15 - 33% with subsequent significant neurological outcome in 35% of the surviving children.

The intention with an EWS system for admitted patients is to ensure that staff identifies a deterioration in the patient's condition in time. Each value releases a "score" and the total score determines if treatment should be initiated. International studies in adult patients have shown that up to 24 hours prior to development of cardiac arrest or life-threatening situations show signs of acute critical illness.

There are national and international models for EWS in adults. Despite this several international reports show that lack of identification and proper actions in patients developing acute and critical illness remains a problem.

A study of 126 deaths among English children showed that 89 deaths occurred in hospital; among those it was estimated that 63 (71%) could have been avoided. Lack of acknowledging the severity of the disease was emphasised as a decisive factor together with absence of measuring and interpreting physical signs correctly. It is paramount for timely care and treatment of children to work systematically with assessing the child's symptoms and the severity of these symptoms.

The quality of the acute treatment has come into focus with a report from the Danish Health and Medicines Authority. Several paediatric departments have establish simulation training in care and treatment of the acute and critically ill child to optimise competences in and ensuring adequate response of healthcare staff. Structured initial assessment of acutely admitted children - the paediatric triage model - comprises systematic assessment of the severity of the child's symptoms within the first few minutes. The implemented triage model determines the order of interventions.

More hospitals have established so-called Rapid Response Teams (RRT), a kind of acute team which can facilitate early intervention in a ward where the patient develops a life-threatening situation or cardiac arrest. The criteria for calling the RRT are a central factor in prevention of acute life-threatening situations and cardiac arrest and early notification of the RRT improves patient outcome. This project shifts focus from response to life-threatening situations and cardiac arrest to early warning of critical illness.

Improvement of patient safety is a focus areas in Denmark; this is shown by initiatives such as the Danish Safer Hospital Programme aimed at reduced mortality and patient harms as well as the Danish Quality Model aimed at improving quality of health services, reduce harms and create better patient pathways. Development of standards for systematic observation of patients and a national clinical guideline for early warning of critical illness in adult patient are specific initiatives at national level

Children are different from adults Care and treatment of critically ill children is a challenge as children's symptoms of critical illness can be uncharacteristic. A critically ill child with sepsis or severely dehydration can seem relatively unaffected and the acute condition is identified by the affected vital parameters. Children's mechanisms to compensate are better than in adults as a child is able to maintain an almost normal blood pressure despite considerable loss of fluid. On the other hand, the child can turn critically ill fast when the limit for the child's ability to compensate is exceeded. Thus, it is important to acknowledge and act on the often subtle signs of acute and critical illness in the child as the prognosis for survival is very poor if the child develops cardiac arrest.

Experiences with EWS in adults show that the system cannot stand alone but should be seen in interaction with several factors. In this protocol Paediatric Early Warning Score (PEWS) is one of several interventions including:

• Education and training of staff
• Standardised monitoring
• Interprofessional guidelines
• Algorithms of action for intervention in the critically ill child.

When health professionals do not systematically observe, interpret and act adequately to changes in the child's condition, it will have serious consequences for the hospitalised child.

A Canadian developed PEWS model is one of the most studied model in larger studies. The model has been developed, tested and subsequently modified and it now consists of seven different parameters. The model includes measurement of blood pressure, which is a specific challenge in a paediatric population. Often children are upset when the blood pressure is measured, which causes the pressure to increase; this increase is not necessarily a sign of clinical exacerbation of disease. At the same time, blood pressure is a late finding in connection with evolving critical illness due to the child's ability to compensate; this could be an argument for not measuring the blood pressure of all hospitalised children. Moreover, the model does not include assessment of level of consciousness in children; the investigators believe this is an important factor for assessing if critical illness is evolving in children. The investigators wish to compare the Canadian model with the investigators "local" model, which includes assessment of level of consciousness but not measuring of blood pressure. Measuring of the blood pressure in the "local" model will be performed as an extra examination in children with a high score i.e. children at risk of evolving critical illness. Measurement of blood pressure will thus not be used as a screening in all hospitalised children as in the Canadian model. Apart from this, the two PEWS models are similar; the underlying action algorithms, age categories and cut off values for vital signs are similar.

In a PEWS review, concluded that there is a considerable variation in purpose, content and threshold for actions as well as lack of evidence for validity, reliability and applicability of the published PEWS models. Reliability is only investigated in one study and no studies report on the experiences of health professionals concerning the applicability of PEWS. In a study of PEWS from Great Britain, there was a large variation in use of parameters; 46 different parameters were used and more hospitals used PEWS models which were not validated or thoroughly investigated. This also applies to EWSs for adults in a Cochrane review, which also draws attention to the lack of randomised controlled studies (RCT) and studies investigating the applicability and clinical relevance according to the clinical staff. There is thus a need and demand for RCT designs to study PEWS but also to study reliability of the clinical staffs' perception of applicability.

This study investigates these core issues using a RCT design, which has not yet been conducted in a paediatric population.

Purpose The purpose of this study was to investigate if PEWS in hospitalised children optimises identification of acute and critically ill children and prevents life-threatening situations and thus reduce number of unplanned admissions to intensive care, cardiac arrest and death. Moreover, the study investigated validation of the clinical applicability of PEWS as well as the acceptability of the tool among healthcare professionals.

Design Multi-center randomised controlled intervention study within a Complex Intervention framework; the study sheds light on the problem, validation of the data collection instrument, testing of the intervention and evaluation.

Randomisation The departments will all have access to web-based randomisation programme, "Trialpartner". Each patient get his or her own number of randomisation. If a patient is readmitted during the inclusion period he or she will obtain the result of the initial randomisation.

Sample size and statistical analysis Data will be analysed on the intention to treat principles. Number of acute critical conditions will lead to consultation by a neonatal specialist or an anaesthesiologist and unplanned transfer to intensive care will be registered. Focus will be on the children's condition in the last 24 hours before documented acute critical or life-threatening condition (as well as observation level and actions in relation to the acute critical condition or unplanned transfer to intensive care). As PEWS must be able to identify children at risk of acute life-threatening condition and at the same time be a "window" for intervention, the data analysis will stop one hour prior to the occurrence of the acute life-threatening condition to avoid overestimating the effect of PEWS.

Background variables are: child's age, gender, discharge diagnosis, length of hospitalisation, triage level at admission, vital signs, respiratory status, temperature, level of consciousness, observation level, initiated interventions, time from observation of deteriorated condition to adequate intervention. This information will be collected from patient records.

Currently, a power calculation has been made based on accessible data from 2011. With a power of 80% and an expected reduction in number of transfers to intensive care of 30%, 4,000 children must be included in each group; the total study population is 26,800 children annually.

An interim analysis is planned half-way through the study. The study will be interrupted if there are significant differences between the two models concerning selected effect parameters.

Ethical considerations The intervention directly affecting hospitalised children will be measurement of pulse and saturation. Measurements are non-invasive and cause no pain or discomfort and are standard procedures in most acutely hospitalised children. It is thus not expected to cause discomfort or give rise to concerns in children and parents. It is important that patients' experience of anxiety, concern or irritation is addressed and dealt with. It is therefore of great importance that ward nursing staff is thought to communicate with and inform patients while they perform the measuring of vital parameters and that nursing staff in a professional and informative way can argue why the measuring is important. Several studies show that bedside observations of vital parameters and simple algorithms based on bedside observations can identify patients at risk of deteriorating. In an ethical perspective it is therefore unethical not to react to these findings and not to implement a new ward practice.

Inclusion Criteria:

• All children admitted to the participating departments
• All children examined at the acute paediatric assessment unit in Central Denmark Region.

Exclusion Criteria:

• Children admitted to neonatal wards
• Children admitted to intensive care units
• Children dead upon arrival to hospital
• Children admitted due to social-interaction problems.