Implementation and adherence to the Bedside Paediatric Early Warning System (BedsidePEWS) in a pediatric tertiary care hospital

Background Clinical deterioration in children admitted to hospital wards often manifests through signs of increasing illness severity that may lead to unplanned Pediatric Intensive Care Unit admissions or cardiac arrest, if undetected. The Bedside Pediatric Early Warning System (BedsidePEWS) is a validated Canadian scoring system used at a large tertiary care children’ hospital to prevent critical illness and standardize the response to deteriorating children on the wards. Methods A 6-month audit was performed to evaluate the use of the BedsidePEWS, escalation of patient observations, monitoring and medical reviews on the wards in 2018. Two research nurses performed weekly visits to the hospital wards to collect data on BedsidePEWS scores, medical reviews, type of monitoring and vital signs recorded. Data were described through means or medians according to the distribution. Inferences were calculated either with Chi-square, Student’s t test or Wilcoxon-Mann–Whitney test, as appropriate (P <0.05 considered as significant).


Introduction
Clinical deterioration in children admitted to hospital wards is often detected through signs of increasing severity of illness, which otherwise may lead to unplanned Pediatric Intensive Care Unit (PICU) admissions or cardiac arrest [1,2]. Pediatric Early Warning Systems (PEWS) have been devised to detect signs of clinical instability with the purpose to activate appropriate and timely interventions to prevent evolving into critical illness. The activation of an intervention is usually recommended when one or more clinical indicator thresholds are breached (as in trigger systems) or if a certain score, which synthetizes the deviation from clinical stability of a set of predefined vital signs or clinical indicators, is reached (as in scoring systems) [3]. In a case control study that evaluates the performance of 18 PEWS, scoring systems performed better than trigger systems, reaching higher specificity and Area Under Roc Curve (AUROC) values than trigger systems. The Bedside Pediatric Early Warning System (BedsidePEWS) ranked second, with an AUROC of 0.88 (CI 0.85 to 0.91) [4].
The BedsidePEWS was developed by Parshuram et al. in 2011 [5] who conducted a multicenter evaluation of its performance, and the first test of its effect on significant patient outcomes through a clustered randomized controlled trial [6]. BedsidePEWS performance was reported to be more effective than the retrospective nurses' ratings of deteriorating patients urgently admitted to the PICU (BedsidePEWS AUROC 0.87, CI 0.85 to 0.89 vs Nurses' ratings AUROC 0.83, CI 0.81 to 0.86, P < 0.0001) [5]. In the Evaluation of Processes of care and Outcomes of Children in Hospital (EPOCH) trial, the BedsidePEWS enabled to significantly reduce the severity of illness of patients urgently admitted to the PICU, but had no significant effect on child mortality [6]. Some evidence of PEWS effectiveness was reported in a systematic review where PEWS were part of a wider systems approach involving a Rapid Response Team (RRT) or Medical Emergency Team (MET) ready to respond to deteriorating patients, with a potential to reduce 31 deaths/10.000 hospital admissions [7]. However, evidence is limited since most studies are single-centered and mainly performed in specialist hospitals [8]. No studies to date have described the implementation of the BedsidePEWS in a tertiary care pediatric hospital. More research is still required to describe compliance to PEWS response algorithms and process outcomes such as the frequency and type of vital signs monitored, the frequency of physician reviews and compliance to the BedsidePEWS score calculation according to the BedsidePEWS score range. Therefore, a 6-month audit was performed to assess the use of the BedsidePEWS and escalation practices on the wards of a tertiary care pediatric hospital in 2018.

Methods
The aim of this audit was to describe compliance with the use of the BedsidePEWS in a large tertiary care children's hospital: 1) Use of the correct chart by age group according to the patient's age; 2) Frequency and type of vital signs, monitoring and documented physician reviews in all children, children with high BedsidePEWS scores (scores ≥ 7 and ≥5 <7) and children with chronic health conditions (CHC). CHC are defined as illnesses that last for 3 months or more or require long term care [9,10]; 3) Presence and adequacy of score calculation. The audit was performed from July to December 2018 by two independent research nurses, members of the research team and experts of BedsidePEWS. The reviewers collected data and performed data evaluation in collaboration with the audit team. A sample of ≥10 charts in every ward, available at the time of the audit, was selected to collect data regarding the last 24 hours of admission.

Analysis
Data were described through means or medians, as appropriate according to the distribution, tested with the Kolmogorov-Smirnov test. Inferences were calculated with chi-square, Student's t test or Wilcoxon-Mann-Whitney test, as appropriate (a p-value <0.05 is considered statistically significant).

Setting
The audit was performed in a large tertiary care pediatric hospital in Central Italy. The hospital has a total of 607 beds across two sites, 25 hospital wards and 4 PICUs/Cardiac Intensive Care Units (CICU) and 1 Neonatal Intensive Care Unit (NICU), with a total of 465 beds at the main site where the audit for the present study was performed. There are also 4 High Dependency Units (HDU) for cardiac, respiratory and hematology-oncology patients with a total of 77 beds for non-invasive ventilation (NIV), inotropic support and isolation if required.
The BedsidePEWS was first introduced in 2014, during the hospital's involvement in the EPOCH randomized cluster clinical trial, because the children's hospital where the current study was conducted was randomly allocated to the intervention study group. A daily hard copy paper, age specific, BedsidePEWS chart was customized to document vital signs and calculate the score. A color system was used to define score thresholds and rapidly identify any deviance of vital signs from normal values, according to five different age groups (<3 months, 4-12 months, >1<4 years, 4-12 years, > 12 years). The BedsidePEWS consists of 7 clinical indicators: heart rate (HR), respiratory rate (RR), systolic blood pressure (SBP), work of breathing (WoB), capillary refill time (CRT), oxygen therapy (OT) and percutaneous oxygen saturation (SpO2). Other vital signs, such as temperature (T) and level of consciousness (LoC) are also required by the hospital Vital Signs (VS) protocol. A score matched response algorithm was embedded in the chart to define: 1) score calculation and VS documentation frequency; 2) timing for medical and nursing review; 3) recommended distribution of high score patients among the nursing team; and, 4) type of monitoring (continuous cardiac monitoring, SpO2 or intermittent). A Vital Signs (VS) protocol with the BedsidePEWS embedded was edited. The documentation of a minimum of 5 clinical indicators of the BedsidePEWS is required to calculate the score. The RRT/MET is called at a score of 7 or more. The score was also validated at the hospital's Bone Marrow Transplant Unit showing a good screening performance in this high-risk patient population [11]. An educational program for all PICU physicians and ward healthcare staff was provided through classroom lectures and case scenarios before implementation. Continuing education for ward staff was provided through morning briefings, case reviews and performance audits on the wards, quality improvement or morbidity and mortality meetings.

Implementation of the BedsidePEWS
The implementation of the BedsidePEWS involved the hospital nursing and medical directors, administrators, the continuing education and research service, and other providers through a multidisciplinary and inter-departmental approach. Since the hospital was introducing the BedsidePEWS within the context of the EPOCH randomized clustered clinical trial, the EPOCH BedsidePEWS implementation team led the implementation process by running weekly meetings during the run-in period followed by two weekly and monthly meetings. Any implementation issues were discussed and resolved: 1) staffing analysis and planning, including nurse/patient ratios, medical roles and consulting processes on the wards; 2) VS monitoring tools and equipment availability, including electrocardiogram monitoring (ECG) and SpO2 monitors on the wards; 3) BedsidePEWS chart customization including additional items relevant to patient care (e.g. pain assessment, care plan, fluid balance); 4) Score matched care recommendations (SMCR) translation and adaptation to the local available nursing and medical levels of response to clinical deterioration on the wards, according to the hospital organization; 5) communication plan and editing of a local VS protocol to standardize monitoring practice on inpatient and day care wards; 6) customization of education modules on BedsidePEWS for all ward and PICU clinicians; 7) maintenance and quality control through monthly performance meetings and documentation audits.
The multidisciplinary research and implementation team was also continuously available to define strategies to facilitate the change process. Focus groups were performed for one month to throw light on issues related to impaired overnight patient monitoring, availability of appropriate monitoring equipment, low nurse staffing ratios, communication between nurses and physicians or among physicians, and PICU involvement in MET calls for non-deteriorating patients with chronic conditions triggering the system because of elevated BedsidePEWS scores. Interventions to improve critical situations were designed. On-the-ward meetings were held and strategies to improve overnight VS monitoring were discussed with local ward teams. Additional SpO2 monitors were obtained. Additional nursing staff was provided to all wards (one additional nurse during the day hours/shift) to facilitate the required compliance to VS monitoring according to patient risk. The role of this team of seven additional nurses was to support patient monitoring, educate ward staff, perform quality checks and act as a reference for the continuing improvement and uptake of the system. The BedsidePEWS was integrated into the handover hospital protocol, to improve communication on high-risk children between day and night shift physicians or the use of the score in nursing communication with the Medical Emergency Team (MET) or the PICU. The importance of using clinical judgement over the BedsidePEWS score was highlighted on the VS protocol and discussed for specific patient populations with respiratory or cardiac chronic diseases, where the score trend is more relevant than the absolute score to detect clinical deterioration.
A total of 259 (99% of the physicians and nurses on the EPOCH wards) healthcare professionals were trained in 2013 through two-hour meetings.

Results
During the study period, a total of 522 VS and score calculations on 177 patient clinical records were observed from 13 wards (4 Pediatric Specialty Units, 4 High Dependency Units, 3 Surgical Pediatric wards, and 2 General Medical Pediatric wards). Children with chronic health conditions (CHC) were 122 (69% of the observed patients), and 25 (14%) children had signs of physiological instability with a BedsidePEWS score ³5. The patient characteristics are shown in Table 1. The appropriate age specific chart was used in 99% of the observations. Median BedsidePEWS scores were significantly different between children with acute illnesses and CHC (P<.001) ( Table 2).  Table 3. Of the 522 score calculations, in 56 (11%) the score was wrongly calculated, and in 27 (5%) no score was documented. In >95% of the observations, heart rate (HR), respiratory rate (RR), SpO2, oxygen therapy (OT), capillary refill time (CRT), work of breathing (WoB) and level of consciousness (LoC) were documented. Temperature (T) was recorded in 79% (n=410) and systolic blood pressure (SBP) in 89% (n=465) of the VS documentation events. (Table 4). Abbreviations: CHC, chronic health conditions; CRT, capillary refill time; HR, heart rate; LoC, level of consciousness; OT, oxygen therapy; RR, respiratory rate; SpO2= percutaneous oxygen saturation; SBP, systolic blood pressure; T, temperature; WoB, work of breathing.
Children at low risk of clinical deterioration with BedsidePEWS scores between 0-2 were reviewed by a physician at least once a day, children with BedsidePEWS scores in the 3-4 score ranges were reviewed within 6 hours in 35% of the observations. Children at higher risk were reviewed within 4 hours in 50% (n=20) in the 5-6 score range and within 2 hours in 42% (n=16) in the ≥7 score range (Table 5).

Discussion
This study explored monitoring and physician review escalation practices matched to BedsidePEWS scores in a large tertiary care pediatric hospital. We found that vital signs monitoring was still suboptimal (≥ 4 hourly) in patients at risk of clinical deterioration with a BedsidePEWS > 4. Nurses tended to use PEWS scores using their own clinical judgement through a process of intuition based on experience, to confirm their assessment of evolving critical illness rather than using solely the score [15]. Repeated exposure to children with chronic illnesses, prevalent in this audit in the higher BedsidePEWS group, might have determined a ≥ 4 hourly frequency of documented patient observations. In addition, workload and nurse/patient ratios ranging from 1/4 in HDUs to 1/10 on pediatric wards may have affected the frequency of documented patient observations, vital signs and monitoring with higher nurse/patient ratios leading to an increased risk of missed care [12,13,14].
We found that compliance with the VS monitoring audit timing thresholds was higher in the group of patients admitted for acute clinical conditions (n = 38/55 complying patients, 69%) compared to children with chronic illnesses (n = 57/122 complying patients, 47%), which constituted the majority of our observations (122, 69%). Patients with chronic diseases, such as congenital heart disease, metabolic diseases or neurological impairment may have altered basal vital signs and higher thresholds due to their condition. In these patients, trends rather than absolute values of the BedsidePEWS score, normally higher than usual, were taken into account [16]. However, proximity to PICU admission, up to 12 hours, has been found to increase with BedsidePEWS scores > 6, making this a population of children at increased risk of clinical deterioration [17]. Patients with chronic illnesses constitute an increasing proportion of the total number of pediatric hospital admissions with a higher technological dependency on digital devices, increased risk of PICU readmissions, hospital length of stay and mortality [18,19,20]. The impact of the use of BedsidePEWS on unplanned PICU admissions of this vulnerable patient group needs to be further explored.
Patient review from a physician is another essential element of escalation of care in relation to BedsidePEWS scores. Less than 50% of the patients with a BedsidePEWS > 2 were reviewed by a physician according to the audit timing thresholds by BedsidePEWS score range. This may be due to several factors, including low situational awareness for deteriorating patients, appropriate clinical judgement applied to stable patients with chronic diseases, workload and production pressure, low nursing empowerment to call for help, and internal hierarchies slowing down the referral process [21,22]. Other staffing layers of intervention including charge nurses, clinical nurse specialists, medical fellows or residents may be considered in this and similar contexts to increase patient review within appropriate timelines.
The score was correctly calculated in 84% (n = 438) of the VS sets observed, 5% (n = 27) were missing, and the rest were incorrect (n = 56, 11%). Missing scores, calculation errors and delay in patient referral have been previously reported to be possibly influenced by low situational awareness and other factors affecting decision making [23,24]. Chapman found rates of calculation errors that reached 17.5% and up to 6.4% of missing scores. Scoring errors were significantly higher in patients urgently admitted to PICU or deaths on the wards [7]. Scoring errors and underscoring often occur during the initial phase of clinical deterioration. The underscoring of deteriorating patients has often been found to reflect nurses' perceptions of patient's risk rather than being the result of a calculation error [25]. In some cases, nurses reported that they may underestimate the vital sign rather than record an abnormal observation that may require, by protocol, an escalation of care [26]. Moreover, errors were reported much more likely in observation sets where some vital signs were missing, or when a patient starts to become unstable, reducing the opportunity to achieve an early detection of deterioration [27]. The advent and use of electronic medical records (EMRs) will overcome this issue by providing an electronic registration of VS and score calculation, possibly also giving prompts to recall VS assessment.

Declarations
Ethics approval and consent to participate: The study was classified as a service evaluation initiative and therefore exempt from ethics committee approval.

Consent for publication: Not required.
Availability of data and materials: All data relevant to the study are included in the article.
Competing interests: None declared.
Funding: This study received no direct funding.
Authors' contributions: OG was involved in the conception and the design of the study, data collection, contributed to data analysis and interpretation and led the writing of this paper. FF was involved in data collection, contributed to data analysis and to drafting this paper. MLCDA was involved in data analysis and the drafting of this paper. IDO, CC were involved in the design and drafting of the paper. ET and MR are the guarantors.