Bedside Pediatric Early Warning System: A Validation Study
Overview
Conducted in an 80-bed pediatric oncology specialty hospital in the Netherlands, this research was initiated to evaluate the predictive utility of the adapted Bedside Pediatric Early Warning System (BedsidePEWS) score. This study aimed to explore the correlation of this score with unanticipated transfers to the pediatric intensive care unit (PICU) and instances of cardiopulmonary resuscitation (CPR) within a group of 1137 hospitalized pediatric oncology patients ranging in age from infancy to 18 years.
Using a Cox proportional hazard model, the study found a significant link between the modified BedsidePEWS score and the likelihood of unplanned PICU admissions or CPR events. Each incremental increase in the score correlated with a hazard ratio of 1.65 for unplanned PICU admission or CPR. The model’s predictive performance was reasonably accurate, with a C-index of 0.83 indicating moderate discriminative ability.
Evaluation of the modified BedsidePEWS score at a common threshold of 8 demonstrated its potential in clinical decision-making. Although the positive predictive value was modest (1.4%), the score’s high negative predictive value (99.9%) suggested that a score below this cutoff strongly indicated the absence of a need for escalated care.
In conclusion, the modified BedsidePEWS score exhibited a meaningful correlation with the likelihood of PICU transfer or CPR events in pediatric oncology patients. This scoring system holds promise as a valuable tool for informing clinical decisions, particularly in determining the optimal timing of PICU transfers, within this specific patient population.
Introduction
Hospitalized pediatric patients, particularly those undergoing oncology treatment, are at heightened risk of sudden and severe clinical deterioration, potentially leading to adverse outcomes like cardiac arrest or mortality. Pediatric oncology patients encounter elevated severity of illness, treatment toxicity, and compromised immunity, rendering them susceptible to rapid declines. Notably, a significant proportion (up to 38%) of these patients require unplanned admission to the PICU, frequently because of situations like sepsis and respiratory failure.
The mortality rates among pediatric oncology patients necessitating PICU admission are notably higher compared to the general PICU population, underscoring the urgency of identifying and addressing deterioration promptly. Clinical deterioration leading to unplanned PICU admissions is frequently linked to the highest PICU mortality rates. Pediatric oncology patients also face a more critical outlook in cases of cardiopulmonary arrest compared to their general pediatric counterparts.
To address these challenges, the Pediatric Early Warning System (PEWS) scores are used to detect early signs of deterioration in hospitalized children. The BedsidePEWS score, a widely used PEWS system, has demonstrated effectiveness in identifying children at risk for cardiopulmonary arrest and has shown promise in predicting clinical deterioration. However, while this score has been implemented in general pediatric care, its performance and validation in the specific context of pediatric oncology patients remain understudied.
This prospective cohort study aims to address this gap by evaluating the predictive ability of a modified BedsidePEWS score in forecasting unplanned PICU admissions or instances of cardiopulmonary resuscitation (CPR) specifically among hospitalized pediatric oncology patients. By assessing the applicability and accuracy of this scoring system, the study aims to contribute valuable insights into improving clinical decision-making and ultimately enhancing patient outcomes within this critical population.
Method
Inclusion Criteria
The study was conducted as a prospective cohort investigation at the Princess Máxima Center, a distinguished national referral institution for pediatric oncology in the Netherlands, spanning from February 2019 to February 2021. The research, which received endorsement from the hospital’s ethical review board (IRB protocol number 16-572/C), specifically targeted patients between the ages of 0 and 18 years who were hospitalized with an International Classification of Diseases in Oncology (ICD-O) diagnosis indicating pediatric malignancy (morphology code 1, 2, or 3).
Exclusion Criteria
Certain groups were excluded from participation, including outpatient cases and patients undergoing routine diagnostic or therapeutic procedures. Additionally, individuals with predefined care limitations, such as those on palliative care or bearing do-not-resuscitate orders that were established during care restriction registration, were not included in the study.
In this study, an enhanced version of the Bedside Pediatric Early Warning System (BedsidePEWS) was integrated, featuring adaptations like the inclusion of temperature and oxygen therapy categorization. This modified protocol was consistently employed and documented by nursing staff in the Electronic Health Records (EHR) of hospitalized patients. Nurses were empowered to manually input subcomponents for automated scoring or directly enter cumulative scores into the EHR, prompting relevant clinical actions. Ensuring protocol adherence involved multifaceted strategies, including recurrent training, onboarding programs, obstacle identification, and BedsidePEWS score review encouragement during rounds and shifts. The scoring process was closely monitored using a weekly dashboard to evaluate nursing performance across shifts.
The primary study objective centered on assessing a composite outcome encompassing unplanned PICU admissions and CPR instances. Notably, a single patient could experience this primary outcome event multiple times during their hospital stay, leading to the use of uninterrupted inpatient ward admissions as the unit of analysis.
The continuity of these ward admissions relied on factors such as primary outcome occurrence, patient discharge, care restriction registration, or reaching 18 years of age. New uninterrupted ward admissions were initiated following scenarios like PICU-to-ward transfers or hospital readmissions.
Secondary outcomes included minor clinical deterioration events necessitating escalated interventions without PICU transfer or CPR. Moreover, the study investigated broader clinical deterioration occurrences, combining significant deterioration requiring PICU transfer or CPR with minor deterioration events.
The study meticulously gathered patient data, hospitalizations, outcomes, vital signs, and Bedside Pediatric Early Warning System (BedsidePEWS) scores from Electronic Health Records (EHRs) using HiX software. A comprehensive data collection and preparation procedure is detailed in the Supporting Information section for further insight. This enhanced BedsidePEWS approach presents valuable insights into proactive patient care and clinical deterioration prediction.
Statistical Analysis
The modified BedsidePEWS score serves as an indicator of the severity of a patient’s illness, reflecting their evolving clinical condition. Recognizing the dynamic nature of patient health and its fluctuations during hospitalization, we employed a time-varying covariate approach. Through a meticulously calculated Cox proportional hazard model, we analyzed the modified BedsidePEWS score’s influence, accounting for the time interval between successive PEWS scores or clinical deterioration events, whichever occurred first. This methodology allowed us to comprehensively incorporate all documented modified BedsidePEWS scores for each patient, acknowledging the score’s varying patterns and potential recurrence of events within individual patients.
Cancer diagnosis groups, including solid tumors, hemato-oncology, and neuro-oncology, were included as significant prognostic factors within the model. Our approach remained consistent when investigating the association between the modified BedsidePEWS score and secondary outcomes, further reinforcing the reliability of our analysis (elaborated upon in the Supporting Information section).
To verify the robustness of our model, we conducted internal validation through Efron’s bootstrap method. In-depth assessment of prediction thresholds was undertaken, considering the score cutoff of 8, a pivotal point necessitating escalated care. This exploration was extended to cover a range of cutoff values, from 5 to 11, utilizing the last modified BedsidePEWS score preceding an event. Prediction measures, such as sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and the number needed to evaluate (calculated as 1/PPV), were systematically estimated, providing a comprehensive evaluation of the model’s predictive capability.
Furthermore, we performed a post-hoc qualitative analysis of the modified BedsidePEWS score within the 24-hour timeframe leading up to primary outcome events, enhancing our insights into its performance during critical periods. All statistical analyses were executed using the R-statistical software, version 3.6.2 (2019-12-12), along with associated packages, underscoring the rigor and reliability of our methodology. In line with best practices, the reporting of our validation study adhered to the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) guidelines
Results
Our analysis encompassed a comprehensive dataset comprising 5628 distinct ward admissions involving 1137 unique patients. A total of 119,813 modified Bedside Pediatric Early Warning System (BedsidePEWS) scores were meticulously considered, providing a robust foundation for our investigation.
The study cohort comprised patients with a median age of 8 years, as indicated by the interquartile range (IQR) of 4 to 14 years. Of the individuals under examination, 43.3% were female. Throughout the course of this study, a noteworthy subset of 103 patients experienced a combined total of 127 unplanned admissions to the PICU and three instances of CPR. Regrettably, within this subset, 14 patients encountered unfavorable outcomes, with their unfortunate demise occurring during PICU admissions or following CPR events. These insights contribute significantly to our understanding of patient demographics and highlight the critical nature of the clinical scenarios under scrutiny.
Scoring Compliance and Physician Involvement
Compliance with modified BedsidePEWS showed variations. Time intervals for score categories 0–3 and 4–5 met intended limits, but category 6–7 exceeded. Physician evaluation occurred in 85% of score ≥8 cases.
Performance of the modified BedsidePEWS: Unplanned PICU admission or CPR
Modified BedsidePEWS significantly associated with unplanned PICU admission or CPR (HR: 1.65, 95% CI: 1.59–1.72). Internal validation indicated a well-calibrated model (slope: 0.99) with a C-index of 0.83 [0.79–0.90]. A cutoff of 8, requiring escalated care, had high negative predictive value (99.9%), low positive predictive value (1.5%), 33.8% sensitivity, and 97.7% specificity. Different thresholds showed varying trade-offs in sensitivity, specificity, and predictive values.
Performance of the modified BedsidePEWS: Minor and any clinical deterioration events
Among the 1137 patients, 234 experienced 463 minor deteriorations, and 276 had 583 clinical deteriorations (including unplanned PICU admissions, CPR, and minor events). Modified BedsidePEWS was significantly linked to both minor and any clinical deteriorations (HR: 1.77 [1.71–1.83] and 1.75 [1.70–1.81]). Discrimination, calibration, and predictive values were akin to the primary outcome, indicating strong negative predictive values (99.6%) and modest positive predictive values (8.3%–9.6%) at a cutoff of 8.
Modified BedsidePEWS scores in the 24 hours prior to non-elective PICU admission or CPR
Common triggers were respiratory failure (35%), sepsis (16%), and cardiovascular failure (15%). Median [IQR] PICU stay was 2 [1–6] days, ranging 0–79 days. Bedside Pediatric Early Warning System (BedsidePEWS) scores showed increasing trends in the 24 hours preceding events, indicating potential clinical deterioration. Notably, 52% of events had scores below 8. Subset analysis revealed varied reasons for PICU transfers, including airway issues, hypertension, neurology, and postoperative care.
Conclusion
This prospective investigation assessed the predictive utility of a modified BedsidePEWS score in detecting clinical deterioration among hospitalized pediatric oncology patients. The study found significant associations between the score and unplanned PICU admission, CPR, and minor clinical deterioration. The widely used cutoff of 8 demonstrated a high negative predictive value of 99.9%, indicating strong accuracy. Notably, nuances emerged, including a moderate discriminative ability and considerations for optimal thresholds. The findings support the score’s role in clinical decision-making for timely interventions. This study stands out for its comprehensive approach, encompassing various patient subgroups and utilizing time-to-event data, providing valuable insights into the score’s performance in this specific context.
Oncology Related Tools
Other
Latest Research
