Validation and Performance of a Geriatric Early Warning Score (Gews) Versus the National Early Waring Score (News) in Predicting Clinical Deterioration in Frail Older Patients

Validation and Performance of a Geriatric Early Warning Score (Gews) Versus the National Early Waring Score (News) in Predicting Clinical Deterioration in Frail Older Patients | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Validation and Performance of a Geriatric Early Warning Score (Gews) Versus the National Early Waring Score (News) in Predicting Clinical Deterioration in Frail Older Patients Hilde Baeyens, Filip Haegdorens, Sven Martens, Marie-Elena Vanden Abeele, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6635964/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract Purpose : Early warning scores, such as the National Early Warning Score (NEWS), are less accurate in detecting clinical deterioration in frail older individuals due to age-related altered physiological responses. A Geriatric Early Warning Score (GEWS) was developed to take into account older adults’ frailty. This study aimed to validate GEWS and compare its predictive accuracy with NEWS, as well as to evaluate the clinical burden of GEWS. Methods : In this prospective multicenter observational study, patients admitted to acute geriatric wards were included. Clinical deterioration was defined as the occurrence of one of four events: (1) unexpected death, (2) ICU transfer, (3) transition to palliative care, or (4) urgent medical/surgical intervention. GEWS and NEWS were compared using C-statistics, performance and clinical burden metrics. Results : Among 511 patients, 348 events were recorded in 302 individuals. GEWS significantly outperformed NEWS across all event types, showing higher AUROC and PR-AUC values ( p <0.0001). At a GEWS threshold ≥5, predictive performance was superior in accuracy (0.940 vs. 0.927), PPV (0.497 vs. 0.365), and specificity (0.977 vs 0.967), all p <0.0001. GEWS ≥5 was also associated with a lower clinical burden (NNE: 2.013 vs 2.738, p <0.0001). For life-threatening events (Type 1-3), GEWS ≥8 provided higher specificity (0.999 vs. 0.995, p <0.0001) and a reduced alarm rate (0.518 vs.1.347, p <0.0001). Conclusion : GEWS offers a more accurate, geriatric-specific alternative to NEWS for detecting clinical deterioration in frail older adults, while reducing clinical workload. A threshold of GEWS ≥5 is recommended for clinical alerting, whereas ≥8 for rapid response team activation. EWS GEWS - frail older people validation clinical burden Figures Figure 1 Figure 2 Key Summary Points Aim: Validation and performance of a Geriatric Early Warning Score (GEWS) in frail older patients. Findings: GEWS is more accurate and geriatric-specific alternative to NEWS for detecting clinical deterioration in frail older patients. A GEWS threshold of ≥ 5 offers the best balance between predictive accuracy and clinical burden, making it the preferred cut-off for routine monitoring. GEWS ≥ 8 is recommended for rapid response team activation. Message: GEWS outperforms the National Early Warning Score (NEWS) in predicting clinical deterioration in hospitalized frail older patients without increasing the clinical burden for healthcare providers. INTRODUCTION The use of an Early Warning Score (EWS) is recommended to reduce preventable in-hospital mortality by facilitating early detection of patient deterioration. An EWS is a composite score derived from deviations in multiple vital parameters, designed to alert healthcare providers to physiological decline that may precede severe adverse events. Nurses calculate EWSs during routine patient assessments, enabling timely escalation of care and potentially preventing critical deterioration, including the need for intensive care unit (ICU) admission[1]. Since its introduction in the United States of America (USA) in 1999, various EWSs have been developed, incorporating different thresholds for escalation, a range of vital parameters, and specific adaptations for selected populations, such as patients with chronic obstructive pulmonary disease (COPD), children and pregnant women [2–4]. As EWSs have been integrated into national hospital care guidelines, they have also been embedded in Electronic Patient Records (EPR) across many European healthcare systems. Currently, older hospitalized patients are typically assessed using the National Early Warning Score (NEWS) (Table 1A) or NEWS2, following guidelines from the Royal College of Physicians in the United Kingdom (UK) [1,5] or the Modified Early Warning Score (MEWS), developed in the USA[6]. However, increasing evidence indicates that existing EWSs may not be well-suited for older patients. For instance, Churpek et al.[7] demonstrated in 2016 that MEWS accuracy decreases with age, resulting in a reduced predictive performance for patients aged 65 and older – the group at highest risk of severe adverse events – compared to younger patients. More recently, in 2022, Vardy et al.[8] suggested that NEWS2 may require modifications for older adults. They recommended using NEWS2 alongside additional clinical assessments, such as the Clinical Frailty Scale (CFS) and the four-A-test (4AT) for delirium, to improve predictive accuracy. Supporting this, Rønningen et al.[9] found in 2023 that NEWS2 poorly predicts in-hospital mortality among patients over 70 with frailty and COVID-19 infection. The limited accuracy of current EWSs in older patients is likely due to age-related physiological and pathophysiological changes that alter vital parameter thresholds for clinical deterioration[10]. Given the high risk of frail older patients to severe adverse events, there is a clear need for an adjusted geriatric early warning score (GEWS) to support nurses in the timely escalation of care. However, despite the widespread use of EWSs, no specific GEWS has yet been developed. In Belgium, acute care for frail older patients is primarily managed by geriatricians, starting in the emergency department (ED) and continuing in the acute geriatric ward. Approximately 90-95% of patients in this acute geriatric ward are admitted through the ED, while the remainder are initially treated by other medical or surgical specialists or in the ICU before being transferred to geriatrics. In 2021, the median length of hospital stay in Belgian acute geriatric wards was 13,5 days[11]. Patients in these wards typically present with acute medical issues and are frail, often with multiple comorbidities and geriatric syndromes[12]. Currently, the decision to alert a geriatrician is based on individual patient parameters and nurse intuition. Some geriatric wards use a standardized ‘track and trigger’ system with predefined cut-off values for vital parameters, based on literature and clinical experience, while others rely on NEWS integrated into their EPR system. Qualitative research among Flemish geriatricians by Geeraert et al.[13] revealed concerns that NEWS generates a high number of false positive alarms, increasing clinical workload without adding value, while also producing false negative alarms that compromise patient safety. They emphasized the need for an EWS specifically tailored to frail older patients – one that is both validated for this population and designed to reduce unnecessary clinical burden[13]. In 2018, a specific GEWS was developed based on expert consensus among Flemish geriatricians (Az Alma Hospital) and existing literature on vital parameter deviations in frail older adults (Table 1b)[10,14–16]. However, this GEWS has not yet been validated for its accuracy to predict three key clinical outcomes: unexpected death, ICU transfer, and transition to palliative care. Additionally, we aimed to assess its value in predicting a fourth outcome particularly relevant in geriatric care: acute clinical deterioration requiring urgent medical or surgical intervention to prevent ICU admission or mortality. These events often provide a critical window for stabilizing the patient and initiate timely conversations about treatment preferences, such as refraining from surgery, ICU transfer, or cardiopulmonary resuscitation - supporting shared decision-making and advance care planning. This study aims to validate GEWS by evaluating its predictive accuracy for these four clinically relevant outcomes in frail older adults admitted to an acute geriatric ward, in comparison with the NEWS. Additionally, we will assess the performance metrics of GEWS and analyse the associated clinical burden across varying threshold levels, relative to NEWS, to identify the optimal threshold that balances predictive accuracy with clinical burden. SUBJECTS AND METHODS 1.Study population and design This multicenter, prospective observational study was conducted between September 1, 2022, and February, 28 2023, in the acute geriatric wards of three Belgian hospitals: Az Alma, St. Trudo and Jessa Hospital. All patients admitted to these geriatric wards with a CFS ≥ 3 were eligible for inclusion. Exclusion criteria were: patients with terminal or palliative care needs at admission and patients undergoing elective surgery. Written informed consent was obtained by the attending geriatrician from either the patient or their designated caregiver. The study protocol was approved by the coordinating medical ethics committee of Jessa Hospital, Hasselt (registration number: B2432022000020). The ethical committees of AZ Alma and St. Trudo hospitals also provided their approval. 2. Data collection 2.1 GEWS The GEWS (table 1b) is a modification of the NEWS, incorporating four key adaptations. First, the cut-off values for vital parameters were adjusted to better account for the frailty of older adults. Frailty reduces physiological reserves to respond to external stressors, as such even minor deviations in vital parameters can lead to severe outcomes. Consequently, GEWS features narrower cut-off ranges of vital parameters compared to NEWS. Second, GEWS quantifies the amount of oxygen therapy more precisely, whereas NEWS only employes a binary classification (oxygen therapy yes/no). Third, the scoring system for assessing consciousness in patients with delirium, whose mental status can fluctuate over time, was refined. In the original NEWS, agitated patients were typically scored as ‘alert’, potentially leading to missed delirium diagnoses. NEWS2 addressed this by introducing the category of ‘new confusion’ as the highest score; however, this adjustment achieved only 4.3% sensitivity, compared to 20.0% when using the 4AT-score[17]. GEWS further improves upon this by assigning the highest score to ‘agitation’ and implementing a more gradual scale for other levels of consciousness, enhancing the detection of delirium-related fluctuations. Finally, GEWS incorporates pain as a sixth vital parameter, assessed using the Numeric Rating Scale (NRS) or the Pain in Advanced Dementia scale (PAINAD). This ensures effective pain monitoring, even in patients with limited communication abilities, addressing a clinical gap where pain – an early indicator of acute medical events – is often overlooked in this patient population. 2.2 Baseline and clinical data The following demographical and clinical information was collected from the medical records: age, sex, CFS, delirium diagnosis (based on the Confusion Assessment Method (CAM) or Delirium Observation Scale (DOS)), Do-Not-Resuscitate (DNR) status at admission and the length of hospital stay. Vital parameters included: temperature, respiratory rate, oxygen saturation, oxygen therapy, pulse rate, systolic blood pressure, pain score (NRS or PAINAD), and level of consciousness (Alert, Voice, Pain, Unresponsive (AVPU), or agitated). 3. Procedure After obtaining written informed consent, demographical and clinical data were collected. Vital parameters were measured twice daily by nursing staff during routine morning and afternoon shifts and manually entered into the EPR. Nurses alerted the physician based on the patient’s condition, specific geriatrician instructions or clinical intuition. Clinical events were recorded by geriatricians with precise timestamps (year/month/day, hour/minutes/seconds) at admission and throughout hospitalization. NEWS and GEWS scores were calculated retrospectively following data export. Clinical deterioration was defined as the occurrence of one of four clinical events, categorized as follows: Type 1 Event Unexpected death Type 2 Event Unanticipated intensive care transfer (i.e. stroke unit, coronary care unit, ICU or medium care unit) Type 3 Event Transition to palliative care Type 4 Event Urgent medical intervention within 12 hours , including: Consciousness related interventions: Intravenous (IV) glucose (> 6 g) for acute hypoglycemic coma Cardiovascular interventions: IV diuretics (for acute pulmonary oedema) IV or oral anti-arrhythmics (for life-threatening arrhythmias) IV nitrates (for unstable angina pectoris) Bleeding or hemostasis interventions: Transfusion of packed red blood cells, platelets (for massive hemorrhage) Synthetic plasma factors (for massive hemorrhage) Anticoagulants (e.g. low-molecular-weight heparin (LMWH)), heparin, or thrombolytics (for embolism, stroke or myocardial infarction) Sepsis and dehydration related intervention: Volume resuscitation with fluid challenge (crystalloids/electrolytes/albumin) IV or oral antibiotics IV corticosteroids (for acute Addison’s disease) Respiratory interventions: Oxygen therapy IV corticosteroids (for COPD exacerbation) Urgent surgical intervention within 12 hours 4. Data analysis Vital parameter data and clinical event timestamps were automatically extracted from the EPR and linked to patient characteristics and clinical event registrations. All data were pseudonymized and centralized by the principal investigator. Since urgent medical conditions – particularly Type 4 clinical events – were often the primary reason for admission to the acute geriatric ward, while Type 1, 2 or 3 clinical events typically occurred later during hospitalization, the hospitalization period was divided into two distinct phases to better capture the latter types. The first phase started upon admission to the ED or the acute geriatric ward. The second phase began either 72 hours after admission – if no clinical event occurred during the first phase – or 72 hours after the last clinical event recorded within the first phase. This division was based on findings by Subbe et al.[6], which demonstrated that only 1.8% of patients still present with critical vital parameters related to the initial reason for admission after 72 hours. Therefore, clinical events occurring after this 72-hour window were considered new, while events occurring within 72 hours of a prior event were considered part of the same episode and were excluded from further analysis. Corresponding vital parameters for these excluded events were neutralized to prevent confounding. Each patient could contribute a maximum of three clinical events to the analysis, with no more than two events included from each of the two defined hospitalization phases. The unit of analysis was a single observation set of vital parameters. Observation sets were categorized as event observation sets (recorded within 12 hours before a clinical event), non-event observation sets (recorded more than 24 hours before an event), and neutralized observation sets (recorded between 24 and 12 hours before an event). Observation sets recorded after an event were also classified as neutralized. However, observation sets linked to Type 1-3 clinical events that occurred 12 hours after a preceding Type 4 event were still considered event observation sets. To minimize bias, data from neutralized observation sets were excluded from analysis. 5. Statistical methods 5.1 Sample size calculation A retrospective pilot study[14] conducted in the geriatric department of AZ Alma during the 2018 influenza season reported a 7.7% mortality rate among 541 admissions. Based on these findings, we estimated an incidence of 5% for Type 1,2, and 3 events and a 60% for Type 4 events, based on clinical experience. To ensure a sufficient number of clinical events for statistical analysis and maintain adequate statistical power, we aimed to include at least 500 patients, allowing for subgroup analyses. 5.2 Statistical analysis Statistical analyses were performed using SPSS Statistics version 28 and R version 4.4. Categorical variables are reported as numbers and percentages, while continuous variables as means ± standard deviation (SD) for normal distributions or medians with interquartile ranges (IQR) for non-normal distributions. Missing values of vital parameters were imputed in two stages. In the first stage, missing values were imputed as 0 if the observation met one of the following criteria: (1) it was the first recorded observation for the patient, (2) no prior values had been documented, (3) it was registered more than 12 hours after the previous observation, or (4) it was a neutralized observation. In the second stage, any remaining missing value was imputed using the last observation carried forward method, whereby missing scores are replaced with the most recently recorded value. Performance comparisons between GEWS and NEWS were conducted using a multitude of performance metrics, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the receiver operating characteristic curve (ROC), along with the area under the ROC curve (AUROC). Corresponding 95% confidence intervals were calculated using a clustered bootstrap percentile approach with 10.000 bootstrap samples. Statistical significance was determined using a p -value of <0.00051 (using Bonferroni correction for multiple testing). However, Romero-Brufau et al.[18] highlighted that C-statistics can be misleading in situations where physiological deterioration occurs with an extremely low prevalence of 0.02 per patient-day in the general population. In such cases, additional performance metrics such as precision-recall (PR) curves, the precision-recall area under the curve (PR-AUC), and the number-needed-to-evaluate (NNE) may be used to assess predictive accuracy. Given the low pretest probability of Type 1- 3 clinical events, the PR-AUC was constructed to better evaluate the PPV versus sensitivity. Following the methodology of Pankhurst et al. [20] clinical burden of different GEWS and NEWS threshold levels was assessed by calculating the NNE, the rate-of-alarms (ROA) (i.e. the number of alarms per 100 patient-days), and the alerted-outcome-event-rate (AOER) (i.e. the number of alerted outcome events per 100 patient-days). Additionally, differences between EWS performance estimates were evaluated against an empirical null distribution, with p -values calculated using a Monte Carlo permutation approach with 9999 random permutations. RESULTS 1. Characteristics of study population and clinical event types A total of 511 patients were included in the study, with a mean age of 85 years (SD=6). The majority were female (60%). Most patients were clinically frail, with a median CFS of 6 [IQR:5-7]. DNR status was documented for all patients. Delirium was identified in 8.2%. The median hospital stay was 11 days (SD=6). A detail overview of participants’ demographic and clinical characteristics is presented in Table 2. Overall, the study population closely reflects the typical patient demographic of acute geriatric wards in Flanders, Belgium. Table 2 also presents the incidence and classification of clinical events (Type 1-4). A total of 348 clinical events was recorded among 297 patients. Of these events, 317 (91%) were classified as Type 4, usually requiring an urgent medical intervention by a geriatrician. Surgical interventions were rare. Type 1, 2 or 3 events occurred in 9%. Additional details on patient characteristics across the participating hospitals as well as descriptions of the interventions administered for Type 4 clinical events, are reported in Addendum 1 . 2.Validation of GEWS vs. NEWS 2.2 For all clinical events (Type 1–4) The AUROC for GEWS was 0.796 [95% CI, 0.770-0.822] compared to 0.732 [95% CI, 0.697-0.765] for NEWS ( p <0.0001). Similarly, the PR-AUC for GEWS was 0.412 [95% CI, 0.350-0.472], whereas for NEWS it was 0.305 [95% CI, 0.249-0.365], ( p <0.0001) (Figure 1). 2.3 For severe clinical events (Type 1–3) In alignment with most EWS studies, which focus on life-threatening events, we also compared GEWS and NEWS for the subset of Type 1–3 clinical events. The AUROC for GEWS was 0.847 [95% CI, 0.779-0.905], compared to 0.806 [95% CI, 0.735-0.882] for NEWS, (p =0.24 ) . The PR-AUC was nearly identical between GEWS (0.246 [95% CI, 0.123-0.394]) and NEWS (0.245 [95% CI, 0.113-0.393]), ( p =0.995) (Figures presented in Addendum 2 ). 3. Performance of NEWS ≥5 S3 and GEWS (at different thresholds), across all clinical events (Type 1–4) GEWS and NEWS differ in the vital parameters they include and their thresholds, leading to variations in alarm activation. A NEWS threshold of 5 or higher, or a single score of 3 on any vital parameter (NEWS ≥5 S3), is widely accepted as providing optimal predictive performance[5]. For GEWS, a clinically determined threshold of ≥4 S3 was initially adopted. Performance metrics for these thresholds across all clinical event types are summarized in Table 3. GEWS ≥4 S3 demonstrated significantly higher sensitivity and NPV compared to NEWS ≥5 S3, ( p’s <0.0001). However, GEWS ≥4 S3 had a significantly lower accuracy and specificity than NEWS ≥5 S3, ( p’s <0.0001), along with a significantly higher ROA and AOER ( p’s <0.0001). Given the significantly higher ROA associated with GEWS ≥4 S3, there may be a concern that this could lead to increased clinical burden on nurses and doctors, alarm fatigue and reduced adherence to EWS protocols. To address this concern, we further examined performance metrics of an alternative GEWS threshold, GEWS ≥5 S3, also presented in Table 3. At similar NPV and sensitivity levels, GEWS ≥5 S3 showed significantly higher accuracy, specificity, and improved PPV compared to NEWS ≥5 S3, ( all p’s <0.0001). Additionally, GEWS ≥5 S3 showed a significantly lower NNE ( p <0.0001), and reduced ROA ( p =0.009), while maintaining a comparable AOER ( p =0.032). Based on these results, GEWS ≥5 S3 was selected for further analysis. 4. Performance of NEWS ≥5 and GEWS ≥5 excluding the S3 score, across all clinical events (Type 1–4) Since only 5 out of 297 observations with an S3 score, were followed by a clinical event (Type 1-4), we also calculated performance metrics for NEWS ≥5 and GEWS ≥5 excluding S3 (Table 4). Excluding S3 had only a minimal impact on the NPV for both NEWS and GEWS. When comparing NEWS ≥5 and GEWS ≥5, the latter consistently showed significantly better results across all performance metrics. 5. Clinical burden of NEWS vs. GEWS, for all clinical events (Type 1–4) From the perspective of clinical burden, the NNE, the ROA and the AOER are particularly important. These metrics are represented in Figure 2 and its corresponding table. The clinical burden over a 24-hour period in a 24-bed ward, representative of acute geriatrics wards in Belgium, is illustrated across the different NEWS and GEWS threshold levels (i.e., NEWS ≥5 with and without S3, and GEWS ≥5 with and without S3). Among the evaluated thresholds, GEWS ≥5 appears to be the preferred option when considering NPV, NNE and ROA. Although GEWS ≥5 with S3 demonstrates a better AOER, with an improvement of 0.102 events per 24 hours in a 24-bed ward compared to GEWS ≥5 alone, this comes at the cost of a substantially higher ROA – an increase of 1.463. In practical terms, detecting one additional true event over a 10-day period would generate approximately 14 alarms, 13 of which would be false positives. 6. Performance of NEWS ≥5 S3 and GEWS ≥5 S3 for severe clinical events (Type 1 – 3) Although the AUROC and PR-AUC differences between NEWS and GEWS for life-threatening events (Type 1–3) were not statistically significant, we further examined their performance and clinical burden in this subset to assess alarm frequency and its impact on clinical workload ( Addendum 3 ). For events Type 1–3, both GEWS ≥5 S3 and NEWS ≥5 S3 had a NPV of 0.996, while for all event type (1-4), NEWS ≥5 S3 had an NPV of 0.958 and GEWS ≥5 S3 had an NPV of 0.961. No statistically significant differences in NPV or sensitivity were found between NEWS ≥5 S3 and GEWS ≥5 S3 for life-threatening events. However, GEWS ≥5 S3 showed significantly higher accuracy, specificity, and lower ROA compared to NEWS ≥5 S3 ( p’s <0.0001). 7. Additional GEWS thresholds The Royal College of physicians recommends a NEWS score ≥7 to trigger a rapid response team (RRT) intervention [5]. As such interventions demand considerable time and manpower, the PPV of the EWS threshold must be sufficiently high to justify activating a RRT beyond standard ward care. To assess this, we compared GEWS ≥7 and ≥8 with NEWS ≥7 ( Addendum 4 ). GEWS ≥7 and NEWS ≥7 showed no significant performance differences. However, GEWS ≥8 showed significantly higher specificity (0.999 vs. 0.995, p <0.0001) and a substantially lower ROA (0.518 vs. 1.347, p <0.0001), without a significant difference in AOER (0.249 vs. 0.415, p =0.0226). Notably, with GEWS ≥8, approximately one in every two alarms corresponded to a true life-threatening event, compared to one in every 2.5 for GEWS ≥7. Practically, GEWS ≥7 would trigger about 3 additional alarms over 25 days to detect one additional true life-threatening event, with two being false positives. 8. Delirium detection Delirium was identified in only 8.2% of patients. Moreover, detection rates varied significantly between the three study centers (10.2 %, 2.5%, and 12.5%, respectively, addendum 1 ). DISCUSSION Currently used EWSs, such as NEWS and MEWS, have shown to be less accurate in predicting clinical deterioration in frail older patients [7–9,19] In this study, we validated a specifically developed GEWS against NEWS and compared their performance metrics for detecting clinical deterioration in frail older patients admitted to an acute geriatric ward. Validation of GEWS Our results indicate that GEWS significantly outperforms NEWS in identifying clinical deterioration in frail older patients across all clinical events (Type 1-4). This was demonstrated by a significantly higher AUROC and PR-AUC compared to NEWS. For life-threatening events (Type 1-3), GEWS performed comparably to NEWS, with no statistically significant differences in AUROC or PR-AUC, supporting a non-inferiority of GEWS in this context. These findings contribute to the limited available literature evaluating the performance of existing EWSs in detecting clinical deterioration in frail older patients. Churpek et al.[7] previously reported that the AUC of MEWS for detecting cardiac arrest was significantly lower in older hospitalized patients compared to younger ones. Likewise, Mitsunaga et al.[20] assessed NEWS and MEWS in the ED for predicting in-hospital mortality in older patients and found an AUC of 0.789 (95% CI [0.747–0.829]) for NEWS and 0.720 (95% CI [0.671–0.765]) for MEWS. Although direct comparisons of our results with the former two studies are not possible due to methodological differences, our findings suggest that GEWS may be the most accurate EWS for predicting clinical deterioration in frail older patients. GEWS threshold and clinical burden We evaluated the performance of GEWS at different thresholds across all clinical events (Type 1-4) to determine the optimal balance between early detection of physiological deterioration and minimizing clinical burden associated with high alarm rates. Haegdorens et al.[21] emphasized that EWSs are primarily designed to effectively rule out serious adverse events and, therefore, should prioritize a high NPV. In this regard, a GEWS ≥4 S3 should be recommended to rule out clinical deterioration, as GEWS ≥4 S3 showed a significantly better NPV than NEWS ≥5 S3, with no significant difference compared to GEWS ≥5 S3. However, lower thresholds lead to increased alarm rates. Indeed, while GEWS ≥4 S3 captured more clinical events than GEWS ≥5 S3, it did so with nearly twice as many alarms, substantially increasing clinical burden without a proportional clinical benefit. When the objective shifts from ruling out to ruling in patients who require escalation of care, a high PPV becomes essential. GEWS ≥5 S3 achieved a significantly higher PPV and lower NNE, while maintaining a comparable NPV and AOER compared to NEWS ≥5 S3. Moreover, GEWS ≥5 S3 tended to produce fewer alarms (ROA, p =0.009). These findings indicate that GEWS ≥5 S3 provides a more favorable trade-off between predictive performance and clinical burden. Interestingly, excluding the S3 component from GEWS ≥5 further improved PPV, and reduced both ROA and NNE. Taken together, across all clinical events (Type 1-4), GEWS ≥5 without S3 offers the best balance of performance and clinical burden, making it the preferred threshold for predicting clinical deterioration in frail older patients. For life-threatening events (Type 1-3), no significant differences in sensitivity or NPV were observed between NEWS ≥5 S3 and GEWS ≥5 S3, indicating that GEWS is neither superior nor inferior in ruling out these events. However, GEWS demonstrated significantly higher accuracy and specificity, along with a lower ROA. While AOER was similar between the two systems, these results suggest that GEWS offers better clinical efficiency than NEWS by generating fewer alerts – thereby reducing alarm fatigue and clinical workload – while maintaining comparable safety in detecting severe events. Finally, since the Royal College of physicians recommends a NEWS ≥7 to trigger escalation of care via rapid response teams (RRTs), we evaluated the optimal GEWS threshold for identifying life-threatening events (Type 1-3) in this context. Compared to NEWS ≥7, GEWS ≥8 showed significantly higher specificity and lower ROA, as well as the lowest NNE and a comparable NPV. Given the high resource demands of RRT interventions, a threshold of GEWS ≥8 appears more appropriate for triggering such responses. Patients scoring ≥5, by contrast, may be more suitably managed by the attending ward physician. Practical recommendations A GEWS threshold of ≥5 is recommended as the standard trigger for nursing staff to notify the ward geriatrician across all clinical event types (1–4), as it offers the most favourable balance between predictive accuracy and clinical burden. To address concerns regarding sensitivity and AOER, it is advised that when GEWS equals 4 or when a single parameter scores 3, nurses closely monitor the patient by reassessing within 2 hours. If any further deterioration is observed during reassessment, the attending physician should promptly be alerted. Additionally, any S3 on a vital parameter should prompt an immediate recheck of all vital signs to confirm or rule out transient abnormalities. For GEWS scores ≥8, an urgent or emergency response by staff with critical care competencies, such as a RRT, should be initiated, including a prompt bedside assessment – unless it has been previously documented that the patient does not wish to be resuscitated or to be transferred to an ICU. Lastly, and most importantly, if a nurse has a strong clinical suspicion or ‘gut feeling’ of deterioration, appropriate action should be taken regardless of the GEWS score. Potential improvements and future perspectives To further refine GEWS, efforts should focus on increasing its PPV while reducing clinical burden and alarm fatigue. A promising direction is the integration of structured clinical intuition, given the recognized impact of nurse intuition on patient outcomes[22]. Haegdorens et al. proposed the Nurse Intuition Patient Deterioration Scale (NIPDS) to quantify such intuition[22]. A Belgian study conducted on medical, surgical and geriatric wards found that combining NIPDS with NEWS improved the prediction of severe adverse events[23]. However, as NIPDS is only validated for the first 24 hours of admission, its utility in longer hospital stays remains unclear. Moreover, some NIPDS features may already be indirectly captured by GEWS. For example, while older adults may underreport pain, NIPDS captures it through facial cues – yet GEWS already includes structured pain assessment (NRS or PAINAD), unlike NEWS. NIPDS also tracks behavior and responsiveness to detect consciousness changes, which GEWS evaluates more gradually than NEWS. These overlaps suggest limited added value from integrating NIPDS into GEWS without further validation. Another key priority is improving the detection of delirium, especially given its prognostic importance in older adults. In our study, delirium was identified in only 8.2% of patients based on CAM or DOS assessments – substantially lower than the expected prevalence of 23.0%[24]. This discrepancy suggests possible under-detection in our study population and limits our ability to draw conclusions about the adequacy of the AVPU scale within GEWS for identifying delirium. We therefore propose that future studies consider evaluating the Richmond Agitation and Sedation Scale (RASS) as an alternative to AVPU within GEWS. RASS provides a graded assessment of consciousness through direct observation and has demonstrated high sensitivity (82.0%) and specificity (85.1%) for identifying delirium in older ED patients[25]. Furthermore, changes in RASS have been associated with improved specificity, up to 92.0%[25]. The scale is time-efficient (requiring less than 10 seconds), does not depend on proxy input, and is applicable across various clinical environments, including the ED, ICU, acute ward and perioperative settings. Strength and limitations This study is the first to validate a GEWS specifically tailored to frail older adults, using cut-off values adjusted to their physiological characteristics. Unlike conventional EWSs, which primarily aim to prevent ICU transfers or cardiopulmonary resuscitation, GEWS focuses on patient-centered outcomes, including Type 4 clinical events. This enables more proactive care that respects individual patient preferences – many of whom prefer to avoid intensive interventions in favor of personalized treatment approaches. GEWS supports timely escalation of care by nurses and geriatricians in alignment with these preferences. A second strength is the study’s real-world, multicenter design, conducted without additional training for nurses or geriatricians. This enhances the external validity of the findings, which are likely generalizable – at least within Belgian geriatric wards and emergency departments. None of the participating centers employed RRTs linked to NEWS or GEWS scores. Instead, nurses acted based on clinical judgement or standing orders. A third strength, is the use of robust methodology, not only including C-statistics, but also performance metrics such as PR-AUC, and clinical burden analysis. These measures are in accordance with best practices for EWS evaluation and are comparable to the approach used by Pankhurst et al.[19] in assessing NEWS2 response thresholds in a UK acute hospital setting. Nonetheless, several limitations should be acknowledged. First, the study population included no robust older adults (i.e. CFS < 3), limiting the applicability of the findings to that subgroup. Since robust and frail older individuals may differ physiologically, NEWS may remain more appropriate for robust patients until further comparative studies are available. Second, the classification of Type 4 clinical events was based on a predefined list of medical interventions, which in hindsight may have led to an underestimation of clinical event rates. Notably, the administration of intravenous analgesia for severe pain (NRS ≥7) was not included as a qualifying Type 4 event. While this omission was unintentional, it may have impacted the measured GEWS performance by overlooking relevant acute clinical response. Third, no item-level analysis was conducted to assess the individual contribution of each vital parameter to the overall score performance. Some parameters may have had limited impact. For instance, respiratory rate – despite being a highly sensitive indicator of deterioration – was often estimated rather than measured, potentially reducing GEWS sensitivity. Improving awareness through training, along with the implementation of validated automated monitoring tools for older adults, could enhance sensitivity. Furthermore, pain was not consistently recorded alongside other vital parameters in two of the three centers, resulting in a higher rate of missing data. Lastly, the level of consciousness and delirium were frequently underreported. These issues in data quality may have impacted the accuracy of GEWS and warrant further investigation. Conclusion This study validated GEWS as a more accurate, geriatric-specific alternative to NEWS for detecting clinical deterioration in frail older patients across all clinical event types (1–4). A GEWS threshold of ≥ 5 was found to provide the optimal balance between predictive performance and clinical feasibility, making it the most appropriate cut-off for routine monitoring. By facilitating earlier and more individualized interventions, GEWS supports care that is better aligned with the complex needs and preferences of frail older patients. For triggering RRT activation, a threshold of ≥ 8 is recommended. Importantly, clinical judgement of nurses and physicians remain indispensable and should guide escalation decisions, regardless of the GEWS score. Declarations Competing interests and funding: The authors declare no financial or non-financial conflict of interest. This study received no external funding. Statistical support was provided through internal resources from the geriatric departments of Ghent University Hospital, St. Trudo Hospital and Jessa Hospital, as well as through personal funding. Author contributions: H.B. and N.V.D.N. were responsible for conceptualization of the study. Methodology was developed by H.B., S.M., M.V.A., N.V.D.N., and S.W., with project administration managed by H.B. H.B., S.M. and M.V.A. led the data curation and investigation. Formal statistical analysis was performed by S.W. Validation of the data was done by F.H. and S.W. H.B., A.B., F.H., and N.V.D.N. contributed to the interpretation of the results and their clinical relevance. The original draft was written by H.B. Supervision was done by N.V.D.N and A.B. All authors contributed to reviewing and editing the manuscript. Internal funding acquisition was coordinated by H.B., S.M., M.V.A., and N.V.D.N. Acknowledgements: The authors would like to sincerely thank St. Trudo Hospital (St. Truiden), Jessa Hospital (Hasselt), and AZ Alma Hospital (Eeklo) for their valuable collaboration and support in conducting this study. References Williams B. The National Early Warning Score: from concept to NHS implementation. Clinical Medicine 2022;22:499–505. https://doi.org/10.7861/clinmed.2022-news-concept. Roland D, Stilwell PA, Fortune P-M, Alexander J, Clark SJ, Kenny S. Case for change: a standardised inpatient paediatric early warning system in England. Arch Dis Child 2021;106:648. https://doi.org/10.1136/archdischild-2020-320466. Umar A, Ameh CA, Muriithi F, Mathai M. Early warning systems in obstetrics: A systematic literature review. PLOS ONE 2019;14:e0217864. https://doi.org/10.1371/journal.pone.0217864. Echevarria C, Steer J, Bourke SC. Comparison of early warning scores in patients with COPD exacerbation: DECAF and NEWS score. Thorax 2019;74:941–6. https://doi.org/10.1136/thoraxjnl-2019-213470. Royal College of Physicians. National Early Warning Score (NEWS): standardising the assessment of acute-illness severity in the NHS. Report of a working party. London: RCP, 2012. London: Royal College of Physisicians; 2012. Subbe CP, Kruger M, Rutherford P, Gemmel L. Validation of a modified Early Warning Score in medical admissions. QJM 2001;94:521–6. https://doi.org/10.1093/qjmed/94.10.521. Churpek MM, Yuen TC, Winslow C, Hall J, Edelson DP. Differences in vital signs between elderly and nonelderly patients prior to ward cardiac arrest. Critical Care Medicine 2015;43:816–22. https://doi.org/10.1097/CCM.0000000000000818. Vardy ER, Lasserson D, Barker RO, Hanratty B. NEWS2 and the older person. Clin Med (Lond) 2022;22:522–4. https://doi.org/10.7861/clinmed.2022-0426. Rønningen PS, Walle-Hansen MM, Ihle-Hansen H, Andersen EL, Tveit A, Myrstad M. Impact of frailty on the performance of the National Early Warning Score 2 to predict poor outcome in patients hospitalised due to COVID-19. BMC Geriatrics 2023;23:134. https://doi.org/10.1186/s12877-023-03842-0. Chester JG, Rudolph JL. Vital Signs in Older Patients: Age-Related Changes. J Am Med Dir Assoc 2011;12:337–43. https://doi.org/10.1016/j.jamda.2010.04.009. author-health. In : Blikvanger: Algemene ziekenhuizen. Naar een gezond België 2025. https://www.gezondbelgie.be/nl/blikvanger-gezondheidszorg/algemene-ziekenhuizen/download-hier-het-volledige-rapport-in-pdf. Accessed April 13, 2025. Koninklijk besluit tot wijziging van het koninklijk besluit van 29 januari 2007 houdende vaststelling eensdeels, van de normen waaraan het zorgprogramma voor de geriatrische patiënt moet voldoen om te worden erkend en, anderdeels, van bijzondere aanvullende normen voor de erkenning van ziekenhuizen en ziekenhuisdiensten. federale overheidsdienst volksgezondheid, veiligheid van de voedselketen en leefmilieu; 2014. https://etaamb.openjustice.be/nl/koninklijk-besluit-van-26-maart-2014_n2014024118. Accessed March 26; 2014. Geeraert G (ugent)02006633. De waarde van de Early Warning Score voor geriaters 2022:21. Baeyens Hilde, Dekoninck Julien, Baeyens Jean-Pierre. P-523 Vital signs in elder patients, urgent need to develop ‘“GEWS”’ (= Geriatric Early Warning System). Eur Geriatr Med 2018;Eur Geriatr Med (2018) 9 (Suppl 1):S1–S367. https://doi.org/10.1007/s41999-018-0097-4. Rodríguez-Molinero A, Narvaiza L, Ruiz J, Gálvez-Barrón C. Normal Respiratory Rate and Peripheral Blood Oxygen Saturation in the Elderly Population. Journal of the American Geriatrics Society 2013;61:2238–40. https://doi.org/10.1111/jgs.12580. Takayama A, Nagamine T, Kotani K. Aging is independently associated with an increasing normal respiratory rate among an older adult population in a clinical setting: A cross-sectional study. Geriatr Gerontol Int 2019;19:1179–83. https://doi.org/10.1111/ggi.13788. Vardy ER, Santhirasekaran S, Cheng M, Anand A, MacLullich AM. NEWS2 shows low sensitivity and high specificity for delirium detection: a single site observational study of 13,908 patients. Clin Med (Lond) 2022;22:544–8. https://doi.org/10.7861/clinmed.2022-0345. Romero-Brufau S, Huddleston JM, Escobar GJ, Liebow M. Why the C-statistic is not informative to evaluate early warning scores and what metrics to use. Crit Care 2015;19:285. https://doi.org/10.1186/s13054-015-0999-1. Pankhurst T, Sapey E, Gyves H, Evison F, Gallier S, Gkoutos G, et al. Evaluation of NEWS2 response thresholds in a retrospective observational study from a UK acute hospital. BMJ Open 2022;12:e054027. https://doi.org/10.1136/bmjopen-2021-054027. Mitsunaga T, Hasegawa I, Uzura M, Okuno K, Otani K, Ohtaki Y, et al. Comparison of the National Early Warning Score (NEWS) and the Modified Early Warning Score (MEWS) for predicting admission and in-hospital mortality in elderly patients in the pre-hospital setting and in the emergency department. PeerJ 2019;7:e6947. https://doi.org/10.7717/peerj.6947. Haegdorens F. Predicting serious adverse events or a safety net – Rethinking the role of early warning scores. Resuscitation Plus 2024;17:100534. https://doi.org/10.1016/j.resplu.2023.100534. Haegdorens F, Wils C, Franck E. Predicting patient deterioration by nurse intuition: The development and validation of the nurse intuition patient deterioration scale. International Journal of Nursing Studies 2023;142:104467. https://doi.org/10.1016/j.ijnurstu.2023.104467. Haegdorens F, Lefebvre J, Wils C, Franck E, Van Bogaert P. Combining the Nurse Intuition Patient Deterioration Scale with the National Early Warning Score provides more Net Benefit in predicting serious adverse events: A prospective cohort study in medical, surgical, and geriatric wards. Intensive and Critical Care Nursing 2024;83:103628. https://doi.org/10.1016/j.iccn.2024.103628. Gibb K, Seeley A, Quinn T, Siddiqi N, Shenkin S, Rockwood K, et al. The consistent burden in published estimates of delirium occurrence in medical inpatients over four decades: a systematic review and meta-analysis study. Age Ageing 2020;49:352–60. https://doi.org/10.1093/ageing/afaa040. Han JH, Vasilevskis EE, Schnelle JF, Shintani A, Dittus RS, Wilson A, et al. The Diagnostic Performance of the Richmond Agitation Sedation Scale for Detecting Delirium in Older Emergency Department Patients. Academic Emergency Medicine 2015;22:878–82. https://doi.org/10.1111/acem.12706. Tables Tables 1 to 4 are available in the Supplementary Files section. Supplementary Files 4TABLES.docx 6SUPPLEMENTARYMATERIALS.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Major revisions 20 Jun, 2025 Reviewers agreed at journal 28 May, 2025 Reviewers invited by journal 14 May, 2025 Editor invited by journal 12 May, 2025 Editor assigned by journal 11 May, 2025 First submitted to journal 10 May, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6635964","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":456717486,"identity":"66a86b5a-c9aa-4d77-8f49-5bd5148c45a3","order_by":0,"name":"Hilde Baeyens","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABD0lEQVRIiWNgGAWjYBACA2bmBgkGhgNQbgWDAZhmbMCnhRFZyxmwFsYGvFoYkLUwthGhxZydsfHGB4Y78vwSycce/pxXZ2xwu/n4A8YdNji1WDYzNlvOYHhmOHNGWrox77bDZgZ3jiU2MJ5Jw+2ww4xt0jwMhxk33Mgxk2bcdsDG4EaOYQNj22H8Wv4wHLbffyP/m+TPOXVALfkfgVr+49fCwHA4cYNEDpsEbwOzGdAWoPfbDuDT0mzZY/AsecaZZ2bSPMcOG0veSDOckXgmGbeW84cP3vhRcce2vz35meSPmjrDvhvJDz583GGHUwtUIxALJEDYCiAnJRDQAAH8UNfLNxClfBSMglEwCkYQAAA0cl749fHThwAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0009-0009-9236-6117","institution":"Ghent University: Universiteit Gent","correspondingAuthor":true,"prefix":"","firstName":"Hilde","middleName":"","lastName":"Baeyens","suffix":""},{"id":456717487,"identity":"4bd181fc-58a7-4205-9445-20827968792d","order_by":1,"name":"Filip Haegdorens","email":"","orcid":"https://orcid.org/0000-0003-1996-8786","institution":"Universiteit Antwerpen Faculteit geneeskunde en gezondheidswetenschappen","correspondingAuthor":false,"prefix":"","firstName":"Filip","middleName":"","lastName":"Haegdorens","suffix":""},{"id":456717488,"identity":"86614d28-079b-44e5-a223-5a005e88e049","order_by":2,"name":"Sven Martens","email":"","orcid":"","institution":"Saint Trudo Hospital: Sint-Trudo Ziekenhuis vzw","correspondingAuthor":false,"prefix":"","firstName":"Sven","middleName":"","lastName":"Martens","suffix":""},{"id":456717489,"identity":"a317a71c-eff0-4487-9b8a-6d3d967d7823","order_by":3,"name":"Marie-Elena Vanden Abeele","email":"","orcid":"https://orcid.org/0000-0003-0509-7414","institution":"Virga Jesseziekenhuis: Jessa Ziekenhuis vwz","correspondingAuthor":false,"prefix":"","firstName":"Marie-Elena","middleName":"Vanden","lastName":"Abeele","suffix":""},{"id":456717490,"identity":"c0666c45-c331-4400-8902-0512c8faf285","order_by":4,"name":"Steven Wallaert","email":"","orcid":"","institution":"Universiteit Gent Faculteit Geneeskunde en Gezondheidswetenschappen","correspondingAuthor":false,"prefix":"","firstName":"Steven","middleName":"","lastName":"Wallaert","suffix":""},{"id":456717491,"identity":"d8d8674f-a3a9-4c52-86a7-91e581465b06","order_by":5,"name":"Nele Van Den Noortgate","email":"","orcid":"https://orcid.org/0000-0001-5546-5380","institution":"State University of Ghent Hospital: Universitair Ziekenhuis Gent","correspondingAuthor":false,"prefix":"","firstName":"Nele","middleName":"Van Den","lastName":"Noortgate","suffix":""},{"id":456717492,"identity":"d8df757c-c657-4195-a8fa-22da9b735d01","order_by":6,"name":"Astrid D.H. Brys","email":"","orcid":"https://orcid.org/0000-0003-2900-5609","institution":"University Hospital Ghent: Universitair Ziekenhuis Gent","correspondingAuthor":false,"prefix":"","firstName":"Astrid","middleName":"D.H.","lastName":"Brys","suffix":""}],"badges":[],"createdAt":"2025-05-10 16:40:42","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6635964/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6635964/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":83124880,"identity":"327e348a-5a19-47e2-90db-508e7441ed8b","added_by":"auto","created_at":"2025-05-20 09:36:25","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":89135,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eReceiver operating characteristic curve (ROC-curve) and precision-recall curve (PR-curve) for GEWS and NEWS, all clinical events (Type 1 – 4)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLeft panel\u003c/strong\u003e: ROC-curve for GEWS (red line) versus NEWS (blue line); \u003cstrong\u003eRight panel\u003c/strong\u003e: PR-curve for GEWS (red line) and NEWS (blue line); dotted lines represent performance levels expected by chance alone.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6635964/v1/6f995e570b2b3173da6d3bdb.png"},{"id":83126226,"identity":"59410b6b-bc01-4f04-97c0-99171d2a9c09","added_by":"auto","created_at":"2025-05-20 09:44:25","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":50837,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eClinical burden of NEWS and GEWS at different thresholds, with and without the S3 score, across all clinical events (Type 1 – 4), during a 24-hours period in a 24-bed acute geriatric ward\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6635964/v1/d78418d408edb3537af6eecd.png"},{"id":83129561,"identity":"f71e25a0-c5e2-4cfa-ba5a-e2ac8d0cf79b","added_by":"auto","created_at":"2025-05-20 10:08:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1213924,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6635964/v1/c03ecd56-ac57-467a-8d0c-729b5109d09a.pdf"},{"id":83124882,"identity":"c2ff0c8c-e88c-41f8-9d7a-9d241cbf906f","added_by":"auto","created_at":"2025-05-20 09:36:25","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":29989,"visible":true,"origin":"","legend":"","description":"","filename":"4TABLES.docx","url":"https://assets-eu.researchsquare.com/files/rs-6635964/v1/4c8a1b0425fa6c6fc8bc0a58.docx"},{"id":83126227,"identity":"53e76ad4-fffd-42e9-a32c-bcfa040615f5","added_by":"auto","created_at":"2025-05-20 09:44:25","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":114537,"visible":true,"origin":"","legend":"","description":"","filename":"6SUPPLEMENTARYMATERIALS.docx","url":"https://assets-eu.researchsquare.com/files/rs-6635964/v1/20c890d2688042910bc888fa.docx"}],"financialInterests":"","formattedTitle":"\u003cp\u003eValidation and Performance of a Geriatric Early Warning Score (Gews) Versus the National Early Waring Score (News) in Predicting Clinical Deterioration in Frail Older Patients\u003c/p\u003e","fulltext":[{"header":"Key Summary Points","content":"\u003cp\u003e\u003cstrong\u003eAim:\u0026nbsp;\u003c/strong\u003eValidation and performance of a Geriatric Early Warning Score (GEWS) in frail older patients.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFindings:\u0026nbsp;\u003c/strong\u003eGEWS is more accurate and geriatric-specific alternative to NEWS for detecting clinical deterioration in frail older patients. A GEWS threshold of \u0026ge; 5 offers the best balance between predictive accuracy and clinical burden, making it the preferred cut-off for routine monitoring. GEWS \u0026ge; 8 is recommended for rapid response team activation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMessage:\u003c/strong\u003e GEWS outperforms the National Early Warning Score (NEWS) in predicting clinical deterioration in hospitalized frail older patients without increasing the clinical burden for healthcare providers.\u003c/p\u003e"},{"header":"INTRODUCTION","content":"\u003cp\u003eThe use of an Early Warning Score (EWS) is recommended to reduce preventable in-hospital mortality by facilitating early detection of patient deterioration. An EWS is a composite score derived from deviations in multiple vital parameters, designed to alert healthcare providers to physiological decline that may precede severe adverse events.\u0026nbsp;Nurses calculate EWSs during routine patient assessments, enabling timely escalation of care and potentially preventing critical deterioration, including the need for intensive care unit (ICU) admission[1].\u003c/p\u003e\n\u003cp\u003eSince its introduction in the United States of America (USA) in 1999, various EWSs have been developed, incorporating different thresholds for escalation, a range of vital parameters, and specific adaptations for selected populations, such as patients with chronic obstructive pulmonary disease (COPD), children and pregnant women [2\u0026ndash;4]. As EWSs have been integrated into national hospital care guidelines, they have also been embedded in Electronic Patient Records (EPR) across many European healthcare systems.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCurrently, older hospitalized patients are typically assessed using the National Early Warning Score (NEWS) (Table 1A) or NEWS2, following guidelines from the Royal College of Physicians in the United Kingdom (UK) [1,5] or the Modified Early Warning Score (MEWS), developed in the USA[6]. However, increasing evidence indicates that existing EWSs may not be well-suited for older patients. For instance, Churpek et al.[7] demonstrated in 2016 that MEWS accuracy decreases with age, resulting in a reduced predictive performance for patients aged 65 and older \u0026ndash; the group at highest risk of severe adverse events \u0026ndash; compared to younger patients. More recently, in 2022, Vardy et al.[8] suggested that NEWS2 may require modifications for older adults. They recommended using NEWS2 alongside additional clinical assessments, such as the Clinical Frailty Scale (CFS) and the four-A-test (4AT) for delirium, to improve predictive accuracy. Supporting this, R\u0026oslash;nningen et al.[9] found in 2023 that NEWS2 poorly predicts in-hospital mortality among patients over 70 with frailty and COVID-19 infection. The limited accuracy of current EWSs in older patients is likely due to age-related physiological and pathophysiological changes that alter vital parameter thresholds for clinical deterioration[10]. Given the high risk of frail older patients to severe adverse events, there is a clear need for an adjusted geriatric early warning score (GEWS) to support nurses in the timely escalation of care. However, despite the widespread use of EWSs, no specific GEWS has yet been developed.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn Belgium, acute care for frail older patients is primarily managed by geriatricians, starting in the emergency department (ED) and continuing in the acute geriatric ward. Approximately 90-95% of patients in this acute geriatric ward are admitted through the ED, while the remainder are initially treated by other medical or surgical specialists or in the ICU before being transferred to geriatrics. In 2021, the median length of hospital stay in Belgian acute geriatric wards was 13,5 days[11]. Patients in these wards typically present with acute medical issues and are frail, often with multiple comorbidities and geriatric syndromes[12]. Currently, the decision to alert a geriatrician is based on individual patient parameters and nurse intuition. Some geriatric wards use a standardized \u0026lsquo;track and trigger\u0026rsquo; system with predefined cut-off values for vital parameters, based on literature and clinical experience, while others rely on NEWS integrated into their EPR system. Qualitative research among Flemish geriatricians by Geeraert et al.[13] revealed concerns that NEWS generates a high number of false positive alarms, increasing clinical workload without adding value, while also producing false negative alarms that compromise patient safety. They emphasized the need for an EWS specifically tailored to frail older patients \u0026ndash; one that is both validated for this population and designed to reduce unnecessary clinical burden[13].\u003c/p\u003e\n\u003cp\u003eIn 2018, a specific GEWS was developed based on expert consensus among Flemish geriatricians (Az Alma Hospital) and existing literature on vital parameter deviations in frail older adults (Table 1b)[10,14\u0026ndash;16]. However, this GEWS has not yet been validated for its accuracy to predict three key clinical outcomes: unexpected death, ICU transfer, and transition to palliative care. Additionally, we aimed to assess its value in predicting a fourth outcome particularly relevant in geriatric care: acute clinical deterioration requiring urgent medical or surgical intervention to prevent ICU admission or mortality. These events often provide a critical window for stabilizing the patient and initiate timely conversations about treatment preferences, such as refraining from surgery, ICU transfer, or cardiopulmonary resuscitation - supporting shared decision-making and advance care planning.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis study aims to validate GEWS by evaluating its predictive accuracy for these four clinically relevant outcomes in frail older adults admitted to an acute geriatric ward, in comparison with the NEWS. Additionally, we will assess the performance metrics of GEWS and analyse the associated clinical burden across varying threshold levels, relative to NEWS, to identify the optimal threshold that balances predictive accuracy with clinical burden.\u003c/p\u003e"},{"header":"SUBJECTS AND METHODS","content":"\u003cp\u003e\u003cstrong\u003e1.Study population and design\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis multicenter, prospective observational study was conducted between September 1, 2022, and February, 28 2023, in the acute geriatric wards of three Belgian hospitals: Az Alma, St. Trudo and Jessa Hospital. All patients admitted to these geriatric wards with a CFS \u0026ge; 3 were eligible for inclusion. Exclusion criteria were: patients with terminal or palliative care needs at admission and patients undergoing elective surgery. Written informed consent was obtained by the attending geriatrician from either the patient or their designated caregiver. The study protocol was approved by the coordinating medical ethics committee of Jessa Hospital, Hasselt (registration number: B2432022000020). The ethical committees of AZ Alma and St. Trudo hospitals also provided their approval. \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2. Data collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e2.1 GEWS\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe GEWS (table 1b) is a modification of the NEWS, incorporating four key adaptations. First, the cut-off values for vital parameters were adjusted to better account for the frailty of older adults. Frailty reduces physiological reserves to respond to external stressors, as such even minor deviations in vital parameters can lead to severe outcomes. Consequently, GEWS features narrower cut-off ranges of vital parameters compared to NEWS. Second, GEWS quantifies the amount of oxygen therapy more precisely, whereas NEWS only employes a binary classification (oxygen therapy yes/no). Third, the scoring system for assessing consciousness in patients with delirium, whose mental status can fluctuate over time, was refined. In the original NEWS, agitated patients were typically scored as \u0026lsquo;alert\u0026rsquo;, potentially leading to missed delirium diagnoses. NEWS2 addressed this by introducing the category of \u0026lsquo;new confusion\u0026rsquo; as the highest score; however, this adjustment achieved only 4.3% sensitivity, compared to 20.0% when using the 4AT-score[17]. GEWS further improves upon this by assigning the highest score to \u0026lsquo;agitation\u0026rsquo; and implementing a more gradual scale for other levels of consciousness, enhancing the detection of delirium-related fluctuations. Finally, GEWS incorporates pain as a sixth vital parameter, assessed using the Numeric Rating Scale (NRS) or the Pain in Advanced Dementia scale (PAINAD). This ensures effective pain monitoring, even in patients with limited communication abilities, addressing a clinical gap where pain \u0026ndash; an early indicator of acute medical events \u0026ndash; is often overlooked in this patient population.\u003c/p\u003e\n\u003cp\u003e2.2 Baseline and clinical data\u003c/p\u003e\n\u003cp\u003eThe following demographical and clinical information was collected from the medical records: age, sex, CFS, delirium diagnosis (based on the Confusion Assessment Method (CAM) or Delirium Observation Scale (DOS)), Do-Not-Resuscitate (DNR) status at admission and the length of hospital stay. Vital parameters included: temperature, respiratory rate, oxygen saturation, oxygen therapy, pulse rate, systolic blood pressure, pain score (NRS or PAINAD), and level of consciousness (Alert, Voice, Pain, Unresponsive (AVPU), or agitated). \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. Procedure\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter obtaining written informed consent, demographical and clinical data were collected. Vital parameters\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003ewere measured twice daily by nursing staff during routine morning and afternoon shifts and manually entered into the EPR. Nurses alerted the physician based on the patient\u0026rsquo;s condition, specific geriatrician instructions or clinical intuition. Clinical events were recorded by geriatricians with precise timestamps (year/month/day, hour/minutes/seconds) at admission and throughout hospitalization. NEWS and GEWS scores were calculated retrospectively following data export.\u003c/p\u003e\n\u003cp\u003eClinical deterioration was defined as the occurrence of one of four clinical events, categorized as follows:\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eType 1 Event\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 497px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eUnexpected death\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eType 2 Event\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 497px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eUnanticipated intensive care transfer\u003c/strong\u003e (i.e. stroke unit, coronary care unit, ICU or medium care unit)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eType 3 Event\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 497px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTransition to palliative care\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eType 4 Event\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 497px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eUrgent medical intervention within 12 hours\u003c/strong\u003e, including:\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eConsciousness related interventions:\u003c/strong\u003e\n \u003cul\u003e\n \u003cli\u003eIntravenous (IV) glucose (\u0026gt; 6 g) for acute hypoglycemic coma\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eCardiovascular interventions:\u003c/strong\u003e\n \u003cul\u003e\n \u003cli\u003eIV diuretics (for acute pulmonary oedema)\u003c/li\u003e\n \u003cli\u003eIV or oral anti-arrhythmics (for life-threatening arrhythmias)\u003c/li\u003e\n \u003cli\u003eIV nitrates (for unstable angina pectoris)\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eBleeding or hemostasis interventions:\u0026nbsp;\u003c/strong\u003e\n \u003cul\u003e\n \u003cli\u003eTransfusion of packed red blood cells, platelets (for massive hemorrhage)\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSynthetic plasma factors (for massive hemorrhage)\u003c/li\u003e\n \u003cli\u003eAnticoagulants (e.g. low-molecular-weight heparin (LMWH)), heparin, or thrombolytics (for embolism, stroke or myocardial infarction)\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eSepsis and dehydration related intervention:\u003c/strong\u003e\n \u003cul\u003e\n \u003cli\u003eVolume resuscitation with fluid challenge (crystalloids/electrolytes/albumin)\u003c/li\u003e\n \u003cli\u003eIV or oral antibiotics\u003c/li\u003e\n \u003cli\u003eIV corticosteroids (for acute Addison\u0026rsquo;s disease)\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eRespiratory interventions:\u003c/strong\u003e\n \u003cul\u003e\n \u003cli\u003eOxygen therapy\u003c/li\u003e\n \u003cli\u003eIV corticosteroids (for COPD exacerbation)\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003e\u003cstrong\u003eUrgent surgical intervention within 12 hours\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e4. Data analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eVital parameter data and clinical event timestamps were automatically extracted from the EPR and linked to patient characteristics and clinical event registrations. All data were pseudonymized and centralized by the principal investigator. Since urgent medical conditions \u0026ndash; particularly Type 4 clinical events \u0026ndash; were often the primary reason for admission to the acute geriatric ward, while Type 1, 2 or 3 clinical events typically occurred later during hospitalization, the hospitalization period was divided into two distinct phases to better capture the latter types. The first phase started upon admission to the ED or the acute geriatric ward. The second phase began either 72 hours after admission \u0026ndash; if no clinical event occurred during the first phase \u0026ndash; or 72 hours after the last clinical event recorded within the first phase. This division was based on findings by Subbe et al.[6], which demonstrated that only 1.8% of patients still present with critical vital parameters related to the initial reason for admission after 72 hours. Therefore, clinical events occurring after this 72-hour window were considered new, while events occurring within 72 hours of a prior event were considered part of the same episode and were excluded from further analysis. Corresponding vital parameters for these excluded events were neutralized to prevent confounding.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEach patient could contribute a maximum of three clinical events to the analysis, with no more than two events included from each of the two defined hospitalization phases. The unit of analysis was a single observation set of vital parameters. Observation sets were categorized as \u003cem\u003eevent observation sets\u003c/em\u003e (recorded within 12 hours before a clinical event), \u003cem\u003enon-event observation sets\u003c/em\u003e (recorded more than 24 hours before an event), and \u003cem\u003eneutralized observation sets\u003c/em\u003e (recorded between 24 and 12 hours before an event). Observation sets recorded after an event were also classified as neutralized. However, observation sets linked to Type 1-3 clinical events that occurred 12 hours after a preceding Type 4 event were still considered event observation sets. To minimize bias, data from neutralized observation sets were excluded from analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e5. Statistical methods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e5.1 Sample size calculation\u003c/p\u003e\n\u003cp\u003eA retrospective pilot study[14] conducted in the geriatric department of AZ Alma during the 2018 influenza season reported a 7.7% mortality rate among 541 admissions. Based on these findings, we estimated an incidence of 5% for Type 1,2, and 3 events and a 60% for Type 4 events, based on clinical experience. To ensure a sufficient number of clinical events for statistical analysis and maintain adequate statistical power, we aimed to include at least 500 patients, allowing for subgroup analyses.\u003c/p\u003e\n\u003cp\u003e5.2 Statistical analysis\u003c/p\u003e\n\u003cp\u003eStatistical analyses were performed using SPSS Statistics version 28 and R version 4.4. Categorical variables are reported as numbers and percentages, while continuous variables as means \u0026plusmn; standard deviation (SD) for normal distributions or medians with interquartile ranges (IQR) for non-normal distributions. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMissing values of vital parameters were imputed in two stages. In the first stage, missing values were imputed as 0 if the observation met one of the following criteria: (1) it was the first recorded observation for the patient, (2) no prior values had been documented, (3) it was registered more than 12 hours after the previous observation, or (4) it was a neutralized observation. In the second stage, any remaining missing value was imputed using the last observation carried forward method, whereby missing scores are replaced with the most recently recorded value.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePerformance comparisons between GEWS and NEWS were conducted using a multitude of performance metrics, including sensitivity,\u0026nbsp;specificity, positive predictive value (PPV), negative predictive value (NPV), and the receiver operating characteristic curve (ROC), along with the area under the ROC curve (AUROC). Corresponding 95% confidence intervals were calculated using a clustered bootstrap percentile approach with 10.000 bootstrap samples. Statistical significance was determined using a \u003cem\u003ep\u003c/em\u003e-value of \u0026lt;0.00051 (using Bonferroni correction for multiple testing). However, Romero-Brufau et al.[18] highlighted that C-statistics can be misleading in situations where physiological deterioration occurs with an extremely low prevalence of 0.02 per patient-day in the general population. In such cases, additional performance metrics such as precision-recall (PR) curves, the precision-recall area under the curve (PR-AUC), and the number-needed-to-evaluate (NNE) may be used to assess predictive accuracy. Given the low pretest probability of Type 1- 3 clinical events, the PR-AUC was constructed to better evaluate the PPV versus sensitivity. Following the methodology of Pankhurst et al. [20] clinical burden of different GEWS and NEWS threshold levels was assessed by calculating the NNE, the rate-of-alarms (ROA) (i.e. the number of alarms per 100 patient-days), and the alerted-outcome-event-rate (AOER) (i.e. the number of alerted outcome events per 100 patient-days). Additionally, differences between EWS performance estimates were evaluated against an empirical null distribution, with \u003cem\u003ep\u003c/em\u003e-values calculated using a Monte Carlo permutation approach with 9999 random permutations.\u0026nbsp;\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cp\u003e\u003cstrong\u003e1. Characteristics of study population and clinical event types\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 511 patients were included in the study, with a mean age of 85 years (SD=6). The majority were female (60%). Most patients were clinically frail, with a median CFS of 6 [IQR:5-7]. DNR status was documented for all patients. Delirium was identified in 8.2%. The median hospital stay was 11 days (SD=6). A detail overview of participants\u0026rsquo; demographic and clinical characteristics is presented in Table 2. Overall, the study population closely reflects the typical patient demographic of acute geriatric wards in Flanders, Belgium.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 2 also presents the incidence and classification of clinical events (Type 1-4). A total of 348 clinical events was recorded among 297 patients. Of these events, 317 (91%) were classified as Type 4, usually requiring an urgent medical intervention by a geriatrician. Surgical interventions were rare. Type 1, 2 or 3 events occurred in 9%. Additional details on patient characteristics across the participating hospitals as well as descriptions of the interventions administered for Type 4 clinical events, are reported in \u003cstrong\u003eAddendum 1\u003c/strong\u003e. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.Validation of GEWS vs. NEWS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e2.2 For all clinical events (Type 1\u0026ndash;4)\u003c/p\u003e\n\u003cp\u003eThe AUROC for GEWS was 0.796 [95% CI, 0.770-0.822] compared to 0.732 [95% CI, 0.697-0.765] for NEWS (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001). Similarly, the PR-AUC for GEWS was 0.412 [95% CI, 0.350-0.472], whereas for NEWS it was 0.305 [95% CI, 0.249-0.365], (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001) (Figure 1).\u003c/p\u003e\n\u003cp\u003e2.3 For severe clinical events (Type 1\u0026ndash;3)\u003c/p\u003e\n\u003cp\u003eIn alignment with most EWS studies, which focus on life-threatening events, we also compared GEWS and NEWS for the subset of Type 1\u0026ndash;3 clinical events. The AUROC for GEWS was 0.847 [95% CI, 0.779-0.905], compared to 0.806 [95% CI, 0.735-0.882] for NEWS, \u003cem\u003e(p\u003c/em\u003e=0.24\u003cem\u003e)\u003c/em\u003e. The PR-AUC was nearly identical between GEWS (0.246 [95% CI, 0.123-0.394]) and NEWS (0.245 [95% CI, 0.113-0.393]), (\u003cem\u003ep\u003c/em\u003e=0.995) (Figures presented in \u003cstrong\u003eAddendum 2\u003c/strong\u003e). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. Performance of NEWS \u0026ge;5 S3 and GEWS (at different thresholds), across all clinical events (Type 1\u0026ndash;4)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGEWS and NEWS differ in the vital parameters they include and their thresholds, leading to variations in alarm activation. A NEWS threshold of 5 or higher, or a single score of 3 on any vital parameter (NEWS \u0026ge;5 S3), is widely accepted as providing optimal predictive performance[5]. For GEWS, a clinically determined threshold of \u0026ge;4 S3 was initially adopted. Performance metrics for these thresholds across all clinical event types are summarized in Table 3.\u0026nbsp;GEWS\u0026nbsp;\u0026ge;4 S3 demonstrated significantly higher sensitivity and NPV compared to NEWS\u0026nbsp;\u0026ge;5 S3, (\u003cem\u003ep\u0026rsquo;s\u0026nbsp;\u003c/em\u003e\u0026lt;0.0001).\u0026nbsp;However, GEWS\u0026nbsp;\u0026ge;4 S3 had a significantly lower accuracy and specificity than\u0026nbsp;NEWS \u0026ge;5 S3, (\u003cem\u003ep\u0026rsquo;s\u0026nbsp;\u003c/em\u003e\u0026lt;0.0001), along with a significantly higher ROA and AOER (\u003cem\u003ep\u0026rsquo;s\u003c/em\u003e\u0026lt;0.0001). Given the significantly higher ROA associated with GEWS \u0026ge;4 S3, there may be a concern that this could lead to increased clinical burden on nurses and doctors, alarm fatigue and reduced adherence to EWS protocols. To address this concern, we further examined performance metrics of an alternative GEWS threshold, GEWS \u0026ge;5 S3, also presented in Table 3. At similar NPV and sensitivity levels, GEWS \u0026ge;5 S3 showed significantly higher accuracy, \u0026nbsp;specificity, and improved PPV compared to NEWS \u0026ge;5 S3, (\u003cem\u003eall p\u0026rsquo;s\u003c/em\u003e \u0026lt;0.0001). Additionally, GEWS \u0026ge;5 S3 showed a significantly lower NNE (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001), and reduced ROA (\u003cem\u003ep\u003c/em\u003e=0.009), while maintaining a comparable AOER (\u003cem\u003ep\u003c/em\u003e=0.032). Based on these results, GEWS \u0026ge;5 S3 was selected for further analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4. Performance of NEWS \u0026ge;5 and GEWS \u0026ge;5 excluding the S3 score, across all clinical events (Type 1\u0026ndash;4)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSince only 5 out of 297 observations with an S3 score, were followed by a clinical event (Type 1-4), we also calculated performance metrics for NEWS \u0026ge;5 and GEWS \u0026ge;5 excluding S3 (Table 4).\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eExcluding S3 had only a minimal impact on the NPV for both NEWS and GEWS. When comparing NEWS \u0026ge;5 and GEWS \u0026ge;5, the latter consistently showed significantly better results across all performance metrics. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e5. Clinical burden of NEWS vs. GEWS, for all clinical events (Type 1\u0026ndash;4)\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFrom the perspective of clinical burden, the NNE, the ROA and the AOER are particularly important. These metrics are represented in Figure 2 and its corresponding table. The clinical burden over a 24-hour period in a 24-bed ward, representative of acute geriatrics wards in Belgium, is illustrated across the different NEWS and GEWS threshold levels (i.e., NEWS \u0026ge;5 with and without S3, and GEWS \u0026ge;5 with and without S3). \u0026nbsp;Among the evaluated thresholds, GEWS \u0026ge;5 appears to be the preferred option when considering NPV, NNE and ROA. Although GEWS \u0026ge;5 \u003cem\u003ewith\u003c/em\u003e S3 demonstrates a better AOER, with an improvement of 0.102 events per 24 hours in a 24-bed ward compared to GEWS \u0026ge;5 alone, this comes at the cost of a substantially higher ROA \u0026ndash; an increase of 1.463. In practical terms, detecting one additional true event over a 10-day period would generate approximately 14 alarms, 13 of which would be false positives.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e6. Performance of\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eNEWS \u0026ge;5 S3\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;and\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eGEWS \u0026ge;5 S3\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003efor severe clinical events (Type 1\u003c/strong\u003e\u0026ndash;\u003cstrong\u003e3)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAlthough the AUROC and PR-AUC differences between NEWS and GEWS for life-threatening events (Type 1\u0026ndash;3) were not statistically significant, we further examined their performance and clinical burden in this subset to assess alarm frequency and its impact on clinical workload (\u003cstrong\u003eAddendum 3\u003c/strong\u003e). For events Type 1\u0026ndash;3, both GEWS \u0026ge;5 S3 and NEWS \u0026ge;5 S3 had a NPV of 0.996, while for all event type (1-4), \u0026nbsp;NEWS \u0026ge;5 S3 had an NPV of 0.958 and GEWS \u0026ge;5 S3 had an NPV of 0.961. No statistically significant differences in NPV or sensitivity were found between NEWS \u0026ge;5 S3 and GEWS \u0026ge;5 S3 for life-threatening events. However, GEWS \u0026ge;5 S3 showed significantly higher accuracy, specificity, and lower ROA compared to NEWS \u0026ge;5 S3 (\u003cem\u003ep\u0026rsquo;s\u003c/em\u003e\u0026lt;0.0001). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e7. Additional GEWS thresholds\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Royal College of physicians recommends a NEWS score \u0026ge;7 to trigger a rapid response team (RRT) intervention [5]. As such interventions demand considerable time and manpower, the PPV of the EWS threshold must be sufficiently high to justify activating a RRT beyond standard ward care. To assess this, we compared GEWS \u0026ge;7 and \u0026ge;8 with NEWS \u0026ge;7 (\u003cstrong\u003eAddendum 4\u003c/strong\u003e). GEWS \u0026ge;7 and NEWS \u0026ge;7 showed no significant performance differences. However, GEWS \u0026ge;8 showed significantly higher specificity (0.999 vs. 0.995, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001) and a substantially lower ROA (0.518 vs. 1.347, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001), without a significant difference in AOER (0.249 vs. 0.415, \u003cem\u003ep\u003c/em\u003e=0.0226). Notably, with GEWS \u0026ge;8, approximately one in every two alarms corresponded to a true life-threatening event, compared to one in every 2.5 for GEWS \u0026ge;7. Practically, GEWS \u0026ge;7 would trigger about 3 additional alarms over 25 days to detect one additional true life-threatening event, with two being false positives. \u003cem\u003e\u003cbr\u003e\u003c/em\u003e\u003cstrong\u003e8. Delirium detection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDelirium was identified in only 8.2% of patients. Moreover, detection rates varied significantly between the three study centers (10.2 %, 2.5%, and 12.5%, respectively, \u003cstrong\u003eaddendum 1\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eCurrently used EWSs, such as NEWS and MEWS, have shown to be less accurate in predicting clinical deterioration in frail older patients [7\u0026ndash;9,19] In this study, we validated a specifically developed GEWS against NEWS and compared their performance metrics for detecting clinical deterioration in frail older patients admitted to an acute geriatric ward.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eValidation of GEWS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur results indicate that GEWS significantly outperforms NEWS in identifying clinical deterioration in frail older patients across all clinical events (Type 1-4). This was demonstrated by a significantly higher AUROC and PR-AUC compared to NEWS. For life-threatening events (Type 1-3), GEWS performed comparably to NEWS, with no statistically significant differences in AUROC or PR-AUC, supporting a non-inferiority of GEWS in this context. These findings contribute to the limited available literature evaluating the performance of existing EWSs in detecting clinical deterioration in frail older patients. Churpek et al.[7] previously reported that the AUC of MEWS for detecting cardiac arrest was significantly lower in older hospitalized patients compared to younger ones. Likewise, Mitsunaga et al.[20] assessed NEWS and MEWS in the ED for predicting in-hospital mortality in older patients and found an AUC of 0.789 (95% CI [0.747\u0026ndash;0.829]) for NEWS and 0.720 (95% CI [0.671\u0026ndash;0.765]) for MEWS. Although direct comparisons of our results with the former two studies are not possible due to methodological differences, our findings suggest that GEWS may be the most accurate EWS for predicting clinical deterioration in frail older patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGEWS threshold and clinical burden\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe evaluated the performance of GEWS at different thresholds across all clinical events (Type 1-4) to determine the optimal balance between early detection of physiological deterioration and minimizing clinical burden associated with high alarm rates. Haegdorens et al.[21] emphasized that EWSs are primarily designed to effectively \u003cem\u003erule\u003c/em\u003e \u003cem\u003eout\u003c/em\u003e serious adverse events and, therefore, should prioritize a high NPV. In this regard, a GEWS \u0026ge;4 S3 should be recommended to rule out clinical deterioration, as GEWS \u0026ge;4 S3 showed a significantly better NPV than NEWS \u0026ge;5 S3, with no significant difference compared to GEWS \u0026ge;5 S3. However, lower thresholds lead to increased alarm rates. Indeed, while GEWS \u0026ge;4 S3 captured more clinical events than\u0026nbsp;GEWS\u0026nbsp;\u0026ge;5 S3, it did so with nearly twice as many alarms, substantially increasing clinical burden without a proportional clinical benefit. When the objective shifts from ruling out to \u003cem\u003eruling in\u003c/em\u003e patients who require escalation of care, a high PPV becomes essential. GEWS \u0026ge;5 S3 achieved a significantly higher PPV and lower NNE, while maintaining a comparable NPV and AOER compared to NEWS \u0026ge;5 S3. Moreover, GEWS \u0026ge;5 S3 tended to produce fewer alarms (ROA, \u003cem\u003ep\u003c/em\u003e=0.009). These findings indicate that GEWS \u0026ge;5 S3 provides a more favorable trade-off between predictive performance and clinical burden. Interestingly, excluding the S3 component from GEWS\u0026nbsp;\u0026ge;5 further improved\u0026nbsp;PPV, and reduced both ROA and NNE.\u0026nbsp;Taken together, across all clinical events (Type 1-4), GEWS \u0026ge;5 \u003cem\u003ewithout S3\u003c/em\u003e offers the best balance of performance and clinical burden, making it the preferred threshold for predicting clinical deterioration in frail older patients.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor life-threatening events (Type 1-3), no significant differences in sensitivity or NPV were observed between NEWS \u0026ge;5 S3 and GEWS \u0026ge;5 S3, indicating that GEWS is neither superior nor inferior in ruling out these events. However, GEWS demonstrated significantly higher accuracy and specificity, along with a lower ROA. While AOER was similar between the two systems, these results suggest that GEWS offers better clinical efficiency than NEWS by generating fewer alerts \u0026ndash; thereby reducing alarm fatigue and clinical workload \u0026ndash; while maintaining comparable safety in detecting severe events.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFinally, since the Royal College of physicians recommends a NEWS \u0026ge;7 to trigger escalation of care via rapid response teams (RRTs), we evaluated the optimal GEWS threshold for identifying life-threatening events (Type 1-3) in this context. Compared to NEWS \u0026ge;7, GEWS \u0026ge;8 showed significantly higher specificity and lower ROA, as well as the lowest NNE and a comparable NPV. Given the high resource demands of RRT interventions, a threshold of GEWS \u0026ge;8 appears more appropriate for triggering such responses. Patients scoring \u0026ge;5, by contrast, may be more suitably managed \u0026nbsp;by the attending ward physician. \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePractical recommendations\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA GEWS threshold of \u0026ge;5 is recommended as the standard trigger for nursing staff to notify the ward geriatrician across all clinical event types (1\u0026ndash;4), as it offers the most favourable balance between predictive accuracy and clinical burden. To address concerns regarding sensitivity and AOER, \u0026nbsp;it is advised that when GEWS equals 4 or when a single parameter scores 3, nurses closely monitor the patient by reassessing within 2 hours. If any further deterioration is observed during reassessment, the attending physician should promptly be alerted. Additionally, any S3 on a vital parameter should prompt an immediate recheck of all vital signs to confirm or rule out transient abnormalities. For GEWS scores \u0026ge;8, an urgent or emergency response by staff with critical care competencies, such as a RRT, should be initiated, including a prompt bedside assessment \u0026ndash; unless it has been previously documented that the patient does not wish to be resuscitated or to be transferred to an ICU. Lastly, and most importantly, if a nurse has a strong clinical suspicion or \u0026lsquo;gut feeling\u0026rsquo; of deterioration, appropriate action should be taken regardless of the GEWS score. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePotential improvements and future perspectives\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo further refine GEWS, efforts should focus on increasing its PPV while reducing clinical burden and alarm fatigue. A promising direction is the integration of structured clinical intuition, given the recognized impact of nurse intuition on patient outcomes[22]. Haegdorens et al. proposed the Nurse Intuition Patient Deterioration Scale (NIPDS) to quantify such intuition[22]. A Belgian study conducted on medical, surgical and geriatric wards found that combining NIPDS with NEWS improved the prediction of severe adverse events[23]. However, as NIPDS is only validated for the first 24 hours of admission, its utility in longer hospital stays remains unclear. Moreover, some NIPDS features may already be indirectly captured by GEWS. For example, while older adults may underreport pain, NIPDS captures it through facial cues \u0026ndash; yet GEWS already includes structured pain assessment (NRS or PAINAD), unlike NEWS. NIPDS also tracks behavior and responsiveness to detect consciousness changes, which GEWS evaluates more gradually than NEWS. These overlaps suggest limited added value from integrating NIPDS into GEWS without further validation.\u003c/p\u003e\n\u003cp\u003eAnother key priority is improving the detection of delirium, especially given its prognostic importance in older adults. In our study, delirium was identified in only 8.2% of patients based on CAM or DOS assessments \u0026ndash; substantially lower than the expected prevalence of 23.0%[24]. This discrepancy suggests possible under-detection in our study population and limits our ability to draw conclusions about the adequacy of the AVPU scale within GEWS for identifying delirium. We therefore propose that future studies consider evaluating the Richmond Agitation and Sedation Scale (RASS) as an alternative to AVPU within GEWS. RASS provides a graded assessment of consciousness through direct observation and has demonstrated high sensitivity (82.0%) and specificity (85.1%) for identifying delirium in older ED patients[25]. Furthermore, changes in RASS have been associated with improved specificity, up to 92.0%[25]. The scale is time-efficient (requiring less than 10 seconds), does not depend on proxy input, and is applicable across various clinical environments, including the ED, ICU, acute ward and perioperative settings. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStrength and limitations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study is the first to validate a GEWS specifically tailored to frail older adults, using cut-off values adjusted to their physiological characteristics. Unlike conventional EWSs, which primarily aim to prevent ICU transfers or cardiopulmonary resuscitation, GEWS focuses on patient-centered outcomes, including Type 4 clinical events. This enables more proactive care that respects individual patient preferences \u0026ndash; many of whom prefer to avoid intensive interventions in favor of personalized treatment approaches. GEWS supports timely escalation of care by nurses and geriatricians in alignment with these preferences.\u003c/p\u003e\n\u003cp\u003eA second strength is the study\u0026rsquo;s real-world, multicenter design, conducted without additional training for nurses or geriatricians. This enhances the external validity of the findings, which are likely generalizable \u0026ndash; at least within Belgian geriatric wards and emergency departments. None of the participating centers employed RRTs linked to NEWS or GEWS scores. Instead, nurses acted based on clinical judgement or standing orders.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA third strength, is the use of robust methodology, not only including C-statistics, but also performance metrics such as PR-AUC, and clinical burden analysis. These measures are in accordance with best practices for EWS evaluation and are comparable to the approach used by Pankhurst et al.[19] in assessing NEWS2 response thresholds in a UK acute hospital setting.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNonetheless, several limitations should be acknowledged. First, the study population included no robust older adults (i.e. CFS \u0026lt; 3), limiting the applicability of the findings to that subgroup. Since robust and frail older individuals may differ physiologically, NEWS may remain more appropriate for robust patients until further comparative studies are available.\u003c/p\u003e\n\u003cp\u003eSecond, the classification of Type 4 clinical events was based on a predefined list of medical interventions, which in hindsight may have led to an underestimation of clinical event rates. Notably, the administration of intravenous analgesia for severe pain (NRS\u0026nbsp;\u0026ge;7) was not included as a qualifying Type 4 event. While this omission was unintentional, it may have impacted the measured GEWS performance by overlooking relevant acute clinical response.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThird, no item-level analysis was conducted to assess the individual contribution of each vital parameter to the overall score performance. Some parameters may have had limited impact. For instance, respiratory rate \u0026ndash; despite being a highly sensitive indicator of deterioration \u0026ndash; was often estimated rather than measured, potentially reducing GEWS sensitivity. Improving awareness through training, along with the implementation of validated automated monitoring tools for older adults, could enhance sensitivity. Furthermore, pain was not consistently recorded alongside other vital parameters in two of the three centers, resulting in a higher rate of missing data. Lastly, the level of consciousness and delirium were frequently underreported. These issues in data quality may have impacted the accuracy of GEWS and warrant further investigation. \u0026nbsp;\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study validated GEWS as a more accurate, geriatric-specific alternative to NEWS for detecting clinical deterioration in frail older patients across all clinical event types (1\u0026ndash;4). A GEWS threshold of \u0026ge;\u0026thinsp;5 was found to provide the optimal balance between predictive performance and clinical feasibility, making it the most appropriate cut-off for routine monitoring. By facilitating earlier and more individualized interventions, GEWS supports care that is better aligned with the complex needs and preferences of frail older patients. For triggering RRT activation, a threshold of \u0026ge;\u0026thinsp;8 is recommended. Importantly, clinical judgement of nurses and physicians remain indispensable and should guide escalation decisions, regardless of the GEWS score.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eCompeting interests and funding:\u003c/strong\u003e\u003c/em\u003e The authors declare no financial or non-financial conflict of interest. This study received no external funding. Statistical support was provided through internal resources from the geriatric departments of Ghent University Hospital, St. Trudo Hospital and Jessa Hospital, as well as through personal funding.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eAuthor contributions:\u003c/strong\u003e\u0026nbsp;\u003c/em\u003eH.B. and N.V.D.N. were responsible for conceptualization of the study. Methodology was developed by H.B., S.M., M.V.A., N.V.D.N., and S.W., with project administration managed by H.B.\u0026nbsp;\u003cbr\u003e\u0026nbsp;H.B., S.M. and M.V.A. led the data curation and investigation. Formal statistical analysis was performed by S.W. Validation of the data was done by F.H. and S.W. H.B., A.B., F.H., and N.V.D.N. contributed to the interpretation of the results and their clinical relevance. The original draft was written by H.B. Supervision was done by N.V.D.N and A.B. All authors contributed to reviewing and editing the manuscript. Internal funding acquisition was coordinated by H.B., S.M., M.V.A., and N.V.D.N.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eAcknowledgements:\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThe authors would like to sincerely thank St. Trudo Hospital (St. Truiden), Jessa Hospital (Hasselt), and AZ Alma Hospital (Eeklo) for their valuable collaboration and support in conducting this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eWilliams B. The National Early Warning Score: from concept to NHS implementation. Clinical Medicine 2022;22:499\u0026ndash;505. https://doi.org/10.7861/clinmed.2022-news-concept.\u003c/li\u003e\n\u003cli\u003eRoland D, Stilwell PA, Fortune P-M, Alexander J, Clark SJ, Kenny S. Case for change: a standardised inpatient paediatric early warning system in England. Arch Dis Child 2021;106:648. https://doi.org/10.1136/archdischild-2020-320466.\u003c/li\u003e\n\u003cli\u003eUmar A, Ameh CA, Muriithi F, Mathai M. Early warning systems in obstetrics: A systematic literature review. PLOS ONE 2019;14:e0217864. https://doi.org/10.1371/journal.pone.0217864.\u003c/li\u003e\n\u003cli\u003eEchevarria C, Steer J, Bourke SC. Comparison of early warning scores in patients with COPD exacerbation: DECAF and NEWS score. Thorax 2019;74:941\u0026ndash;6. https://doi.org/10.1136/thoraxjnl-2019-213470.\u003c/li\u003e\n\u003cli\u003eRoyal College of Physicians. National Early Warning Score (NEWS): standardising the assessment of acute-illness severity in the NHS. Report of a working party. London: RCP, 2012. London: Royal College of Physisicians; 2012.\u003c/li\u003e\n\u003cli\u003eSubbe CP, Kruger M, Rutherford P, Gemmel L. Validation of a modified Early Warning Score in medical admissions. QJM 2001;94:521\u0026ndash;6. https://doi.org/10.1093/qjmed/94.10.521.\u003c/li\u003e\n\u003cli\u003eChurpek MM, Yuen TC, Winslow C, Hall J, Edelson DP. Differences in vital signs between elderly and nonelderly patients prior to ward cardiac arrest. Critical Care Medicine 2015;43:816\u0026ndash;22. https://doi.org/10.1097/CCM.0000000000000818.\u003c/li\u003e\n\u003cli\u003eVardy ER, Lasserson D, Barker RO, Hanratty B. NEWS2 and the older person. Clin Med (Lond) 2022;22:522\u0026ndash;4. https://doi.org/10.7861/clinmed.2022-0426.\u003c/li\u003e\n\u003cli\u003eR\u0026oslash;nningen PS, Walle-Hansen MM, Ihle-Hansen H, Andersen EL, Tveit A, Myrstad M. Impact of frailty on the performance of the National Early Warning Score 2 to predict poor outcome in patients hospitalised due to COVID-19. BMC Geriatrics 2023;23:134. https://doi.org/10.1186/s12877-023-03842-0.\u003c/li\u003e\n\u003cli\u003eChester JG, Rudolph JL. Vital Signs in Older Patients: Age-Related Changes. J Am Med Dir Assoc 2011;12:337\u0026ndash;43. https://doi.org/10.1016/j.jamda.2010.04.009.\u003c/li\u003e\n\u003cli\u003eauthor-health. In : Blikvanger: Algemene ziekenhuizen. Naar een gezond Belgi\u0026euml; 2025. https://www.gezondbelgie.be/nl/blikvanger-gezondheidszorg/algemene-ziekenhuizen/download-hier-het-volledige-rapport-in-pdf. Accessed April 13, 2025.\u003c/li\u003e\n\u003cli\u003eKoninklijk besluit tot wijziging van het koninklijk besluit van 29 januari 2007 houdende vaststelling eensdeels, van de normen waaraan het zorgprogramma voor de geriatrische pati\u0026euml;nt moet voldoen om te worden erkend en, anderdeels, van bijzondere aanvullende normen voor de erkenning van ziekenhuizen en ziekenhuisdiensten. federale overheidsdienst volksgezondheid, veiligheid van de voedselketen en leefmilieu; 2014. https://etaamb.openjustice.be/nl/koninklijk-besluit-van-26-maart-2014_n2014024118. Accessed March 26; 2014. \u003c/li\u003e\n\u003cli\u003eGeeraert G (ugent)02006633. De waarde van de Early Warning Score voor geriaters 2022:21.\u003c/li\u003e\n\u003cli\u003eBaeyens Hilde, Dekoninck Julien, Baeyens Jean-Pierre. P-523 Vital signs in elder patients, urgent need to develop \u0026lsquo;\u0026ldquo;GEWS\u0026rdquo;\u0026rsquo; (= Geriatric Early Warning System). Eur Geriatr Med 2018;Eur Geriatr Med (2018) 9 (Suppl 1):S1\u0026ndash;S367. https://doi.org/10.1007/s41999-018-0097-4.\u003c/li\u003e\n\u003cli\u003eRodr\u0026iacute;guez-Molinero A, Narvaiza L, Ruiz J, G\u0026aacute;lvez-Barr\u0026oacute;n C. Normal Respiratory Rate and Peripheral Blood Oxygen Saturation in the Elderly Population. Journal of the American Geriatrics Society 2013;61:2238\u0026ndash;40. https://doi.org/10.1111/jgs.12580.\u003c/li\u003e\n\u003cli\u003eTakayama A, Nagamine T, Kotani K. Aging is independently associated with an increasing normal respiratory rate among an older adult population in a clinical setting: A cross-sectional study. Geriatr Gerontol Int 2019;19:1179\u0026ndash;83. https://doi.org/10.1111/ggi.13788.\u003c/li\u003e\n\u003cli\u003eVardy ER, Santhirasekaran S, Cheng M, Anand A, MacLullich AM. NEWS2 shows low sensitivity and high specificity for delirium detection: a single site observational study of 13,908 patients. Clin Med (Lond) 2022;22:544\u0026ndash;8. https://doi.org/10.7861/clinmed.2022-0345.\u003c/li\u003e\n\u003cli\u003eRomero-Brufau S, Huddleston JM, Escobar GJ, Liebow M. Why the C-statistic is not informative to evaluate early warning scores and what metrics to use. Crit Care 2015;19:285. https://doi.org/10.1186/s13054-015-0999-1.\u003c/li\u003e\n\u003cli\u003ePankhurst T, Sapey E, Gyves H, Evison F, Gallier S, Gkoutos G, et al. Evaluation of NEWS2 response thresholds in a retrospective observational study from a UK acute hospital. BMJ Open 2022;12:e054027. https://doi.org/10.1136/bmjopen-2021-054027.\u003c/li\u003e\n\u003cli\u003eMitsunaga T, Hasegawa I, Uzura M, Okuno K, Otani K, Ohtaki Y, et al. Comparison of the National Early Warning Score (NEWS) and the Modified Early Warning Score (MEWS) for predicting admission and in-hospital mortality in elderly patients in the pre-hospital setting and in the emergency department. PeerJ 2019;7:e6947. https://doi.org/10.7717/peerj.6947.\u003c/li\u003e\n\u003cli\u003eHaegdorens F. Predicting serious adverse events or a safety net \u0026ndash; Rethinking the role of early warning scores. Resuscitation Plus 2024;17:100534. https://doi.org/10.1016/j.resplu.2023.100534.\u003c/li\u003e\n\u003cli\u003eHaegdorens F, Wils C, Franck E. Predicting patient deterioration by nurse intuition: The development and validation of the nurse intuition patient deterioration scale. International Journal of Nursing Studies 2023;142:104467. https://doi.org/10.1016/j.ijnurstu.2023.104467.\u003c/li\u003e\n\u003cli\u003eHaegdorens F, Lefebvre J, Wils C, Franck E, Van Bogaert P. Combining the Nurse Intuition Patient Deterioration Scale with the National Early Warning Score provides more Net Benefit in predicting serious adverse events: A prospective cohort study in medical, surgical, and geriatric wards. Intensive and Critical Care Nursing 2024;83:103628. https://doi.org/10.1016/j.iccn.2024.103628.\u003c/li\u003e\n\u003cli\u003eGibb K, Seeley A, Quinn T, Siddiqi N, Shenkin S, Rockwood K, et al. The consistent burden in published estimates of delirium occurrence in medical inpatients over four decades: a systematic review and meta-analysis study. Age Ageing 2020;49:352\u0026ndash;60. https://doi.org/10.1093/ageing/afaa040.\u003c/li\u003e\n\u003cli\u003eHan JH, Vasilevskis EE, Schnelle JF, Shintani A, Dittus RS, Wilson A, et al. The Diagnostic Performance of the Richmond Agitation Sedation Scale for Detecting Delirium in Older Emergency Department Patients. Academic Emergency Medicine 2015;22:878\u0026ndash;82. https://doi.org/10.1111/acem.12706.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 4 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"european-geriatric-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"EGEM","sideBox":"Learn more about [European Geriatric Medicine](https://www.springer.com/journal/41999)","snPcode":"41999","submissionUrl":"https://www.editorialmanager.com/egem/default2.aspx","title":"European Geriatric Medicine","twitterHandle":"","acdcEnabled":false,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"EWS, GEWS - frail, older people, validation, clinical burden","lastPublishedDoi":"10.21203/rs.3.rs-6635964/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6635964/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose\u003c/strong\u003e: Early warning scores, such as the National Early Warning Score (NEWS), are less accurate in detecting clinical deterioration in frail older individuals due to age-related altered physiological responses. A Geriatric Early Warning Score (GEWS) was developed to take into account older adults’ frailty. This study aimed to validate GEWS and compare its predictive accuracy with NEWS, as well as to evaluate the clinical burden of GEWS.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e: In this prospective multicenter observational study, patients admitted to acute geriatric wards were included. Clinical deterioration was defined as the occurrence of one of four events: (1) unexpected death, (2) ICU transfer, (3) transition to palliative care, or (4) urgent medical/surgical intervention. GEWS and NEWS were compared using C-statistics, performance and clinical burden metrics.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: Among 511 patients, 348 events were recorded in 302 individuals. GEWS significantly outperformed NEWS across all event types, showing higher AUROC and PR-AUC values (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001). At a GEWS threshold ≥5, predictive performance was superior in accuracy (0.940 vs. 0.927), PPV (0.497 vs. 0.365), and specificity (0.977 vs 0.967), all \u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001. GEWS ≥5 was also associated with a lower clinical burden (NNE: 2.013 vs 2.738, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001). For life-threatening events (Type 1-3), GEWS ≥8 provided higher specificity (0.999 vs. 0.995, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001) and a reduced alarm rate (0.518 vs.1.347, \u003cem\u003ep\u003c/em\u003e\u0026lt;0.0001).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e: GEWS offers a more accurate, geriatric-specific alternative to NEWS for detecting clinical deterioration in frail older adults, while reducing clinical workload. A threshold of GEWS ≥5 is recommended for clinical alerting, whereas ≥8 for rapid response team activation.\u003c/p\u003e","manuscriptTitle":"Validation and Performance of a Geriatric Early Warning Score (Gews) Versus the National Early Waring Score (News) in Predicting Clinical Deterioration in Frail Older Patients","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-20 09:36:21","doi":"10.21203/rs.3.rs-6635964/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Major revisions","date":"2025-06-20T10:06:46+00:00","index":"","fulltext":""},{"type":"reviewerAgreed","content":"","date":"2025-05-29T00:04:54+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-05-14T19:29:36+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"European Geriatric Medicine","date":"2025-05-12T18:14:55+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-05-11T19:35:21+00:00","index":"","fulltext":""},{"type":"submitted","content":"European Geriatric Medicine","date":"2025-05-10T12:40:25+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"european-geriatric-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"EGEM","sideBox":"Learn more about [European Geriatric Medicine](https://www.springer.com/journal/41999)","snPcode":"41999","submissionUrl":"https://www.editorialmanager.com/egem/default2.aspx","title":"European Geriatric Medicine","twitterHandle":"","acdcEnabled":false,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"43aa5cf4-9050-4cde-a8ac-429eb7161ffa","owner":[],"postedDate":"May 20th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-09-17T07:55:29+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-20 09:36:21","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6635964","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6635964","identity":"rs-6635964","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}