The global population is experiencing a speedy demographic transition characterized by a significant increase in the proportion of older population (1, 2). Advancements in healthcare, public health interventions and living standards have contributed to prolonged lifespan worldwide, resulting in a rapidly expanding geriatric population (3). The global population aged 60 years and older is expected to rise substantially, with one in six people reaching this age group by 2030 and approximately 2.1 billion older adults by 2050 (4). This demographic shift has important implications for healthcare systems, especially for emergency departments (EDs), where older population constitutes a growing proportion of patient visits. In Pakistan, where population surpasses 240 million, approximately 7% individuals are currently aged >60 years and this population is expected to further increase substantially in the coming decades (5).

Older adults frequently attend EDs with complex medical needs. Multimorbidity, polypharmacy, frailty and age related physiological decline often contribute to unusual clinical presentations and increased healthcare utilization among geriatric patients (6). These factors can further complicate the clinical assessment and increase the risk of adverse outcomes, including prolonged hospital stay, intensive care unit admission, and mortality. As a result, early diagnosis of high risk elderly patients in emergency settings is essential for improving clinical outcomes and optimizing healthcare resource allocation (7).

Triage is a critical part of emergency care that allow rapid assessment and prioritization of patients according to the severity and complexity of their condition and urgency of treatment. (8) Several standardized triage systems have been developed to support clinical decision making in EDs globally. Among the most extensively used triage systems are the Emergency Severity Index (ESI), the Canadian Triage and Acuity Scale (CTAS) the Manchester Triage System (MTS), and the Australasian Triage Scale (ATS), which categorize patients based on illness severity and anticipated resource utilization (9, 10). The Emergency Severity Index (ESI), a five-level triage algorithm, has become one of the most commonly used triage systems globally and has demonstrated acceptable reliability and validity in various EDs. (9, 11).

The performance of conventional triage systems may be compromised in elderly patients due to atypical presentations and multimorbidity. Older adults frequently presents with unusual manifestations, subtle physiological changes, and multiple co morbid conditions that can complicate clinical assessment and lessen the accuracy of triage classifications. (12) Studies evaluating the effectiveness of ESI in older populations have reported decreased predictive accuracy for important clinical outcomes in geriatric ED populations. In specific, research has demonstrated that elderly patients may be at high risk of under-triage, where seriously ill patients are assigned lower acuity categories, probably delaying timely recognition and intervention for critical conditions (13, 14).

In addition to conventional triage systems, physiological early warning scores have been found to assist physicians in diagnosing patients at high risk of clinical deterioration. The National Early Warning Score (NEWS) is one of the most extensively used physiological scoring systems and incorporates routinely measured vital signs such as respiratory rate, heart rate, blood pressure, oxygen saturation, temperature and level of consciousness (15). Several studies have demonstrated that higher values of NEWS are associated with increased risk of adverse outcomes among ED patients. Elevated NEWS scores at presentation have been linked to increased likelihood of hospital admission, transfer to intensive care units, and mortality in both general and geriatric emergency department populations (16, 17).

Most recently, geriatric specific risk assessment tools have been developed to enhance risk stratification among elderly population presenting to ED. One such tool is Screening Tool Risk Score Assessment in the Emergency Department for Geriatric patients (S-TRIAGE),which combine physiological parameters with clinical risk factors more observed in older adults to predict short term morality among geriatric emergency patients (18).

Despite the widespread use of triage systems and early warning scores in ED, accurately predicting adverse outcomes among geriatric population remains a significant challenge. Many existing studies have evaluated the performance of traditional triage systems and physiological scoring tools in general emergency department populations. However, comparatively limited evidence is available regarding the effectiveness of geriatric-specific screening tools in predicting clinical outcomes among elderly patients. Consequently, there is a growing need to identify more reliable tools for early risk stratification in geriatric emergency care. Therefore, the present study aims to compare the diagnostic accuracy of the Emergency Severity Index (ESI), National Early Warning Score (NEWS), and Screening Tool Risk Score Assessment in the Emergency Department for Geriatric Patients (S-TRIAGE) in predicting 30-day adverse outcomes among older adults presenting to the emergency department.

A longitudinal observational study was conducted in the emergency department of Ziauddin University Hospital (North Campus), Karachi. The study was conducted from Geriatric patients aged ≥65 years of either gender who presented to the ED during the study period and underwent triage assessment. Eligible patients were recruited using a non-probability consecutive sampling technique. Patients presenting with cardiac arrest, trauma, accidental injuries, or those unwilling to participate were excluded. The sample size was calculated using the Wan Nor Arifin sample size calculator based on a reported S-TRIAGE sensitivity of 91.2% and specificity of 42.3% for mortality prediction, assuming a 27.5% incidence of geriatric emergency visits, a 95% confidence level,10% margin of error, yielding a required sample of 130 participants. (18) Following ethical approval from the Ethical Review Committee of Ziauddin University Hospital and written informed consent from participants or their legal representatives, demographic characteristics, medical history, and clinical parameters were collected for all participants. Clinical variables included heart rate, blood pressure, respiratory rate, body temperature, oxygen saturation, and Glasgow Coma Scale (GCS). Each patient was assessed using three scoring systems including Emergency Severity Index (ESI), National Early Warning Score (NEWS), and Screening Tool Risk Score Assessment in the Emergency Department for Geriatric patients (S-TRIAGE). Patients were followed for 30 days after their ED visit to determine clinical outcomes, including hospital admission, intensive care unit admission, or mortality. Data were analyzed using Statistical Package for Social Sciences (SPSS) version 27. Continuous variables were reported as mean ± standard deviation, while categorical variables were presented as frequencies and percentages. Diagnostic performance of the scoring systems was evaluated using receiver operating characteristic (ROC) curve analysis. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC) were calculated to compare the predictive accuracy of ESI, NEWS, and S-TRIAGE in identifying adverse outcomes.

The enrolled participants demonstrated mean age of 77.82 ± 10.29 years. Among 130 patients, n=72 (55.4%) were males and n=58 (44.6%) were females. The baseline physiological parameters included a mean heart rate of 94.28 ± 19.09 bpm, systolic blood pressure of 144.78 ± 34.96 mmHg, diastolic blood pressure of 79.50 ± 17.20 mmHg, respiratory rate of 20.92 ± 10.60 breaths/min, temperature of 37.40 ± 0.54°C, SpO₂ of 93.56 ± 5.88%, and GCS score of 14.66 ± 0.72. Hypertension (73.1%) and diabetes mellitus (56.9%) were the most common comorbidities.

More than two-thirds of the patients (71.5%) had been hospitalized during the six months preceding presentation. The median S-TRIAGE, ESI, and NEWS scores were 3 (IQR, 2–4), 3 (IQR, 1–4), and 4 (IQR, 1–6), respectively. Based on S-TRIAGE, nearly half of the cohort (46.9%) was classified as intermediate risk, whereas 60.0% were categorized as low risk according to the NEWS risk stratification. During the index ED visit, 60 (46.2%) patients were admitted to the ward and 50 (38.5%) required ICU admission. At 30-day follow-up, 89 (68.5%) patients survived, while 14 (10.8%) died. These findings are summarized in Table 1.

Table 1: Baseline Demographic and Clinical Characteristics

Adverse 30-Day Outcome and Associated Factors

A composite adverse 30-day outcome (death, readmission, or ED revisit) occurred in 41 of 130 patients (31.5%). Patients with an adverse outcome were significantly older (p = 0.009) and had higher diastolic blood pressure (p = 0.034), lower SpO₂ (p = 0.003) and GCS (p < 0.001), higher NEWS (p < 0.001) S-TRIAGE (p < 0.001) scores, and lower ESI levels (p < 0.001) compared with those without an adverse outcome. Heart rate, systolic blood pressure, respiratory rate, and temperature were not significantly different between groups (all p > 0.05).

Among categorical variables, adverse outcomes were significantly associated with age group (p = 0.0007), S-TRIAGE risk category (p < 0.0001), ESI level (p < 0.0001), recent hospitalization within six months (p = 0.0006), and malignancy (p = 0.0298), whereas sex (p = 0.289), hypertension (p = 0.5127), diabetes mellitus (p = 0.1127), atrial fibrillation (p = 0.1759), coronary artery disease (p = 0.7768), chronic obstructive pulmonary disease (p = 0.5512), previous cerebrovascular accident (p = 1.000), and NEWS risk band (p = 0.1357) showed no significant association.

Table 2: Association Between Demographic, Clinical, and Risk-Stratification Variables and 30-Day Adverse Outcome (N = 130)

†Percentages are calculated within each row (i.e., the proportion of patients in that category who did, or did not, experience the adverse outcome). Bold p-values denote statistical significance (p ≤ 0.05).

Diagnostic Accuracy: ROC Analysis

Receiver operating characteristic (ROC) analysis demonstrated that S-TRIAGE had the highest discriminatory accuracy for predicting the composite adverse 30-day outcome, with an AUC of 0.810 (95% CI: 0.726–0.894), followed by ESI (AUC: 0.700, 95% CI: 0.601–0.800) and NEWS (AUC: 0.629, 95% CI: 0.527–0.731). Pairwise comparisons demonstrated that the AUC of S-TRIAGE was significantly greater than that of ESI (ΔAUC = 0.109, 95% CI: 0.035–0.184, p = 0.0037) and NEWS (ΔAUC = 0.181, 95% CI: 0.080–0.281, p = 0.0004), whereas no significant difference was observed between ESI and NEWS (ΔAUC = 0.071, 95% CI: −0.055 to 0.198; p = 0.2703).This is shown in table 3.

Table 3: Overall Discriminatory Accuracy of S-TRIAGE, ESI, and NEWS for 30-Day Adverse Outcome, and Pairwise Comparisons

Figure 1:  ROC curves for S-TRIAGE, ESI, and NEWS predicting the composite 30-day adverse outcome (N = 130).

Diagnostic Performance at Optimal Cut-offs

Diagnostic performance analysis demonstrated that S-TRIAGE (cut-off ≥4) achieved a sensitivity of 65.9% (95% CI: 50.5–78.4), specificity of 85.4% (95% CI: 76.6–91.3), positive predictive value (PPV) of 67.5% (95% CI: 52.0–79.9), negative predictive value (NPV) of 84.4% (95% CI: 75.6–90.5), and an overall accuracy of 79.2%. ESI Level 1 yielded a sensitivity of 56.1% (95% CI: 41.0–70.1), specificity of 85.4% (95% CI: 76.6–91.3), PPV of 63.9% (95% CI: 47.6–77.5), NPV of 80.9% (95% CI: 71.7–87.5), and accuracy of 76.2%. Using the optimal Youden cut-off (NEWS ≥3), NEWS demonstrated the highest sensitivity (80.5%, 95% CI: 66.0–89.8) but low specificity (37.1%, 95% CI: 27.8–47.5), with a PPV of 37.1% (95% CI: 27.8–47.5), NPV of 80.5% (95% CI: 66.0–89.8), and an accuracy of 50.8%. At the clinically recommended threshold (NEWS ≥5), sensitivity was 51.2% (95% CI: 36.5–65.7), specificity 65.2% (95% CI: 54.8–74.3), PPV 40.4% (95% CI: 28.2–53.9), NPV 74.4% (95% CI: 63.7–82.7), and overall accuracy 60.8% as shown in table 4.

Table 4. Diagnostic Performance of S-TRIAGE, ESI, and NEWS at Optimal Cut-offs

The findings of this study demonstrated that S-TRIAGE outperformed both ESI and NEWS in discriminating patients at risk of experiencing a composite adverse outcome within 30 days of emergency department presentation. The observed AUC of 0.810 exceeded those of ESI (0.700) and NEWS (0.629), supporting the superior predictive performance of S-TRIAGE in geriatric population. These findings are consistent with previous study, where S-TRIAGE demonstrated an AUROC of 0.824-0.826 for predicting 28-day mortality and the requirement for life-saving interventions (18). The comparable discriminatory performance observed in this study relative to the original derivation study provides preliminary support for the generalizability of S-TRIAGE. The discriminatory performance of NEWS observed in this study was lower than that reported in several previous studies involving older emergency department patients. Kim et al. (2020) reported that NEWS achieved an AUC of 0.820 for predicting in-hospital mortality in ED patients over 65 years, substantially outperforming the Korean Triage and Acuity Scale in the same cohort (17). Similarly, a systematic review and meta-analysis by Wei et al. (2023) reported pooled NEWS2 AUCs of 0.85-0.90 for short-term mortality across prehospital and ED settings (19). The comparatively lower performance of NEWS may reflect differences in outcome definitions, with prior studies focusing predominantly on mortality rather than composite adverse outcomes. The composite outcome included 30-day readmission and ED revisit alongside mortality. These additional outcomes are also influenced by functional status, social determinants, and chronic disease management beyond acute physiological derangement. Consequently, the lower performance of NEWS likely reflects its design for detecting acute clinical deterioration rather than predicting broader 30-day adverse outcomes. These findings are in agreement with an expanding body of literature reporting suboptimal performance of early warning scores in older adults. Kemp et al. (2020) reported that NEWS2 showed only modest discrimination for adverse outcomes in frail ED patients (20). Similarly, another study reported that NEWS2 accuracy for 30-day mortality drops sharply in patients over 75 years (AUC= 0.70) compared with younger patients (AUC= 0.88) (21). Age-related physiological alterations, including attenuated tachycardic and febrile responses, lower baseline blood pressure, and the effects of polypharmacy on cardiovascular parameters, may reduce the sensitivity of conventional vital-sign thresholds in older adults, (22, 23) consistent with the findings of the present study. The moderate performance of ESI is consistent with the heterogeneous findings reported in previous studies. Kemp et al. (2022) reported that ESI demonstrated good performance for predicting 3-day mortality in patients aged ≥65 years but showed lower accuracy for 30-day mortality and hospital admission (13). This pattern is consistent with our findings, highlighting that ESI is effective in initial triage but carries limited utility for predicting 30-day adverse outcomes. The observed performance of ESI is likely attributable to its intended purpose as a triage tool. Geriatric emergency medicine demonstrated that general-purpose triage systems, optimized for immediate clinical prioritization and resource utilization, are less effective in identifying older adults at risk of subsequent multifactorial clinical decline (24). These challenges have encouraged the development of geriatric-specific triage instruments such as ISAR-based adaptations (25) and machine-learning-derived admission-risk models, reflecting a broader shift toward purpose-built tools for this population rather than continued reliance on age-agnostic scores (26). The study limitations includes single-center design and relatively small sample size which may limit the generalizability of the findings. Moreover, reliance on routine clinical documentation introduces the potential for measurement error and information bias. Therefore, future studies should include larger, prospective, multicenter studies using standardized data collection and clinically homogeneous outcomes.

-TRIAGE exhibited superior discriminatory performance compared with ESI and NEWS in predicting adverse outcomes in geriatric ED patients, supporting the value of geriatric-specific risk stratification tools. The findings support integrating geriatric-specific clinical and physiological determinants into emergency triage frameworks to improve early risk recognition and timely management. Such instruments may enhance risk-adjusted decision-making and optimize outcomes among older adults presenting to the ED.