Risk stratification tool for older emergency department patients: the 6-items brief geriatric assessment Authors: Prabakar VaittinadaAyar, Dorothee Riedlinger, Sara Fakhiri, Marc-Florent Tassi, Mehmet A Karamercan, Defne Z Dündar, Mehmet Ergin, Andrea Fabbri, Lars P Bjornsen, Sándor Somodi, Visnja N Adam, Effie Polyzogopoulou, Avni Demir Huseyin, Anna Slagman, Said Laribi Published: 2026 ABSTRACT AbstractBackgroundOlder patients are increasingly presenting to Emergency Departments (EDs) and therefore, outcome prediction tools are important to provide adequate care. The study aimed to explore the value of the 6-item brief geriatric assessment (BGA) as a rapid screening tool for identifying the potential risk of adverse outcomes in older patients.MethodsThis was an observational prospective cohort study, conducted in 36 European EDs and included patients aged 65 years and older with any complaint for seven consecutive days. The 6-item BGA was performed at admission and classified the population into three levels of risk (low (LR), moderate (MR) or high (HR)). Generalised linear mixed models were used to assess the association of end points (admission, length of stay (LOS) and death) and BGA risk groups.ResultsAmong 4657 patients, 48% were men and the median age was 77 years. The association of ward admission was similar in BGA LR and MR groups but higher in the HR group (risk ratio RR = 1.22 [95%-confidence interval (CI) 1.13–1.31]). Compared to the LR group, the hospital LOS was similar in MR but higher in the HR group with a LOS ratio at 1.31 [95%-CI 1.20–1.42]. Mortality was higher in the MR and HR groups with RR = 2.02 [95%-CI 1.20–2.83] and 3.12 [95%-CI 2.47–3.77] respectively.ConclusionThe study showed that higher BGA risk levels were associated with adverse outcomes in older ED patients. The 6-item BGA may help support rapid risk stratification and complement existing ED triage systems for identifying frailer patients. Further prospective studies are needed before routine implementation.Trial registrationThe study was registered with the U.S. National Library of Medicine Clinical Trials (trial number NCT04680299; registration date: 2020-12-22).Supplementary InformationThe online version contains supplementary material available at 10.1186/s12877-026-07825-9.Keywords: Brief geriatric assesment, Emergency department, Risk stratification Introduction Older patients are increasingly presenting to Emergency departments (ED) due to the world’s ageing population [1–3]. Older adults currently make up about one-quarter of all ED patients [4, 5]. Due to increased frailty and a higher burden of comorbidities, this group faces a significantly elevated risk of adverse outcomes and mortality [6].Previous studies have shown that patients aged 65 years or older require more medical care, were more often hospitalised, had a longer hospital length of stay (LOS) [7–9]. Extended wait times for hospital admission have been linked to missed medications, immobilisation and disrupted sleep, all increasing the risk of adverse outcomes [10]. To improve and initiate adequate care in this aged population, outcome prediction tools are important. Many tools have been developed to assess frailty in older patients: Frailty Index (FI), the FRAIL scale, Clinical Frailty Scale [CFS], or Tilburg Frailty Indicator [TFI] [11]. The Comprehensive Geriatric Assessment (CGA) is the gold standard for evaluating older adults, covering medical, functional, cognitive, psychological, and social domains to guide personalised care. However, its use in routine practice remains limited because of time, staffing, and logistical constraints, especially in busy settings such as primary care and emergency departments [12, 13]. A recently developed clinical tool, the 6-item Brief Geriatric Assessment (BGA), has shown promise as a straightforward method for stratifying older adults in the ED according to their risk of adverse outcomes [14]. The Brief Geriatric Assessment (BGA) was developed as a rapid screening approach using validated, time-efficient tools to identify high-risk older adults. The 6-item BGA provides a classification into three levels, the highest level being predictive of prolonged hospital stay, higher readmission rate after ED visit, and higher in-hospital and long-term mortality [7, 15–19]. This BGA tool has been validated in large cohorts in Canada [19] and France [14, 16] but to our knowledge not at a multinational level. Recently a large European cohort, the EGERS study, enrolled patients over 65 years in 36 EDs of 9 European countries. This study showed an hospital admission rate of 51% with an overall in-hospital mortality of 8% [20]. The aim of this planned sub-study of the EGERS cohort was to explore the value of the 6 item BGA on a large European cohort as a rapid screening tool for identifying the potential risk of adverse outcomes in older patients. Methods The study design, population, variables and outcomes measures were presented in detail in the EGERS study [20]. Briefly it was an international, multicenter, observational, prospective cohort study conducted in 36 EDs in 9 European countries over a span of seven consecutive days for patient recruitment. All consecutive patients aged 65 years or older, presenting to participating EDs with any complaint were considered for inclusion, between October 19th and November 30th 2020. Data collection encompassed various aspects, including demographic information, timing and day of admission, primary symptoms, vital signs at ED presentation, existing medical conditions, regular medications, post-ED outcomes (including patient disposition, in-hospital outcome, length of stay and final hospital diagnoses). The primary outcome was in-hospital mortality 30 days after ED visit, stratified by risk levels (low, moderate, high) based on the 6-item BGA. The secondary objectives based on BGA risk levels included assessing hospitalization rates and hospital LOS. The 6-item BGA was performed at admission. It listed the following items: age category (> 85), male gender, polypharmacy (≥ 5 different medications per day), receipt of home support, past 6-month history of falls, and temporal disorientation (inability to give the current month and/or year). These items were coded “yes” or “no”. Temporal disorientation and a history of falls have been identified as key predictors of in-hospital adverse outcomes compared to other BGA items [7]. Therefore, these two major factors were assigned a score of 5 points each, while each of the four minor items received a score of 1 point [21]. Scores ranged from 0 (the lowest risk) to 14 (highest risk). The 6-item BGA stratifies the risk in 3 levels: low risk (0 to 3, (LR)), moderate risk (4 and 5, (MR)), and high risk (> 5, (HR)) (supplementary material Table S1). Ethics The study adhered to the Declaration of Helsinki, and was approved by the ethics committee of TC GAZI University (Institutional Review Board (IRB) n° 91610558-604.01.02; Research code number: 2019 − 263). As this study involved no change in clinical care, no study-specific interventions for participants and no collection of any personally identifiable information, the IRB approved the study with a waiver of written informed consent. Statistical analysis Patients’ demographics, characteristics and outcomes are presented according to their BGA risk level. Categorical variables are reported as counts and percentage and continuous variables as median, first and third quartile (Q1, Q3). For each variable, non-adjusted comparison across BGA risk level groups was done using chi-square test for categorical variables or Kruskal–Wallis test for continuous variables. Generalised linear mixed models were used to account for clustering effect within each country, and quantify the magnitude of the association between risk level and hospitalization, length of stay (LOS) or mortality. Also, BGA variables were not included in multivariable analyses, in order to avoid collinearity and overadjustment. Risks for hospitalization or mortality were modelled using binomial distribution and logit link. For the in-hospital LOS, because most patients stayed for less than one day, we used a zero-inflated generalised Poisson model to estimate correctly the association of risk level and hospital LOS [22]. ED LOS and observation unit LOS were modelled with lognormal distribution. Results are reported as conditional risk ratios for hospital admission and mortality and as conditional length ratio for LOS. Initially, the analyses were performed using complete-case analysis, meaning that for each model, all patients with missing data for the BGA score or the outcome were excluded from the analysis. In the imputed version, 50 datasets were created in which missing values were replaced by predicted values derived from all available data. The models were developed using these imputed datasets. Fifty imputed datasets were generated using chained equations with the predictive mean matching method. For each variable, the predictors included demographic characteristics, medical and medication history, clinical data collected during emergency department management, as well as a country-specific random effect. As a result, there were no remaining missing data in any of the 50 datasets. The models were then fitted separately in each of the 50 datasets, including all subjects in every analysis. The final estimates were obtained by pooling the results across the 50 imputed models. Statistical analyses were performed using R version 4.5.1 (download date 2025-06-31) [23]. Results Population characteristics The EGERS cohort included 5767 patients. Missing data to calculate the BGA score resulted in the exclusion of 1110 cases from the EGERS dataset (19%). Analyses were conducted in 4657 patients for whom a complete variable set for the calculation of the BGA was available (Fig. 1). Flow chart Of 4657 patients, 3220 (69%) patients were classified as LR according to the BGA score (Table 1). The median age was 77 years (Q1-Q3: 71–84). As expected due to the construction of the BGA score, HR patients were older, 84 years (76–89) as compared to low-risk patients 75 years (70–81) and to MR patients 75 (70–82); p-value < 0.001. Men represented 48% (n = 2231) of the study population. The distribution of sex across risk groups was not independent, but directly influenced by the score construction. This explained why the moderate-risk group was predominantly composed of female patients. France, Germany, Ireland, Italy and Norway had a higher proportion of HR patients. The day or the time of admission had no influence on the risk level. Traumatic complaints represented about 25% of patients in the MR and HR groups. Patients in the high risk group more frequently presented with traumatic complaints. LR and MR groups had respectively 14% and 15% of patients with no comorbidity vs. 4% in the high risk group (p-value < 0.001). They also had 19% and 17% of their population with no chronic medication vs. 9.5% in the high risk group (p-value < 0.001). The HR group had higher proportion for cardiovascular diseases and risk factors (prior stroke, hypertension, heart failure, dyslipidemia, alcohol abuse). They also had a higher proportion of hypertension, statins and antiplatelet treatment. Table 1Patients demographics and characteristicsCharacteristic6-items Brief Geriatric Assessment (risk)P-valuealow (0–3)N = 3220moderate (4–5)N = 226high (> 5)N = 1211Sex*< 0.001 Female1615, 50%(48–52)209, 92%(88–95)584, 48%(45–51) Male1605, 50%(48–52)17, 8%(5–12)627, 52%(49–55)Age (median, (Q1-Q3))75 (70–82)75 (70–81)84 (76–89)< 0.001Country< 0.001 Croatia114, 3% (3–4)9, 4% (2–8)42, 4% (3–5) France1092, 34% (32–36)77, 34% (28–41)526, 43% (41–46) Germany58,2% (1–2)6, 3% (1–6)38, 3.1% (2–4) Greece107, 3% (2–4)3, 1% (0–4)34, 3% (2–4) Hungary146, 5% (4–5)13, 6% (3–10)50, 4% (3–5) Ireland235, 7% (6–8)13, 6% (3–10)89, 7% (6–9) Italy193, 6% (5–7)23, 10% (7–15)124, 10% (9–12) Norway129, 4% (3–5)7, 3% (1–7)73, 6% (5–8) Turkey1146, 36% (34–37)75, 33% (27–40)235, 19% (17–22)Time admission0.4 00:00–08:00328, 10% (9.2–11)29, 13% (8.9–18)135, 11% (9.5–13) 08:01–16:001735, 54% (52–56)117, 52% (45–58)640, 53% (50–56) 16:01–20:00739, 23% (22–24)44, 19% (15–25)261, 22% (19–24) 20:01–23:59417, 13% (12–14)36, 16% (12–22)174, 14% (12–17)Day admission0.7 Monday516, 16% (15–17)30, 13% (9.3–19)219, 18% (16–20) Tuesday512, 16% (15–17)33, 15% (10–20)182, 15% (13–17) Wednesday504, 16% (14–17)35, 15% (11–21)176, 15% (13–17) Thursday484, 15% (14–16)45, 20% (15–26)175, 14% (13–17) Friday434, 13% (12–15)30, 13% (9–19)169, 14% (12–16) Saturday419, 13% (12–14)31, 14% (10–19)162, 13% (12–15) Sunday350, 11% (10–12)22, 10% (6–15)127, 10% (9–12)Major complaint< 0.001 Non trauma2791, 87% (85–88)156, 69% (62–75)841, 70% (67–72) Trauma387, 12% (11–13)54, 24% (19–30)300, 25% (22–27) Both42, 1% (1–2)16, 7% (4–11)69, 6% (5–7)Comorbidity2754, 86% (84–87)193, 85% (80–90)1159, 96% (94–97)< 0.001 Hypertension1799, 56% (54–58)121, 54% (47–60)750, 62% (59–65)< 0.001 Diabetes mellitus773, 24% (23–26)61, 27% (21–33)315, 26% (24–29)0.3 Coronary artery disease684, 21% (20–23)36, 16% (12–22)258, 21% (19–24)0.2 Dyslipidemia523, 16% (15–18)25, 11% (7–16)206, 17% (15–19)0.081 Active Malignancy375, 12% (11–13)16, 7% (4–11)128, 11% (9–12)0.090 Asthma/COPD402, 13% (11–14)17, 8% (5–12)125, 10% (9–12)0.027 Prior stroke251, 8% (7–9)21, 9% (6–14)242, 20% (18–22)< 0.001 Dementia/Alzheimer0, 0% (0–0)29, 13% (9–18)427, 35% (33–38)< 0.001 CKD with dialysis48, 2% (1–2)6, 3% (1–6)21, 2% (1–3)0.4 Left ventricular failure236, 7% (7–8)22, 10% (6–13)128, 11% (9–12)0.002 Right ventricular failure66, 2% (2–3)6, 3% (1–6)23, 2% (1–3)0.8 Coronary revascularisation230, 7% (6–8)11, 5% (3–9)81, 7% (5–8)0.4 Chronic Liver Disease62, 2% (2–3)6, 3% (1–6)36, 3% (2–4)0.10 Smoking145, 5% (4–5)8, 4% (2–7)55, 5% (4–6)0.8 Alcohol Abuse74, 2% (2–3)4, 2% (1–5)55, 5% (4–6)0.002Chronic Medication2598, 81% (79–82)185, 82% (76–87)1096, 91% (89–92)< 0.001 ACE Inhibitors +/- Arbs1130, 35% (33–37)70, 31% (25–38)429, 35% (33–38)0.4 Betablockers996, 31% (29–33)64, 28% (23–35)408, 34% (31–36)0.11 Antiplatelet969, 30% (29–32)57, 25% (20–31)430, 36% (33–38)< 0.001 Diuretics784, 24% (23–26)46, 20% (15–26)391, 32% (30–35)< 0.001 Statins736, 23% (21–24)38, 17% (12–22)283, 23% (21–26)0.088 Calcium antagonists584, 18% (7–20)31, 14% (9.6–19)232, 19% (17–22)0.2 Anti diabetics485, 15% (14–16)32, 14% (10–20)195, 16% (14–18)0.6 NSAID +/- Analgesics360, 11% (10–12)26, 12% (7.8–17)188, 16% (14–18)< 0.001 Antidepressant256, 8% (7–9)28, 12% (9–18)265, 22% (20–24)< 0.001 Inhaled beta2mimetics314, 10% (9–11)14, 6% (4–10)79, 7% (5–8)0.004 Oxygen+/- NIV at home45, 1% (1–2)0, 0% (0–2)20, 2% (1–3)0.14Polypharmacy (n ≥ 5) *698, 22% (20–23)32, 14% (10–20)378, 31% (29–34)< 0.001Home help services*539, 17% (15–18)17, 8% (5–12)646, 53% (50–56)< 0.001Fall within last 6 months*0, 0% (0–0)93, 41% (35–48)690, 57% (54–60)< 0.001Temporary disorientation*0, 0% (0–0)116, 51% (45–58)780, 64% (62–67)< 0.001Data are presented as n, % (95% confidence interval) or as Median (IQR)Abbreviations: COPD Chronic Obstructive Pulmonary Disease, CKD Chronic Kidney Disease, ACE Angiotensin Converting Enzyme, ARB Angiotensin Receptor Blocker, NSAID Non-steroidal Anti-Inflammatory Drugs, NIV Noninvasive Ventilation* Brief geriatric assessment itemsa Pearson’s Chi-squared test with simulated p-value ; Kruskal-Wallis rank sum test Patients demographics and characteristics Data are presented as n, % (95% confidence interval) or as Median (IQR) Abbreviations: COPD Chronic Obstructive Pulmonary Disease, CKD Chronic Kidney Disease, ACE Angiotensin Converting Enzyme, ARB Angiotensin Receptor Blocker, NSAID Non-steroidal Anti-Inflammatory Drugs, NIV Noninvasive Ventilation * Brief geriatric assessment items a Pearson’s Chi-squared test with simulated p-value ; Kruskal-Wallis rank sum test Outcomes Admission to the emergency department observation unit (EOU) proportion was higher in moderate and high risk groups respectively 26% and 26% vs. 19%; p < 0.001 (Table 2). But admission to a ward was more common in high risk group (53%) compared with the low- (41%) and moderate-risk (42%). Also for admission in intensive care unit, high risk patients were more likely to be admitted (13%) in comparison of low (8.4%) and moderate risk group (11%); p-value < 0.018. The length of stay (LOS) in EOU was longer for high risk group (median hours 8 [3 ;18]). The stay in EOU for low and moderate risk group was repectively (5.0 [2.0;13.0] and 4 [2.0;12]); p-value < 0.001. The median LOS in hospital days increased respectively according the risk score (low (4 [1;10]), moderate (5.0[1;12]), high (6.0[2;12]); p-value < 0.001). The proportion of in-hospital mortality also increased according the risk score ((low (6.0%), moderate (12%), high (16%); p-value < 0.001)). Table 2OutcomesCharacteristic6-items Brief Geriatric Assessment (risk)P-valuealow (0–3)N = 3220moderate (4–5)N = 226high (> 5)N = 1211EOU admission608, 19% (18–20)59, 26% (21–32)315, 26% (24–29)< 0.001Ward admission1304, 41% (39–42)94, 42% (35–48)638, 53% (50–56)< 0.001ICU admission263, 8% (7–9)25, 11% (8–17)152, 13% (11–15)0.018ED LOS (hours)5 (3–8)5 (2–9)6 (3–9)< 0.001EOU LOS (hours)5 (2–13)4 (2–12)8 (3–18)< 0.001Missing data 2614 168 897 Hospital LOS (days)4 (1–10)5 (1–12)6 (2–12)< 0.001In hospital mortality177, 6.0% (5.2–6.9)25, 12% (7.8–17)182, 16% (14–18)< 0.001Data are presented as n, % (95% confidence interval) or as Median (IQR)Abbreviations: ED Emergency Department, EOU Emergency Observation Unit, ICU Intensive Care Unit, LOS Length of Staya Pearson’s Chi-squared test with simulated p-value; Kruskal-Wallis rank sum test Outcomes Data are presented as n, % (95% confidence interval) or as Median (IQR) Abbreviations: ED Emergency Department, EOU Emergency Observation Unit, ICU Intensive Care Unit, LOS Length of Stay a Pearson’s Chi-squared test with simulated p-value; Kruskal-Wallis rank sum test Association between risk level and outcomes The admission to the EOU was more often associated with moderate and high risk groups respectively, Risk Ratio (RR) = 1.42 [95%-confidence interval (CI)1.02–1.81] and RR = 1.63 [95%-CI 1.36–1.90]. Ward admission was similar in LR and MR groups but higher in the HR group (RR = 1.22 [95%-CI 1.13–1.31]). ICU admission rate was higher in the HR group (RR = 2.05 [95%-CI 1.62–2.47]) (Fig. 2A forest plot). Forest plot. Risk ratios for admission (A) length of stay (B) and mortality (C) adjusted for comorbidities according 6-Item brief geriatric assessment risk group with multiple imputation analyses. Imputed cases (blue) vs. Complete cases (green) The probability of patients to stay one day or more was significantly higher in the HR group with a length ratio at 1.31 [95%-CI 1.20–1.42] but not in the MR group. HR BGA was associated with a longer ED LOS (in hours), with a Length Ratio = 1.11 [95%-CI 1.07–1.14). Among patients who stayed in the EOU, LOS was higher in the high risk group LR = 0.1.47 [95%-CI 1.26–1.68] (Fig. 2B forest plot). Hospital mortality was higher in moderate and high risk groups with respectively RR = 2.02 [95%-CI 1.20-2,83] and RR = 3.12 [95%-CI 2.47–3.77] (Fig. 2C forest plot). After multiple imputations for missing data (Fig. 2A, B, C), the results were the same between imputed cases (blue lines) vs. complete case (Green lines). Discussion This study showed that 6-item BGA score is a promising tool to stratify older ED patients in terms of hospital admission, length of stay and mortality. To our knowledge it is the largest study population in which the 6-item BGA was assessed. The 6-item BGA score could enable early identification and targeted management of frail patients at a higher risk of adverse outcome in the older ED population. Admission This study showed a higher risk of ward hospitalisation in the HR group. BGA score was assessed for the risk of unplanned hospital admission for older patients in primary care [24]. The risk of unplanned hospital admission for the high-risk group was significantly higher (adjusted OR = 3.71, P = 0.032, and adjusted hazard ratio (HR) = 2.81; P = 0.035). However, except for HR level patients, its performances were poor. Regarding ED readmission, BGA prognostic value was poor [17]. Other tools were assessed to predict or prevent unplanned admission or readmission. The Identification of Seniors at Risk (ISAR) [25, 26], the Triage Risk Screening Tool (TRST) [25–27], the eight-item questionnaire of Runciman [25], the seven-item questionnaire of Rowland [25, 27] and the Emergency Admission Risk Likelihood Index (EARLI) [28] had a high negative predictive value but predicted modestly unplanned readmission. A large retrospective study showed an association between emergency admission or readmission and age > 80 years and > 2 admissions during the previous year [29]. Frailty was associated with hospital admission [30]. Models developed to predict preventable hospitalisations tended to have better predictive performance than models to predict all-cause hospitalisations [31, 32]. Length of stay This cohort showed longer hospital stays in moderate- and high-risk groups, especially when LOS < 1 day was excluded. BGA 6-item was used to predict the hospital LOS in many studies [7, 16, 19, 21, 33]. In most of these studies the BGA had a good prediction capacity for prolonged hospital LOS. Another interesting point in the present study was that in high risk patients ED LOS was higher. Increased ED length of stay in high-risk patients aligns with prior evidence showing that prolonged ED stay in older adults is associated with worse in-hospital outcomes, including increased morbidity and mortality [10]. Mortality The concept of frailty had already demonstrate its importance to identify patients which were at risk of adverse outcomes compared to none-frail patients [30]. BGA was used to predict mortality in geriatric patients [18, 34]. Clinical frailty scale (CFS) was used in many studies to assess the risk of mortality in frail older ED patients [11, 35, 36] and particularly mortality at 90 and 180 days [36]. FRAIL (fatigue, resistance, ambulation, illnesses, and loss of weight) and CFS are simple frailty measures that may identify older adults at highest risk of adverse outcomes of hospitalisation. FRAIL performed better in predicting in-hospital mortality [35]. Frailty tools appear most useful for triage; BGA could be integrated similarly to the Clinical Frailty Scale [36]. BGA stratification and clinical signification This score was assessed in this prospective study on all consecutive patients older than 65 years. An interesting fact to observe was that there were few patients in the moderate risk group with a large proportion of women. This moderate-risk group should perhaps be managed as carefully as the high-risk group. They might also require longer hospitalisation or more frequent outpatient follow-ups after ED discharge. In the literature the moderate risk group was not underrepresented but it can be explained by the fact that in most studies patients were selected after ward admission [19, 33, 34] or in outpatient care to predict unplanned admission [24]. While the CFS primarily evaluates baseline frailty status through global clinical judgement of functional reserve and dependency, the BGA incorporates structured clinical and acute vulnerability parameters frequently encountered in the ED setting, including falls, temporal disorientation, medication burden, and social support requirements. We therefore emphasise that the BGA should not be considered a replacement for the CFS, but rather a potentially complementary tool aimed at rapid risk stratification in emergency care. In particular, the BGA may facilitate early identification of patients requiring closer monitoring or geriatric-oriented management during the initial ED assessment. The CFS is a well-validated frailty measure widely used to predict adverse outcomes in older adults, including hospitalisation and mortality [37]. Prolonged ED length of stay in older patients has been independently associated with adverse outcomes [10]. In some centres, the BGA was used to facilitate collaboration with geriatric teams and support early geriatric-oriented management for high-risk patients. In this context, the BGA helped support early targeted management and treatment, defined as clinical interventions tailored to patient risk stratification based on the BGA score, including prioritization for monitoring, diagnostic testing, or early therapeutic interventions before ward admission, particularly during periods of hospital bed shortages. It could also support triage prioritisation in older ED patients. However, further prospective studies are needed to evaluate its impact on patient outcomes and resource allocation. The BGA shares several characteristics with other rapid frailty screening tools, including the Scottish THINK Frail Tool [38] and the ProAGE score developed in Brazil in 2020 [39] and validated in 2025 [40] in both Brazilian and US cohorts, highlighting the overlap between existing frailty instruments and the need for greater harmonisation across clinical settings and countries. Limitations There are several limitations to this study. First, some data were collected during the COVID-19 pandemic, which may have affected healthcare system organisation and patient outcomes, including hospital admission rates, ED length of stay, and mortality. In addition, differences in healthcare access and organisation between participating EDs or regions may have influenced local ED utilization patterns and the types of medical conditions encountered. Factors such as geographic location, availability of primary care services, socioeconomic status, and regional healthcare infrastructure may also have affected ED use patterns (Supplementary Results and Figures S1–S3), disease severity at presentation, and clinical outcomes including admission, LOS, and mortality. Second, the study population was predominantly recruited from two countries (France and Turkey), resulting in a geographic imbalance that may limit the generalizability of the findings to other healthcare systems or populations. Although statistical analyses accounted for clustering effects, residual differences between countries and centers may persist. Third, more than 19% of the initial study population was excluded because of missing BGA data. No specific pattern or explanation for the missingness could be identified. While the remaining cohort remained large and heterogeneous, missing data may still have introduced selection bias. To address this issue, we performed multiple imputation analyses, which yielded results consistent with the primary analyses, supporting the robustness of our findings. Nevertheless, the possibility of residual bias related to missing data cannot be completely excluded.Another limitation concerns the relatively small size of the moderate-risk group, particularly among male patients. Although confidence intervals were not markedly wide, this subgroup imbalance may reduce statistical power and limit the robustness of subgroup analyses. Findings related to the moderate-risk category should therefore be interpreted cautiously and considered exploratory.In addition, the observational nature of the study exposes the analyses to potential residual confounding despite adjustment strategies and mixed-effect modelling. Associations identified in this study should therefore not be interpreted as causal relationships. Moreover, outcomes were limited to short-term in-hospital measures, including admission, ED LOS, and 30-day mortality. No long-term outcomes such as readmission rates, post-discharge functional decline, institutionalization, or long-term mortality were available. Future prospective studies are needed to better evaluate the long-term prognostic and clinical utility of the BGA in older ED patients. Finally, we acknowledge that the moderate-risk group represents a relatively small proportion of the cohort. This distribution was not predefined but reflects the actual distribution of patients when applying the BGA score in a large, real-world population. While the size of this subgroup is limited, we note that the associated estimates did not show markedly wide confidence intervals, suggesting that the precision of the main findings was not substantially compromised. Nevertheless, the smaller sample size may still reduce statistical power and limit the robustness of subgroup comparisons, particularly for less frequent outcomes. Conclusion The EGERS study showed that higher BGA risk levels were associated with increased hospital admission, longer ED length of stay, and higher short-term mortality among older ED patients. These findings suggest that the 6-item BGA score may be useful for rapid risk stratification in routine ED practice and could complement existing triage systems for identifying higher-risk older patients. However, given the observational design and study limitations, further prospective studies are needed to evaluate its added value compared with existing frailty tools and its impact on clinical outcomes and resource allocation. Supplementary Information Supplementary Material 1. Acknowledgements We want to thank Dr Pradeebane Vaittinada Ayar for his help in graphics elaboration. Abbreviations Angiotensin Converting Enzyme Angiotensin Receptor Blocker Brief Geriatric Assessment Chronic Kidney Disease Chronic Obstructive Pulmonary Disease Emergency Admission Risk Likelihood Index Emergency Department Emergency Observation Unit High Risk Hazard Ration Intensive Care Unit Identification of Seniors at Risk Length of Stay Low Risk Moderate Risk Non Iinvasive Ventilation Non-steroidal Anti-Inflammatory Drugs Odd Ratio Risk Ratio Triage Risk Screening Tool Authors’ contributions Conceptualization: Prabakar VaittinadaAyar, Saïd Laribi Formal analysis: Prabakar Vaittinada Ayar, Dorothee Riedlinger, Sara Fakhiri, Anna Slagman, Saïd Laribi Methodology and validation: All authors. Writing—original draft: Prabakar Vaittinada Ayar Writing—review and editing: All authors. All authors meet the ICMJE criteria, and provided final approval to submit the manuscript for publication. Funding None. Data availability All study data is available on reasonable request to the corresponding author, on approval of the EUSEM research network chair. Declarations Ethics approval and consent to participate The study adhered to the Declaration of Helsinki, and was approved by the ethics committee of TC GAZI University (IRB n° 91610558-604.01.02 ; Research code number: 2019 − 263). As this study involved no change in clinical care, no study-specific interventions for participants and no collection of any personally identifiable information, the IRB approved the study with a waiver of signed informed consent. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement All study data is available on reasonable request to the corresponding author, on approval of the EUSEM research network chair.