This retrospective observational study included patients ≥ 65 years of age visiting one of the two ED in Region Halland (RH) between 1st of January 2018 and 31st of December 2018. RH is located on the southwest coast of Sweden with a population of approximately 330,000, where 23% were older adults aged ≥ 65 years. The region had 468 hospital beds during the study period. Patients who arrive at the ED routinely present their chief complaint and are triaged according to RETTS, which is commonly used in Sweden [19]. In RETTS, triage levels are color-coded to indicate clinical urgency. Red represents life-threatening conditions requiring immediate attention, orange denotes very urgent cases, yellow indicates urgent but stable conditions, green is used for less urgent issues, and blue applies to non-urgent cases. The classification is based on vital signs and presenting symptoms to guide prioritization of care [20].
During the study period, all patients aged over 65 years who visited either of the two EDs in RH, Sweden were included if they had at least one recorded vital sign—SBP, heart rate (HR), SpO₂, respiratory rate (RR), body temperature, or LOC. Patients with no documented vital signs were excluded from the study.
Data was generated from the Regional Healthcare Information Platform (RHIP) [21]. RHIP links pseudo-anonymized data regarding healthcare encounters and healthcare utilization in the region using routinely collected data during standard care. Variables extracted were age, gender, comorbidities before visits to the ED. The included individuals were grouped by age into three categories: 65–74 years, 75–84 years, and 85 years or older. Comorbidities were calculated using Charlson Comorbidity Index (CCI). Comorbidities was based on all primary and secondary diagnoses across visits to all caregivers and divided into mild (CCI score 1–2), moderate (CCI score 3–4) and severe (CCI score ≥ 5) [22]. All diagnoses were registered according to the International Classification of Disease-10 (ICD-10). In addition, chief complaint, mode and time of arrival along with triage level and vital signs were collected for each patient. Chief-complaint for ED-visits were collected from electronic medical records. Fatigue, confusion, non-specific complaints, generalized weakness and risk of falling was defined as non-specific complaint (NSC) [23]. Outcomes following ED arrival were obtained from RHIP, including triage level, ED length of stay (LOS), hospital admission rate, total bed-days, 72-hour revisit rate, and 7-day mortality. Data on 7-day mortality were collected for all patients, regardless of whether they were admitted to the hospital or discharged from the ED. Number of bed-days was calculated on patients in the population that was admitted to the hospital after visiting the ED. A 72-hour revisit was defined as a return visit to one of the two ED in RH within 72 h of the initial visit, regardless of the presenting complaint.
Vital signs were registered in triage on arrival at the ED according to RETTS, only one set of vital signs were collected for each patient. Vital signs recorded were SBP (mmHg), HR, (beats/min), SpO2 (%), RR (/min), temperature (°C) and LOC. In the ED setting for this study, LOC was assessed using the Reaction Level Scale (RLS), which is routinely used in many EDs across Scandinavia. RLS ranges from 1 (fully alert) to 8 (no reaction), with higher scores indicating more severe impairment [24]. To improve international comparability, RLS scores were converted to Glasgow Coma Scale (GCS) categories based on previously published evidence [25, 26]. The following approximation was applied: RLS 1 → GCS 13–15 (mild or no impairment), RLS 3 → GCS 9–12 (moderate impairment), and RLS 4–8 → GCS 3–8 (severe impairment). This stratification enabled classification of consciousness levels using internationally recognized GCS thresholds while preserving clinically meaningful distinctions relevant to emergency care.
The primary outcome measure was 7-days mortality after visiting the ED. Secondary outcome measures were ED LOS, admission and in-hospital LOS.
Descriptive statistics were used to summarize patient characteristics and vital signs. Continuous variables were presented as means with standard deviations (SD) and analysed using one-way ANOVA, while categorical variables were reported as frequencies and percentages.
Normality of continuous variables (SBP, HR, SpO₂, RR, temperature, and LOC) was assessed through visual inspection of histograms and by evaluating skewness and kurtosis. These methods provided insight into data symmetry, variability, and the presence of outliers. Although some variables deviated from a normal distribution, the large sample size justified the use of parametric methods. According to the Central Limit Theorem, the sampling distribution of the mean approximates normality as sample size increases. Therefore, t-tests and ANOVA were considered appropriate and robust for analysing group differences in this dataset [27].
To explore the predictive performance of individual vital signs for 7-day mortality, Receiver Operating Characteristic (ROC) curve analysis was performed. The Area Under the Curve (AUC) quantified each variable’s ability to discriminate between survivors and non-survivors, with higher values indicating greater predictive accuracy [28].
To enhance interpretability and reflect clinical practice, vital signs were categorized based on clinically established thresholds and cut-offs used in previous studies [11]. Categories were defined as follows: SBP (≤ 80, 81–100, 101–120, 121–140, 140–160, ≥ 160 mmHg), HR (≤ 50, 51–75, 76–100, 101–125, > 125 bpm), SpO₂ (≤ 80, 81–85, 86–90, 91–95, 96–100%), RR (≤ 9, 10–19, 20–29, ≥ 30 breaths/min), temperature (≤ 30 °C, 31–34 °C, 35–37 °C, 38–39 °C, ≥ 40 °C), and LOC (GCS 3–8, 9–12, 13–15). A reference category was defined for each vital sign. Reference categories for vital sign variables were chosen to support a logical, stepwise progression across risk levels in the regression analysis, rather than using clinically “normal” ranges, in order to better reflect the relationship between physiological extremes and mortality risk [29]. Missing values were categorized as “missing data”.
Binary logistic regression was used to assess associations between categorized vital signs and 7-day mortality. Univariable models estimated crude effects, while multivariable models were adjusted for age, sex and comorbidities. Results were reported as odds ratios (OR) and adjusted odds ratios (AOR) with 95% confidence intervals (CI). To ensure model validity, multicollinearity among independent variables was assessed using Variance Inflation Factor (VIF) and Tolerance statistics. VIF values < 5 and Tolerance values > 0.1 were considered acceptable. A p-value of < 0.05 was considered statistically significant. The analyses were executed with IBM SPSS Statistics 27, Armonk, New York, USA. There were no missing values besides from vital signs which are presented in Table 1.
