Study design
This was a retrospective cohort study in ED patients.
Study setting and population
The locations were two university hospitals of Stockholm, Karolinska University Hospital Huddinge and Solna. Total (Huddinge and Solna) annual ED visits were around 150 000 patients. Between 1/1/2010 and 31/12/2016, 641 314 visits at the two hospitals ED’s were included in this study and analysed. The patient data were extracted from the hospital administrative system. Patient data was excluded if there was no full documentation of all variables, if patients died upon arrival to the ED or if the patient had an ED LOS >4000 minutes (values above 4000 minutes were assumed not to be probable and interpreted as typing errors) leaving 639 385 patients that were included in the present analyses.
Study protocol
At arrival to the ED at both hospitals, the patient visit is immediately registered and triage according to RETTS-A is performed by nurses [28]. RETTS-A Triage priority is determined by using a combination of the patient’s presenting symptoms and signs in addition to vital sign values. The RETTS-A triage scale priorities are: red, orange, yellow, green, and blue, in declining priority of acuity. The two highest levels of acuity (red and orange) represent potential life-threatening conditions whereas the other two levels (yellow, green) represent stable patients in need of acute care. Blue priority, representing non-urgent complaints. The patient’s presenting symptoms are matched to one of 59 Emergency Symptoms and Signs (ESS) algorithms in accordance with RETTS-A. The vital signs for each triage level have specific cut off values indicating different levels of acuity. The chief complaint algorithms are known as emergency symptoms and signs for emergency care. Each emergency symptom and sign includes one or more chief complaints and is classified according to the International Statistical Classification of Diseases and Related Health Problems, 10th Revision, 2007 (ICD-10), and a RETTS-A logistic process is attached to each algorithm. The more urgent of either the vital signs or presenting symptoms and signs becomes the patient’s final triage priority. Patients with blue priority, representing non-urgent complaints and minor injuries, were referred to a primary acute health care centre from the ED and therefore not included in this study. Patients with green priority have vital signs in, or close to, normal range and thus less urgent complaints than yellow, orange, and red patients.
Data collection and variables
The collected variables were age, sex, any of the ten most common chief complaints pre-defined by RETTS-A (abdominal pain, chest pain, shortness of breath, painful or swollen extremity, malaise, dysrhythmia, allergic reaction, syncope, intoxication, fever and undefined), triage priority at arrival, if the patient was given prehospital care given by ambulance or not, if the patients were admitted to in-hospital care or not, if the patient presented to the ED in the weekend or not. The chief complaints can be seen as a crude proxy for comorbidity and should eliminate some confounding associated with complaint.
Outcomes
Primary outcomes were 7 and 30-days mortality, counted from registration to the ED. Information on patient survival as dependent variable was extracted from the Swedish population register, administrated by the Swedish Tax Agency, which includes every Swedish resident and has a high validity and completeness. Thus there was a near complete follow-up of 7 and 30-days mortality for every patient visiting the EDs included in this study.
Statistical analyses
We present descriptive data on the study cohort including mean and standard deviation for baseline characteristics. Patients were categorized into different quintiles of ED-LOS (Table 1) in order to provide further insights into potential threshold effects on these associations. ”Pearson Chi-square test and One-way ANOVA was used for comparison across groups. Multiple logistic regression models were performed to investigate the relationship between ED-LOS, continuous model and quintile model, and mortality. The model included age, sex and triage-priority, the ten most common chief complaints, prehospital care given or not, in-hospital care, when the patient presented to the ED (day of the week and time of the day) and the diagnosis at the ED. We tested multiplicative interactions and stratified instead of adjusting when needed. Odds ratios (OR) and 95% confidence intervals (CI) are presented. P-values < 0.05, two-sided, were considered significant. Statistical analyses were performed using the software STATA version 13.
Ethical considerations
A permit was issued from the Ethics Review Board in Stockholm, reference number: 2017/1252-31/1.