Study design and setting
We conducted a cohort study. We identified all admissions to the emergency department at Viborg Regional Hospital, Regional Hospital Central Jutland, between 1 January 2014 and 31 December 2015 using administrative data from the EPR.
In Denmark, a free, tax-funded health care system ensures that all citizens have unrestricted access to general practitioners (GPs) and hospital care [18, 19]. Every Danish citizen is affiliated with a GP who in daytime refers the patient to the hospital. In evening, at nighttime, and on weekends, GPs rotate staffing of regional out-of-hours service centers, where they receive all patient calls. In case of a life-threatening condition or injury, patients can dial 1–1-2 and present by ambulance to the emergency department.
The Emergency Department at Viborg Regional Hospital is one of five emergency hospitals in Central Denmark Region. The emergency department employs 145 nurses and physicians, including eight senior physicians. With backing from physicians from other departments, all acute patients with a referral diagnosis covering general surgery, orthopedic surgery, and internal medicine are diagnosed and treated in the emergency department. Children, other than those with minor injuries, are received in the pediatric department, and patients with serious heart-related events bypass the emergency department are directed to the department of cardiology. Patients with psychiatric diseases are admitted to a psychiatric hospital. The emergency department, Viborg Regional Hospital, has previously been described in detail [20].
Data source
We have used data from MidtEPJ, the EPR developed by Systematic and used by all somatic and psychiatric hospitals in Central Denmark Region. Viborg Regional Hospital, as a part of Regional Hospital Central Jutland, was one of the last hospitals in the region to implement MidtEPJ in 2013. The MidtEPJ is a work, communication and documentation tool. It is accessible by multiple authorized users (e.g., nurses, physicians, and secretaries) and supports the clinical workflow across groups of health professionals, departments, and hospitals. The MidtEPJ documents patient morbidity, treatment, and care over time. It contains both administrative data on hospital admission, including date (hours and minutes), department, source of admission, and clinical data such as age, gender, primary and secondary diagnoses, and triage score. The MidtEPJ is linked to the unique Civil Personal Registration number (CPR number) every Danish citizen is assigned at birth and to residents upon immigration. The CPR number is a 10-digit number that contains embedded information on birth date and sex. The CPR number ensures unambiguous patient identification. When using MidtEPJ data for research, the data is extracted directly from the source system itself and is stored in a regional data warehouse (named the Business Intelligence Portal). The data warehouse gathers data from a number of the region’s different electronic systems, including MidtEPJ for quality assurance and health statistics purposes.
To compute the CCI score for each patient, data were obtained from the Danish National Patient Registry (DNPR), which is a central medical registry that has recorded information on all hospital admissions since 1995 [21]. The record of each admission is linked to the CPR number.
Information on all-cause mortality within 30 days following the admission date was captured by linking the patient’s CPR number to the Danish Civil Registration System (CRS). Established in 1968, the CRS stores complete and daily updated information on vital events, and can be retrieved for research purposes while protecting the Danish citizens’ anonymity [22]. CRS thus contains complete information on vital events of all patients included in this study. Patients were followed from date of admission until the date of death from any causes, the 30th day after discharge, or emigration, whichever occurred first.
Study population
We included all admissions to the emergency department between 1 January 2014 and 31 December 2015 (a flowchart for the patient visits included in the study is available in see Additional file 1: Appendix I). This time interval was chosen because the emergency department became an independent department (independent leadership, budget, unique administrative department code) on 1 January 2014. The emergency department started receiving a wider range of patient categories, and the emergency room, formerly a part of the department of orthopedic surgery, became a part of the emergency department. On 1 January 2016 the organization of the emergency department changed again, as the department began to receive more medical patients who earlier would have been admitted to a nearby hospital.
To ensure complete follow-up, we only included patients with a CPR number.
A patient’s visit at the hospital may consist of admissions to one or more departments. A patient admitted to the emergency department may be transferred to the ICU and afterward transferred to an internal medical department. During one hospital visit, some patients may be admitted to the same department more than once. In this study, we included hospital visits with up to five consecutive admissions. Furthermore, if more than four hours elapsed between two admissions, we considered it as two different hospital visits.
We excluded those patients treated at two clinics, that are organizationally part of the emergency department but are physically located in the cities of Skive and Silkeborg. Moreover, we excluded patients with missing information about date of finishing treatment within the emergency department. For the 30-day mortality analysis, we excluded nine patients due to invalid date of death.
Time of admission
In this study, the exposure was the time of admission. We defined six time periods: daytime (from 7:00 a.m. to 2:59 p.m.), evening (from 3:00 p.m. to 10:59 p.m.), and nighttime (from 11:00 p.m. to 6:59 a.m.) on weekdays and on weekends. Patients were considered weekend admissions if they were admitted between 3:00 p.m. on Friday and 6:59 a.m. on Monday. Patients admitted on all other days and times were considered weekday admissions. Public holidays were coded as weekend. We chose the time periods based on knowledge about how the emergency department was staffed and organized on weekdays and on weekends [20]. Classifying time of admission into six periods, including daytime, evening, and nighttime on weekdays and on weekends, is an attempt to provide a more subtle description of the weekend effect.
Explanatory variables
For each of the six time periods we described age, gender, comorbidity, triage score, and source of admission (GP, other hospital departments, self-referral, or other). Patient age was described based on five groups: 0–19, 20–39, 40–59, 60–79, and > 80. The department (and hospital) the patients were transferred to after initial treatment within the emergency department, as well as primary diagnosis reflecting the reason for admission and identified at the time of discharge, was examined too. According to Danish guidelines and the guidelines of World Health Organization, the primary diagnosis assigned at hospital discharge should be the main reason for a patient’s hospitalization [23]. We coded the diagnoses according to the International Classification of Diseases, 10th revision (ICD-10). However, we combined the infectious diseases in one group and merged other non-infectious diseases into a single diagnostic group, leaving us with fourteen diagnostic groups (details are outlined in see Additional file 2: Appendix II).
The CCI score was computed for each patient. This index reflects the number and seriousness of comorbid diseases. In this study, we collected data based on admissions recorded within the 10 years prior to admission. Three groups were created: Low (index score 0), Moderate (index score 1–2) and High (index score > =3), categorical based on ICD-10 codes.
As a proxy for disease severity, we included the triage score. All emergency departments in Central Denmark Region use the tool Danish Emergency Process Triage system (DEPT). DEPT is a five-step triage system that prioritizes patients according to the degree of life or truancy threat and thereby is indicative of how fast they are to be seen by a physician. It is based on triage using vital signs (airway, oxygen saturation, respiratory rate, pulse, blood pressure, Glasgow Coma Score, and temperature), which are collected by nurses as an integral part of the initial process of care, combined with pre-defined attention points related to the symptoms the patient had when admitted. The DEPT score is categorized by five groups of triage scores: blue (minor injury, only used in the emergency room), green (not urgent), yellow (less urgent), orange (urgent) and red (life-threatening).
Outcomes
For consistency with previous studies of the weekend effect, the main outcome in this study was 30-day mortality [10, 12]. LOS was secondary outcome. Emergency department and hospital LOS was calculated as the number of minutes from admission to the emergency department to final discharge or transfer to another hospital department from the emergency department. During the two-year study period, several patients had multiple emergency department visits. In the analysis of 30-day mortality, we included the last admission to the emergency department for each patient. In total, 21,736 patient-visits were included.
Statistical analysis
We calculated the proportions of patients admitted during daytime, evening, and nighttime on weekdays and on weekends and characterized them according to patient characteristics. For each time group, we collected data on age, gender, CCI score, source of admission, primary diagnosis, triage score, transfer to another department, and LOS. Relative risks (RRs) for being triaged orange or red, for having a stay longer than 24 h within the emergency department, and for being transferred to ICU, comparing weekday admissions with weekend admission, were calculated. RRs were displayed with their 95% confidence intervals (CIs) and P values to indicate precision of estimate.
To facilitate comparisons, direct standardization adjusted for age and gender was used to compute the 30-day mortality rate for patients admitted to the emergency department in each of the six time periods (daytime, evening, and nighttime on weekdays and on weekends) [24]. As standard population for the mortality analysis, we applied the patients admitted in daytime on weekdays. Thus, for each time period, we estimated what would have been the 30-day mortality rate in this time period, if the population in that particular time period was equal according to age and gender with the one in our standard population. Mortality rates were displayed with their 95% confidence intervals (CIs) to indicate precision of the estimate. Subgroup analyses were performed for each triage score (red, orange, yellow, green and blue). Mortality rates for patients with missing triage score data were added. In our analysis, we used the triage score as stratification variable, because the triage score is used within the department as a proxy of severity of disease. Thus, the 30-day mortality rate may differ between these groups. More subgroup analyses were performed, comparing the mortality rates for patients discharged to home from the emergency department and the mortality rates for patients transferred to other departments after initial treatment within the emergency department.
A Cox proportional hazards regression with 95% confidence intervals [25, 26] was performed to test differences in the 30-day mortality rate for patients admitted to the emergency department in daytime, evening, and nighttime on weekdays and on weekends. As reference population, we applied the patients admitted in daytime on weekdays. To correct for differences in patient characteristics, we included confounders as following: age groups, gender, comorbidity burden, triage score, and whether the patients had or had not previous admissions to the emergency department within the two-year period included in the study. For the statistically tests, a P value less than 0.05 was considered significant.
Analyses were performed using the statistical software package STATA (version 11, Stata Corp, College Station, Texas, USA).