Aim
The aim is to describe how the Huddinge site at the Karolinska university hospital (KH) responded to the COVID-19 crisis, and how ED crowding, and important input, throughput and output factors for ED crowding developed at KH during a 30-day baseline period followed by the first 60 days of the COVID-19 outbreak in Stockholm Region.
Study design
This is a retrospective descriptive study of KH response to the COVID-19 crisis. In the study, the development of the crisis was divided into six different phases separated by five major events that changed the conditions for the ED (Fig. 1). The five major events were defined retrospectively by the research team. The first event was when patients with COVID-19 started to arrive at the ED. The next major event was the implementation of new working methods to cope with this. The last three events were based on sudden changes in the inflow of ambulances due to ambulance diversion from other hospitals following regional decisions. ED crowding together with important input, throughput and output factors for ED crowding were studied for each phase using existing data from the hospital data warehouse.
Study setting
KH is the southern site at Karolinska university hospital. KH’s assignment is to deliver emergency care, specialized and in some areas highly specialized care in combination with research and education. KH has 760 beds and the ED had 53,508 visits in 2019. The ED has a low proportion of non-urgent and non-complex patients as these will be sorted to a co-located ED with imaging capability led by general practitioners. The department for infectious diseases at KH is the largest in the country and the hospital is the primary receiver of patients with suspected highly infectious diseases in the region. The department for perioperative medicine and intensive care is organized in a single organization responsible for intensive care and the operating theatres supporting all surgical specialties with perioperative care. A brand-new operating theatre with 23 operating rooms was just completed at KH and was meant to be inaugurated during the study period. The department of Emergency and Internal Medicine manage the ED and the emergency wards. In the last few years there has been a growing focus on emergency medicine at KH and there are currently 30 emergency medicine residents in training to become specialists. The ED is mainly staffed with physicians from this department and the ED is divided into medical and surgical patient flows. There is a long-term trend at KH with increased ED crowding and boarding. Unpublished data from standard internal reports show that the mean ED LOS for admitted patients increased by 55% from 4.4 h in the year 2013 to 6.8 h in 2019.
Detailed description of the six phases in the study
Phase 1, “baseline” (Feb 1 to Mar 1)
This phase begins at the start of the study and ends after the winter school holiday. This phase represents a situation where no, or very few suspected COVID-19 patients arrived in the ED.
Phase 2, “Early phase, normal working methods” (Mar 2 – Mar 13)
This phase starts when many in the population of Stockholm return from the winter school holiday. Many families returned from travels to countries where the disease was already widely disseminated, which accelerated the spread of the disease in the Stockholm Region. The following week on March 9, a regional decision confirmed that all patients with suspected COVID-19 should be directed to KH. According to standard protocol, patients were received in an isolation room where investigations would be completed before it was decided whether to discharge or admit the patient. After a few days, it became evident that this process would not cope with the volumes of patients and new routines and practices were developed. On March 12 at 14:00, the hospital raised the alertness level to level two out of three according to the Hospital Emergency Operations Plan. This means a partial mobilization of hospital resources and includes cancelling all planned treatments that can wait and the establishment of a Hospital Command Group [16] working according to European/NATO guidelines [17]. During this phase there was a ramp-up of the regional call-center “1177” giving advice on when and where to seek care and the population in Stockholm was asked to always call before seeking care. An online tool for self-triage was also launched. Guidelines for seeking care at the ED’s were strict and patients that did not experience rapid deterioration or had respiratory disorders at rest were directed to self-care [18].
Phase 3, “Intermediary phase, new working methods” (Mar 14 – Mar 29)
On March 14, major changes in routines and practices were implemented. These were fine-tuned during the following phases using an agile approach with high presence from first-line management. All patients were now received in a tent outside the ED staffed by at least a resident physician in emergency medicine, an experienced nurse together with a nurse assistant. An initial assessment was performed, and patients were sorted depending on if COVID-19 was suspected or not. Patients with suspected COVID-19 that had mild symptoms and did not belong to a high-risk group were referred to the co-located GP-led ED or diverted to self-care and isolation at home with instructions if the situation deteriorated. Inside, the ED was separated into two equally sized sections where patients with suspected COVID-19 were sorted to one section while all other patients were sorted to the other section. The exception was orthopaedic patients that were now treated in the elective department during office hours. The inflow was initially much lower in the COVID-19 section, while the other section struggled with maintaining a high patient flow, now with half the number of available ED rooms and fewer resources in relation to the number of patients. Competence in the ED was strengthened as all residents in emergency medicine where called back from external rotations and replaced interns that were instead transferred to the emergency wards. Staffing was reinforced during evenings and nights and the surgical specialist position was now staffed 24/7 instead of only weekdays 10–18. Practices were also changed, after an initial assessment and work-up of unstable patients, an early decision on admission to inpatient care was made based on basic lab tests, point-of-care blood gases, and the overall clinical picture. Imaging and further diagnostics were kept to a minimum. The early decision to quickly admit patients with suspected COVID-19 to inpatient care without complete diagnostics was enabled by the transformation of one of the emergency wards with 22 single rooms that performed further investigation and diagnostics that in normal cases would have been performed in the ED. As soon as a patient had a result on their COVID-19 test they would be transferred to an infection ward if positive and to another ward if negative unless they were stable enough to be discharged home. During this phase, the hospital increased the number of beds in the infection wards and started the ramp-up of intensive care capacity. KH was also no longer the primary receiver in the region as in phase 2 and the inflow of patients arriving with ambulance was shared among all of seven hospitals in the region based on geographical segmentation.
Phase 4, “Exponential phase with controlled ambulance inflow” (Mar 30 – Apr 2)
This phase starts with the implementation of a regional decision to rebalance the load across the seven hospitals in the region. In the earlier phases, ambulances were directed to the closest hospital, but clustered outbreaks in some of the suburbs led to a congestion of COVID-19 patients in three of the sevens hospitals in the region. The decision led to an increased inflow of ambulances at KH. During this period, the intensive care capacity was further increased at KH to manage the increased inflow of critically ill patients.
Phase 5, “Exponential phase with extreme ambulance inflow” (Apr 3 – Apr 8)
This phase starts with a regional decision of directing all ambulances with patients showing respiratory symptoms in Region Stockholm towards KH. The rapid build-up of intensive care capacity in KH now enabled the hospital to relieve the other hospitals in the region to handle about 50% of the regional projected patient flow for intensive care. Even if there was now available intensive care capacity in the hospital, the large inflow of critically ill patients during a short time-period was extremely challenging for the ED. Due to the impact that the first decision had on KH, a new regional decision was taken April 5 that reduced the ambulance diversion for all priority 1 ambulances, and all ambulances from two of the six other hospitals.
Phase 6, “Plateau, stabilization of inflow” (Apr 8 – Apr 30)
This phase starts with a regional decision to further reduce the ambulance inflow to KH back to the level in phase 4 again. The number of new cases and patients admitted into intensive care in the Stockholm region also peaked and reached a plateau at this time [19]. Intensive care capacity was now increased to its planned maximum capacity. During this phase there was a shift in patient flows where many patients who had been cared for in the intensive care units had recovered enough to come out of the ventilator but were still very weak, many of them suffering from hypoxemia. To care for post-intensive care patients, the capacity of the high-dependency emergency care unit was increased gradually from 6 to 18 beds by closing down other emergency care wards and relocate the staff. This increased the level of care but reduced the number of available beds.
Variables
As a proxy for the outcome ED crowding we used the mean ED LOS for each phase. Factors potentially impacting crowding were grouped into input, throughput and output. Input factors were the number of patient visits and the number of patient visits arriving by emergency medical services (EMS). Throughput factors indicating the share of diagnostics and treatment performed in the ED were the proportion of patient visits that included medical imaging during the ED stay and the proportion of patient visits that resulted in an admission to inpatient care. Output factors were the bed occupancy level in the emergency wards and the number of care episodes at the emergency wards, together indicating the available capacity for the ED to admit patients to inpatient care without a finalized diagnosis. Another output factor was the total hospital bed capacity indicating the general ability to admit patients to inpatient care. To provide a view on this, the average number of staffed inpatient beds together with the average number of patients by type of ward, phase and date were presented. The wards were grouped in four categories: intensive care, emergency care, infection, and other wards (excluding pediatric and obstetric wards). Basic demographic information on the distribution of sex and age in intervals < 40, 40–59, 60–79 and 80+ were also presented.
Data sources/measurement
Data used in the study was based on statistics extracted from KH data warehouse. Crosstabs of data were delivered to the research team using Tableau Desktop 2018.3.0 and imported into RStudio 1.1.463 and R version 3.6.1 to create the table, the analysis of correlations and high-resolution graphs. ED LOS is measured from the first registration at the front desk or in the later stages in the tent outside the ED, until the patient leaves the ED. Bed occupancy is measured as the daily average of observations on number of staffed inpatient beds and number of admitted patients that is registered at each ward. The number of staffed inpatient beds is entered manually by each ward as soon as the status changes. The number of patients in each ward is updated automatically as patients are admitted and discharged in the electronic healthcare records. The Data warehouse stores information on beds and patients with 15 min resolution so each daily average is based on 96 observations.
Statistical methods
For the outcome ED crowding, mean ED LOS was estimated with 95% confidence interval for each phase. Input, throughput and output factors previously known to impact ED crowding together with age and sex were described for each phase using distributions, means and proportions. Pearson correlations with P-value and 95% confidence intervals where each phase represented a single observation were estimated between ED LOS and the key input, throughput and output factors. The input factor used was average visits per day, the throughput factor was proportion of imaging performed and the output factor was bed occupancy in the emergency wards. Development of staffed inpatient beds and average number of patients were visualized in line graphs by the type of ward, phase and date.