Impact of varying wave periods of COVID‐19 on in‐hospital mortality and length of stay for admission through emergency department: A territory‐wide observational cohort study

Abstract Background The COVID‐19 pandemic has been associated with excess mortality and reduced emergency department attendance. However, the effect of varying wave periods of COVID‐19 on in‐hospital mortality and length of stay (LOS) for non‐COVID disease for non‐COVID diseases remains unexplored. Methods We examined a territory‐wide observational cohort of 563,680 emergency admissions between January 1 and November 30, 2020, and 709,583 emergency admissions during the same 2019 period in Hong Kong, China. Differences in 28‐day in‐hospital mortality risk and LOS due to COVID‐19 were evaluated. Results The cumulative incidence of 28‐day in‐hospital mortality increased overall from 2.9% in 2019 to 3.6% in 2020 (adjusted hazard ratio [aHR] = 1.22, 95% CI 1.20 to 1.25). The aHR was higher among patients with lower respiratory tract infection (aHR: 1.30 95% CI 1.26 to 1.34), airway disease (aHR: 1.35 95% CI 1.22 to 1.49), and mental disorders (aHR: 1.26 95% CI 1.15 to 1.37). Mortality risk in the first‐ and third‐wave periods was significantly greater than that in the inter‐wave period (p‐interaction < 0.001). The overall average LOS in the pandemic year was significantly shorter than that in 2019 (Mean difference = −0.40 days; 95% CI −0.43 to −0.36). Patients with mental disorders and cerebrovascular disease in 2020 had a 3.91‐day and 2.78‐day shorter LOS than those in 2019, respectively. Conclusions Increased risk of in‐hospital deaths was observed overall and by all major subgroups of disease during the pandemic period. Together with significantly reduced LOS for patients with mental disorders and cerebrovascular disease, this study shows the spillover effect of the COVID‐19 pandemic.


| BACKGROUND
The coronavirus outbreak (COVID-19) was first described in late 2019 and has brought unprecedented challenges to patients and health care systems around the world. 1 Fear of becoming infected with COVID-19 or a delay in seeking health care during the evolving pandemic has contributed to a significant decline in emergency hospital visits. 2 As a result, there have been fewer emergency admissions. The decrease in emergency admissions during the early stage of the COVID-19 pandemic has been widely reported in the United States and Europe. [3][4][5][6] Delays in admission and in the treatment received while in hospital may exacerbate morbidity and increase in-hospital mortality, especially, for example, in the presence of acute cardiovascular disease. 7,8 Emergency departments (EDs) are responsible for triaging patients who need hospitalization for acute treatment or safely discharged. They also play a vital role in facilitating efficient use of resources by reducing pressure on in-patient beds while preventing treatment delay and reducing mortality with limited resources. Before the pandemic, studies have been reported that ED crowding increased length of stay (LOS) and in-hospital mortality. 9 During the pandemic, the associations among ED admission, in-hospital stay, and mortality showed changed patterns. For example, a multicenter retrospective cohort study indicated that fewer acute admissions to ED for cardiovascular emergencies but four times higher risk of death at ED and shorter LOS. 10 Given the necessity to maximize efficient use of resources and quality of patient care in hospital, information on patterns of ED admission, LOS, and mortality, 11 in particular, understanding the impact of COVID-19 on these parameters over time, is critical in decision making and allocating medical resources in the most effective manner. However, there are as yet few population-based studies comparing LOS before and during the COVID-19 pandemic. Several studies 7,10,12,13 found patients with acute cardiovascular conditions in 2020 had shorter LOS, but a significant change in in-hospital mortality in 2020 compared with 2019 has not been consistently reported.
Previous research has established that a decreased hospital admission and a higher mortality rate for non-COVID medical conditions during the pandemic compared with the pre-pandemic period, especially for patients admitted with respiratory diseases, cancer, pneumonia, and sepsis. 14 However, none of the published studies has examined the in-hospital mortality and LOS of emergency admissions. Therefore, we conducted a territory-wide retrospective cohort study to evaluate the association of 28-day in-hospital mortality and LOS with emergency admission during the pre-pandemic, pandemic wave, and interwave periods of COVID-19 in Hong Kong, a densely populated city with prior experience of combating severe acute respiratory syndrome (SARS). Our study included all emergency admissions in January to November 2020 relative to the equivalent period in 2019.
We aimed to compare the differences in a set of characteristics associated with in-hospital mortality and LOS before and during the COVID-19 pandemic waves. Subgroup analyses were done by patients with different medical conditions. A better understanding of the spillover effects of the COVID-19 pandemic across waves would provide meaningful implications for healthcare leaders to help at-risk populations during pandemic.  [16][17][18] Extensive research has shown highly coding accuracy in CDARS. [16][17][18] This study analyzed territory-wide retrospective cohort data from all 18 EDs through CDARS, ensuring its representativeness.

| Study population
All patients admitted as emergencies in Hong Kong between January 1, 2019, and November 30, 2019, and between January 1, 2020, and November 30, 2020, were included for analysis. We excluded hospital episodes with missing data for LOS, which represented a limited proportion of total ED admissions (0.4%). Patients who visited EDs (EDs visit) and were then admitted to hospitals were regarded as ED admissions and were included in our study as identified by records of destination of discharged at EDs. Patients who were hospitalized in public hospitals, whether transferred or directly admitted to other departments of that hospital from EDs, were included.
Our study included a list of patients' covariates, comprising age, gender, race (Chinese and non-Chinese), living regions (Hong Kong Island, Kowloon, and New Territories), living in a residential care home for the elderly, and recipients of comprehensive social security assistance (CSSA). Severity of cases was defined by the mode of arrival, time of attendance (shift: night; day; evening), triage category, total time spent in ED (≤4 h or not), and disease diagnosis. The International Statistical Classification of Disease and Related Health Problems, Ninth Revision (ICD-9-CM) diagnosis codes were used to identify primary diagnosis of the following diseases: lower respiratory tract infection (LRTI), airway disease, hypertension, coronary heart disease (CHD), cerebrovascular disease and myocardial infarction (MI), diabetes, mental disorders, sepsis, chronic kidney diseases, cancer, and COVID-19-related conditions. The specific diagnosis for admitted patients was recorded during hospitalization other than diagnosis from ED. A social deprivation index (SDI), an indicator of socioeconomic status, was measured for each district by averaging six related variables. 19 A set of hospital-related variables included hospital academic status and hospital size.
The first confirmed COVID-19 case in Hong Kong was announced on January 23, 2020. 20 We specified different wave periods in our analysis: first-wave period from January 23 to February 29, second wave from March 17 to April 21, third wave from July 5 to August 31, and interwave period from March 1 to March 16, April 22 to July 4, and September 1 to November 30, 2020. Cohort data were extracted on February 3, 2021 ( Figure 1). Cohort data were extracted on February 3, 2021. The end of each wave was defined as the local cases of the day less than 5. The beginning of the second and third waves was determined as emerge of the first local cases after the last waves and the local cases increased consecutively afterward.

| Outcome measures
The primary outcome was the time from emergency admission to inhospital death within the 28-day follow-up period. The original definition of 28-day in-hospital mortality included all deaths that occurred during hospitalization within 28 days but excluding deaths that occurred after discharge from hospital. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) diagnosis codes were used to identify the primary cause of death. Diagnosis codes are listed in Table S1. The secondary outcome of this study was LOS, which was counted as the number of complete or partial days spent in hospital after attendance at an ED.

| Data analysis
Descriptive statistics of patients' covariates including demographic and clinical characteristics, and the LOS of each hospital episode in 2019 and 2020 were displayed separately. Balance of patients' covariates between the 2 years was evaluated using the absolute standardized mean difference (SMD). An SMD <0.1 implied a covariate balance between the years. 21 The incidence rate was a more precise estimate of the true risk of death in the dynamic populations. 22 The weekly in-hospital mortality rate was calculated by using the weekly number of in-hospital deaths divided by person-days at risk of each week during the period of hospitalization or 28 days, whichever was the earlier. Besides, to show the daily risk of death after admission by years, incidence rates from Day 0 to Day 28 were calculated.
In order to examine the heterogeneous effects of COVID-19 on in-hospital mortality across the years and waves, the primary end point of time from emergency admission to in-hospital death before and after pandemic or during different pandemic waves in 2020 was assessed by a Kaplan-Meier plot and compared with a log-rank test.
Cox proportional hazard models were performed to assess the impact of COVID-19 on incidence of in-hospital death within 28 days. Each hospital episode was observed from the date of emergency admission to one of the following events: discharge or 28 days as right censoring. Hazard ratio (HR) with 95% confidence interval (CI) for 2020 versus 2019 adjusted by individual and hospital characteristics was F I G U R E 1 Number of COVID-19 cases by date of reporting and timelines of interventions in Hong Kong. The epidemic curve shows the reported cases and key interventions that have occurred to date from January 1 to November 30, 2020 (gray shading areas for the three waves of COVID-19 pandemic). The first confirmed COVID-19 case in Hong Kong was announced on January 23, 2020: first wave from January 23 to February 29; second wave from March 17 to April 21; third wave from July 5 to August 31, 2020. Y axis refers to numbers of confirmed cases. The definition of local cases according to whether the patient had a travel or residence history in other regions within 14 days before diagnosis and likely exposure to pathogens in those regions; the word "regions" here refers to areas outside of Hong Kong  reported for the overall study population and the different subgroups.
Proportional hazard assumptions were tested by using a log-log plot for categorical variables. There was no evidence to reject the proportional hazards assumption.
For the secondary outcome measurements, the mean differences in LOS between the 2 years were generated by linear regression after adjusting for individual and hospital characteristics. Patients who died on the day of hospital admission would have a follow-up time and an LOS of 0.5 days for consistency with another study, 23 assuming that death occurred in the middle of the day that death was registered.
A p value < 0.05 was regarded as statistically significant. All data management and statistical analyses were performed by using STATA version 16.0 (College Station, TX: StataCorp LLC).   The increase in 28-day in-hospital mortality in 2020 was significant,    have been prevented to some extent because physicians have to rethink which treatment is truly necessary at circumstance of limited medical resource during pandemic. 28 With reductions in emergency admissions of À28.4%, À31.1%, À25.0% during the first, second, and third wave, respectively, and more information being available about COVID-19 patient care, it was expected that the in-hospital mortality rate would be higher. However, an increase in in-hospital death during pandemic waves (Table S4) implies that the higher in-hospital mortality rate was due not only to the decrease in emergency admissions but also to more critically ill patients being admitted during the pandemic. 13 This was borne out in our study during the pandemic year with a higher proportion of patients being admitted through EDs with critical and urgent conditions. On the other hand, imposed social restrictions mean that people who live alone are less likely to be aware of mild complications. The Hong Kong government implemented a work from home policy and from January 29, 2020 ( Figure 1). The advantages of early treatment are lost by any delay in seeking medical care and further increase in-hospital mortality and the burden on hospitals. 8 The trend towards a significant increase in in-hospital mortality in the third wave is the more worrying. Unlike the similar increase patterns for in-hospital mortality and confirmed COVID-19 cases for the first wave, the in-hospital mortality grows gradually after the beginning of the third wave in all major diagnosis groups (Figure 2). A U.S. study also showed an increase in non-COVID-19 mortality in F I G U R E 5 Mean difference in length of stay (LOS) (with 95% confidence interval) of emergency admissions in 2020 compared with 2019. Mean difference in LOS were estimated using linear regression models adjusting for age, gender, race, region, living in residential care home for elderly, comprehensive social security assistance (CSSA), social deprivation index (SDI), arrive mode, time of attendance, triage category, total time spend in emergency department (ED), hospital size, and academic status. Average LOS and standard deviation of each year are shown in columns countries with higher COVID-19 mortality rates, probably resulting from weak public health infrastructure. 29 A previous report warned that later presentation of infection will result in higher acuity of disease and, therefore, an increase in-hospital mortality in later periods of the pandemic. 13 Additionally, the observed higher in-hospital mortality rate during the third wave is likely to be the result of overcrowding in the ED and a resulting increased hospital occupancy, 30 which is reflected in our data with a higher proportion of patients spending more than 4 hours in EDs during the third wave (Table S3).

| RESULTS
As the surge in confirmed COVID-19 cases put increased pressure on hospital capacity during the third wave, 31 the AsiaWorld-Expo Centre in Hong Kong was opened as a temporary isolation facility on August 1, 2020 to help cope with the increase in cases (Figure 1). 32 Notably, our findings show an increased risk of death in patients with non-COVID-related respiratory disease and mental disorders in 2020. Previous studies have reported that patients with cardiovascular conditions, cancer, and respiratory disease were considered at higher risk of death during the pandemic. 7,8,25,26,33 A nationally representative study in China also reported additional deaths from pneumonia occurred in first 3 months of COVID-19 pandemic in Wuhan. 26 One local study from Hong Kong highlighted the winter peak in incidence of influenza A (H1N1) from late December 2019 to late February 2020 as coinciding with the first wave as defined in our study. 34 However, in our study, there was an increase in the number of 28-day inhospital deaths from non-COVID pneumonia during both the first and third waves (Table S4). This suggests that when patients presented to EDs with respiratory conditions during the pandemic period, the condition was more severe than in the previous year. We speculate, but this is likely due to fear of going to hospital and repeated delays to seek medical care. Meanwhile, a higher in-hospital mortality risk of respiratory disease may result from a reduction in denominator given that very few respiratory viruses were circulating in 2020. 35,36 Patients admitted with mental disorders also had a significantly higher in-hospital mortality rate. Mental health problems have been reported in other studies, 37,38 and a previous Hong Kong study 39 also reported a mental health crisis after a lengthy period of social restrictions.
Our study found statistically significant decrease in LOS during pandemic year, and the slight difference of mean LOS before and after pandemic suggests this result might not be practically significant.
However, the large reduction in LOS among patients who die in hospital is most marked in those who died within 2 days after admission ( Figure S2). These are to be expected, mainly as a result of a higher daily in-hospital mortality rate within the first 2 days of admission during the pandemic year ( Figure 3).

| CONCLUSIONS
In 2020, patients admitted as emergencies experienced a higher 28-day in-hospital mortality rate and shorter LOS in Hong Kong. The trend towards a significant increase in in-hospital mortality in the third wave shows the spillover effect of COVID-19.

PATIENT AND OTHER CONSENTS
This study was approved and a waiver of the need for patient consent allowed by the institutional review board of the University of Hong Kong/Hospital Authority West Cluster (UW 20-112).

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1111/irv.12919.