COVID-19 pandemic: impact of lockdown measures on presentations to a Singapore tertiary care hospital emergency department

INTRODUCTION

Coronavirus disease 2019 (COVID-19) has affected the healthcare systems worldwide. First reported as a cluster outbreak of pneumonia of unknown cause on 31 December 2019 in Wuhan, China, COVID-19 has since spread globally. The World Health Organization declared COVID-19 a pandemic on 11 March 2020.

Singapore confirmed its first COVID-19 patient on 23 January 2020. Subsequent rise in COVID-19 cases resulted in public health measures to curb the local spread of disease. Singapore first raised its Disease Outbreak Response System Condition[1] level from yellow to orange on 7 February 2020, followed by implementation of an ‘elevated set of safe-distancing measures’, termed the ‘circuit breaker (CB) measures’.[2] Similar to the lockdown measures in other countries, CB measures included government-mandated closure of all nonessential services, take-away meals only at food and beverage establishments, telecommuting, home-based learning for school-going children, and safe-distancing and mask-wearing requirements. These measures were effective from 7 April 2020 to 1 June 2020, followed by a ‘phased approach’ to resuming socioeconomic activities.[3]

Downstream effects on emergency department (ED) utilisation in pandemics have been studied.[4,5,6,7,8,9] Despite the rise in infection rates, overall ED attendance declined worldwide,[10,11,12,13,14] with several studies reporting significant reduction in attendances for life-threatening conditions[15] such as acute coronary syndrome (ACS)[16] and stroke.[17] These observations echo those made in previous outbreaks such as severe acute respiratory syndrome (SARS)[6] and Middle East respiratory syndrome–related coronavirus (MERS-CoV).[9]

This study analyses the impact of the COVID-19 pandemic lockdown measures on the utilisation of an adult tertiary hospital ED in Singapore, by comparing to a corresponding nonpandemic (non-CB) period in 2019. Understanding the dynamics of ED utilisation in the COVID-19 pandemic can help guide decisions for planning healthcare responses for future pandemics.

METHODS

This was a retrospective observational study of ED patient visits to Singapore General Hospital, the largest tertiary teaching hospital in the country, with approximately 130,000 visits annually (350 visits per day in 2019[18]). It comprised electronic medical record review of ED patients who presented during the CB period (7 April 2020–1 June 2020) and a comparable non-CB period in the preceding year (7 April 2019–1 June 2019).

Patients were triaged into ‘clean’ and ‘fever’ cases, and those who were identified with epidemiologically significant infectious risks were seen in isolation-precaution facilities. Before the CB period, patients triaged to the ‘fever’ group included suspected/active pulmonary tuberculosis, varicella zoster lesions, known carriers of drug-resistant bacteria (e.g., vancomycin-resistant Enterococcus), MERS-CoV and Ebola suspects. During the CB period, COVID-19 suspects (patients with fever/acute respiratory illness or significant travel/contact histories) and confirmed cases were included in this group.[19] Triage acuity levels ranged from the most to the least severe [footnote of Table 1].

Table 1.

Comparison of total number/median daily number of emergency department (ED) visits between non-circuit breaker (non-CB) period and circuit breaker (CB) period^a (N=34,127).

Characteristic	n (%)/median (IQR)		Median and proportional differences		P †
	Non-CB period (n=19,132)	CB period (n=14,995)	Estimate*	95% CI*
Total no. of ED visits
Fever and clean cases (n=34,065)					<0.001
Fever	514 (2.7)	3,216 (22)	19	18, 20
Clean	18,611 (97)	11,724 (78)	−19	−20, −18
Fever cases (n=3,730)					<0.001
P1F	40 (7.8)	65 (2.0)	−5.8	−8.2, −3.3
P2F	218 (42)	365 (11)	−31	−36, −27
P3F	256 (50)	2,786 (87)	37	32, 41
Clean cases (n=30,335)					<0.001
P1	2,570 (14)	1,492 (13)	−1.1	−1.9, −0.3
P2+	4,762 (26)	4,371 (37)	12	11, 13
P2	5,773 (31)	3,591 (31)	−0.39	−1.5, 0.7
P3	5,400 (29)	2,243 (19)	−9.9	−11, −8.9
P4	106 (0.6)	27 (0.2)	−0.34	−0.48, −0.19
Daily no. of ED visits
Main ED	342 (322, 358)	240 (230, 258)	−96	−106, −85	<0.001
Main ED and FSA	342 (322, 358)	256 (246, 290)	−74	−88, −61	<0.001
Fever cases	9 (7, 11)	47 (38, 72)	37	33, 43	<0.001
P1F	1 (0, 1)	1 (0, 2)	0	0, 1	0.1
P2F	4 (2, 5)	6 (5, 8)	2	2, 3	<0.001
P3F	4 (3, 6)	39 (31, 61)	34	30, 40	<0.001
Clean cases	331 (312, 349)	204 (196, 218)	−123	−132, −113	<0.001
P1	45 (38, 52)	26 (22, 31)	−18	−12, −2	<0.001
P2+	85 (76, 93)	78 (72, 83)	−7	−12, −2	0.003
P2	102 (88, 115)	65 (56, 73)	−39	−45, −32	<0.001
P3	98 (88, 107)	37 (33, 48)	−57	−62, −53	<0.001
P4	2 (1, 3)	0 (0, 1)	−1	−2, −1	<0.001

^aNon-CB period is from 7 Apr 2019 to 1 Jun 2019 and CB-period is from 7 Apr 2020 to 1 Jun 2020; both were over a period of 56 days. *Statistical tests performed: Wilcoxon rank-sum test, two-sample test for equality of proportions. ^†Statistical tests performed: Wilcoxon rank-sum test, Pearson’s chi-square test. CI: confidence interval, FSA: flu screening area, IQR: interquartile range, P1: critically ill patient, P1F: critically ill ‘fever’ patient, P2: ambulatory patient with urgent condition, P2+: nonambulatory patient with urgent condition, P2F: nonambulatory ‘fever’ patient, P3: ambulatory patient with nonurgent condition, P3F: ambulatory ‘fever’ patient, P4: nonemergency case

The flu screening area (FSA) was a temporary off-site extension in the ED launched during the COVID-19 pandemic to deal with the surge from walk-in and primary healthcare provider referrals. It also served as a secondary national screening centre to support the National Centre for Infectious Diseases, following health authority directives for swab, isolation and containment.[20] Patients seen at the FSA were compared based on the daily and total number of visits between CB and non-CB periods. Data extracted included daily patient volume at the ED, patient demographics (including age, gender, nationality), waiting time to consultation (WTC), triage acuity, ED reattendances, location of consultation (main ED or FSA), primary diagnoses, transfers to high dependency (HD)/intermediate care area (ICA)/intensive care unit (ICU) from the general ward (GW) within 72 h of admission, ED and inpatient length of stay (LOS), and causes of death. ED reattendances were classified into 0–24 h, 24–48 h and 48–72 h after the first ED discharge, as per the Ministry of Health’s performance indicator for public hospital EDs.

Continuous and categorical variables were expressed as median (interquartile range [IQR]) and frequency (percentage), and compared using Wilcoxon rank-sum test and chi-square test, respectively. A P value <0.05 was considered significant. The median and proportional differences were also presented with 95% confidence intervals (CIs). All statistical analyses were performed using R software version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

We reviewed a total of 19,132 visits to the ED by 16,932 patients in the non-CB period and 14,995 visits to the ED by 13,356 patients in the CB period (over 56 days with an equal number of weekends). [Table 1]. Median number of daily visits in the non-CB period was 342 (IQR 322–358), which reduced significantly to 256 (IQR 246–290) in the CB period. Despite a decline in the overall number of ED visits, the proportion of ‘fever’ cases increased in the CB period (22% vs. 2.7%). This was mainly due to ambulatory ‘fever’ cases (P3F), which contributed 87% of ‘fever’ cases in the CB period, up from 50% in the non-CB period (P < 0.001). Total ‘clean’ cases decreased from 18,611 to 11,724 (−37%). P3 and P2 cases had the largest decreases in daily visits, while P1 and P2+ had minimal decreases in daily visits (all P < 0.05).

The characteristics of our study population are shown in Table 2. Proportions of ED reattendances were similar across the two periods. The admission rate increased from 48% to 53% during CB, with increasing admissions reported in the isolation wards (from 267 to 609). The proportions of ICU beds used were similar (1.4%–1.5%). During CB, there was a smaller proportion of discharge against medical advice cases (1.9% vs. 2.5%) and patient abscondment fell, while prehospital cardiac arrest (dead on arrival) numbers remained unchanged.

Table 2.

Demographics and outcomes for patients presenting during non-circuit breaker (non-CB) period and circuit breaker period (CB) (excluding flu screening area cases) (N=32,879).

Characteristic	Median (IQR)/n (%)		Median and proportional differences		P †
	Non−CB period (n=19,132)	CB period (n=13,747)	Estimate*	95% CI*
Age (yr)	56 (36, 70)	59 (39, 71)	2.0	2.0, 2.0	<0.001
<16	212 (1.1)	47 (0.3)	−0.77	−0.95, −0.58
16–65	12,509 (65)	8,723 (63)	−1.9	−3.0, −0.9
>65	6,411 (34)	4,977 (36)	2.7	1.6, 3.7
Male	9,731 (51)	7,482 (54)	3.6	2.5, 4.7	<0.001
Singapore citizen	15,382 (80)	11,008 (80)	−0.32	−1.2, 0.6	0.5
Overall ED reattendance	471 (2.5)	311 (2.3)	−0.20	−0.54, 0.14	0.3
0–24 h	154 (0.8)	114 (0.8)	0.02	−0.18, 0.23	0.9
24–48 h	163 (0.9)	106 (0.8)	−0.08	−0.28, 0.12	0.5
48–72 h	154 (0.8)	91 (0.7)	−0.14	−0.33, 0.05	0.2
Disposition type (n=32,864)					<0.001
Treated and discharged	6,504 (34)	4,423 (32)	−1.8	−2.9, −0.8
Admitted	9,181 (48)	7,286 (53)	5.0	3.9, 6.1
Discharge against medical advice	477 (2.5)	262 (1.9)	−0.59	−0.91, −0.26
Absconded/triaged without consultation	310 (1.6)	158 (1.1)	−0.47	−0.73, −0.21
Dead on arrival	38 (0.2)	32 (0.2)	0.03	−0.07, 0.14
Transferred/referred to other hospital	136 (0.7)	105 (0.8)	0.05	−0.14, 0.25
Others	2,479 (13)	1,473 (11)	−2.2	−3.0, −1.5
First care level at inpatient admission (n=15,301)					<0.001
General ward	7,139 (89)	6,233 (86)	−2.6	−3.6, −1.5
Monitored ward (HD/ICA/ICU)	651 (8.1)	402 (5.5)	−2.5	−3.3, −1.7
Isolation ward	267 (3.3)	609 (8.4)	5.1	4.3, 5.9
Next care level for general ward patients (n=8,561)					0.4
General ward	3,904 (94)	4,138 (94)	−0.70	−1.7, 0.3
Monitored ward (HD/ICA/ICU)	196 (4.7)	234 (5.3)	0.56	−0.4, 1.5
Isolation ward	40 (1.0)	49 (1.1)	0.14	−0.31, 0.59
Inpatient length of stay (n=14,882)	3.0 (2.0, 6.0)	4.0 (2.0, 7.0)	0	0.0, 1.0	<0.001
P1F (n=82)	4.5 (3.0, 8.0)	4.0 (2.0, 8.0)	0.0	−2.0, 1.0	0.5
P2F (n=346)	5.0 (3.0, 11.0)	4.0 (3.0, 6.8)	−1.0	−2.0, 0.0	0.004
P3F (n=707)	2.0 (2.0, 5.0)	4.0 (2.0, 6.0)	1.0	0.0, 1.0	<0.001
P1 (n=3,024)	3.0 (2.0, 6.0)	4.0 (2.0, 7.0)	0.0	0.0, 0.0	0.003
P2+ (n=6,055)	4.0 (2.0, 7.0)	4.0 (2.0, 8.0)	1.0	0.0, 1.0	<0.001
P2 (n=3,302)	3.0 (2.0, 5.0)	3.0 (2.0, 6.0)	0.0	0.0, 1.0	<0.001
P3 (n=1,360)	3.0 (2.0, 5.0)	3.0 (2.0, 5.0)	1.0	0.0, 1.0	<0.001
P4 (n=4)	1.50 (1.25, 1.75)	3.00 (3.00, 3.00)	1.5	1.0, 2.0	0.2
Waiting time to consultation (mins) (n=32,361)	22 (8, 56)	22 (9, 50)	0	0, 0	0.7
Fever case (n=2,533)	34 (17, 60)	19 (11, 34)	−10	−13, −8	<0.001
P1F (n=101)	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.0	0.0, 0.0	0.2
P2F (n=523)	29 (16, 57)	19 (12, 31)	−9	−13, −6	<0.001
P3F (n=1,909)	40 (24, 65)	20 (11, 37)	−17	−20, −14	<0.001
Clean case (n=29,824)	21 (8, 56)	22 (9, 54)	0	0, 1	0.045
P1 (n=4,045)	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.0	0.0, 0.0	<0.001
P2+ (n=9,094)	15 (9, 28)	19 (10, 35)	3	2, 3	<0.001
P2 (n=9,241)	25 (11, 61)	32 (12, 77)	4	3, 5	<0.001
P3 (n=7,322)	57 (27, 110)	56 (27, 106)	−1	−3, 2	0.7
P4 (n=122)	65 (33, 140)	73 (46, 130)	3	−24, 31	0.8
ED length of stay (h) (n=32,844)	3.81 (2.57, 5.52)	3.88 (2.76, 5.37)	0.03	−0.01, 0.08	0.2
Fever case (n= 2543)	4.29 (2.81, 5.92)	4.32 (3.07, 6.03)	0.14	−0.01, 0.36	0.2
Characteristic	Median (IQR)/n (%)		Median and proportional differences		P †
	Non−CB period (n=19,132)	CB period (n=13,747)	Estimate*	95% CI*
P1F (n=101)	4.5 (3.6, 6.3)	4.6 (3.4, 7.2)	0.07	−0.90, 1.02	>0.9
P2F (n=526)	4.85 (3.68, 6.81)	4.67 (3.27, 6.49)	−0.24	−0.64, 0.16	0.2
P3F (n=1,916)	3.53 (2.31, 5.14)	4.24 (3.02, 5.91)	0.73	0.46, 1.01	<0.001
Clean case (n=30,292)	3.80 (2.56, 5.51)	3.82 (2.71, 5.25)	−0.03	−0.08, 0.02	0.2
P1 (n=4,055)	4.12 (2.83, 6.10)	3.63 (2.63, 4.99)	−0.52	−0.65, −0.39	<0.001
P2+ (n=9,114)	4.3 (3.1, 6.4)	4.1 (3.1, 5.5)	−0.28	−0.37, −0.20	<0.001
P2 (n=9,359)	3.78 (2.57, 5.31)	3.81 (2.69, 5.25)	0.01	−0.08, 0.09	0.9
P3 (n=7,632)	3.20 (1.97, 4.87)	3.26 (2.05, 4.68)	−0.01	−0.11, 0.08	0.8
P4 (n=132)	2.25 (1.28, 3.59)	2.66 (1.49, 5.07)	0.34	−0.36, 1.18	0.3
Death	38 (0.20)	32 (0.20)	0.03	−0.07, 0.14	0.6
Cause of death (n=70)					0.6
Cardiac arrest	4 (11)	5 (16)	5.1	−14, 24
AMI/IHD/CAD	16 (42)	17 (53)	11	−15, 37
Pneumonia	6 (16)	3 (9.4)	−6.4	−25, 12
Others	12 (32)	7 (22)	−9.7	−33, 14
Specific primary diagnoses
Fever	454 (2.4)	608 (4.4)	2.0	1.6, 2.5	<0.001
Upper respiratory tract infection	425 (2.2)	1198 (8.7)	6.5	6.0, 7.0	<0.001
Pneumonia	393 (2.1)	733 (5.3)	3.3	2.8, 3.7	<0.001
Diarrhea/gastroenteritis/food poisoning	505 (2.6)	205 (1.5)	−1.1	−1.5, −0.8	<0.001
Alcohol intoxication	84 (0.4)	26 (0.2)	−0.25	−0.37, −0.13	<0.001
Road traffic accident	37 (0.2)	39 (0.3)	0.09	−0.02, 0.21	0.12
Assault	34 (0.2)	26 (0.2)	0.01	−0.09, 0.11	>0.9
Chest pain	1,043 (5.5)	716 (5.2)	−0.24	−0.74, 0.25	0.3
Non-specific/non-immediately life-threatening chest painb	888 (4.6)	583 (4.2)	−0.4	−0.86, 0.06	0.088
Specifically serious/life-threatening chest painc	130 (0.7)	104 (0.8)	0.08	−0.12, 0.27	0.5
Cardiac arrest	25 (0.1)	29 (0.2)	0.08	−0.02, 0.18	0.1
Stroke	164 (0.9)	103 (0.7)	−0.11	−0.31, 0.09	0.3
Pulmonary embolism, deep vein thrombosis	28 (0.1)	16 (0.1)	−0.03	−0.11, 0.05	0.6
Abdominal pain complaint	541 (2.8)	369 (2.7)	−0.14	−0.51, 0.22	0.5

^aNon-CB period is from 7 Apr 2019 to 1 Jun 2019 and CB-period is from 7 Apr 2020 to 1 Jun 2020. ^bExamples include musculoskeletal chest pain and atypical chest pain. ^cExamples include acute coronary syndrome, acute myocardial infarction, ischaemic heart disease and coronary artery disease. *Statistical tests performed: Wilcoxon rank-sum test, two-sample test for equality of proportions. ^†Statistical tests performed: Wilcoxon rank-sum test, Pearson’s chi-squared test. AMI: acute myocardial infarction, CAD: coronary artery disease, CI: confidence interval, DOA: dead on arrival, ED: emergency department, HD: High dependency, ICA: intermediate care area, ICU: intensive care unit, IHD: ischaemic heart disease, IQR: interquartile range, P1: critically ill patient, P1F: critically ill ‘fever’ patient, P2: ambulatory patient with urgent condition, P2+: nonambulatory patient with urgent condition, P2F: nonambulatory ‘fever’ patient, P3: ambulatory patient with nonurgent condition, P3F: ambulatory ‘fever’ patient, P4: nonemergency case, URTI: upper respiratory tract infection

Most patients admitted to the GW were not transferred to other wards (e.g., isolation ward) or a monitored bed such as HD/ICA/ICU) until discharge (94%) in both periods. Median inpatient LOS of ED admissions increased significantly from 3 (IQR 2–6) to 4 (IQR 2–7) days during the CB period.

Waiting time to consultation increased for P2+ (median difference +3 min) and P2 (+4 min) cases. The converse was observed in ‘fever’ cases, specifically P3F and nonambulatory ‘fever’ (P2F) patients. P3F cases saw the biggest decrease in WTC from a median of 40 (IQR 24–65) min to 20 (IQR 11–37) min, while P2F cases saw a decrease in WTC from 29 (IQR 16–57) min to 19 (IQR 12–31) min. ED LOS for P3F cases increased from a median of 3.53 (IQR 2.31–5.14) h during the non-CB period to 4.24 (IQR 3.02–5.91) h (P < 0.001) during the CB period.

The proportion of deaths was stable at 0.2%. Deaths due to acute myocardial infarction, ischaemic heart disease and coronary artery disease accounted for more than half (53%) of the deaths in the CB period, up from 42% in the non-CB period (P = 0.358).

The proportion of the top three diagnoses in the CB period increased by 12% points: fever (+2.0%), upper respiratory tract infection (URTI) (+6.5%) and pneumonia (+3.3). There was no change in life-threatening conditions such as ACS and strokes.

DISCUSSION

Recent reports on the COVID-19 pandemic have shown an overall fall in ED attendances and in presentations of life-threatening conditions.[10,11,12,14] Our study shows that the effects may not be universal and may vary depending on factors such as public health measures, policies and behaviour of the community.

Our study noted decreased ED attendance during the CB period, similar to that described in the past and present pandemics globally.[4,5,6,7,9,10,11,12,13,14] The CB measures coupled with increased education on hand hygiene likely led to the decrease (−1.1%, P < 0.001) of diarrhoea, gastroenteritis and food poisoning cases,[10] similar to previous pandemics.[4] Cessation of night entertainment services (pubs, clubs) and dine-in at restaurants may have contributed to a decrease in alcohol intoxication cases.[21] This reduction in avoidable conditions may be an unintended benefit of ‘lockdown’ measures on ED attendances. Whether these measures have similar impact to reduce the healthcare burden beyond the pandemic deserves further evaluation.

In contrast, attendance for assault remained unchanged, which we postulate could be due to psychological distress from physical isolation, in particular of domestic abuse victims with their perpetrators,[22] and loss of job security.[23] These numbers may possibly even be underreported, given the propensity of domestic violence to be well hidden.[10,24]

The rate of venous thromboembolic events such as deep vein thrombosis and pulmonary embolism remained stable despite contrary reports[25] and anecdotal accounts of more such cases during the lockdown. This is reassuring considering the anticipated higher risk of venous thromboembolic event from decreased physical activity during the lockdown confinement and prothrombotic risk from COVID-19 infections, although the incidence of the latter was low in Singapore in comparison to other countries.[26]

The largest median decreases in visits were in P2 and P3 patients, suggesting an overutilisation of emergency resources during the nonpandemic periods. Potentially, public education can influence right siting of care (to polyclinics, general practitioners [GPs]) for nonurgent conditions, allowing the ED to focus on its primary mission of managing critically ill patients. Other factors contributing to decreased ED attendance include public advisories to stay at home and fear of nosocomial spread of disease,[27] similar to those observed in previous pandemics.[4,5,6,7,9] The finding that the majority of local COVID-19 infections during the CB period were in foreign worker dormitories rather than in the community[28] suggests that public perception and fear of the pandemic may have contributed to the decreased volume.[14] Community perceptions of COVID-19 were compounded by uncertainty, fear and anxiety. Timely and clear communications between the government and the public have been shown to effectively calm and encourage better social cooperation.[29]

It would be disastrous for a severely ill patient to avoid or delay hospital visits due to such concerns, risking worse consequences.[6,14,15,16,17] Fortunately, this was not seen. Contrary to studies demonstrating evidence of the COVID-19 pandemic having a deterrent effect on patients’ health-seeking behaviour,[2,14,15,16,17] we found no significant reduction in presentations for life-threatening conditions including ACS, strokes and critical sepsis-related conditions during the CB period. Similarly, an Australian study found that there were decreases in ED attendances across many diagnoses during the COVID-19 pandemic, but presentations for acute myocardial infarction and burns remained unchanged,[21] and a Taiwan study found no reduction in ST-elevation myocardial infarction admissions during this pandemic.[30] However, a multihospital, multistate US study reviewing electronic health records found an increase in all-cause mortality during the pandemic, amidst decreased ED volumes that occurred despite the low number of community COVID-19 cases.[14] Reports on the experiences from previous pandemics were similarly divided in their conclusions.[4,5,6,8,9]

Despite a fall in ED attendances, surge capacity is still required due to the longer turnaround times for patients in view of the additional measures that are needed for safe hospital operations during a pandemic. Critical to the ED operations in a pandemic, the surge in ‘fever’ cases seen during the CB period did not compromise ED resources. The setting up of the FSA provided surge capacity for primary care referrals (largely comprising providers without swab capabilities, patients staying in congregate settings such as dormitories who required referral to ED before transportation to isolation facilities) and decanted P3F patients from main ED to FSA.[19,20] A significant (40%) load of ‘fever’ patients, and 8.3% of the total ED load [Table 3], was borne by the FSA.

Table 3.

Number of visits at the main ED and the FSA during the circuit breaker period.

Characteristic	n (%)		Overall
	Main ED	FSA
All cases	13,747 (92)	1,248 (8.3)	14,995
Fever cases	2,035 (63)	1,181 (37)	3,216
P3F cases	1,666 (60)	1,120 (40)	2,786
P3F cases per daya	26 (22, 36)	12 (9, 24)	39 (31, 61)

^aData presented as median (interquartile range). ED: emergency department, FSA: flu screening area, P3F: ambulatory ‘fever’ patient

A significant proportion of the local outbreak occurred among foreign workers living in dormitories.[19,28] Medical teams were deployed to the dormitories to treat and swab patients, referring only those who required hospital care to the ED.[31] This strategy of forward triaging and treatment in the community prevented hospitals from being overwhelmed. Pandemic healthcare response plans could adopt a strategy of ‘forward’ triage and treatment posts at the prehospital stage as an anticipatory measure to avoid surges in ED attendances. Telemedicine consults[32] and ‘drive-by’ swabbing facilities[33] for stable patients could further spare ED resources. This seems feasible, given that in our study, more than half of these cases were eventually treated and discharged (60%) [see Supplemental Digital Appendix].

Despite a decrease in total ED attendance, median WTC significantly increased for critically ill patients (P2+ and P2), and this could be due to several reasons. Manpower for P2+ and P2 areas were often redeployed to supplement the ‘fever’ area. At peak hours, during the non-CB period, the department may have up to four senior doctors and 16 junior doctors on shift. During CB, the manpower was required to be split into five segregated teams taking turns on 12-hour shifts to prevent cross-infections, resulting in six senior doctors and nine junior doctors on shift at any one time. Additional manpower was seconded from other clinical departments, with two junior doctors supplementing each team. Despite the manpower augmentation, there were fewer staff and reduced redeployability of personnel during peak attendance.[18] Seconded manpower were not ED trained, and hence were not as familiar with ED workflows. More time could have been spent taking COVID-19-directed screening history. In contrast, median WTC significantly decreased for ‘fever’ cases. This could be due to more dedicated staffing in the ‘fever’ area and the opening of an FSA to deal with P3F patient load. Robust COVID-19 hospital workflows could also have resulted in smoother processes for this patient group.

However, P3F patients were found to have increased ED LOS during the CB period, and this was attributed to thorough epidemiological history-taking and the wait for swab results before disposition. The increased proportion of P3F admissions to the isolation ward was attributed to the decision to admit certain COVID-19 suspect cases for containment due to lack of community isolation facilities in the early pandemic period. This aspect of hospital resource utilisation should change as the pandemic response evolves towards that of managing an endemic disease. It is easy to overlook the needs of non-COVID-19 illness patients during a pandemic. Hence, holistic pandemic response plans are needed to ensure that appropriate manpower and resources are allocated to care for all who are critically ill during a pandemic.

The ways in which a pandemic and its lockdown measures can interact with a population are undoubtedly complex. Being a single-centre study, our findings may not reflect the experiences at other local centres. As this was a retrospective review of records, we lacked information regarding ED attendees’ attitudes and beliefs regarding ED utilisation during a pandemic. A prospective multicentre study in future could provide useful insights on the attitudes and health-seeking behaviour of ED patients.

In conclusion, the study gives us a glimpse of how the lockdown measures in response to COVID-19, and their downstream effects, could impact ED resource utilisation. Despite a decrease in attendances, hospital EDs must invest in building surge capacity to deal with the screening and containment of emerging infectious disease cases in the initial stages of a pandemic. Along with public health measures and robust hospital resource management, the enabling of community and primary care services would help protect limited ED resources and alleviate the healthcare burden in hospitals during lockdowns. This provides a basis for improving our healthcare response plans from a tertiary hospital perspective.

Conflicts of interest

There are no conflicts of interest.

Supplemental digital content

Appendix at http://links.lww.com/SGMJ/A287

APPENDIX

Table S1.

Number of visits at the main ED and the FSA during the circuit breaker period

Characteristic	n (%)		Overall
	Main ED	FSA
All cases	13,747 (92)	1,248 (8.3)	14,995
Fever cases	2,035 (63)	1,181 (37)	3,216
P3F cases	1,666 (60)	1,120 (40)	2,786
P3F cases per day*	26 (22, 36)	12 (9, 24)	39 (31, 61)

^aData presented as median (interquartile range). ED: emergency department, FSA: flu screening area, P3F: ambulatory ‘fever’ patient

Table S2.

Characteristics of P3F cases seen at the flu screening area during the circuit breaker period (N=1120).

Characteristic	n (%)
Nationality
Singaporean	299 (27)
Non-Singaporean	821 (73)
Disposition type
Treated and discharged	677 (60)
Admitted	346 (31)
Referred to other government hospitals	3 (0.3)
Others	94 (8.4)

P3F: ambulatory ‘fever’ patient.

Funding Statement

Nil.