Introduction
The ongoing COVID-19 pandemic has posed a major challenge to the delivery of typical
healthcare services worldwide. For gastroenterologists, the COVID-19 pandemic has
compromised the availability of routine and even emergency endoscopy services because
upper endoscopy is considered to be an aerosol-generating procedure [1]. A study from New York, one of the epicenters of the COVID-19 pandemic, showed that
cases of upper gastrointestinal bleeding (UGIB) were more severe after the COVID-19
outbreak [2]. Another Austrian study showed that the number of cases of non-variceal UGIB declined
after the lockdown for COVID-19 [3].
Unlike many other countries, in Hong King, the number of COVID-19 cases remains relatively
low and so far, there has been no large-scale community outbreak. However, we previously
reported that the overall endoscopy volume has been reduced by more than 50 % since
the outbreak of COVID-19, resulting in delay in diagnosis of gastric and colorectal
cancer in more than 37 % of cases [4]. The aim of this study was to investigate changes in the number and characteristics
of patients who were hospitalized for UGIB after the first wave of COVID-19 in Hong
Kong.
Patients and methods
Data source and patients
The first local case of COVID-19 was reported on January 23, 2020. From then until
June 30, 2020, 1,205 COVID-19 cases were reported (the end of follow up of the present
study). We identified all adult patients, aged 18 years or above, who presented with
UGIB to all local public hospitals in Hong Kong between October 1, 2018 and March
31, 2020. Data were retrieved from the Clinical Data Analysis and Reporting System
(CDARS), a territory-wide electronic health care database of the Hong Kong Hospital
Authority. The Hospital Authority is the sole public healthcare provider for the 7.3
million local population, which manages over 85 % of all hospital beds in 43 public
hospitals. All local residents are eligible for heavily subsidized primary, secondary,
and tertiary care provided by the Hospital Authority, which is directly funded by
the government. The CDARS captures important clinical information including patient
demographics, diagnosis, medication prescription and dispensing records, laboratory
results, and surgery and endoscopy performed. All data in the CDARS are anonymized.
The International Classification of Diseases, Ninth Revision (ICD-9), was used for
disease coding and the accuracy of the coding of gastrointestinal bleeding had been
verified in our previous studies [5]
[6]. Patients with upper gastrointestinal bleeding (UGIB) and associated endoscopic
interventions were identified with relevant ICD-9 codes (Supplementary Table 1). Other relevant clinical information including baseline demographics, endoscopic
diagnosis, endoscopic intervention for hemostasis, laboratory parameters, blood transfusion
requirement, 7-day rebleeding, mortality, and overall length of hospital stay were
reviewed. Due to the difficulty in defining rebleeding based on the electronic database,
we defined 7-day rebleeding as undergoing two upper endoscopies with the same diagnostic
coding for UGIB within 7 days. Similarly, we used all-cause mortality, rather than
bleeding-related mortality, in this study.
The daily number of COVID-19 new case was obtained from the Center for Health Protection
of the Department of Health of Hong Kong [7]. The study was approved by the Institutional Review Board of the University of Hong
Kong and the Hong Kong West Cluster of the Hospital Authority (UW 20–279).
Time trend analysis of newly diagnosed upper gastrointestinal tract bleeding during
COVID-19
The week that COVID-19 started was defined as the week when the first local case of
COVID-19 was diagnosed in Hong Kong, which was January 23, 2020. The end date of this
study was June 30, 2020. Time trend analysis was performed on the hospitalization
rates for UGIB during this period as compared with preceding months up to October
1, 2018. Because of daily and seasonal variations in UGIB, the autoregressive integrated
moving average model (ARIMA) was constructed to predict the number of patients with
UGIB, causes of UGIB, as well as other clinical outcomes on a weekly basis. The corrected
Akaike’s information criterion (AICc) was used to choose the ARIMA models with different
combinations of parameters. The model parameter used were a seasonal ARIMA(p, d, q)(P, D, Q)
s
whereas p,d,q referred to the autoregressive, difference, and moving average term
of the ARIMA model and P,D,Q was the corresponding part in the seasonal part and s
the seasonal period. The data were first used to fit the model with adjustment for
the holiday effect of Christmas and Lunar new year. The mean percentage error (MPE)
was used to test the accuracy of the forecast. Sensitivity analysis was performed
with Petitt’s test to look for a turning point in the trend of UGIB during the whole
COVID-19 study period.
Statistical analysis
The difference between the ARIMA prediction and actual number of patients diagnosed
with UGIB was compared with a two-sided Student’s t-test. The upper and lower limit of the 95 % confidence interval was used to define
the worst- and best-case scenarios in the ARIMA prediction, respectively. Other clinical
parameters were compared with a Student’s t-test when appropriate. Pearson correlation was used to assess the relationship between
new COVID cases and UGIB cases with hospitalization. P < 0.05 was considered statistically significant.
Results
Observed changes in the number of UGIB cases during COVID-19
Between October 2018 and June 2020, there were a total of 2,416 hospitalizations for
UGIB, including 824 hospitalizations during the COVID-19 period. Since the first local
case of COVID-19, there was a dynamic change in the number of UGIB hospitalizations,
with an initial decline follow by a rebound. There was a significant negative association
between the number of COVID-19 cases per week and the number of hospitalizations for
UGIB per week (Pearson correlation: –0.53, P < 0.001). Time series analysis showed there was a turning point at week 14 from the
first COVID-19 case on the trend for UGIB cases (Petitt’s test, P < 0.001), which also coincided with the end of the first COVID-19 wave in Hong Kong
([Fig. 1]).
Fig. 1 Time series analysis of the number of UGIB cases after COVID-19. The upper panel showed
the number of upper endoscopies (OGDs) performed between October 2018 and June 2020.
The shaded area refers to the pre-COVID-19 period. The lower panel shows the number
of UGIB patients hospitalized. The blue line refers to the actual number of UGIB patients
observed whereas the green line is the predicted number of UGIB hospitalizations,
based on the ARIMA model with 95 % CI (in light blue). The blue column in the top
right panel shows the number of COVID-19 cases per week during the same period.
Number of UGIB hospitalizations: Observed vs ARIMA model prediction
Given potential daily and seasonal variations in UGIB incidence, we compared the observed
number of UGIB hospitalizations per week with the number predicted by the ARIMA(0,1,1)(0,1,1)52. The mean percentage error for the ARIMA model was 1.5 %.
During the initial 14-week COVID-19 period, the number of UGIB hospitalizations was
significantly less than predicted (mean 29.8 vs 35.5 per week, P = 0.05). However, there was a rebound in the number of UGIB hospitalizations after
week 14 (observed vs predicted per week: mean 39.8 vs 26.7, P < 0.01). ([Fig. 1] and [Table 1])
Table 1
Actual and predicted number of patients with UGIB hospitalized per week after the
first local COVID-19 case.
|
Actual
|
Predicted
|
P value
|
|
Overall (from Jan to Jun 2020)
|
34.6 [30.8–38.4]
|
31.3 [28.1–34.5]
|
0.17
|
|
Before week 14
|
29.8 [25.4–34.2]
|
35.5 [34.5–38.1]
|
0.05
|
|
After week 14 till June 2020
|
39.8 [34.5–45.1]
|
26.7 [23.6–29.8]
|
< 0.01
|
UGIB, upper gastrointestinal bleeding.
[ ], 95 % confidence interval.
Clinical characteristic and causes of UGIB cases during COVID-19
The characteristics of UGIB patients admitted at different periods is shown in [Table 2]. When compared to the pre-COVID-19 period, patients admitted during the COVID-19
period were significantly older (mean age 65.8 vs 67.8 years, P = 0.01). For patients admitted during the COVID-19 period, the hemoglobin level was
significantly lower (mean 8.3 vs 7.5 g/dL, P < 0.01) and the blood transfusion rate was higher (50.2 % vs 64.5 %, P < 0.01). There was no significant difference in the timing of endoscopy after admission
(mean 1.02, 95 %CI:0.98–1.05 vs 1.06, 95 %CI:0.98–1.14 days, P = 0.332) or the percentage of patients requiring endoscopic hemostasis (77.3 %, 95 %CI:
74.8–79.7 vs 76.3 %, 95 %CI: 72.4–80.2, P = 0.68) before and after COVID-19.
Table 2
Characteristics of patients with UGIB admitted before and after the outbreak of COVID-19
in Hong Kong.
|
Before COVID-19
|
After COVID-19
|
P value
|
|
Mean patient age in years (± SD)
|
65.8 ± 0.5
|
67.8 ± 1.8
|
0.01
|
|
Male sex
|
68.4 %
|
67.8 %
|
0.77
|
|
Hemoglobin in g/dL (± SD)
|
8.3 ± 0.1
|
7.5 ± 0.2
|
< 0.01
|
|
Urea in mmol/L) (± SD)
|
13.7 ± 1.0
|
12.5 ± 0.6
|
0.90
|
|
Creatinine in umol/L (± SD)
|
137 ± 13
|
163 ± 25
|
0.33
|
|
Blood transfusion
|
50.2 % [47.6–52.9 %]
|
64.5 % [59.3–69.7 %]
|
< 0.01
|
|
Endoscopic hemostasis
|
77.3 % [74.8–79.7 %]
|
76.3 % [72.4–80.2 %]
|
0.68
|
UGIB, upper gastrointestinal bleeding; SD, standard deviation.
[ ], 95 % confidence interval.
Regarding patient outcomes ([Table 3]), there was no significant difference in rates of 7-day rebleeding (5.1 %, 95 %CI:
3.8–6.3 vs 6.7 %, 95 %CI: 5.2–8.3; P = 0.11) or all-cause mortality (7.1 %, 95 %CI: 5.7–8.6 vs 6.9 %, 95 %CI: 4.8–8.9;
P = 0.82) after COVID-19. However, the average length of stay was significantly shorter
after COVID-19 (11.4 vs 9.8 days, P < 0.001). Peptic ulcer bleeding remained the most common cause of UGIB both before
(66.0 %) and after (66.1 %) COVID-19. Notably, there was a significant increase in
the proportion patients with UGIB with variceal bleeding after COVID-19 (5.3 % vs
10.5 %, P < 0.01).
Table 3
Outcomes and causes of UGIB in patients admitted during the COVID-19 pandemic.
|
Before COVID-19
|
After COVID-19
|
P value
|
|
Rebleeding within 7 days
|
5.1 % [3.8–6.3 %]
|
6.7 % [5.2–8.3 %]
|
0.11
|
|
All-cause mortality
|
7.1 % [5.7–8.6 %]
|
6.9 % [4.8–8.9 %]
|
0.82
|
|
Length of stay in days (± SD)
|
11.4 ± 0.5
|
9.8 ± 0.5
|
0.02
|
|
Causes of UGIB (Mean number of cases per week)
|
|
Peptic ulcer
|
20.4 (66.0 %)
|
23.0 (66.1 %)
|
0.13
|
|
Variceal
|
1.6 (5.3 %)
|
3.6 (10.5 %)
|
< 0.01
|
|
Hemorrhagic gastritis
|
2.3 (7.4 %)
|
1.9 (5.5 %)
|
0.49
|
|
Hemorrhagic duodenitis
|
0.2 (0.8 %)
|
0.1 (0.3 %)
|
0.27
|
|
Esophageal hemorrhage
|
0.2 (0.7 %)
|
0.2 (0.5 %)
|
0.45
|
|
Mallory Weiss tear
|
4.8 (15.5 %)
|
4.2 (12.2 %)
|
0.25
|
|
Angiodysplasia
|
0.5 (1.7 %)
|
0.8 (2.3 %)
|
0.22
|
|
Dieulafoy lesion
|
0.6 (1.9 %)
|
0.8 (2.3 %)
|
0.29
|
|
Others
|
0.2 (0.5 %)
|
0.03 (0.1 %)
|
0.40
|
UGIB, upper gastrointestinal bleeding; SD, standard deviation.
[ ], 95 % confidence interval.
Discussion
Based on the territory-wide electronic health database for Hong Kong and ARIMA prediction,
we noted a dynamic change in hospitalizations for UGIB in Hong Kong. There was a small
but significant reduction in hospitalizations for UGIB in Hong Kong during the initial
phase of COVID-19, followed by a rebound in the number of UGIB hospitalizations. Patients
admitted during the COVID-19 outbreak were found to be older with lower hemoglobin
levels. There was also a higher proportion of variceal bleeding during this period.
In contrast to New York and Austria [2]
[3], Hong Kong was never locked down during this pandemic. However, there was still
a significant reduction in the volume of endoscopy performed during COVID-19 (> 50 %),
particularly elective procedures [4]. This study further showed that even the number of hospitalizations for UGIB was
affected during the initial phase of COVID-19 outbreak. We showed that there was an
inverse association between the number of COVID-19 cases and the number of UGIB hospitalizations.
During the first 14 weeks of the COVID-19 outbreak in Hong Kong, there was a significant
decrease in patients presenting to the hospital with UGIB compared to the prediction
from our ARIMA modelling. Notably, we observed a rebound in the number of patients
with UGIB during the later phase of COVID-19. The reason for this rebound is unknown
but may be due to the delay in presentation of some patients with minor bleeding.
In fact, the proportion of patients with variceal bleeding actually increased during
the COVID-19 period, which supports this speculation.
Although outcomes in patients who were not admitted to the hospital cannot be determined,
there was no significant difference in rates of observed 7-day rebleeding and all-cause
mortality, before and after COVID-19. Notably, there was a significantly shorter length
of stay for patients with UGIB after COVID-19. In addition to creating space in the
hospital for patients with COVID-19, the significantly shorter stays may also have
been due to patient or physician preference, possibly based on previous adverse experiences
associated with the SARS outbreak in Hong Kong in 2003, when hospital-acquired infection
played a major role.
The strength of this study was the use of the comprehensive local public health care
database, which captured information on all patients admitted to public hospitals
in Hong Kong. Moreover, as the incidence of UGIB can vary with the season and even
the day of the week, we used the ARIMA model based on weekly data from 2018 to derive
the predicted numbers of patients hospitalized for UGIB.
This study has limitations. First, there was a lack of proper control, particularly
absence of information on patients with UGIB not hospitalized during this period or
admitted to private hospitals. Second, we only looked at short-term outcome and not
longer-term data. Third, limited information was available on baseline characteristics
due to the use of an electronic healthcare database and the risk of misclassification
based on ICD-9 codes alone. However, the same electronic database has been validated
in our previous GIB studies [6]
[7]. Finally, the accuracy of adjudication of outcomes such as rebleeding and bleeding-related
death cannot be verified in the electronic database.
Conclusion
In conclusion, there was a dynamic change in the number of patients hospitalized for
UGIB during the first wave of COVID-19 in Hong Kong with an initial drop followed
by a rebound in the number of cases. Patients admitted during COVID-19 were older
with lower hemoglobin levels. There was also a higher proportion of bleeding varices
during this period, but there were no observable changes in major clinical outcomes
such as rebleeding and death. Our study findings suggest that the health-seeking behavior
of patients who presented with UGIB may have been affected by the COVID-19 pandemic,
which was inversely associated with the number of local COVID-19 cases and the trend
was more prominent in the initial phase of the outbreak.
