Datasets
This study utilised Korea’s National Health Insurance (NHI) claims data. Because this dataset includes data of the entire national population and all medical institutions are designated as providers, the NHI claims database represents the population’s utilisation of medical care and has been used to estimate the prevalence and incidence of diseases in Korea.
This study targeted data from January 2020 to July 2020, and for comparison, a dataset containing data from January 2017 to December 2019 was constructed. It takes a certain amount of time for a hospital's claims to be registered in the NHI data system, and the larger hospitals tend to claim multiple cases at once. Therefore, there is a delay between the actual date of medical care and the date in the system that indicates utilisation. To address this problem, we extracted the data after six months later (when the care was received on by July 31, 2020, and the data was extracted on December 18th, 2020) since almost 95% of claims are reported within six months29. The statistical analysis was conducted using STATA SE/14.2.
Disease categories
For patients who visited the medical institutions taking ambulatory services during those periods, we extracted the claims with acute upper respiratory infections (AURIs), cancer, diabetes, or high blood pressure as the main and first sub-diagnosis. We identified the patient visit records pertaining to AURIs according to the code J00-J06, from the International Statistical Classification of Diseases and Related Health Problems (ICD). We considered diabetes (ICD-10 codes: E10, E11, E12, E13, E14) and high blood pressure (ICD-10 codes: I10, I11, I12, I13, I14, I15) as two representative chronic diseases. Lastly, we use the data regarding cancer (denoted as “C” in the ICD-10 codes).
Treated and Control Group
We chose AURIs as the treatment group for an interrupted time series analysis (ITSA) because respiratory diseases have similar symptoms as those of COVID-19 and the level of "infectiousness" is usually inferred based on the amount of viral shedding from the upper respiratory tract in patients 30. The fear and uncertainty around COVID-19 might have led people to refrain from physical and social interaction, which would help to reduce the spread of AURIs. Moreover, the anxiety of visiting a hospital for a minor illness might have kept the elderly population from seeking hospital services. Therefore, a substantial reduction in the incidence of AURIs could have implications for COVID-19 responses 31.
However, the incidence of chronic disease is independent of the COVID-19 pandemic and social distancing measures. Furthermore, individuals with chronic diseases visit the hospital regularly for prescriptions and check-ups. Healthcare utilisation according to chronic disease type can indicate whether hospital visits from elderly people have been affected by the COVID-19 pandemic due to the fear of infection or social distancing. In addition, there might be decreases in diagnoses because mandatory health check-ups have been delayed or deferred. Considering this, our results might underestimate the hospital visits associated with chronic diseases. The medical needs of patients with cancer might not be affected by the COVID-19 pandemic, but because of the fear of acquiring COVID-19, patients may avoid in-person visits to the hospital and postpone their appointments 32.
Statistical Analysis
Empirical methodology
Single ITSA
We set two interventions; the first intervention time as the first week of February 2020(week 6, 2020), when the first COVID-19 case in Korea was reported in week 5, 2020; because of the high fear of acquiring it and uncertainty on COVID-19 in Korea, there was a shortage of N95 masks and hand sanitisers; the second intervention time was set at the fourth week of April, 2020(week 17, 2020), when the level of social distancing was relaxed after the number of new COVID-19 cases decreased to less than 20[1]. We performed a single-group ITSA using Newey-West robust errors with one lag, on the year-over-year(YOY) growth rate in the number of patients per week related to each disease as the dependent variable.
Outcome measures
Outcome measures of interests in this study is the year-over-year(YOY) growth rate in the number of patients per week. To begin with, we examine the level change of the YOY growth rate by two interventions. Furthermore, we focus on the trend change of the YOY growth rate as well as the difference of post intervention trend change of the YOY growth rate between the treated and the control group in multiple group ITSA, in order to estimate the intervention effect after implementation to the counterfactuals.
Multiple group ITSA
We extended a single ITSA to a multiple group ITSA, including the control group. We compared the interruption effect of hospital visits from AURIs and other diseases (high blood pressure, diabetes, cancer, and total ambulatory services). First, we defined the treatment group as those with AURIs because they have a similar transmission mechanism as that of COVID-19 and the patients might be affected by the widespread behavioural changes related to the pandemic. Second, we identified the control group as those with chronic diseases because their incidence has little association with the COVID-19 pandemic, but routine medical care is essential. Lastly, we added cancer outpatient visits as the control group because their appointments should not be deterred but might be deferred due to the fear of infection during hospital visits. In multiple group ITSA, we use the data from 2017 to 2020, since the trend and the difference in level before the intervention should be similar among treated and control groups.
Summary statistics
Table 1 presents the summary statistics for hospital visits by 65–84-year-olds from January to July in 2017-2020. Panel (a) shows the summary statistics for the period of January 1, 2020, to July 31, 2020, which represents 1) the number of patients per week 2)the year-over-year(YOY) growth in the number of patients per week for 65–84-year-olds at the onset of the COVID-19 pandemic crisis. Panel (b) presents ones between January 1 and July 31, 2017, 2018, and 2019. On one hand, the patients visit from total ambulatory services, and chronic diseases did not change much, while the number of patients with AURIs decreased during the COVID-19 pandemic period by 18.7% (with a standard deviation of 0.313) between the same week in 2019 and 2020. On the other hand, the number of patients from most of diseases increased during 2017-2019.
Table 1. Summary statistics
Diseases
|
|
- January 1st-July 31st, 2020
|
- January 1st-July 31st
(2017, 2018, 2019)
|
|
|
Mean
|
Standard Dev
|
Mean
|
Standard Dev
|
Total
|
Number of patients per wk
|
5,128,322
|
1,103,284
|
5,201,306
|
1,010,169
|
|
Change in the number of patients per wk
|
0.031
|
0.278
|
0.05
|
0.313
|
High blood pressure
|
Number of patients per wk
|
513,163
|
22,902
|
542,579
|
40,086
|
|
Change in the number of patients per wk
|
-0.029
|
0.156
|
-0.007
|
0.108
|
Diabetes
|
Number of patients per wk
|
767,402
|
34,972
|
718,213
|
59,682
|
|
Change in the number of patients per wk
|
0.033
|
0.164
|
0.054
|
0.116
|
Cancer
|
Number of patients per wk
|
75,990
|
6,133
|
54,665
|
10,238
|
|
Change in the number of patients per wk
|
0.178
|
0.24
|
0.221
|
0.191
|
AURIs
|
Number of patients per wk
|
111,645
|
54,308
|
135,613
|
38,837
|
|
Change in the number of patients per wk
|
-0.187
|
0.313
|
0.03
|
0.158
|
Note: This table presents the summary statistics for hospital visits for ambulatory services by 65–84-year-olds. Panel(a) covers the period between January 1st and July 31st in 2020 while panel (b) the same period of panel (a) in year 2017, 2018, and 2019. Mean represents the average number of weekly patients from each disease. The change in the volume of patients is calculated as the difference between the same week from the previous year (i.e., (week 6(2020)-week6(2019)]/week6(2019)). The number of patients only include ones taking ambulatory services.