Yearly Spatiotemporal Patterns of COVID-19 During the Pandemic Period: An In-Depth Analysis of Regional Trends and Risk Factors in the Republic of Korea

Abstract

Background: South Korea was one of the first countries to experience the Coronavirus disease (COVID-19) epidemic, and the regional-level trends and patterns in the incidence and case-fatality rates have been observed to evolve with time. This study established yearly spatiotemporal evolution patterns of COVID-19 by region and identified possible regional risk factors accounting for the observed spatial variations. Methods: COVID-19 data between 20 January 2020 and 31 August 2023 were collected from the Korean Centers for Disease Prevention and Control (KCDA). We generated epidemic curves and calculated the yearly incidence and case-fatality rates for each region. In addition, choropleth maps for the location quotient of cases and deaths to visualize yearly regional intensities were generated and the Moran’s I calculated. Associations between the incidence and case-fatality rates with regional risk factors were estimated using regression models. All analyses were performed in R version 4.4.2. Results: We noted a significant difference in the incidence rate by year, with 2022 recording the highest for all regions. A consistent and significant spatial autocorrelation for cases and deaths across all years was observed with Moran I values above 0.4 (p < 0.05). There was a positive association of COVID-19 incidence rates with the population density (RR = 0.02, CI: 0.01–0.04, p = 0.03), percentage aged 60 years and above (RR = 0.03, CI: 0.01–0.05, p = 0.01), smoking prevalence (women) (RR = 0.79, CI: 0.54–1.04, p = 0.01), and diabetes prevalence (women) (RR = 0.51, CI: 0.32–0.71, p = 0.04). Conclusions: The spatiotemporal evolution patterns of COVID-19 in Korea consisted of oscillating hot and cold spots across the pandemic period in each region. These findings provide a useful reference to the government as it continues with the routine surveillance of COVID-19 across the country.

1. Introduction

Since the confirmation of the index case of SARS-CoV-2 in South Korea in January 2020, the virus has caused nearly 35 million reported infections and over 35,000 deaths [1]. Since then, the COVID-19 incidence rates have surged and waned, with different patterns observed across different periods and regions in South Korea [2]. The empirical literature demonstrates that COVID-19 incidence rates exhibit yearly fluctuations in intensity across different regions of South Korea. Although the COVID-19 spatiotemporal dynamics have been assessed in South Korea [3,4,5,6], most studies were carried out at the beginning of the pandemic when available COVID-19 data were limited, leading to some key limitations. Firstly, the early studies relied on incomplete or underreported COVID-19 case data due to delays in testing and reporting. Secondly, the studies focused mainly on short-term trends, making it difficult to capture the evolving spatial patterns of transmission over an extended period in Korea. Thirdly, many analyses used aggregated data at the city or regional level, which may have masked finer-scale variations.

According to Kim and Castro, 2020, the spatiotemporal analysis of daily confirmed COVID-19 cases from 250 districts in South Korea between 20 January and 31 May 2020, showed strong spatial autocorrelation, and the spatial pattern of clusters changed over time, with the duration of clusters becoming shorter [3]. Additionally, these studies identified several hot and cold spots, revealing that, while the first wave was initially contained in the Daegu and Gyeongbuk areas, it rapidly spread across the entire country [4,5,6].

Furthermore, some studies have suggested that the patterns of COVID-19 incidence, spread, severity, and mortality are significantly associated with factors including demographic factors such as population density [7,8,9,10], sex ratio [11,12,13], the percentage of individuals aged ≥60 years [14,15], and per capita income [16,17,18]; health-related factors such as obesity prevalence [19,20], smoking prevalence [21,22], and diabetes prevalence [23,24,25]; and environmental factors including average temperature [26,27,28,29], average rainfall [29,30], and average humidity [26,31].

Understanding the spatiotemporal evolution patterns of COVID-19 across different geographical regions, as well as the risk factors for these spatial trends is critical for a comprehensive understanding of COVID-19 epidemiology. These insights are vital for designing effective public health preparedness plans to mitigate future infectious disease emergencies.

In this study, we analyzed the yearly spatiotemporal incidence patterns of COVID-19 during the pandemic, along with its associated risk factors, across different regions of South Korea. This will provide a detailed and data-driven analysis of COVID-19 transmission dynamics in South Korea by integrating regional epidemiological data with key demographic, human-related, and environmental factors.

2. Methods

2.1. Study Design and Data Collection

We carried out a cross-sectional study involving the analysis of COVID-19 data in South Korea. Data on COVID-19 were extracted from the day of confirmation of the index case (20 January 2020) to August 2023. We collected the daily number of cases and deaths stratified by gender, region, and age group from the COVID-19 official website hosted by the South Korean Centers for Disease Control and Prevention [1]. In addition, the yearly population for each region was collected from the South Korean Statistical Information Service [32]. Based on the review of the relevant literature, we selected ten risk factors that could account for the observed differences in the spatial distribution of COVID-19 in South Korea. We collected data on the population density (people per km²); sex ratio (number of males to females); percentage aged ≥60 years; per capita income (Won); obesity prevalence (for both men and women); smoking prevalence (for both men and women); and diabetes prevalence (for both men and women) from the South Korean National Statistics Agency [33] and data on the average temperature (°C), average rainfall (mm) and average humidity (%) from the Korean Metrological Agency [34].

2.2. Data Analysis

2.2.1. Spatiotemporal Pattern Analysis

We generated the COVID-19 epidemic curves for South Korea between 2020 and 2023. In addition, we calculated the yearly incidence and case-fatality rates for each region, and the statistical differences between the incidence and case-fatality rates across years were estimated using the Friedman test. Using the number of cases and deaths per region, we generated spatial maps for COVID-19 cases and deaths between 2020 and 2023. We also generated choropleth maps for the location quotient of cases and deaths to visualize regional intensities across the different years. To evaluate the spatial clustering of the cases and deaths, the Moran’s I index was calculated. The Moran’s I index is suitable for evaluating spatial clustering as it measures the degree to which COVID-19 cases in one region are similar to those in the neighboring regions. A high positive Moran’s I value indicates clustering, suggesting that regions with high (or low) case counts are surrounded by others with similar case counts, whereas a negative value implies a dispersed pattern, meaning cases are more evenly spread. If Moran’s I approaches zero, it suggests a random distribution, helping policymakers determine whether targeted interventions should focus on specific high-risk areas or if transmission is more evenly spread across regions.

2.2.2. Risk Factor Analysis

The unit of analysis was the region. We applied multiple linear regression to estimate the best-fit regression equations and to assess the amount of variation that can be explained by the risk factors, assuming independent noise terms, all with an identical normal distribution. We built several main effect multivariable regression models to identify the factors significantly associated with COVID-19 cases and deaths per 100,000 people. To reduce overfitting caused by the limited sample size (n = 17 regions), the potential predictors for model development were first identified by a univariable screening process with a pre-set p-value of 0.05. Then, we used a backward stepwise elimination approach, based on a likelihood ratio test, to select the final set of covariates for retention in the COVID-19 outcome models. All models were robust to heteroskedasticity (Breusch–Pagan test), and multicollinearity was not observed, as measured by the variance inflation factor (VIF < 3). All the statistical analyses were conducted using R version 4.4.2, and the statistical level of significance was set at p < 0.05.

2.2.3. Ethical Consideration

This study was approved by the Institutional Review Board of Konyang University (2023-03-022).

3. Results

Between 20 January 2020 to 31 August 2023, the cumulative number of confirmed COVID-19 cases in South Korea was 34,572,554, with 35,605 deaths. The time series plot of the daily confirmed cases showed an exponential surge in the number of daily new cases between February and June 2022 corresponding to the period of the Omicron variant predominance (Figure 1).

Overall, Seoul recorded the highest cumulative incidence rate (IR: 71,606.58), while Gyeongsangbuk-do had the lowest cumulative incidence rate per one hundred thousand of the population (IR: 60,804.15). On the contrary, the highest case-fatality rate of 0.14% was recorded in Busan, Daegu, Gangwon, and Gyeongsangbuk-do. Sejong reported the lowest case-fatality rate (0.02%) (

Table 1. Cumulative incidence rate and case-fatality rate of COVID-19 by region in South Korea.

SN	Region	Abbreviation Used in Spatial Maps	Population (2022)	Cumulative Number of New Cases (2020–2023)	Cumulative Incidence Rate/100,000 Population	Cumulative Number of Deaths (2020–2023)	Cumulative Case-Fatality Rate (2020–2023)
1	Seoul	Se	9,428,372	6,751,335	71,606.58	6587	0.10
2	Busan	Bus	3,317,812	2,092,642	63,072.95	2911	0.14
3	Daegu	Dg	2,363,691	1,516,421	64,154.79	2050	0.14
4	Incheon	Inch	2,967,314	1,991,892	67,127.78	1941	0.10
5	Gwangju	Gwa	1,431,050	1,018,499	71,171.45	868	0.09
6	Daejeon	Dj	1,446,072	1,013,275	70,070.85	1003	0.10
7	Ulsan	Uls	1,110,663	738,128	66,458.32	543	0.07
8	Sejong	Sej	383,591	273,413	71,277.22	61	0.02
9	Gyeonggi-do	Gye-gi	13,589,432	9,266,797	68,191.20	8585	0.09
10	Gangwon-do	Gang	1,536,498	1,005,836	65,462.89	1406	0.14
11	Chungcheongbuk-do	Chbuk	1,595,058	1,075,474	67,425.39	1093	0.10
12	Chungcheongnam-do	Chnam	2,123,037	1,390,798	65,509.83	1638	0.12
13	Jeollabuk-do	Jbuk	1,769,607	1,167,948	66,000.42	1261	0.11
14	Jeollanam-do	Jnam	1,817,697	1,142,483	62,853.32	1083	0.09
15	Gyeongsangbuk-do	Gye-buk	2,600,492	1,581,207	60,804.15	2177	0.14
16	Gyeongsangnam-do	Gye-nam	3,280,493	2,075,991	63,282.90	2065	0.10
17	Jeju	JJ	678,159	451,523	66,580.70	317	0.07

).

There was a significant difference in the incidence rate per 100,000 of the population by year, with the year 2022 recording the highest incidence rate for all regions (

Table 2. Annual incidence rate of COVID-19 by region in South Korea (2020–2023).

SN	Region	Population (2022)	Incidence Rate 2020/100,000 Population	Incidence Rate 2021/100,000 Population	Incidence Rate 2022/100,000 Population	Incidence Rate 2023/100,000 Population	p-Value
1	Seoul	9,428,372	201.43	2186.05	57,661.02	11,558.08	<0.001
2	Busan	3,317,812	56.27	699.47	50,983.42	11,333.79	<0.001
3	Daegu	2,363,691	330.03	648.27	52,701.05	10,475.44	<0.001
4	Incheon	2,967,314	95.68	1141.77	55,593.75	10,296.58	<0.001
5	Gwangju	1,431,050	75.54	485.31	58,769.85	11,840.75	<0.001
6	Daejeon	1,446,072	58.43	788.69	57,834.12	11,389.61	<0.001
7	Ulsan	1,110,663	60.50	558.23	55,040.01	10,799.59	<0.001
8	Sejong	383,591	38.84	519.56	59,452.39	11,266.43	<0.001
9	Gyeonggi-do	13,589,432	106.33	1264.82	56,519.83	10,300.22	<0.001
10	Gangwon-do	1,536,498	78.56	728.80	55,153.28	9502.26	<0.001
11	Chungcheongbuk-do	1,595,058	72.91	647.94	56,475.94	10,228.59	<0.001
12	Chungcheongnam-do	2,123,037	77.86	774.41	54,434.19	10,223.37	<0.001
13	Jeollabuk-do	1,769,607	47.30	493.39	54,432.82	11,026.91	<0.001
14	Jeollanam-do	1,817,697	30.81	303.57	51,792.85	10,726.10	<0.001
15	Gyeongsangbuk-do	2,600,492	93.21	495.68	50,428.57	9786.69	<0.001
16	Gyeongsangnam-do	3,280,493	40.39	589.88	52,196.33	10,456.29	<0.001
17	Jeju	678,159	61.34	622.57	53,278.95	12,617.84	<0.001

Note: This table presents the yearly incidence rate of COVID-19 across different regions of South Korea from 2020 to 2023. The incidence rate is expressed per 100,000 people. The p-values are estimated using the Friedman Test and indicate the statistical significance of the observed differences across the years.

). A similar trend was observed by age and gender (

Table 3. Annual incidence rate of COVID-19 by age and gender in South Korea (2020–2023).

		Cumulative Number of New Cases (2020–2023)	Incidence Rate 2020/100,000 Population	Incidence Rate 2021/100,000 Population	Incidence Rate 2022/100,000 Population	Incidence Rate 2023/100,000 Population	p-Value
Age	0–9	3,270,282	59.19	1331.64	84,004.27	7178.17	<0.001
	10–19	4,246,977	80.14	1260.19	76,608.37	12,351.11	<0.001
	20–29	5,001,143	151.28	1320.11	64,542.98	11,919.27	<0.001
	30–39	5,077,726	116.79	1268.93	62,727.14	12,642.00	<0.001
	40–49	5,237,546	106.79	1034.23	53,722.32	10,013.03	<0.001
	50–59	4,531,012	132.29	930.62	42,741.18	8808.28	<0.001
	60–69	3,898,836	129.99	1083.39	41,205.28	10,243.13	<0.001
	70–79	2,056,083	124.63	879.39	40,596.89	12,142.82	<0.001
	80+	1,252,949	133.95	774.75	41,684.90	12,943.17	<0.001
Gender	Male	18,680,325	120.63	1051.94	59,171.59	12,054.34	<0.001
	Female	15,892,229	115.45	1164.41	51,319.84	9389.84	<0.001

Note: This table presents the cumulative number of COVID-19 cases and incidence rates (per 100,000 people) across different age groups and gender in South Korea from 2020 to 2023. The p-values are estimated using the Friedman Test and indicate the statistical significance of the observed differences across years.

). The year 2021 had the highest case-fatality rate for all regions, and no significant difference was observed in the case-fatality rates of COVID-19 by year for each of the regions (

Table 4. Annual case-fatality rate of COVID-19 by region in South Korea (2020–2023).

SN	Region	Population (2022)	CFR 2020	CFR 2021	CFR 2022	CFR 2023	p-Value
1	Seoul	9,428,372	0.91	0.77	0.08	0.05	0.718
2	Busan	3,317,812	2.62	0.94	0.14	0.07	0.213
3	Daegu	2,363,691	2.58	1.06	0.12	0.08	0.233
4	Incheon	2,967,314	0.92	0.68	0.09	0.07	0.744
5	Gwangju	1,431,050	0.56	0.55	0.09	0.06	0.865
6	Daejeon	1,446,072	0.83	1.40	0.09	0.08	0.558
7	Ulsan	1,110,663	3.87	0.56	0.07	0.04	0.030
8	Sejong	383,591	0.67	0.15	0.02	0.03	0.724
9	Gyeonggi-do	13,589,432	1.80	0.90	0.08	0.05	0.405
10	Gangwon-do	1,536,498	1.08	0.79	0.13	0.13	0.732
11	Chungcheongbuk-do	1,595,058	1.98	0.82	0.10	0.08	0.359
12	Chungcheongnam-do	2,123,037	1.39	0.73	0.11	0.08	0.575
13	Jeollabuk-do	1,769,607	1.31	1.08	0.10	0.07	0.583
14	Jeollanam-do	1,817,697	0.89	0.56	0.09	0.11	0.780
15	Gyeongsangbuk-do	2,600,492	2.48	0.76	0.14	0.09	0.229
16	Gyeongsangnam-do	3,280,493	0.30	0.44	0.09	0.10	0.948
17	Jeju	678,159	0.00	0.31	0.06	0.09	0.92633

Note: This table presents the yearly case-fatality rate of COVID-19 across different regions of South Korea from 2020 to 2023. The p-values are estimated using the Friedman Test and indicate the statistical significance of the observed differences across the years.

).

Figure 2 and Figure 3 show different spatiotemporal patterns across Korea for cases and deaths, respectively. Compared to other years, 2022 had the highest number of hotspots specifically in Seoul, Gyeonggi-do, Chungcheongnam-do, Gyeongsangbuk-do, and Gyeongsangnam-do regions (Figure 2). Similarly, compared to other years, 2022 showed the highest clusters of deaths in the Seoul/Gyeonggi-do area (Figure 3).

The relative concentration of COVID-19 in each region compared to the country was performed using the LQ analysis. The LQ case analysis (

Table 5. Yearly COVID-19 location quotient (LQ) by region in South Korea.

Region	LQ Cases 2020	LQ Cases 2021	LQ Cases 2022	LQ Cases 2023	LQ Deaths 2020	LQ Deaths 2021	LQ Deaths 2022	LQ Deaths 2023
Busan	0.49	0.63	0.92	1.05	0.85	0.73	1.40	1.05
Chungcheongbuk-do	0.65	0.59	1.02	0.96	0.84	0.60	1.05	1.10
Chungcheongnam-do	0.69	0.71	0.99	0.96	0.63	0.63	1.22	1.12
Daegu	2.90	0.59	0.95	0.97	4.90	0.77	1.22	1.18
Daejeon	0.52	0.72	1.05	1.06	0.28	1.23	0.96	1.18
Gangwon	0.70	0.66	1.00	0.89	0.49	0.65	1.41	1.74
Gwangju	0.66	0.44	1.06	1.10	0.24	0.30	0.98	0.97
Gyeonggi-do	0.94	1.16	1.03	0.97	1.11	1.28	0.88	0.68
Gyeongsangbuk-do	0.82	0.45	0.91	0.91	1.33	0.42	1.33	1.21
Gyeongsangnam-do	0.35	0.53	0.94	0.97	0.07	0.29	0.96	1.46
Incheon	0.85	1.05	1.02	0.97	0.51	0.87	0.98	1.01
Jeju	0.54	0.57	0.97	1.18	0.00	0.22	0.65	1.60
Jeollabuk-do	0.42	0.45	0.98	1.02	0.36	0.59	1.11	1.12
Jeollanam-do	0.27	0.27	0.93	1.00	0.16	0.19	0.88	1.65
Sejong	0.35	0.49	1.11	1.09	0.16	0.09	0.24	0.41
Seoul	1.77	1.98	1.04	1.07	1.06	1.88	0.87	0.87
Ulsan	0.53	0.50	0.99	1.00	1.35	0.35	0.76	0.57

, Figure 4 and Figure 5) showed that, in 2020 and 2021, except for Seoul, Incheon, and Gyeonggi-do, all regions had the LQ < 1 indicating a low degree of COVID-19 concentration in most of the area of the regions compared to the nation. However, in 2022 and 2023, over eight regions had the LQ > 1 indicating a high degree of COVID-19 concentration in most of the area of the regions compared to the nation.

The Moran index results for testing the spatial autocorrelation of COVID-19 in South Korea from 20 January 2020 to 31 August 2023 are shown in

Table 6. Results of the Moran index analysis.

Year	Variable	Moran I	Z Score	p-Value
2020	Cases	0.75	5.17	<0.001
	Deaths	0.48	3.21	0.0006
2021	Cases	0.89	6.26	<0.001
	Deaths	0.88	6.09	<0.001
2022	Cases	0.78	5.72	<0.001
	Deaths	0.68	4.73	<0.001
2023	Cases	0.80	5.63	<0.001
	Deaths	0.71	4.57	<0.001

. The results demonstrated a consistent and significant spatial autocorrelation for both cases and deaths across all years with Moran’s I values above 0.4 (p < 0.05). Moran’s I values for cases were higher than for deaths each year, indicating stronger clustering for cases (Table 6).

The findings of the multivariable regression analysis to identify factors associated with COVID-19 incidence and case-fatality rates are presented in (

Table 7. Linear and negative binomial regression to identify factors associated with COVID-19 cases and deaths.

Risk Factors	Cases		Deaths
	Linear Regression	Negative Binomial Regression	Linear Regression	Negative Binomial Regression
	β (95% CI), p-Value	RR (95% CI), p-Value	β (95% CI), p-Value	RR (95% CI), p-Value
Population density	0.02 (0.00–0.04), 0.03	0.02 (0.01–0.04), 0.03	0.01 (0.01–0.03), 0.47	0.02 (0.01–0.05), 0.99
Sex ratio	5.18 (4.85–11.89), 0.01	0.21(0.1–0.30), 0.01	0.03 (0.01–0.06), 0.50	0.03 (−8.40–8.74), 0.99
Percentage aged 60 years and above	0.02 (0.01–0.03), 0.03	0.03 (0.01–0.05), 0.01	0.04 (0.01–0.05), 0.59	0.04 (0.03–0.06), 0.97
Obesity prevalence (men)	8.10 (6.20–12.40), 0.21	0.05 (0.02–0.12), 0.18	0.01 (−0.03–0.05), 0.44	0.08 (−3.22–4.77), 0.96
Obesity prevalence (women)	15.78 (12.44–39.00), 0.10	0.11 (0.09–0.19), 0.89	−0.02 (−0.06–0.03, 0.20	−0.23 (−6.01–5.11), 0.90
Smoking prevalence (men)	29.02 (23.29–0.35),0.05	0.11 (0.07–0.12), 0.02	0.00 (−0.06–0.05), 0.83	−0.05 (−4.58–6.47), 0.98
Smoking prevalence (women)	106.50 (41.67–171.34), 0.02	0.79 (0.54–1.04), 0.01	0.05 (−0.07–0.18), 0.21	0.46 (−11.31–19.09), 0.92
Diabetes prevalence (Men)	42.13 (36.52–44.74) 0.02	0.27 (0.26–0.28), 0.01	0.00 (−0.05–0.06), 0.74	0.15 (−4.89–6.31), 0.94
Diabetes prevalence (Women)	74.57 (20.64–128.50), 0.03	0.51 (0.32–0.71), 0.04	0.04 (−0.06–0.15), 0.22	0.40 (−8.07–13.22), 0.91
Per capita income (Won)	113.79 (106.19–119.40), 0.04	0.65 (0.59–0.74), 0.04	0.04 (−1.54–1.62), 0.92	0.46 (−11.31–19.09), 0.92
Average temperature	17.45 (15.56–50.47), 0.15	0.09 (0.03–0.21), 0.15	0.01 (−0.05–0.08), 0.46	0.17 (−5.33–7.49), 0.94
Average rainfall	0.09 (0.05–0.14), 0.10	0.10 (0.08–0.12), 0.28	0.00 (0.00–0.00), 0.66	0.00 (−0.02–0.02), 0.97
Average humidity	1.54 (0.09–1.89), 0.56	0.02 (0.01–0.06), 0.14	0.00 (−0.02–0.02), 0.83	−0.01 (−1.97–1.95), 0.99

). There was a positive association of COVID-19 incidence rates with the population density (RR = 0.02, CI: 0.01–0.04, p = 0.03), percentage aged 60 years and above (RR = 0.03, CI: 0.01–0.05, p = 0.01), smoking prevalence (women) (RR = 0.79, CI: 0.54–1.04, p = 0.01), and diabetes prevalence (women) (RR = 0.51, CI: 0.32–0.71, p = 0.04) where higher values of these variables were associated with a higher number of detected cases. In contrast, the sex ratio (RR = 0.21, CI: 0.1–0.30), p = 0.01), smoking prevalence in men (RR = 0.11, CI: 0.07–0.12, p = 0.02), diabetes prevalence in men (RR = 0.27, CI: 0.26–0.28, p = 0.01), and per capita income (RR = 0.65, CI: 0.59–0.74, p = 0.04) were negatively associated with the COVID-19 incidence rate. No significant association was found between the risk factors and the COVID-19 case-fatality rate in the linear and negative binomial regression.

4. Discussion

To date, most studies on the spatiotemporal analysis of COVID-19 in South Korea were based on the number of cases and deaths over a short period [4,35]. This study analyzed the spatial evolutionary characteristics of COVID-19 for the entire pandemic period for the 17 regions in Korea. In addition, these results reveal the regional risk factors associated with the spatial COVID-19 spread.

The results show that the COVID-19 epidemic in South Korea was spatially clustered to different degrees between 2020 and 2023. This was similar to a previous study that also reported the spatial clustering of COVID-19 in different regions [2]. For the overall pandemic period, hotspots were observed to be in Seoul and the Gyeonggi-do area. This could be because Seoul and Gyeonggi province have high population density compared to other regions. The empirical literature demonstrates that person-to-person contact increases the risk of COVID-19 infection, and, once COVID-19 is transmitted in a region, agglomeration is more likely to occur in densely populated areas [3,36,37].

The three-year epidemic curve demonstrated that, despite the control measures implemented between 2020 and 2023, the highest number of reported cases occurred in 2022, coinciding with the rapid spread of the Delta and Omicron variants across the country. Empirical studies attribute this surge in cases to the high virulence and transmissibility of these [38].

The LQ for cases and deaths also evolved with time, from just one city in 2020 having an LQ > 1 to more than eight cities having an LQ > 1 in 2023. This evolution was similar to previous reports [39]. In addition, the Moran index results for testing the spatial autocorrelation of COVID-19 cases and deaths showed an existing significant correlation between the regions for all years. These results were similar to a previous study that showed no spatial autocorrelation of COVID-19 deaths across different periods [40]. This could be due to the timely availability of clinical support conditions for critically ill COVID-19 patients; available testing capacities and the stringent implementation of the COVID-19 measures with the establishment of various strategic interventions reportedly had a positive impact in avoiding deaths from COVID-19 in various cities [41].

Our study found a positive association between the COVID-19 incidence rate and population density, the percentage of adults aged 60 years and older, smoking prevalence in women, diabetes prevalence in women, and the sex ratio and a negative association with smoking prevalence in men, diabetes prevalence in men, and per capita income when we looked at regional level risk factors.

Several studies have concluded that population density was an important factor that influenced the spread of COVID-19. According to a study in the literature [42], for every unit increase in population density (persons per km²), a 14.5% rise in the COVID-19-infected case count could be expected. Another study also claimed that population density was a major factor in increasing the transmission of COVID-19 [43]. This finding suggested that continuous interventions (e.g., social distancing and quarantine) could be recommended in regions with high population densities even as activities return to normal. The underlying mechanism for the association with population density is related to the increased transmission of saliva, respiratory droplets, and/or aerosol between individuals when people are in close physical contact [44,45]. Because of the notable influence of population density on COVID-19, the incidence rate rather than the count should be used as the dependent variable in COVID-19-related studies.

Another important population demographic variable that significantly affected the COVID-19 incidence in South Korea was the percentage of people aged 60 years and older. This was similar to studies that showed that older adults were significantly more at risk of COVID-19 infection compared to younger individuals [46,47,48,49]. Older adults often have weaker immune systems and pre-existing conditions, which make them more susceptible to infection and prolonged viral shedding [49].

Health-related behavioral factors, such as smoking prevalence (women) and diabetes prevalence (women), were significantly positively associated with the COVID-19 incidence rate, while a significant negative association was noted with the sex ratio, smoking prevalence (men), and diabetes prevalence (men). Smoking and diabetes have been identified as one of the key risk factors associated with the COVID-19 incidence rate [50]. Our finding is consistent with a previous study that suggested a stronger risk of COVID-19 mortality in obese women than men [51]. However, we recommend further research to investigate how these predictors affect the COVID-19 incident rate.

Furthermore, we found no associations between COVID-19 case-fatality rates in regions with behavioral factors. This finding is consistent with research showing that COVID-19 deaths were not associated with any factor and contrary to previous studies suggesting that mortality due to COVID-19 was significantly higher in older people [47]. However, further studies using individual-level data are needed to confirm how age, sex, and behavioral factors and their interactions contribute to the variations in COVID-19 outcomes.

The main strength of this study was the use of reliable COVID-19 data that were freely accessible on the website of the KCDC. Using all pandemic data, our study, in detail, examined the dynamics of cases and deaths in all regions of Korea.

As limitations, our study used regional-level risk factors. Therefore, the results of this study can only suggest associations between risk factors and COVID-19 outcomes at the regional level but cannot be interpreted as associations at the individual level. In addition, COVID-19 patient-level information such as obesity, smoking, and diabetes status was unavailable. Furthermore, due to the evolution of vaccination campaigns and public health policies, the influence of certain risk factors on the pandemic may have changed over time. Government-imposed public health and social measures including social distancing, wearing masks, quarantine, and improved treatment options may have contributed to the annual differences in incidence rates. Also, changes in testing availability and reporting standards may have improved case detection, potentially affecting comparisons across different years.

5. Conclusions

COVID-19 continues to be a challenge to public health around the world as it has become endemic in most regions. Since September 2023, COVID-19 has been included in routine surveillance together with other respiratory viruses.

Our findings showed a decline in incidence rates for each region post-2022, which corresponded to the year of implementation of both non-pharmacological and pharmacological interventions of COVID-19. This suggests that vaccination campaigns, early detection, and targeted public health interventions played crucial roles in managing COVID-19 incidence rates. We recommend that the government should continue promoting booster vaccinations, particularly for vulnerable populations, while investing in real-time epidemiological surveillance to detect and contain any emerging or re-emerging variants early. On the other hand, the case fatality also showed an increase from 2020 to 2021 and a decrease in 2022 and 2023. The distribution pattern of COVID-19 cases between 2020 and 2023, obtained through a spatial approach in this research, shows different cluster patterns by year and region. The regional risk factors associated with the COVID-19 incidence in both the linear and negative binomial regressions were population density, smoking prevalence (men), smoking prevalence (women), diabetes prevalence (men), diabetes prevalence (women), and per capita income. Understanding the complete yearly regional spatiotemporal patterns and risk factors is crucial to supporting the government in its routine surveillance of the SARS-CoV-2 virus and effective policymaking for its prevention and control. We recommend that this spatial approach is absorbed into routine surveillance outputs for COVID-19.