Spatio-temporal Analysis of COVID-19 Hotspots in India Using Geographic Information Systems

Kanav, A.,¹ Yadav, B.,¹ Sharma, R.²and Kumar, J.^1*

¹Department of Geography, School of Basic Sciences, Central University of Haryana, India

²Department of Geography, KLP College, Rewari, India,

*Corresponding Author

Abstract

The aim of this study is to identify hotspot regions of COVID-19 in India from March 2020 to August 2023. Identifying hotspots is essential for effective pandemic management, as it helps policymakers understand the dynamics of virus spread and allows for more precise public health campaigns. The present study is a district level analysis of India at five different points in time, where we calculate the cumulative incidence rate (CIR), cumulative fatality rate (CFR) and recovery rate (RR) for COVID-19. Further, we apply Global Moran's I, Getis-Ord Gi* and Anselin local Moran's I index to identify COVID-19 hotspots by using Geographic Information Systems (GIS) technology. The results show that the spatial and temporal variation of the CIR is very high across India. The CIR was recorded lower in May 2020 as the affected people were immobilized due to the lockdown. However, the CFR was high and RR was low due to inadequate medical facilities and treatment. The findings revealed that mainly two hotspot regions existed in India until May 2021, the National Capital Region, Haryana, Punjab, Rajasthan, Uttar Pradesh and Maharashtra in the south. However, this scenario has entirely changed since January 2022, when northern India has changed into a cold-spot and the southern coastal states have become the pandemic hot-spot region. Combining hot-spot analysis with Global & Anselin local Moran's I offers a precise method for locating statistically significant COVID-19 case cluster areas and identifying high-risk areas.

Keywords: Anselin Moran's I, COVID-19, Getis-Ord Gi*, India, Spatial Autocorrelation

1. Introduction

COVID-19 has a vast geographic reach and has affected millions of people globally, leading to numerous fatalities [1]. Designated a pandemic by the WHO on March 11, 2020 [2], it prompted a global call for unity to safeguard the vulnerable populace. India, ranking second in total infected persons after the United States, faced challenges due to factors such as social distancing, demographics, and health infrastructure [3]. Additionally, the government's approach to mitigation strategies played a crucial role in influencing the spread of virus [4]. Even amidst lockdowns and closures, millions of people traveled back to their hometowns, navigating the challenges posed by the pandemic.

In August 2023, the global tally for COVID-19 infections stood at over 769 million, with a sobering count of more than 6.95 million fatalities. Throughout the course of the epidemic, the media outlets and online platforms have broadcast various narratives to depict the crisis in diverse ways [5]. In this discourse, the discipline of geography emerges as a vital player in understanding the spread of the pandemic. According to Bissell [6], geographical concepts regarding location and movement empower us politically to identify new forms of injustice and suffering, prompting reflections on how we collectively shape the future of a place. Furthermore, Schueller et al., [7] established connections between place, mobility, and the geopolitical dimensions of the COVID-19 pandemic. The application of GIS and spatial statistical techniques has proven instrumental in comprehending the spatial distribution and pinpointing the risk zones of the disease. Concurrently, the pandemic has prompted geographers to reassess their relationship with place and space, giving rise to the creation and propagation of knowledge [1].

Different methods, including autocorrelation, hotspot analysis, space-time scan and clustering analysis have been employed to illustrate the areal distribution of the COVID-19 pandemic.

Through these spatial statistical techniques, researchers gain insights into the spatial patterns, enabling planners to devise effective policies for minimizing the risk of the COVID-19 pandemic. The widely used Getis-Ord method serves as a valuable tool for portraying the spatial distribution of infectious diseases like hand-foot-and-mouth disease and dengue [8] and [9]. It proves effective in identifying clusters of high and low concentrations in geographical areas and has been utilized in simulating outbreaks of various diseases [10] and [11]. Numerous studies have delved into identifying areas affected by COVID-19 at global, regional, and local levels. The Getis-Ord method was recently widely used in hot spot detection of COVID-19 at the global level like in USA [12], China [13] and [14], Kazakhstan [15], Brazil [16], Iran [17], Bangladesh [18], Malatya Province, Turkey [19], Indonesia [20] and Finland [21].

In India, a number of studies spanning across different disciplines of epidemiology, public health and social sciences have been conducted to understand the spread of COVID-19 virus and its impact on people. A comprehensive study in the states of Tamil Nadu and Arunachal Pradesh [22] found that children and adults in resource-limited settings may transmit COVID-19 more readily to same-age individuals. Kumar [23] proposed a network-based model to predict the spread of COVID-19, considering human mobility patterns through migration and air travel data. The model could be used to estimate resource needs and identify high-risk routes. Some studies in India concerning the epidemic trend of COVID-19 transmission have been regarding the prediction and analysis of COVID-19 positive cases using deep learning [24] and data-driven models [25]. Acharya and Porwal [26] prepared a district-level vulnerability index for prioritizing resource allocation and increasing management efficiency across the country.

Other studies pertaining to the pandemic included studying the effect of lockdown on air quality [27] and an assessment of the short-term decline in anthropogenic emission of CO2 in India [28]. Country-wide lockdown as a product of COVID-19 was also deemed as a necessary evil by Shehzad et al., [29] as it proved to be beneficial in mitigating air pollution for India and surrounding countries. One study also found that people residing in areas riddled with air pollution were more susceptible to respiratory infections [30]. Scholarly interest was also shown in studies of migrants and livelihoods. As the lockdowns in India caused widespread economic hardship, 80% of households experienced reduced food intake, with 60% lacking essential funds and 30% resorting to loans [31]. The stranded migrant laborers had to suffer from a loss of livelihood and resultant unsafe living conditions and an increased risk of COVID-19 infection [32]. The study also found that most migrant laborers failed to benefit from government assistance. Arora et al. also found that there was a reduced willingness and prevalence of fear among male migrants to resume work [33].

Since gaining independence, India has grappled with various health challenges, including cholera, dengue, measles, smallpox, and recently, COVID-19. In response to this historical backdrop, the Government of India crafted national health policies in 1983, 2002, and 2017 [34]. The most recent policy, established in 2017, aimed at achieving universal healthcare and "health for all" [35]. The current COVID-19 pandemic has posed a formidable threat in terms of infections and fatalities in India. The existing healthcare policies, shaped by past experiences, have proven to be insufficient in addressing the unique challenges posed by COVID-19. Recognizing the need for additional measures, the Government of India implemented a sweeping lockdown to curb community transmission. The first lockdown was initiated on 24th March 2020, for 21 days, including restrictions on travel and social gatherings in both public and private spaces. Given the gravity of the situation, the lockdown was extended until 31st May 2020, across three subsequent phases. The affected areas were categorically classified into green, orange, and red zones based on the severity of the outbreak.

The swift implementation of a stringent lockdown had an instant impact, effectively slowing the spread of the virus in India. This proactive measure provided valuable time to prepare critical medical infrastructure, strengthen human resources and leverage technological advances. Despite the relatively low severity of the COVID-19 pandemic in India compared to many other developing countries, the crisis has exacerbated pre-existing economic risks [36]. Researchers expect that lockdown will have an adverse and lasting effect on the country's economy [37]. The underfunded public healthcare system has posed challenges to the country's pandemic management strategy, highlighting the need for comprehensive measures in the face of such global health emergencies. The Government of India established a National Task Force (NTF) for COVID-19, with a primary focus on initiating prompt pandemic research and investigations.

Recognizing the significance of technological innovations in policymaking, the government engaged researchers at all levels to mitigate the impact of the COVID-19 pandemic by launching funding through Department of Science and Technology [38]. The Prime Minister's Office (PMO) took the lead in establishing the 'Prime Minister's Citizen Assistance and Relief in Emergency Situations Fund' (PM CARES Fund) to address the challenges posed by the COVID-19 pandemic. PM-CARES was specifically designed to encourage small donations from people, showcasing the strength of collective participation in alleviating emergency situations. The Prime Minister also encouraged businesses and higher-income community to shoulder the economic needs of individuals from lower-income groups. In doing so, he urged them to refrain from reducing the salaries of those unable to render services due to their inability to come to the workplace on certain days. This call emphasized a sense of shared responsibility and solidarity during these challenging times.

2. Objective

The present study analyses the COVID-19 trends, cumulative incidence rate (CIR), cumulative fatality rate (CFR), recovery rate (RR) of the states and union territories of India (Figure 1). The study also evaluates the spatial distribution and spatial clustering patterns of confirmed and deceased COVID-19 cases at different five periods at the district level using Global Moran's I, Getis-Ord Gi* and Anselin local Moran's I index method to investigate the high-risk and low-risk clusters.

3. Materials and Methods

3.1 Data Extraction

The study covers the period May 31, 2020 to August 02, 2023 for based on Government of India data source available at www.incovid19.org, that is updated on daily confirmed, deceased, and recovered cases of COVID-19 in each state and union territory at the district level as depicts in Figure 2. The data was cross-checked with state government data from different states in India, where necessary.

3.2 Data Analysis/Methods

Four peaks of COVID-19 have been identified in Figure 1 based on confirmed. In addition to the peaks, we also analyze the latest situation, hence covering five different time-periods (Table 1). Cumulative incidence rate (CIR), cumulative fatality rate (CFR) and recovery rate (RR) are defined by equations 1 to 3:

Equation 1

Equation 2

Equation 3

3.3 Spatial Analysis

3.3.1 Global Moran's I (Spatial Autocorrelation)

Spatial autocorrelation is a statistical technique that can be used to unveil the spatial patterns of phenomena, such as the occurrence of COVID-19. Global Moran's I provides a single value for the entire dataset, measuring the overall degree of spatial clustering [39]. The analysis offers both graphical and numerical outputs. The graphical output visually represents the distribution pattern of the data, showcasing whether it's scattered, clustered, or follows a random pattern. On the numerical side, the analysis provides five key values: Moran's I Index, Expected Index, P-value, Z-score, and Variance. The I Index is calculated using equations 4 to 6:

Equation 4

Equation 5

Table 1: Hotspot identification and mapping justification

Figure 1: Map of India

Figure 2: Daily total number of confirmed cases of COVID-19 in India from January 30, 2020 to July 31, 2023

Equation 6

Where:

I is the Moran's I.

s_o is a normalization factor.

w_{i, j} is the spatial weights between locations i and j.

z_i and z_j are standardized values of the variable of interest at locations i and j.

E(I) is the expected value of Moran's I.

V(I) is the variance of Moran's I

Global Moran's I index values fall within the range of -1.0 to +1.0; a positive value (>0) indicates clustering i.e. similar features are clustered together; a negative value (<0) indicates a random distribution; and a value of zero (=0) indicates a dispersed pattern. This can be crucial in identifying high and low-risk clusters for further analysis, providing valuable insights into the spatial patterns of the data [40]. As an inferential statistic, the conclusions drawn from the analysis are contingent on the assumption being tested and any observed patterns or relationships are considered in light of the null hypothesis [41]. The null hypothesis in our study states that COVID-19 is distributed randomly across the country. Thus, Global Moran's I can be used to confirm the overall trend of COVID-19 spread. This initial assessment provides a broader picture of potential hotspots.

3.3.2 Getis-Ord Gi* (Hotspot analysis)

A hotspot is a region with a higher incidence concentration than a random distribution. Getis-Ord Gi* method was performed for hot spot and cold spot analysis of COVID-19 in India using Arc GIS 10.8 software. Hot spot analysis employs the Getis-Ord Gi* statistic to evaluate each feature in a dataset, identifying statistically significant spatial clusters. This technique assesses each feature by considering its proximity to neighboring features [42]. Z-scores and p-values serve as indicators of where these clusters of high and low values reside. While a feature with a high z-score may initially attract attention, it may not hold statistical significance given its surrounding features. To be deemed a statistically significant hotspot, a feature must not only exhibit a high score but also be surrounded by other features with similarly high scores. A statistically significant z-score arises when the sum of a feature and its adjacent features deviates considerably from the expected sum.

This deviation is substantial enough to rule out the possibility of a random chance. In the case of positive z-scores, higher values indicate more intense clusters, referred to as hot spots. On the contrary, negative z-scores with smaller values represent a cold spot i.e. clustering of low values [42].

To perform hotspot analysis, confirmed cases and death cases data from 31 May, 2020 to 02 August, 2023 were used. For each feature, the resulting z-score and p-value reveal where high and low values cluster. High positive values indicate hotspots, while high negative values indicate cold spots. High positive and negative z scores indicate more and less clustering, respectively. The Getis-ord local statistic is given by equation 7:

Equation 7

Where x_iis the attribute value for feature j, w_i,j is the spatial weight between feature i and j, n is equal to the total number of features. and S are defined in equations 8 and 9:

Equation 8

Equation 9

The G_i* statistic is a Z score so no further calculations are required.

3.3.3 Anselin local Moran's I (Spatial Cluster Outlier)

By employing Anselin Local Moran's I, researchers can visualize clusters and outliers of COVID-19 cases on a map [43]. This method assigns a z-score and p-value to each feature, indicating the statistical significance of its spatial clustering. A positive z-score signifies that a feature is surrounded by similar features (e.g., high COVID-19 incidence areas clustered together), while a negative z-score indicates that a feature is surrounded by dissimilar features (e.g., high COVID-19 incidence areas surrounded by low incidence areas).

The results are typically classified into four categories:

• High-High Clusters: These areas are characterized by a high concentration of COVID-19 cases.
• Low-Low Clusters: These areas exhibit a low incidence of COVID-19 cases.
• High-Low Outliers: These areas have a surprisingly high incidence of COVID-19 cases amidst a generally low-incidence region.
• Low-High Outliers: These areas have a surprisingly low incidence of COVID-19 cases amidst a generally high-incidence region.

Table 2: Cumulative Incident Rate (CIR)

Figure 7: Anselin local Moran's I analysis of deceased cases

5. Discussion

This study provides an evaluation of spatial distribution of COVID-19 in India from the year 2020 to 2023, at five specific peaks. The Global Moran's I analysis confirmed the existence of clustering in India, although varying between different points in time. After the verification of the presence of clusters in COVID-19 confirmed and deceased cases, the study proceeded to ascertain the spatial positioning of these clusters. Getis-Ord Gi* statistic proved to be useful in computing the location of coldspots and hotspots of COVID-19 confirmed and deceased cases. At the beginning of the pandemic, north and south-western India was a major hotspot. By mid-2023 north India was majorly a coldspot whereas south India remained a hotspot, extending to southern states of Kerela and Tamil Nadu. It was also established that Anselin's Local Moran's I is helpful in ascertaining the nature of spatial clustering.

COVID-19 spread rapidly throughout India soon after its initial identification in Wuhan (China) in late December, 2019. Shortly thereafter, India recorded its first case in January, 2020 from Kerala and by May, 2020 hotspot had formed over the southern states. Eventually, the hotspot of COVID-19 also emerged near the northern capital region. Throughout the course of the pandemic in India, the north eastern states remained cold spots with LL clustering.

This observation can be attributed to the geographical isolation of this region, once the lockdown restrictions were executed. The region is also marked by lower population densities, which can prevent the easy transmission of the virus.

The maps of analyses showed hotspots and clustering which kept on changing with the passage of time. This evolving distribution pattern can be attributed to different reasons like the differences in risk factors of different areas, ability of each state to mitigate the propagation of the virus, and effective execution of lockdown measures. In addition to the social distancing norms and restriction on travel imposed by the central government, some states adopted their own safety measures. Kerela adopted a decentralized approach, involving local communities in COVID-19 control efforts. A network of dedicated COVID-19 hospitals was established, in addition to quarantine measures and contact tracing. Maharashtra, the state with high COVID-19 activity acts accordingly and increased the health infrastructure capacity, including beds, ventila-tors, and medical personnel. ‘eSanjeevani' initiative for telemedicine consultations was also launched. Madhya Pradesh initiated a ‘Roko-Toko' campaign meaning ‘stop & alert' to promote the wearing of face masks among people.