Spatial epidemiology of COVID-19 infection through the first outbreak (March-May 2020) in the city of Mashhad, Iran

: Background: Since December 2019, SARS-CoV-2 infection has converted to a severe threat to global health. It is now considered as the fifth worldwide pandemic problem. This study aims to explore spatial-time distribution of COVID-19 in the first outbreak of COVID-19 in the second major city of Iran (Mashhad). The results will pave the way for better tracking of COVID-19. Methods: Data were collected from two tertiary hospitals in Mashhad in June 2020. They included demographic findings and residential address of the patients with confirmed COVID-19 disease by polymerase chain reaction test. The univariate logistic regression model was used to assess the influence of age and sex on mortality. For spatial-time analysis, after calculating empirical Bayesian rate for every neighborhood, the local Moran's I statistic was used to quantify spatial autocorrelation of COVID-19 frequency at the city neighborhood level. Results: Of 1,535 confirmed cases of COVID-19 included in this study, 951 (62%) were male. Odds of death for patients over 60 years of age was more than three times higher (odds ratio [OR]: 3.7, CI [2.8-4.8]) than for those under the 60 years. In addition, the ratio of relative mortality for male patients was significantly higher than the female (OR: 1.58, CI [1.2-2]). The univariate regression model also revealed that odds of death increased along with increase in duration of hospitalization secondary to COVID-19 disease (OR: 1.02, IQR [1.01-1.02]). The downtown area had a significant high-high cluster throughout much of the study period (March-May 2020). Conclusions: Collection of geographic information system (GIS) map data on SARS-CoV-2 provides insight into clusters of infection and high risk places for COVID-19 transmission. GIS-based map could potentially be used to predict future places of involvement for health systems.


Background:
The 7th member of the coronavirus family began spreading across the world in December 2019 after being identified in a seafood market in Wuhan, China (1). The World Health Organization (WHO) named the newly emerged virus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and the disease that it causes, coronavirus disease 2019  in Jan 7, 2020 (2).
The citizens who traveled from Wuhan City were leading sources of transmission of  to the rest of China (3). Restriction measures introduced at Wuhan to control the spread of COVID-19 included rapid isolation of confirmed and suspected cases, clinical management of affected patients, enhanced medical testing, closure of entertainment facilities, prohibition of public meetings, and, eventually, restriction of movement (4,5). Despite meeting these restriction measures, SARS-CoV-2 virus spread much faster than previously anticipated. Iran was one of the earlier affected countries might be due to high economic relations with China (6). The religious city of Qom was among the earlier affected cities in the country (7) which have direct air flight and high religious relations with Mashhad City.
The number of infected cases and deaths is still increasing rapidly. In developing countries, where there is high population density, accelerated urbanization, and shortage of medical services, inhibition of COVID-19 spread is difficult (8). Furthermore, an unprecedented health issue, economic challenges, and feeling of discontent and sentiments created for global population (9).
In the other hand, COVID19 spread through the body's processes 100 times faster than related viruses (10).
In 1694, the association between place and health (containment of plagues) developed in Italy (11).
The map's vital position in the monitoring of infectious diseases in pandemics was subsequently accepted (12). The development of geographical information systems (GIS) allowed further study and identification of disease patterns (13). In fact, a revolution has arisen in the area of infectious diseases, by engaging health geography with web-based resources. We assume that GIS will help the numerous countries fight more effectively against COVID-19 through finding high-risk areas in each city. GIS can play a leading role in managing the COVID-19 epidemic as seen earlier (8).
Given that, knowledge of the distribution of COVID-19 in highly endemic regions is important for provision of health care services and for planning the residence of the region. Here, we explore and examine the temporal and spatial distribution of COVID-19 in the first outbreak of COVID-19 in the second major city of Iran (Mashhad). We hope that in future these data resulting from the study of GIS applications will pave the way for better tracking of COVID-19, but also for planning health services, medications, hospital and intensive care unit (ICU) capacity, etc.

Methods:
Study Area: This study was conducted in the city of Mashhad, the capital of Khorasan-Razavi Province and the second most populous city in Iran. Mashhad, located in northeastern Iran, has a estimated metro population of 3,208,000 according to 1.78% increase from 2019 (14). Mashhad has the highest number of tourists in Iran every year because it is the religious center of Iran. Figure 1 indicates the region being investigated.

Statistical Methods:
Continuous variables were represented as median and interquartile range (IQR) and compared by the Mann-Whitney U test. Categorical variables were expressed as number (%) and compared once between males and females and once between survivors and non-survivors by χ2 test or Fisher's exact. The univariate logistic regression model was used to assess the influence of age and sex on mortality. Un-adjusted odds ratios (OR) with 95% confidence intervals (CI) were calculated and reported. All tests were two-sided, and α of less than 0.05 was considered statistically significant.
All statistical analyses were performed using R software, version 4.0.5 (R Foundation for Statistical Computing) and Microsoft Excel version 16.

Spatial Time Analysis:
The core concept behind the hierarchical Bayesian model that used in the current study is that in each area-specific incidence is based on pooling information from neighboring areas (prior distribution). Methodological aspects of the Bayesian analysis applied to geographical clustering have been reported in a previous study (15). After calculating the empirical Bayesian rate (EBR) for every neighborhood, the local Moran's I statistic (16) was used to quantify spatial autocorrelation of COVID-19 frequency at the city neighborhood level. This test calculates a zscore and p-value to determine whether the apparent similarity (spatial clustering of either high or low values) or dissimilarity (presence of spatial outliers) is more pronounced than predicted in a random distribution. The null hypothesis suggests that the COVID-19 cases are randomly distributed throughout the sample region. An extremely positive z-score for a feature indicates that the surrounding features have similar values (either high values or low values). However, a low negative z-score for a feature suggests a statistically significant spatial data outlier (16). We used a 95% confidence level (CL), and all the clusters and outliers found in this study were significant at this CL. We used ArcGIS, v. 10.5 (ESRI, Redlands, CA, USA) and GeoDa (https://spatial.uchicago.edu/geoda) for spatial analyses.

Results:
In this section, the statistical results are first reported and then the spatio-temporal results are discussed.
Statistical results: Considering 1,535 confirmed cases of COVID-19 included in this study, 951 (62%) and 584 (38%) were male and female, respectively. Table 1 shows a comparison between sex groups and in addition reveals a comparison based on patients' health outcomes. As a result, two classes of nonsurvivors (n=356, 23.2%) and survivors (n=1,179, 76.8%) were compared.

Age and sex
The median age of all patients was 62 years (IQR [47-73]). Sixteen cases (1.0%) were younger than 20 years and 842 (54.8%) of them were older than 60 years. Women were marginally older (one year on average) and they had no significant differences in terms of age-groups and length of stay (LOS), which is the span from hospital entry to death or discharge, compared to men. More men (n=951) than women (n=584) were admitted to these two tertiary hospitals, because of SARS-CoV-2 infection. In fact, among the 356 non-survivor cases of COVID-19, the ratio of male to female was 70% (n=249) to 30% (n=107).

Risk factors
By using the variables in the univariate logistic regression model, it was found that older age (≥60 years), male sex, and longer LOS were associated with increased risk of death (      The four categories shown in Figure 4 correspond to the four quadrants in the Moran scatter plots as shown in Figure 5. If nearby or neighboring areas are more identical, this is understood as positive spatial autocorrelation. Negative autocorrelation describes patterns that vary from neighboring areas. For example, the first scatter plot (which shows the neighborhood level Covid-19 for 04 -17 February) has a Moran's value of -0.017 that should be interpreted as a spatial random pattern. Note that the EBR values have been standardized and are specified in standard deviational units (the mean is zero and the standard deviation is 1). Similarly, the EBR spatial lag was computed for these standardized values ( Figure 5).

Limitations:
During data mining from the medical records, there were 25 records with false address of residence that were not included in the report. In fact, the total number of individuals registered with COVID-19 was 1560 in two hospitals. There is shame in cultures regarding infection with COVID-19 (22,23). This stigma and fear of future follow-up led a minority of people to write incorrect addresses in their medical records. Another limitation is a lack of data on underlying medical comorbidities and how this influenced mortality. In addition, in the current study only using address of residence was utilized as a risk factor for infection and death but we do not know where people have traveled within Mashhad or outsides. Hence, it is not possible to attribute acquisition of infection to the neighborhood of residence.

Conclusion:
The collection of GIS-based mapping data is essential for understanding the distribution of SARS-CoV-2 infection within a city or region. These data can be used to show fronts of infection and high risk places for transmission of COVID-19 infection. GIS based maps will predict future places of involvement for health systems. It seems this would be considered as a necessary step in the current global health challenge.

List of abbreviations:
Geographic

Consent for publication:
Not applicable

Availability of data and materials:
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests:
The authors declare that they have no competing of interest.

Funding:
The study was funded by Vice Chancellor of Research in Mashhad University of Medical Sciences (grant number 981796).

Authors' contributions:
H