Introduction
In December 2019, a novel coronavirus, the Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2), was diagnosed in China. SARS-CoV-2 is the case of Coronavirus disease-2019 (COVID-19) that quickly spread to all countries around the world turning into a massive worldwide health concern [
1,
2]. This disease commonly causes severe respiratory deficiency resulting in increased admissions into Intensive Care Units (ICUs) with high mortality rates [
3,
4].
Since the outbreak of COVID-19, research has been conducted with unparalleled speed to investigate the clinical and epidemiological features of the disease. Suleyman et al. [
5] conducted a study in metropolitan Detroit in the U.S. that included 463 patients with COVID-19 showing that 94% of the patients had at least one comorbidity and their most common symptoms were cough, fever and dyspnoea, respectively. Age above 60 years, male sex, pronounced obesity and chronic renal disease were soon seen as strongly associated with ICU admission and an above-average Case Fatality Rate (CFR) [
5]. In Wuhan, a retrospective study performed by Zhou et al. [
6] involving 191 hospitalized patients, who had been discharged or died by 31 Jan 2020, showed that 50% of them had co-morbidities; the death rate was about 28% (
n = 54). Being old and having a higher sequential organ failure assessment score (SOFA) was significantly associated with increased CFR [
6]. Another retrospective study in Wuhan, conducted on the 102 COVID-19 cases hospitalized between 31 Jan and 5 Mar 2020, reported a mortality level of 15%, with most fatalities being old [
7]. Docherty et al. conducted a cohort study in the UK involving 20,133 hospital in-patients between 6 Feb and 19 Apr 2020, which showed that 60% of those admitted were men and 77% had at least one co-morbidity, while 26% of the patients (
n = 5165) eventually died [
8].
Iranian studies report similar results, e.g., a retrospective study in Tehran [
9] presented the infection rate as twice as common in men as in women, a CFR of 8%, and a co-morbidity rate of 15.9% in those above 60 years. In another study conducted in Shiraz [
10], 63% of the patients were male, with fatigue, cough and fever as the most common symptoms. Here, 8% overall mortality and a significant association between ICU admission and death rate, were reported [
10]. Although previous studies have shown that most COVID-19 cases have a promising prognosis [
11], patients with underlying chronic co-morbidities would experience more serious health consequences, including a substantial CFR [
12].
This new member of the coronavirus family is highly contagious. More than one and half a year after it first emerged, as of June 10, 2021, the number of confirmed cases has reached 175 million with close to 4 million deaths globally. Of these, close to 3 million cases and more than 81,000 deaths are Iran’s share and it had the highest number of daily death caused by COVID-19 in November 2020 (> 450) [
13]. In order to generate the best response strategies, analysing and representing the spatial and temporal spread of the virus is crucial for epidemiologists and health policymakers [
14]. Health organizations, especially World Health Organization (WHO), have increasingly applied spatial analysis to represent and control disease outbreaks. Geographic Information Systems (GISs) have shown successful results concerning contagious diseases, and applications are been shown highly useful in mapping geographical distribution of disease prevalence as well as visualising transmission trends and modelling spatial environmental aspects of disease occurrence [
15‐
18]. Accordingly, the GIS toolbox is highly useful for decision-making, as well as understanding the spatiotemporal dynamics and control of COVID-19 [
19,
20].
Most spatial analyses of the COVID-19 outbreak have been conducted in China. For example, Tang et al. [
21] reviewed the daily data flow of new cases and identified hotspots in the areas where the virus originated in the study period. Furthermore, Fan et al. [
22] showed that the spatial distribution pattern of confirmed COVID-19 cases followed a particular, geographical pattern. They found hotspots to be mainly restricted to the outbreak areas, especially in densely populated areas. In another study conducted in Hubei Province of China, researchers found significant spatial autocorrelation and clustering at the local level in the study area [
23]. In India, spatiotemporal analysis of confirmed COVID-19 cases at the provincial level showed significant differences in disease incidence across the Indian provinces. As well, the potential capacity of a heavy COVID-19 outbreak in India in the future, was predicted by this study [
24]. In contrast, a comprehensive GIS-based study in Catalonia, Spain showed a random distribution without a clear spatial pattern and any local autocorrelation based on Global Moran’s
I [
25].
In Iran, the work of Mazar et al. [
26], in one of the first studies on the spatiotemporal distribution of COVID-19 outbreaks, examined the effects of travel on the spatial distribution of the infection. Their research focused on identifying areas with high prevalence rates, especially in the provinces where the virus originated. From the GIS-based maps, it was clear that the disease spread from the north-central provinces Tehran and Qom, known as the administrative and religious centres of the country, respectively, which both are important foci for travellers [
26].
In the context of spatial analysis of the COVID-19 outbreak, most previous studies have been conducted at the macro-level (world, country or province), generally without addressing the epidemiological features of the pandemic. Here, we aimed to perform a spatiotemporal analysis of new infections in a big city. To the best of our knowledge, no such study has been done in Mashhad, the second-most populous city in Iran. Besides, this study was conducted while the study area was in a vigilant status with a high epidemic alert. Our approach can be used as a basis for future spatial modelling of the disease and provide valuable knowledge for preventive measures in the study area or other similar metropolitan areas.
Discussion
To the best of our knowledge, this is the first retrospective study conducted in Mashhad City with a simultaneous focus on statistical and spatiotemporal analysis. Our results show a mortality rate of 17.7% and the median age of patients was higher than that found in Chinese and Australian studies [
37,
71], but lower compared to American and European reports [
8,
38]. In accordance with most other studies [
8,
39‐
41], infected women were found to be slightly older than men in our study (2 years on average). We also found the most frequent COVID-19 infections (36.6%) occurring in the 51–60 and 61–70 age groups (Fig.
3), findings closely in line with studies conducted in Italy and China [
42,
43]. However, in contrast to studies in other countries [
8,
38,
44], the frequency of infection in younger age groups (31–40 and 41–50) in our study was considerable (25.2%), a fact that could be due to the lower median age of the Iranian population, which means that a larger proportion of younger people in Iran compared to many other countries are exposed in their daily activities [
45]. Following the studies comparing the characteristics of survivors and non-survivors of COVID-19 patients [
6,
7,
46,
47], we also found those who did not survive were on average much older than those who did. The CFR in elderly patients (age ≥ 60) investigated by OR was almost five times more than the younger patients, indicating that senility might be a risk factor for mortality. This is hardly surprising as older patients are more vulnerable to serious complications due to a less active immune system and, particularly because they commonly have various underlying diseases. In order to predict poor prognosis in these patients, Ma et al.’s [
48] suggested frailty assessment in the early stage of disease, which could help to identify potentially severe pneumonia.
The findings related to pre-existing medical conditions are in line with Emami et al.’s study [
45], which reviewed the prevalence of underlying diseases in COVID-19 cases. As seen in Table
2, all investigated comorbidities, except liver disorders, were significantly higher more common in non-survivors compared to those who survived (
p < 0.05). Although most studies [
6,
46,
47] show a higher prevalence of underlying diseases in the deceased patients, all did not find any significant differences regarding the prevalence of co-morbidities in non-survivors compared to survivors [
7,
10]. However, although there is a high variability of the prevalence rates for CVD (7.5–40.0%) and diabetes (8.0–38.0%), it is well-known that these maladies are the most prevalent pre-existing medical conditions among COVID-19 patients [
5,
6,
8,
10,
11,
39,
43,
46,
49]. As shown on the upper side of Fig.
2b, CVD was unaccompanied with any other co-morbidity in nearly 15% of those who died in Mashhad, which indicates that CVD is a serious condition that potentially can lead to the poor prognosis of COVID-19 [
50,
51]. The suggestion that can be made considering this high risk for CVD patients is the use of tele-rehabilitation, which has been well proven in previous studies [
52,
53]. In agreement with our results, a nationwide analysis conducted in England has shown that diabetes (both types) is associated with a significantly increased CFR by COVID-19 [
54]. Our results revealed that patients with malignancies also have a much higher risk, which increases the CFR almost four times. Immunosuppressive therapies and intrinsic frailty are possibilities that potentially put this population at risk [
46].
According to the results of a meta-analysis [
55] and the report from the Centres for Disease Control and Prevention (CDC) in the US [
56], fever, cough, dyspnoea and fatigue are the most commonly manifested symptoms in COVID-19 patients. Regardless of the order, our findings agree with this. We found a significant difference in the frequency of main manifested symptoms between the deceased and the survivors (Table
2), this difference was not significant in other studies [
6,
47]. Fever has been reported as the most common symptom in most studies [
5,
6,
11,
12,
39,
57], while others, similar to our results (70%), report dyspnoea as the most common symptom [
47,
58,
59]. As shown, the combination of dyspnoea-cough-fever was the most commonly manifested symptom among the patients who did not survive (23.5%) (Fig.
2a). Since 82% of deceased patients had dyspnoea as one of the manifested symptoms and also 16.5% of them reported it as the only symptom of disease, it can be claimed that suspicious cases presenting with this symptom should be taken seriously even if it is the sole symptom.
In agreement with the results of other studies [
6,
7,
9‐
11,
46], our results showed that COVID-19 infects males more often than females (Table
1). However, contrary to our finding that the CFR in the male sex was not significantly higher than that in the female (OR: 1.05, CI [0.88–1.25], a majority of studies maintain that sex may influence mortality due to COVID-19 [
5,
8,
9]. This is an important finding that must be further studied.
In contrast to reports in influential journals [
8,
38] claiming that mechanical respiration results in very high CFRs, our study found only a 38% (209/549) mortality in patients receiving mechanical ventilation due to COVID-19 induced shortness of breath. However, we found admission to an ICU as a risk factor that increased the CFR up to three times. The median duration of symptoms manifestation before admission was 2 days, i.e. slightly lower than that of English patients (4 days, IQR [1-8]) and much lower than that of Chinese patients (11 days, IQR [8–14]) [
6,
8]. The 5-day median LOS in our study agrees quite well with that by Richardson et al. [
38] (4 days, IQR [2-7]), but it was less than half of that in Bhatraju et al.’s study [
58] (12 days, IQR [8-18]). It is noteworthy that, based on the results of regression analysis, increased LOS values are associated with higher CFRs. The relatively low LOS rate in our study may be due to the sudden increase of infected patients leading to a shortage of medical equipment and active hospital beds at the early stages of the COVID-19 pandemic in Iran.
Lockdown of cities can be a great solution to control and prevent the spread of the disease [
60‐
62]. In Iran, the lack of public awareness promoted the rapid spread of COVID-19 and the rate of hospital admission and mortality increased dramatically. However, by raising public awareness about the seriousness of the COVID-19 threat, imposing travel restrictions and closing schools and universities as mentioned by other studies [
63,
64] as an effective factor in reducing the burden of COVID-19, the first peak of the outbreak was eventually largely controlled (Fig.
4). But perhaps the most important factor in this control was the annual Iranian Nowruz two-week holiday. Starting on March 20, it turned out to be a useful excuse for a national lockdown, e.g., already on March 14, 1 week before the holiday, the largest commercial and pilgrimage centre in the CBD of the city was closed due to the outbreak of COVID-19. This shows that a timely urban policy can be very effective. But due to the economic fragility of the Iranian community, the closure of the central parts of the cities cannot be sustained for the longer term. Due to the complex nature of the COVID-19, short-term temporal and spatial policies can fail. Hence, more efficient space-time policies are needed.
Using hotspot analysis, we identified statistically significant transmission areas in terms of the spatial autocorrelation of COVID-19 incidence and deaths in the two prominent areas of the city characterized by high traffic and interchange, i.e. the central part of the city or CBD and the industrial area in north-western Mashhad. The number and density of pilgrimage, commercial, and tourist services centres (as well as hotels and inns) are high in the former, while the latter consists of industrial areas surrounded by rural areas from where large numbers of people commute to work. In addition, as concluded in Mazar et al.’s study [
26], in the frequently visited cities where travel back and forth is high and permanent, the prevalence of the COVID-19 is also high. Figure
1 reveals that the density of infected people living in areas close to metro lines was much higher than the rest of the city. Accordingly, as proven in the previous studies [
65‐
68], in congested places such as business centres and public transportation, a quicker spread of the virus is expected. In agreement with the findings of Pourghasemi et al. [
69], it can be concluded that the areas considered public spaces of the city (such as subway stations, parks, commercial areas, and pilgrimage centres) are high-risk areas since adherence to health protocols there has been weaker than in the residential areas. However, we should not forget that the marginal areas (e.g. north-western areas) of the city have less potential accessibility compared with central urban, due to the high density of hospitals and other health care centres [
15,
70,
72]. Most of these neighbourhoods are low-income parts of the city and they are very vulnerable to the pandemic.
The downtown of Mashhad is no exception from other areas with a high prevalence of COVID-19 cases, but in line with the points made by Pourghasemi et al. [
69], it presents a multitude of places where high transmission of the virus can occur. Because the COVID-19 infection did not necessarily form hotspots only where the population density is high in our study, the results of the current study are neither are in line with the findings of Tang et al. [
21], who showed the hotspots where the virus originated in China nor do they correspond to the findings of Fan et al. [
22], who found the hotspots to be mainly restricted to densely populated or developed areas. Accordingly, we instead found it useful to follow other patterns and variables in each district of the city, such as the household’s socioeconomic status, urban built environment and air quality index. As the spatiotemporal maps show (Fig.
6), we found the spatiotemporal pattern of the COVID-19 to be very dynamic and unpredictable, which is in agreement with the findings of Arauzo-Carod [
25]. Indeed, in agreement with other spatial studies [
23,
24] on COVID-19 distribution and trends, we found that the spatial patterns of the infection in the city of Mashhad were clustered rather than random. Identifying such clusters over time would help urban health planners to implement tailored lockdown strategies. For example, HH and HL areas should be lock downed first (Fig.
7).
Limitations
Only 33% of all included cases were approved by RT-PCR testing and the rest of the cases were clinically confirmed. This is because of the retrospective study design and including a large number of COVID-19 cases. Interestingly, however, the obtained results were in line with other studies. The presence of such data can make the results more generalizable. As previously shown in other studies, hypertension affects the prognosis of the COVID-19 patients negatively, but we had no separate data about it and it was therefore collected under the heading CVD. As well as, the absence of data on patients who were still hospitalized at the time of reporting may have biased the findings. In order to address these types of data deficiencies in the future, using structured forms for clinical data gathering and also developing registry programs are highly recommended [
73]. Due to the short period (3 months) covered, the advantages of spatial analysis cannot tell us more now. The spatial and temporal dynamics of the disease need to be studied over a longer period in order to provide more effective solutions. However, due to our discovery of hotspots and coldspots, city policymakers can concentrate their solutions in the central, eastern, and north-eastern parts of the city. In particular, central sectors need more sustainable solutions. Finally, it was possible that some patients had not gone to the hospitals, thus we did not have their data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.