Georeferencing of COVID-19 Positive Nasopharyngeal Swabs to Support Emergency Management in an Area of Northern Italy

Abstract

Spatial distribution heterogeneity of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been observed in several countries. While previous studies have covered vast geographic areas, detailed analyses on smaller territories are not available to date. The aim of our study was to understand the spatial spread of SARS-CoV-2 in a province of Northern Italy through the analysis of positive nasopharyngeal (NP) swabs. The study was conducted on subjects who lived in the province of Alessandria with at least one positive NP swab between 2 March and 22 December 2020. To investigate if clustering occurred, the proportion of SARS-CoV-2 positive subjects over the total number of residents in each small administrative subregion was calculated and then mapped. A total of 17,260 subjects with at least one positive NP swab were included; the median age was 54 years (Interquartile range 38–72) and 54.9% (n = 9478) of our study population were female. Among the 192 towns scanned, 26 showed a prevalence between 5% and 7.5%, one between 7.5% and 10% and two with more than 10% positive swabs. The territories with a higher prevalence of positive subjects were located in areas with at least one nursing home and potential clusters were observed within these structures. The maps produced may be considered a useful and important monitoring system to identify areas with a significant and relevant diffusion of SARS-CoV-2.

1. Introduction

21 February 2020 will be remembered in the history of this pandemic as marking the first Italian case of COVID-19 that had not been traced back to contact with China [1]. This date is important to frame and define in a timely manner the beginning of the SARS-CoV-2 epidemic in our country; indeed, since that moment the virus spread quickly from northern Italy (particularly from Lombardy and Piedmont) to other regions with a significant and relevant geographical/territorial heterogeneity. Geographical and territorial differences in incidence and mortality rates have been associated with different factors, such as the arrival time of the virus in specific areas, population age structure, urban development, population density, health system amenities and anti-contagion policies, and strategies adopted by local authorities [2].

In June 2020, the Italian National Institute of Statistics (ISTAT), in collaboration with the Italian National Institute of Health (ISS), published preliminary data on the daily number of deaths in Italy during the first wave of the pandemic, stratified by any cause of death (including SARS-CoV-2 infection). The Piedmont region showed standardized age mortality rates related to the virus that were slightly higher than national rates (39.8 and 39.2 per 100,000, respectively). In particular, the highest mortality rate was found in the province of Alessandria (72.1 × 100,000), the third in Piedmont for case incidence after Cuneo and Turin [3], and among the top ten provinces by incidence rate of COVID-19 in Italy during the first phases of the COVID-19 epidemic [4].

To our knowledge, previous studies evaluating the spatial spread of the pandemic have analyzed vast geographic areas and generally considered entire nations [2,5,6,7]. To be more specific, in Italy, the “Civil Protection Department”, the national institution that predicts, prevents and manages emergency events, provides data updated daily about the number of positive nasopharyngeal (NP) swabs and deaths, stratified by region/area. These data, derived from administrative health databases, are useful to describe and compare the spread of SARS-CoV-2 and its impact among regions/provinces as well as to draw national analytic and synthetic maps [8]. Despite this, a focus on small administrative entities was less frequent. Among the Italian regions that use the data of positive NP swabs in order to generate maps at a local level, considering the single town as an administrative unit, it is possible to include Piedmont. However, these data are not aggregated for time periods; thus, it is not possible to ascertain and understand temporal-spatial trends. In addition, no information on the presence of nursing homes or hospitals in the areas taken into consideration is provided and assured. Finally, only data without interpretation are reported.

Consequently, it is necessary to identify and describe, through the geospatial statistics, the territories with a major risk of contagion/diffusion in order to give an adequate answer to the question of which are the most vulnerable neighborhoods or communities. This information can suggest where there is a need for greater monitoring, and it helps administrations to allocate resources across the territory. Particularly, in Italy, the COVID-19 sanitary emergency is managed at different administrative levels, not only national and regional but also local, through the local health authorities (LHAs). A LHA is an administrative territorial entity of the National Health Service in Italy introduced by the National Health Service Reorganisation Act 1992. LHAs in Italy represent government departments that are responsible for health issues and health services, playing a central and pivotal role in providing essential public health services to communities. Present in each LHA is a Department of Prevention, competent for a specific territory, that coordinates and supervises interventions for the safeguard of public health, of the environment and of life conditions and the welfare of people and animals. In addition, this entity enacts all the measures needed to control the spread of pandemic/endemic/epidemic infections.

In this context, health geography plays a supportive role, highlighting the interactions of public health officials, affected subjects and first responders to improve estimations of disease propagation and the probability of new outbreaks [9].

One of the most frequently used exploratory data analysis and geo-visualization tools in the fields of public health and epidemiology, that constantly use the interactions mentioned above, is represented by choropleth maps. During the first waves of the pandemic, there was a significant and relevant production of web-based maps and online mapping platforms that are available on interactive webpages or applications globally. These tools showed several peculiar characteristics of the global diffusion of COVID-19. In addition, these maps were helpful to explore the geographical distribution of case sites, divide geographical areas and color them in relation to health outcomes, such as the number of subjects with at least one positive NP swab.

As mapping absolute counts can be misleading because spatial units vary in area and number of observations, one approach is standardization by calculating the proportion of subjects testing positive for COVID-19 [10]. The primary aim of our geospatial study was to ascertain and understand extensively the geographical spread of SARS-CoV-2 through the detailed analysis of positive NP swabs from the LHA area of Alessandria. In addition, our secondary purposes were to establish comparisons between the first and the second pandemic wave in Italy and to identify areas/territories more heavily affected by the diffusion of this respiratory virus,

2. Materials and Methods

This study focused on the LHA of the Province of Alessandria (Piedmont Region, Italy) and the surrounding area, including 192 towns and municipalities of different sizes (including Moncalvo and Trino Vercellese). The population on 31 December 2019 was 428,535. A map of northern Italy with a focus on Alessandria province is shown in Figure 1. The names of the towns included in the spatial analysis are reported in the Appendix A (Figure A1).

We considered subjects who lived in the area with at least one positive NP molecular or antigen-test swab between the 2 March to the 22 December 2020. For each subject included in the analysis, demographic data, such as gender and age, along with residence address and hospitalization information, were available. The data from the swabs’ results were recorded in the Piedmont Regional platform to manage quarantine and contact tracing of COVID-19 positive subjects while the hospitalization data were reported in the hospital discharge records of the LHA. Linkage between these databases was performed using an anonymous form. Additionally, the locations of nursing homes were obtained to evaluate the possible impact of residential structure on the spread of SARS-CoV-2.

Finally, to standardize the absolute number of subjects and to evaluate the proportion of subjects with at least one positive NP swab, the resident population in each administrative area was obtained from the ISTAT records [11].

Statistical analysis:

Descriptive statistics of the subjects included in the analysis were presented considering the entire time period and then separately for the two SARS-CoV-2 waves: March–June and July–December, 2020. Normally distributed data were reported as mean and standard deviation (SD), whereas data following a non-normal distribution were indicated as median and interquartile range (IQR). The categorical variables were summarized as numbers and percentages.

The individual address records and the position of nursing homes were geocoded as Universal Transverse Mercator (UTM) geographic coordinates, and latitude and longitude coordinates were recorded in order to perform and conduct territorial analysis. Moreover, the geographical coordinates of Alessandria and its LHA boundaries were obtained and derived from the ISTAT records, considering the town as the unit of interest [8]. Then, for each administrative unit (town), we counted the number of subjects with at least one positive NP swab. To describe the spatial distribution of SARS-CoV-2 in the area of the Alessandria LHA, we calculated the proportion of SARS-CoV-2 positive subjects over the total number of residents in each small administrative sub-region separately for the variables of age and gender. These data were also plotted using choropleth maps. Specifically, we categorized the prevalence in 5 different groups, using a blue scale as: <2.5, 2.5|−5, 5|−7.5, 7.5|−10 and ≥10 [12]. The choice to report 5 fixed classes of prevalence, rather than the quintiles or percentiles (more frequently used), was justified by the fact that it allowed us to easily compare different maps and identify specific areas of low, medium and high prevalence of positive subjects. In the graphical visualization, we also added the positions of nursing homes to evaluate if towns with higher numbers of positive subjects were related to presence of residential structures. Finally, a more detailed exploratory analysis in the places identified as high risk was conducted considering the specific addresses and the time period of the NP swabs to evaluate if the outbreak was of family origin.

All analyses were performed using SAS 9.3 (SAS Institute Inc., Cary, NC, USA.), R version 3.4.1 (R Core Team, Vienna, Austria), STATA 11 (StataCorp LLC, College Station, TX, USA) or QGIS 3.4.3.

3. Results

In the period from 2 March to 22 December 2020, 173,117 NP swabs were performed: 61,101 (35.3%) during the first months of the pandemic (until June) and 112,016 (64.7%) in the second part of the year, in the middle of the second wave (July–December). Among the study population, 17,260 subjects with at least one positive NP swab were included. The median age was 54 years (IQR 38–72) and 54.9% (n = 9478) of the subjects were female. A frequency distribution of the total residents in the area, taking into consideration the positive subjects is reported in

Table 1. Gender and age distribution for the total population and subjects with at least one positive NP swab. Prevalence of positive NP swabs is reported in brackets.

	ISTAT Population   n = 428,535	Positive NP Swabs (%)   n = 17,260
Gender
Female	220,289	9478 (4.3%)
Male	208,246	7782 (3.7%)
Age, years
<50	206,659	7139 (3.4%)
50–65	101,732	4440 (4.4%)
65+	120,144	5681 (4.7%)

. Out of the total residents, 4% showed at least one positive NP swab.

Relevant differences were observed in the variables of gender and age: females showed a higher prevalence of positive NP swabs, compared to males (4.3% vs. 3.7%, respectively). In addition, the probability of testing positive increased in direct proportion to age (<50 years 3.4%, 50–65 years 4.4% and 65 + 4.7%). Moreover, males with a positive NP swab revealed a higher probability of hospitalization, compared to females (28.5% vs. 20.9%, respectively).

The distribution of positive NP swabs showed significant differences when the two first waves of the pandemic in our country were compared. Four thousand one hundred sixty-nine swabs were performed in the period of March–June, while 13,091 were performed in the period of July–December, with peaks of 400 or more positive swabs in one day in November (Figure 2). Furthermore, four thousand two hundred and three (24.3%) subjects needed to be hospitalized, and there was a significant difference between the two waves: 43.8% (n = 1826) of subjects with a positive NP swab were hospitalized in the first months of the pandemic while only 18.2% (n = 2377) were hospitalized in the second wave.

Among the 192 towns analyzed, 65 (33.8%) showed a prevalence lower than 2.5%, 98 (51.0%) between 2.5% and 5%, 26 (13.5%) between 5% and 7.5%, one between 7.5% and 10% and two (1.0%) with more than 10% positive subjects. In Figure 3 the prevalence of subjects with positive NP swabs, with the indication of the presence of nursing homes, is reported. Interestingly, the three towns with a higher prevalence of positives were located in areas with at least one residential nursing home structure. In Casal Cermelli (prevalence 12.2%), 109 (75.69%) out of 144 subjects with positive NP swabs lived in one of the two nursing homes, and almost all were positive in April 2020. Similar considerations can be made for San Sebastiano Curone (prevalence 10.6%) and Odalengo Grande (prevalence 7.8%) in which 50% (n = 29) and 70.59% (n = 24) of positive subjects lived in nursing homes, respectively.

Figure 4 and Appendix A Figure A2 and Figure A3 display the geographical distributions of positive NP swabs by age and gender. The number of towns/areas with a higher prevalence of positive subjects increased with age, and same the consideration can be made for females versus males. Focusing our attention on towns with a high prevalence of positive NP swabs among older subjects (more than 65 years old), we observed the presence of a potential cluster of cases that occurred in Sale and Castellazzo Bormida (April 2020), Villanova Monferrato (September 2020), Quargnento (October 2020), Basaluzzo and Bassignana (November 2020) and Valmacca (December 2020). All of these can be associated with outbreaks in nursing homes present in these towns.

Finally, we observed four towns/areas that were COVID-19 free and territories able to contain the spread of SARS-CoV-2. Paradigmatic was the case of the towns of Spigno Monferrato and Parodi Ligure, in which a psychiatric community and a nursing home are located, respectively, with only one positive subject identified.

4. Discussion

This geo-epidemiological study, conducted in an Italian homogenous area in terms of political and health decisions, reveals and highlights a territorial heterogeneity in spatio-temporal diffusion trends among people testing positive for SARS-CoV-2 through a RT-PCR or antigen-test. Particularly, in the LHA of Alessandria province in the Piedmont region, we found not only an expected difference in the number of positive NP swabs in time (first vs. second wave) but also, in our geospatial study, a peculiar and noteworthy location distribution of SARS-CoV-2 infection emerged. Specifically, the performed analysis showed that towns with a higher proportion of subjects with at least one positive NP swab were generally characterized by the presence of nursing homes or long-term care facilities.

The total number of swabs performed can explain the difference of positive NP swabs observed in time. During the first phase of pandemic, the number of SARS-CoV-2 tests were lower than during the second phase due to the limited availability of diagnostic kits/tests and to significant organizational problems. The general population were surely tested less during the first wave, because of a very low testing capability. Despite this, an underestimation of positive subjects can be assumed, and we hypothesized that this bias was not territorially related. In addition, in the first wave the percentage of hospitalizations among positive subjects was very high, as generally only people with the worst clinical condition were tested. On the other hand, in the second wave, more swabs were performed; thus, it was possible to identify even asymptomatic subjects, who are less frequently hospitalized. Differences can also potentially be explained by various and multiple lockdown strategies adopted by the Italian government. The first Italian decree of “Further implementing provisions of the Urgent measures for the containment and management of the COVID-19 epidemiological emergency” was issued on February the 23^rd (number six); however, it was followed by numerous others that were less or more restrictive. Subsequent laws were approved during the course of the year, with different measures and procedures for limiting infection by SARS-CoV-2.

Differences in prevalence were also observed in terms of gender and age. We observed that females presented a higher probability to test positive for SARS-CoV-2, compared to males; in addition, older people were shown to be more susceptible to infection by SARS-CoV-2. However, men, despite the low prevalence, seemed to be more frequently hospitalized than women. In the literature, in line with our results, gender-based differences in COVID-19 were observed; emerging global data suggest that men appear to be at a higher risk of contracting this infection, when compared to women. This gender-based disparity has resulted not only in the incidence, but also in hospitalization rates [13]. Although the results that emerged from our analysis did not reproduce the findings from similar literature, we might assume that these differences and discrepancies could be related to a lack of symptom data. Indeed, we collected and registered only NP swab positivity, without knowing the real presence and the level of severity of COVID-19. Nevertheless, the age-dependent effects in the transmission of COVID-19 shown in our study are completely supported by previous literature [14].

Interestingly, the observed territorial heterogeneity can almost be explained by the presence of nursing homes/long-term facilities; in cities/towns where a high number of positive cases was observed, at least one was present. Moreover, when specifically considering the residential addresses and the date of swab, evident clusters of cases can be hypothesized. In Italy, between March and June 2020, the ISS conducted a survey on the contagion of COVID-19 in nursing homes. Three thousand two hundred ninety-two residential and social and health structures (96% of the total) were contacted, but only 41.3% of them completed the questionnaire. Interestingly, Alessandria was counted among the 15 provinces with a higher mortality rate among residents in nursing homes (5.3 per 100 person-year). The main reasons that partially explain the high mortality values could be the low number of NP swabs performed, the scarcity of personal protective equipment (PPE; e.g., surgical masks or FFP2/FFP3 masks, nitrile gloves, single-use gowns, etc.) and the difficulty in managing those patients (e.g., transfer of residents affected by COVID-19 in hospital facilities, isolation of residents affected by COVID-19, etc.) [15].

Other studies have underlined that the COVID-19 pandemic has disproportionately affected care home residents internationally, and nursing homes are considered one of the most vulnerable and high risk settings for the transmission of SARS-CoV-2 in the elderly population [16,17]. Indeed, in several countries (Canada, Sweden and the UK), more than 20% of all reported COVID-19 deaths took place in long-term care settings [18].

The majority of similar geospatial epidemiological studies were conducted in China, the United States of America and Brazil during the first wave of the pandemic [19]. In addition, a significant production of web-based maps has been registered globally, showing several different characteristics of the global diffusion of COVID-19. Web-based maps can be described as online mapping platforms that are available on interactive webpages or applications (e.g., web-based maps produced by the German Koch Institute that describe the spread of SARS-CoV-2 infection into each municipality in Germany through the spatial and temporal analysis of administrative data, such as infection rates, testing sites and population dynamics). Despite their multiple limits (e.g., inconsistent utilization of scales and units of aggregation, production of non-graduated and non-normalized choropleth maps, the scarce analysis of local data etc.), these web-based maps represent an excellent and useful tool for analyzing and ascertaining the spread of infectious diseases, such as COVID-19 [20].

Collectively, the results that emerged from our study will serve as the basis for the future characterization of a COVID-19 outbreak evolution; in particular, the care homes of the area should be analyzed in detail, using several variables, such as type of care home, number of beds, etc. Comparisons with other regions was somewhat complex to perform due to the specific and peculiar way that the COVID-19 pandemic affected each area, as has emerged from the available literature on the theme. Moreover, different containment measurements put into place by the local authorities of the individual regions in separate times could contribute to the challenging and tricky task of making comprehensive confrontations.

This study presents some important limitations and weaknesses that must be taken into account. Firstly, this study was carried out only on the population of the Alessandria LHA, representing a small part of the Italian population. Indeed, our results may not be completely generalizable to all populations, as there could be differences in the criteria considered for negativity and positivity, in the subjects tested and in the availability and use of tests.

Moreover, by considering only positive NP swabs, other real cases of COVID-19 could be excluded from the analysis. In addition, as suggested by other authors, the wide heterogeneity in the diffusion of COVID-19 is not entirely attributable to individual demographic and clinical characteristics (e.g., age, sex state of health, etc.), and the risk of contagion is also deeply influenced by socio-economic factors, such as educational disadvantages, housing crowding, mobility and population density [21]. These data were not available at the time of the analysis; however, it would be interesting to include them in future considerations. Finally, the area units we adopted may be too small, and it would be useful to reproduce the study by relying on areas of intermediate dimension. Nevertheless, among the strengths of this work, it is possible to cite the use of administrative health data with robust and valid statistical analyses. The major strength of this study is that it contributes to the monitoring of epidemiological data from the LHA of Alessandria by identifying potential hotspot areas for COVID-19 and areas in which there were no positive subjects through the use of choropleth maps. In addition, this study could serve as the basis for future in-depth analyses to evaluate the factors that favor or inhibit the spread of the virus. Finally, we will propose the use of these maps (updated weekly) to the Prevention Department to contain the spread of SARS-CoV-2.

5. Conclusions

By analyzing the prevalence of subjects with at least one positive NP swab, we were able to monitor a specific area and identify towns where the spread of SARS-CoV-2 was over- or under-expressed. This could represent a tool for administrators to be prepared to respond to current and future epidemic waves by identifying areas where the spreading possibility of the virus is highest. Spatiotemporal analysis and disease mapping represent useful instruments of health and social geography.

Understanding and ascertaining the spatial-temporal diffusion of COVID-19 is critical to curbing its spread.

Through the creation of analytic and synthetic maps of the prevalence of subjects with at least one positive NP swab, it is possible to monitor a specific region, identifying towns and areas at higher risk of spreading of SARS-CoV-2 infection.

This study, despite its limitations that must be taken into account, could represent a potentially beneficial tool for local administrators and health authorities in decision-making in order to give adequate and satisfactory responses to current and future epidemics. In addition, these maps could drastically reduce the frequency with which sidelined communities are left unprotected.

Lastly, the identification and the understanding of the spreading of infectious diseases, through geospatial information, could have a critical and social impact, entailing clinical and epidemiological implications in public health; for this reason, similar studies are warranted.