Study design
This is an ecologic study based on the geographical unit of the 119 neighborhoods in the municipality of Fortaleza (Figure 1). In order to define the main transmission networks (and chains), we assumed that the spatial diffusion of the COVID-19 epidemic is influenced by population mobility. We also assumed that the spatial spread of the COVID-19 epidemic follows the hierarchical model based on networks and population mobility that plays a key role in the constitution of the transmission chains.
The rationale of the study assumed that based on the incidence rate of the initial phase of the epidemic (infectivity load), the flow of passengers moving between neighborhoods can influence the infection load in each neighborhood of the municipality. If one considers the epidemic vulnerability attributed to populations living in these neighborhoods, combined with the mentioned infection burden, one can estimate the propensity for a serious COVID-19 epidemic.
To measure the propensity of a severe epidemic, the following events were initially calculated: (i) infectivity burden - ItyB; (ii) infection burden - IonB; and (iii) population epidemic vulnerability index (PEVI). Then, the (iv) Propensity score for a severe epidemic in the neighborhoods of the city of Fortaleza was estimated.
Infectivity Burden
Infectivity burden was calculated using the epidemiological surveillance records of COVID-19 in Fortaleza, so the number of confirmed cases were denoted as infectivity burden. These data were formally obtained from the Municipal Health Department on April 7, 2020. The suspected cases of COVID-19 had been investigated according to the recommendations of the Ministry of Health. All confirmed cases (with laboratory confirmation) of COVID-19 reported to the Municipal Health Department of Fortaleza until March 12, 2020, residing in the municipality, were included. This period corresponds to the initial phase of the epidemic, when predominantly imported cases (travelers) initiated the transmission; indigenous cases would be detected later after being in contact with the initial cases through community transmission [6]. In this phase, the Fortaleza City Council adopted more general measures for surveillance, prevention, and control of COVID-19 [7].
Infection Burden
The Infection burden was measured by combining the infectivity burden and population mobility between the neighborhoods of Fortaleza. The mobility burden - MtyB, was evaluated through the daily travel flows, by looking at public transport use with work motivation between the two neighborhoods. Notably, the measurement of the flow of people between the neighborhoods of Fortaleza (excluding displacements within the same neighborhood) only became possible after a strategic study was conducted by the city of Fortaleza and other institutions on this theme in 2019 called Home Origin-Destination Survey (OD survey).
The OD survey comprised a sample survey performed through interviews in households, providing the values of the variables in this analysis with a detailed matrix of the trip patterns and travel choices. Data were collected in a database, which describes the various attributes of the activities and trips of the city’s inhabitants, as well as the respective socioeconomic status and characteristics of individuals and their families, in order to describe a pattern of displacement of people and the chain of their activities throughout a typical business day. The information from the OD Matrix is of immense importance in the analysis of transport systems, comprising fundamental elements for planning and decision making, and has therefore been integrated in this study. The infection burden indicator was calculated using the following formula: (see Formula 1 in the Supplementary Files)
PEVI
The PEVI was constructed according to the Urban Health Index approach recommended by WHO [8], to demonstrate the population attributes that best represent, from a collective point of view, the vulnerability of the population to COVID-19. This index comprises seven (7) sociodemographic indicators, based on the 2010 Brazilian census of the Brazilian Institute of Geography and Statistics (IBGE). The indicators that make-up the PEVI are: (i) proportion of households with more than two residents per bedroom, (ii) illiteracy proportion, (iii) proportion of the population in extreme poverty, (iv) proportion of households without running water and sanitation, (v) proportion of unemployment, (vi) Gini of family income, and (vii) proportion of people living in subnormal agglomerations.
IBGE classifies subnormal agglomerates groups as consisting of 51 or more housing units, characterized by the absence of ownership titles, and at least one of the following characteristics: irregularity of circulation routes, size and shape of the lots, and lack of essential public services (such as garbage collection, sewage, water, electricity, and public lighting).
The data for each of the indicators above were previously obtained and translated into the Human Development Units (HDUs) by the Institute for Applied Economic Research (IPEA), with the exception of the proportion of subnormal agglomerations. The HDUs represent units of analysis with relatively homogeneous socioeconomic characteristics, and the original data were used to produce the Metropolitan Region Human Development Atlas. The HDUs were designed to generate more homogeneous areas, based on socioeconomic conditions, than the weighted areas of the IBGE. To calculate the proportion of subnormal agglomerations, data from the IBGE were used, considering the population living under these conditions by the total population of the neighborhood, thereby obtaining the percentage of people living in subnormal agglomerations per neighborhood.
After the values of the indicators were stored in a database, the accuracy, completeness, and consistency of the data were verified for calculation of the summary measure. There are two main steps in calculating the index: (1) standardization of indicators, and (2) amalgamation of standardized indicators. Each of these steps can be performed in a mathematically straightforward manner.
The standardization of the values of each indicator is performed using the following formula: (see Formula 2 in the Supplemental Files)
where is the standardized value of I, max (I) is the highest value of I among all observations, and min (I) is the lowest value of I among all observations.
Since the values are obtained for all indicators and units, the next step was to integrate into a single composite index, here called VEPI. VEPI is calculated for each unit, using the geometric mean of the values for each unit. Considering that there are j indicators, the formula used for this calculation was:(see Formula 3 in the Supplemental Files)
Where, is the standardized value of the seven indicators for a given neighborhood, and j corresponds to all other neighborhoods.
In addition to the point estimate, VEPI variance and standard error were also calculated, with respective confidence interval estimates based on these measures. Considering that the purpose of this index is to identify the “geographical” disparities of the studied phenomenon, a diagnosis of its distribution, evaluating the differences between the highest and lowest values as well as its visual comparison with the homogeneous distribution of these values, was carried out through the “qqnorm” graph (Figure 2).
Severe Epidemic Population Propensity
The main study outcome is the Severe Epidemic Population Propensity. The estimated score of the propensity for a severe epidemic amongst the population of the Fortaleza neighborhoods was calculated by combining the infection burden with PEVI. The calculation was carried out using the multiplicative approach, involving the multiplication of these two indexes.