Spatial analysis of road traffic accident deaths in the state of Paraná, 2007-2016

Brazil is in fifth place among countries with the highest number of land transport accidents. The state of Paraná, Brazil, was the object of this study which conducted spatial analysis with the aim of identifying areas where this phenomenon occurs more and their time series over a 10-year period. This was an ecological and exploratory observational study covering the period 2007 to 2016 in 39 micro-regions of the state of Paraná. Data of road traffic accident deaths as per the International Classification of Diseases (ICD-10, codes V01 to V89) held on the Mortality Information System, were analyzed. Relative risk rates were calculated and choropleth maps were built. A total of 31,651 deaths from the causes examined were recorded according to municipality of occurrence. The most frequent ICD-10 items found were those involving automobile occupants, motorcyclists, pedestrians and cyclists in road traffic accidents. An overall falling trend was found with effect from 2012. The rate by area did not show pronounced spatial dependence and there was considerable variation, whereby the Cerro Azul micro-region had the lowest relative risk in the period, while in Campo Mourão deaths were around 53.3% above the expected level. The estimated average annual trend for the Curitiba micro-region had the steepest fall in the period, while Campo Mourão had the highest rising trend. The trend analysis indicated places where more robust public policy interventions and enforcement actions need to be reviewed.


Introduction
The World Health Organization has estimated that globally in 2016, 1.35 million people died as a consequence of land transport accidents -LTA (WHO, 2018); i.e. approximately 3,700 people a day. The road traffic injury control is a public health problem (Mohan, 2008), since the road traffic violence is in eighth place among the leading causes of death worldwide and leaves some 50 million people injured, according to the Global Status Report on Road Safety 2018 (WHO, 2018). Also, the number of occurrences is disproportionate in developing countries (Nantulya & Reich, 2002), revealing inequalities (Nantulya & Reich, 2003), with worse conditions for populations from India (India State-Level Disease Burden Initiative Road Injury Collaborators, 2020), and Latin America (Martínez, Contreras, & Moreno, 2020), and Africa (Adeloye, Thompson, Akanbi, Azuh, Samuel, & Omoregbe, 2016)  According to a report by the Brazilian Institute of Applied Economic Research -Instituto de Pesquisa Econômica Aplicada (IPEA), the cost of road traffic accidents in terms of public expenditure is estimated to be around BRL 40 billion per annum for highway accidents and BRL 10 billion per annum for urban roads and street accidents (IPEA, 2015). The greatest cost relates to lost productivity, due to time off work or death, with impacts on Social Security and families, followed by health expenditure, overburdening emergency, hospital and rehabilitation services (Andrade & Mello-Jorge, 2016).
Alcohol intake is one of the main risk factors associated with road traffic accidents (Barone et al., 2019;Borges et al., 2017;Das, Gjerde, Gopalan, & Normann, 2012;Valen et al., 2019). Thus, in order to be minimized the consequences of this interaction, public policies were adopted by governments, through the reduction of the permitted level of blood alcohol, or by the prohibition of any blood alcohol content (Fell & Voas, 2014;Nagata, Setoguchi, Hemenway, & Perry, 2008;Nistal-Nuño, 2017).
In the Brazilian context, some positive evidence of these strategies has also been identified (Andreuccetti, Carvalho, Cherpitel, Ye, Ponce, Kahn, & Leyton., 2011;Malta, Berna, Silva, Claro, Silva Júnior, & Reis, 2014). The Brazilian Traffic Code, instituted by Law No. 9503/1997(Brasil, 1997, allowed, with no penalties, up to 6 decigrams of alcohol per litre of blood, which is the equivalent of two measures (80ml) of distilled drink. In order to come into line with the global reality of greater restraint, the so-called Dry Law (Law No. 11705/2008) altered the Highway Code and established zero tolerance of blood alcohol content, stiffer penalties for drivers and restrictions on the sale of alcoholic beverages on federal highways (Brasil, 2008a). In 2012, the New Dry Law (Law No. 12760/2012) introduced innovations regarding some topics provided for by the previous Dray Law, such as proving alcohol consumption by the use or otherwise of the blood alcohol measurement device (breathalyzer) (Brasil, 2012). Finally, in 2016, Law No. 13281 introduced a new article for the offense of refusing to have a breathalyzer test (Brasil, 2016).
The Life in the Traffic Project (Projeto Vida no Trânsito) began in 2012, being a version of the Road Safety in Ten Countries (RS-10) project funded by the Bloomberg Philanthropies Foundation (Bloomberg Philanthropies, 2013). and adapted to Brazilian reality, coordinated by the Ministry of Health and involving other government bodies (Brasil, 2017;Curitiba, 2010;UN Road Safety Collaboration, 2020). Its objective is to plan and enhance road traffic safety strategies, which involves actions at the federal, state and municipal levels (Morais Neto et al., 2013). The project was initially implemented in five state capitals, one in each of the country's five macro-regions. In 2013 it was expanded to municipalities with more than one million inhabitants and to all the country's remaining state capitals (Curitiba, 2010;UN Road Safety Collaboration, 2020). In the state of Paraná, Curitiba was the first city to take part in the program and by 2017 a further 12 of the state's cities were taking part (Paraná, 2019).
Paraná has a high LTA mortality rate, of 30.3, occupying the first position in Brazilian's South and Southeast regions (Ladeira et al., 2017). At the national level, occupies the third position in the register of accidents with victims (CNT, 2018).
Also, it is a state located in a cluster considered to be of high risk, as it has a large number of municipalities bordering other states in the country, which has a statistically significant impact on the number of deaths from road traffic accidents (Morais Neto et al., 2012).
Thus, the aim of this study was to identify the spatial distribution of deaths due to road traffic accidents in Paraná, from 2007 to 2016, according to the state's micro-regions.

Methodology
This was an ecological and exploratory observational study covering the period 2007 to 2016 in 39 micro-regions of the state of Paraná. Data of road traffic accident deaths as per the International Classification of Diseases (ICD-10, codes V01 to V89) (WHO, 2016) held on the Mortality Information System (http://sim.saude.gov.br/default.asp), were analyzed.
Relative risk (RR) rates were calculated and choropleth maps were built. Annual trend was analyzed using a statistical model considering the neighboring structures of the micro-regions. Data on all causes covered by ICD-10 (codes V01 to V89) Research, Society andDevelopment, v. 10, n. 11, e431101119942, 2021 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v10i11.19942 4 were also studied by micro-region and by year. The population in each micro-region as per the 2010 census (IBGE, 2010) was used to calculate mortality rates.
RR is a measurement commonly used to map diseases, being the ratio between the number of deaths recorded and the number of deaths expected, calculated according to a given criterion (Pereira, 2008). It is common to consider the expected number of deaths as being the result of the death rate in a period and total area multiplied by the population in each study area and period. As such, the overall rate in Paraná during the entire analysis period and the population as counted by the 2010 census (IBGE, 2010) were recorded. In order to model the Standardized Mortality Rate, it was considered that the rate in each micro-region varies, so that neighboring micro-regions tend to have similar rates. In addition, it was considered growth of the rate in each micro-region assuming this occurrence similar to that of its neighboring micro-regions. This is the model validated by Castro et al. (2004).
As such, the formula for occurrence in micro-region in time , is: where is the expected number of deaths in micro-region and is RR. The statistical model considers that the risk logarithm is defined by: where is an intercept or reference rate (for the first year) and expresses annual relative growth. This means that there is a rate for each area and time which is being modeled by a rate in each area in the first year plus annual growth particular to each micro-region.
When specifying the statistical model for and , the proposal recommended by Riebler et al. (2016) was taken into account. In this case, for example, the following formula for was: where has two terms: and . The term is assumed to be similar for neighboring areas in order to capture possible regional variations, and the term is not structured in space in order to capture variations particular to each municipality. Term has variance of 1, that is to say, , while is standardized in order to obtain marginal variance equal to 1 (one); however, each depends on the values in the neighboring areas. These two terms are weighted by the parameter, this being a parameter that measures the proportion of the spatial effect structured in space and is associated with the degree of spatial correlation. The term measures the extent of the influence of the effect, whereby the greater this value, the greater the variation of , indicating greater variation in the rate between micro-regions.
Similarly, there is: where and have properties like those of and , respectively. In order to complete the specification of the model in accordance with the paradigm of the Bayesian model, it was considered a priori distribution for the , ,

Results
The map of Paraná's micro-regions, with their respective resident population (expressed in thousands) in 2010 are shown in Figure 1. In 2010, Paraná's population was estimated to be 10,444,526 inhabitants. As such, the annual road traffic accident death rate in the same year was 33.0 per 100,000 inhabitants. There were 31,651 deaths in Paraná according to the data analyzed. The number of deaths each year and respective crude annual rate per 100,000 inhabitants can be seen in Table 1, showing a reduction in values from 2014. The rates in the maps shown in Figure 2 were categorized. This made it possible to organize the micro-regions by rate groups for each year studied.  Table 2).  Figure 3A shows the map of the crude rate per 100,000 inhabitants, found in each micro-region for the entire period (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). Figure 3B shows estimated RR per micro-region. The micro-region with the lowest estimated RR in the period was Cerro Azul (65.9% fewer deaths), while Campo Mourão had the highest estimated RR (53.3% more deaths). Figure 3C shows the estimated average annual mortality rate rising (or falling) trend. The Curitiba micro-region can be seen to have had the largest fall in the period, with an average annual falling trend of 5.4%, while Campo Mourão had the highest rising trend in the period, with an average annual increase of 1.3% over the 10-year period. It can be seen that there are no micro-regions with annual growth greater than 3%, while the rate fell by more than 3% in three micro-regions (Cerro Azul, Curitiba and Rio Negro).

Discussion
This study aimed to evaluate the RR and to conduct a spatial analysis of the areas of the state of Paraná, according to road traffic accident data, from 2007 to 2016. The most frequent ICD-10 items found involved automobile occupants, motorcyclists, pedestrians, and cyclists. The spatial analysis revealed a variation, with areas with low RR, such as Cerro Azul, Research, Society andDevelopment, v. 10, n. 11, e431101119942, 2021 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v10i11.19942 8 and with high RR, such as Campo Mourão.
The justification for choosing the state of Paraná is due to that this state is among the worst in terms of road accident indicators in Brazil, occupying the third national position and the first in the southern region of the country (CNT, 2018;Ladeira et al., 2017). The state is also located in a cluster of municipalities considered to be at high risk for accidents, since many of these municipalities border with municipalities in other states, which determines an extensive road network (Morais Neto et al., 2012).
In 2010 the United Nations Organization defined the period from 2011 to 2020 as the Decade of Action for Road Safety, creating a plan with sustainable targets to be met in order to reduce by half the number of deaths caused by road traffic accidents. This initiative aims to save lives by building and maintaining good highway infrastructure, developing safer vehicles, improving driving behavior and optimizing post-accident care (UN Road Safety Collaboration, 2020).
It is important to mention that road traffic accidents burden the governments, with expenses that can vary, in high income countries from 0.5% to 6.0% of the GDP, while in low and middle income countries, this value varies from 1.1% to 2.9% (Wijnen & Stipdonk, 2016).
In Brazil, public policies, implementation and enforcement of national restrictive legal measures on drink-driving (Brasil, 2000(Brasil, , 2008b, appear to have been effective over the years, especially after the Dry Law (Brasil, 2008a) was implemented, this being a possible hypothesis for the fall in the RR rates, especially between 2013 and 2016, in most of the micro-regions of the state of Paraná.
However, it is necessary to discuss the role that these alcohol regulatory measures play in coping with traffic deaths and injuries. Although in the Brazilian context the impact was perceived after the implementation of such measures, this occurred in the periods immediately after the restrictive measures. In the following months, there was a trend of stability and increase in traffic deaths in the country, which indicates an inability to fully control it, which becomes a controversial topic (Morais Neto et al., 2013). In Chile, for example, the same occurred, with a more pronounced decline after the law in relation to alcohol-related deaths, but these differences were not statistically significant (García-Echalar & Rau, 2020).
Some countries have also identified that blood alcohol regulation has not resulted in a reduction in accidents. A natural experiment, using data on road traffic accidents and alcohol consumption in Scotland (interventional group -the blood alcohol concentration limit for drivers was reduced from 0·08 g/dL to 0·05 g/dL in 2014) and England and Wales (control group). Authors found that lowering alcohol concentration in Scotland was not associated with a reduction in road traffic accidents, but it was associated with a small reduction a reduction in the purchase of alcoholic beverages (Haghpanahan et al., 2019). Ferris et al. (2017) analyzed a 10-year period of data of serious road injuries, and concluded that policies and procedures implemented to decrease drink-driving have not reduced alcohol-related serious road injuries rates.
These differences maybe can be explained by the fact that Brazil is a developing country, and so regulatory measures appear to be more effective, contrary to what is found in some developed countries. However, other measures are identified as capable of positively impacting the reduction of accidents such as education campaigns (Bon de Sousa, Santos, Mateus, Areal, Trigoso, & Nunes, 2016), improvement of the road network, and traffic management (Sun, Liu, Chen, & He, 2019).
In the present study, analysis on three levels is needed in order to gain a more complete panorama of the RR results in the 39 micro-regions of Paraná, between 2007 and2016, were: the main highway network, demographic density and an indicator (e.g. estimated RR). It is suggested that the existence of main highways increases the likelihood of LTA-related deaths. This hypothesis would be sufficient to provide an explanation for the greater part of Paraná's territory, such as the central and western regions. However, this two-dimensional analysis of the problem is not sufficient to explain situations such as the east and part of the northeast regions of the statewhich have considerable main highways but low relative risk when compared to the other regions of the state. Especially in the micro-region where the state capital, Curitiba, is located, the Life in the Traffic Project has been in operation since 2010. This factor may by the differential for it being the micro-region with the biggest drop in the mortality rate in the period studied, when considering the falling trend rate data (CNT, 2018).
Another level to be considered is demographic density, which is inversely proportional to RR, possibly because of the presence of more resources and law enforcement. Indicators, such as RR, are directly proportional to the presence of main highways, and inversely proportional to demographic density (IBGE, 2017).
The city of Cerro Azul achieved the best overall indicators, while Campo Mourão had the worst results. Two arguments are key to supposing why the first one was successful: not having high population density, but not being distant from large urban centers (located at around 2 hours from Curitiba and 3 hours from Ponta Grossa) and not being on any main highway. These are factors that must have contributed to success in reducing deaths over time following the implementation of the Dry Law in 2008 (Brasil, 2008a).
The result found for the Campo Mourão micro-region is probably related to it being crossed by the BR-158 highway, known in the period studied as being one of Brazil's most dangerous highways and having the highest number of deaths per 100 accidents in Paraná. It should be emphasized that this stretch of highway is part of the Paraná Integration Ring Road, formed by the state's main highways. The Integration Ring Road is divided into six sections, administered by six outsourced companies.
Some of this study's limitations which may act as confounding factors include: the specific accident (event) site, whether in urban areas or on open highways, given that the governmental database does not detail this information and, above all, death certificates do not provide information on the site of the event and death may occur in places other than the site of the accident; periods of heavier traffic such as summer holidays and the presence of one-off actions such as police inspection road blocks, like "Summer Operation" when these actions are organized more frequently at year-end, whereby the seasonality of such programs is not included in the time unit used.
Finally, the data was not submitted to standard correction considering the increase in the fleet of vehicles in Paraná, nor standard correction of the specific rates of type of vehicle used at the time of LTAs. Moreover, population projections for each year studied were also not used.
These data demonstrate the need to understand and act on the phenomenon, because identifying regions of greater vulnerability enables the establishment of more efficient public policies, with the aim of contributing to minimizing harm and meeting globally agreed targets. The benefits generated by a set of appropriate public interventions not only result in less costs for the health system, but fundamentally save lives (Banstola & Mytton, 2017;Wijnen & Stipdonk, 2016).

Conclusion
The spatial analysis data on deaths due to road traffic accidents in the state of Paraná are particularly relevant for the Federal Administration, as evidence of weak points that need to be reviewed, as well as the need to conduct new studies in this area. Moreover, it is a pioneer study in the state, which contributes to analysis of the annual trend and provides important considerations about the panorama for forthcoming years, serving as an indicator for new public interventions. Future research can be done in order to strengthen relations with agencies related to traffic control, as well as the expansion and applicability of the method in other Brazilian municipalities.