Iberoamerican analysis and classification of labor accidents in the civil construction industry Análisis y clasificación iberoamericana de la accidentalidad laboral en la industria de la construcción civil

Millions of accidents and thousands of deaths resulting from work activity occur every year and 30% of this comes from civil construction. As to this sector is one of the engines of the Ibero-American economies, it becomes important to identify the behavior of the occupational injuries in the region. For this reason, the main objective of this article is the development of an analysis of Ibero-American occupational accidents in the construction sector, based on the results obtained from official statistical sources for the period 2013-2017. This article was carried out through quantitative surveys, statistical observations, determination of correlations and use of the multicriteria methods AHP and TOPSIS. The main results of this investigation were the characterization of the problem, an Ibero-American ranking of accidents and an analysis of the basic aspects required to the employers of each country involved. It was also possible to identify that in the period of analysis the region had an improvement in the fight against this problem, as there was a successive reduction in the accident figures over time.


Introduction
According to the International Labor Organization (OIT, 2015), it was estimated that approximately one worker dies every 90 seconds and 895 suffer accidents as a consequence of their work. This represents 350,000 deaths and 313 million accidents annually worldwide. Around 30% of these accidents occurred in the construction sector.
Similarly, taking into account the information from this organization, it is known that personnel working in the construction industry are 3 to 4 times more likely to die from occupational accidents than those working in other industries. This is mainly due to the extreme working conditions to which workers are exposed (Solís, 2017). Given the above, it is clear that the issue of occupational injuries in the civil construction industry generates significant human losses, affecting society and the economy. Therefore, it is a serious global problem that deserves to be studied (Zhang, 2013).
It is important to mention that construction is one of the industries with the greatest influence on the progress and financial growth of nations around the world because of its strong influence on other sectors that are predominant in the economy, mainly as a result of the close relationship between infrastructure and development (Pérez-Foguet et al., 2007); (Martínez-Aires et al., 2015). Based on this financial influence, it is noteworthy that construction provides opportunities for low-skilled or unskilled workers, helping to reduce unemployment with low investment by eliminating training costs (Galindo and Sosvilla, 2012). However, the positive factors mentioned above are contradictory since they also lead to labor informality and accidents (Del Águila, 2015), thus affecting the productivity, quality, and reputation of companies (Enshassi et al., 2009). This produces an additional public health problem and, at the same time, a loss of public and private funds. In the case of Europe, only considering working hours lost due to accidents at work, 2.6% or more of the gross domestic product (GDP) is lost annually (Oliveira et al., 2012). Globally, according to (Takala et al., 2014), the economic costs to countries from occupational injuries including work-related diseases range from 1.6% to 6.0% of total national GDP.
It is noteworthy that one of the key points for success in the construction industry is occupational safety since investing in this parameter reduces costs, increases production efficiency and, most importantly, preserves life. (González et al., 2016).
For the development of this study, some of the most important countries in Ibero-America 1 were chosen, including Argentina, Brazil, Chile, Colombia, Spain, Mexico, Peru, Portugal, and Uruguay. The focus of this study was the safety and health at work (SHAW), having as main objectives the analysis of the figures of accidents in the Ibero-American civil construction industry for the period 2013-2017 and the creation of annual Ibero-American rankings of occupational accidents in the sector through the application of the Analytic Hierarchy Process (AHP) and Technique for order preference by similarity to ideal solution (TOPSIS) methods. For the creation of the classifications, the accident rate was related to other important figures, such as the GDP and the human development index (HDI).
This study is justified by three fundamental aspects: (i) Although the topic is well identified in the literature, effective monitoring of results is necessary to identify trends and make the necessary decisions for their prevention, considering that it evaluates the statistics that demonstrate the risk that still exists in Ibero-American construction companies; (ii) it addresses an issue that, due to the significant financial and social costs, requires sound management of construction works, so it is necessary to establish a risk management model consistent with its importance; (iii) it provides essential knowledge to manage work-related risks, which is extremely useful to develop strategies to mitigate accidents that generate risk in construction.
It is important to emphasize that the relevance of this study lies in the fact that the results obtained can help the health authorities of the countries involved to have a clearer vision of the behavior in the fight against occupational accidents, providing a basis for the orientation of future preventive activities and SHAW policies, specifically for the construction sector. Also, this study will help to fill a gap that exists in the region regarding the use of TOPSIS and AHP multi-criteria models together, to address the existing problems in construction.

Methodology
Because of its characteristics, this study is considered as a quantitative, correlative and descriptive nonexperimental research, in which the problem of occupational injuries in the civil construction industry of Ibero-America was described. The research was developed by observing and relating different factors without manipulating the natural conditions or the variables analyzed, aiming to observe the performance of the countries in terms of accident rates and mortality. This can also be classified as longitudinal research since data were analyzed over time for annual periods between the years and 2017(Hernández et al., 2010. This study was carried out with secondary data (Malhotra, 2008). The selection of the sample was projected and then defined by the availability of the data, which were mostly collected from the national institutes of statistics. Taking this into account, it is important to point out that a forced selection of the sample was carried out since only 9 of the 22 countries that were considered to be included in the research showed relevant data for the development of the study. From the beginning, it can be seen that many of the Ibero-American nations show little concern for collecting and disseminating the accident figures, despite the knowledge that we have today about the socio-economic impact that the problem represents. The sample was made up of 9 countries, 7 from Latin America and 2 from the Iberian Peninsula. These data are considered to be significant since they represent an average of 80.0% of the population and 85.3% of the GDP (current prices) of all Ibero-America for the period under analysis. (Table 1) shows the data sources used.
In addition, for the development of the study, a database was initially created to collect the information and observation of outliers was carried out, to later conduct an analysis using basic descriptive statistics with the intention of identifying behaviors, trends and generalities in the variables associated with accidents and deaths.  (

Accident rate analysis
To begin with this analysis, it is important to define what an occupational accident is. An occupational accident is any sudden event occurring in the workplace, or resulting from work, where the worker contracts a physical injury that requires medical attention and results in temporary or permanent disability or death (Acevedo and Yánez, 2016). From this definition and remembering that the problem of occupational injuries generates various consequences in both social and economic areas, where those most affected are the workers and their families, it becomes crucial to analyze the behavior of the accident figures in civil construction.
It is important to mention that the data on occupational accidents and deaths used in this study include accidents in the workplace and also in itinere 3 since the figures obtained from official agencies in most countries do not discriminate between the types of accidents.

Number and percentage of accidents of workers in civil construction
Regarding  Finally, it is emphasized that the behavior of the countries, regarding the quantity and percentage of occupational accidents in construction, differs, mainly because the quantity of accidents is an independent variable and the percentage of accidents depends inversely on the number of workers which in turn is strongly correlated with the labor force (PEA) of each country.

Number and percentage of deaths of workers in civil construction
The

Regarding the percentage of deaths in occupational accidents based on the total number of construction workers in each country (PMT), there was an average in the evaluated countries in the period 2013-2017 of 0.01114%, a quite low figure, which indicates an excellent behavior, showing in a certain way an advance in SHAW by all the countries. Such a percentage represents approximately 11 deaths per 100,000 workers. For this data set, a decreasing trend was observed. However, Argentina, Mexico, Brazil and Portugal have the highest percentage of deaths, while Peru and Uruguay again show the lowest values, which makes them the two countries with the best behavior according to the figures analyzed so far (Figure 4). After reviewing the analyses of accident rates and mortality, it is important to mention that, in the countries under study, especially those in Latin America
, it is very difficult to record 100% of the accidents and deaths, since these data are directly related to the accident reporting systems of each country, which sometimes involve numerous paperwork, thus impeding the recording of accidents (Carlos, 2009). Also, it is well known that many of the accidents in the construction industry are left in the shadows because of the evident lack of transparency, especially in low-level positions. Given the above, it should be noted that the conditions of the study could be affected, resulting in some inaccuracy in the estimates.   (Rojas, 2007). Usually, two well-known methods are used. On the one hand, the Pearson's method specifically studies the linear correlations, for those quantitative variables that comply with the condition of normality and continuity (Filho and Júnior, 2009). On the other hand, the Spearman's method studies linear and non-linear correlations, between any type of quantitative variable, without considering special mathematical conditions, being a less restrictive method in its application (Restrepo and González, 2007).

Correlations between variables Establishing correlations between two or more variables is used to determine whether a mathematical association exists in their behavior and to know to what extent this interaction occurs
Specifically, this study uses the Pearson correlation when the two variables to be associated were set as normal. On the contrary, where only one or none of the variables to be correlated were set as normal, the Spearman correlation was used. In order to detect possible non-linear correlations between the variables that showed normality, the Spearman's method was also applied to them. Given the above, first, the normality of the variables mentioned in the methodology was determined through the Shapiro-Wilk test. This method was used because the samples had less than 30 observations (Lopes et al., 2013). More precisely, 5 normality analyses per variable were carried out, one for each year under study. For this test, a 95% confidence interval for the mean was considered and only the variables that met this condition in all years under study were assumed to be normal.
In this sense, the APIB, the HDI, the percentage of accidents (PAT) and the percentage of deaths due to occupational accidents (PMT) were established as normal variables. Therefore, the Pearson's method was initially used to determine correlations between them. Based on (Table 3), values in the ranges -1.0 to -0.40 and 0.40 to 1.0 were considered to be relevant correlations. (Table 4) (Campos et al., 2009). (Adapted by the authors).

However, by using Spearman's method, two of the correlations already analyzed through the Pearson's method were determined to have a higher correlation index. This indicates that the correlation between them is better adjusted to a form other than the linear method. These correlations were between the HDI and the percentage of accidents of workers (PAT) and between the HDI and the APIB.
Other correlations were also found through the Spearman's method, where a strong and direct correlation between PEA and companies and employers in the construction industry (CEC) stands out, pointing out that the two variables grow together, but not at equal rates. (Table 5) shows the results of the variables with significant correlations found through the Spearman's method. Again, based on the values in (Table 3), moderate, strong and very strong correlations were considered as correlations to be taken into account.

Ibero-American classification of occupational accidents in the civil construction industry
This section presents 5 classifications of occupational accidents in the civil construction industry; one for each year of the evaluated period. Also, there is an average classification that summarizes the general behavior for the topic in the 5-year period.
Four variables were considered in the creation of the classifications: APIB, HDI, percentage of accidents (PAT) and the percentage of deaths from occupational accidents (PMT). These criteria were taken into account because they were the most significant criteria for occupational accidents and mortality in the civil construction industry, according to the analysis made in section 3.2.
As mentioned in chapter 2, the TOPSIS method was used to establish the classifications. Bearing in mind that the implementation of this method requires assigning a percentage weight to each of the considered criteria, the decision was made to apply the AHP method to determine the value of each of the referred weights.
The AHP method, developed by Professor Thomas Saaty, is one of the most widely used multi-criteria methods in decision-making support for problems with various perspectives (Marins at al., 2009). Initially, for the implementation of the AHP method in this study, a comparative matrix (CM) was created in order to establish the importance of each of the 4 aforementioned criteria. The CM was established under the parameters indicated in (Saaty, 2008) fundamental scale, in which values from 1 to 9 are given to each position of the matrix, depending on the association of the criteria. Here, the values assigned to each CM position were determined by the authors of this study, based mainly on the degree of correlation between variables.
With the CM ready, a normalized comparative matrix (NCM) was created which resulting from dividing each component of the CM into the sum of its respective column. Consequently, the priority eigenvector was calculated by finding the arithmetic mean of each row of the NCM. The value of each of the components of the eigenvector represented the percentage weight of each parameter involved. (Table 6) shows these values.
To complete the application of the AHP method, the consistency of the CM proposed by the authors was verified through the consistency index (CI), using the eigenvalue as a basis. The calculation of the CI was given by an equation proposed by (Saaty, 2008), in which the eigenvalue and the number of criteria evaluated are related. The application of this equation resulted in a CI of 0.0468. Finally, the CI had to be transformed into the consistency rate (CR), obtained by dividing the CI by a constant depending on the size of the matrix. The value of this constant in this case was 0.9, giving a result of 0.0520 for the CR. Based on the above, according to (Saaty, 2008), the CM is considered to be consistent, as well as the results from the application of the AHP method, if the CT is less than 0.10. In this case, the consistency of the results has been verified (Gomede and De Barros, 2012).  The TOPSIS method was developed by Ching-lai Hwang and Paul Yoon to evaluate the performance of different alternatives according to a series of criteria. This technique is mainly based on the approach to the ideal solution and the distance to the non-ideal solution, providing a good basis for the creation of classifications without the need to use advanced software (Valladares, 2011).

Correlated Variables Correlation Value Correlation Type
The process to determine each of the classifications with the TOPSIS method is detailed as follows. First, the decision matrix (DM) was established, where the 4 criteria initially chosen were placed in the columns (HDI, APIB, PAT, PMT) and the alternatives were placed in the rows, which in this case were represented by the countries under analysis. Then, the corresponding data were entered in each position of the matrix, resulting in a 9x4 matrix. Subsequently, the column vector norm was calculated to establish the normalized decision matrix (NDM), obtained by dividing the components of the DM into the norm column for each component. After that, each column was assigned the corresponding percentage weight, according to each criterion. These weights were previously determined using the AHP method. Each component of the NDM was then multiplied by the percentage weight to obtain the weighted normalized decision matrix (WNDM) (Da Costa and Junior, 2013).
In the same sequence, with the determination of the WNDM, the ideal solution and the non-ideal solution could be obtained, assuming that the results closest to the ideal solution were associated with the countries with the best accident behavior, which translates into fewer accidents and deaths. For this reason, the lowest values associated with the percentage of accidents (PAT) and the percentage of deaths (PMT) criteria were considered for the ideal solution. The same applies to the HDI because of its direct correlation with the percentage of accidents at work (PAT). For APIB, the highest value was considered to be the best result for the ideal solution as this is inversely correlated with accident rates, which means that the higher the APIB, the fewer accidents, and deaths, probably because the higher the production, the greater the investment in safety in the construction sector. As for the nonideal solution, the opposite values to the positive ideal solution were considered.
Based on the ideal and non-ideal solution, it was possible to determine the distances to these solutions presented by each country, which in turn allowed the calculation of the relative proximity (RP) to the ideal solution, which is a value dependent on the previously calculated distances.
To complete the application of the TOPSIS method, the classification was made with the resulting value of the RP, with the countries in the top positions showing the best performance in terms of occupational injuries. In contrast, the countries in the last positions were the ones that showed the worst behavior, presenting the highest rate of deaths and accidents (Krohling and De Souza, 2011). Based on the above, (

Conclusions
Based on the results obtained in this study, it can be concluded that, although the countries analyzed show some differences in their policies, in their level of development and in the way they collect the data, in general, all of them show concern for the problem of occupational injuries and present strategies to reduce them. This is reflected in the research results, where a decrease in the number of accidents and deaths was observed.
However, it should be clarified that despite the analyzed data were obtained from official agencies, it may not fully represent the reality of the industry because of the great informality and outsourcing existing in the construction industry that to some extent distorts the collection of data by the sources. For the specific case of Argentina, Chile and Portugal, the organization and the greater number of criteria presented in their statistics should be recognized.
Considering all of the above, it is recommended that the quality of data associated with occupational accidents be standardized and enhanced in all countries since the same parameters would be available for studies, which would provide a more uniform picture. Similarly, it is suggested that SHAW entities in each country should be stricter in verifying compliance with the laws to maintain and continue to decrease the number of contingencies since there are still significant gaps between what the law recommends and what is actually applied in the workplace.
It is important to remember that the problem of accidents is partly a result of the great risk implicit in the work of civil construction. However, the vast majority of accidents can be avoided if companies comply with the law and provide the basic means for prevention, if the state intensifies controls and if workers follow preventive measures, respect safety standards and are aware that physical integrity and life are at stake.
Finally, it is suggested to continue this work including more countries of the region and segmenting the sample in the construction sub-sectors, aiming at identifying the sub-sectors that present more influence on the problem. It is also recommended to extend the period of study to check whether the improvement seen in the period 2013-2017 continued in the following years.