4.1. Compared with Previous Studies
To improve safe production in the construction industry, many scholars worldwide have focused on conducting statistical analysis of construction accidents. Park et al. collected 675 cases of fatal construction accidents in the Korean construction industry from 2007 to 2013 [
41]. Chong and Low reported 42,775 accidents in the Malaysian construction industry from 2000 to 2009 [
20]. Soltanzadeh et al. selected 500 accidents in 13 of the largest Iranian construction projects from 2009 to 2013 [
42]. Lombardi et al. extracted and analyzed a sample of 116 fatal construction accidents in Italy from 2002 to 2015 to investigate accidents due to electric shocks in this industry [
43]. From 2012 to 2016, more than 2850 construction workers lost their lives due to construction activities in China, with an average value of 1.57 deaths per day [
10]. Moreover, from 2010 to 2019, approximately 7275 construction employees lost their lives due to construction activities in China, with an average value of 1.99 deaths per day, indicating that safety within the construction industry has deteriorated. However, no previous studies have provided an explanation of this phenomenon [
10]. Under the wave of urbanization in China, many construction projects are necessary to satisfy citizens’ expanding demands, particularly their residential needs. These activities are accompanied by fatal construction accidents. Importantly, we focused on the characteristics of major and severe construction accidents, for example, by providing statistical descriptions of major and severe construction accidents by year, month, day of the week, and accident type. The study by Shao et al. [
10] did not include these analyses.
The numbers of construction accidents and deaths during the study period were 6005 and 7275, and the ratio of deaths to construction accidents was 1.21. According to the National Bureau of Statistics [
1], the numbers of traffic accidents and deaths in China in 2019 were 247,646 and 62,763, and the ratio of deaths to traffic accidents was just 0.25. Compared with drivers and passengers, construction workers consistently face more dangerous situations. In other words, the construction industry is more dangerous than other industries [
40,
44].
Once an accident has occurred in a company, the company faces fines and the suspension of work and production, which greatly affects the company’s image and efficiency. Although accident reporting is mandated by the government, it is impossible for all accidents to have been reported [
45]. According to Heinrich theory, there are 29 minor-injury and 300 no-injury unreported accidents for each major-injury accident reported [
46]. There may be more no-injury and minor-injury construction accidents that remain unreported, but we cannot find any records on such accidents.
Previous researchers have reported that most construction accidents involve falls from a high place. In the US, most construction workers who lost their lives fell from a high place [
6], which is in line with the findings of our study (
Figure 6). Moreover, Moniruzzaman and Andersson reported that falls from a high place were more common in Sweden among private companies than among state-owned companies [
47]. Most construction activities occur in high places, and employees have to carry out their corresponding operations in these places where more hazards exist, leading to falls. As stated in the Introduction, the causes of construction accidents may vary from country to country [
32]. In the case of South Australia and Malaysia, falling from a high place is not the main accident type [
20,
44]. Previous studies have focused only on the type of construction accidents, and no accident type analysis of major and severe accidents in the construction industry has been conducted [
6,
10]. Collapses are the most common type of major and severe accidents in the construction industry (
Figure 10), indicating that collapses can easily lead to mass deaths and injuries.
Although the location of accidents is an important factor in improving safe production in the construction industry, previous studies have not paid attention to this issue [
10,
25]. In this study, we found that the greatest number of construction accidents in China occurred along openings and edges (
Figure 7). Interestingly, vehicles such as trucks cause the most accidents in Malaysia [
20]. However, no explanation for this phenomenon was presented. The reasons for this difference are as follows. First, the statistical methods and dimensions involved in analyzing construction accidents in China and Malaysia are different. Second, the statistical periods used in Chong and Low’s study [
20] and in our study are from 2000 to 2009 and from 2010 to 2019, respectively. Since the science and technology related to construction activities have developed rapidly, the construction industry’s mode of production has also changed.
In general, the more economically developed the provinces (such as Jiangsu, Zhejiang, and Guangdong) are, the higher the number of fatal accidents these provinces experience, which is in line with the findings of Moniruzzaman and Andersson [
47]. To investigate the geographical distribution of construction accidents in China, previous scholars have adopted similar methods, but their analyses have been based on different factors. Shao et al. first determined the geographical distribution of the number of fatal accidents [
10]. To measure the sustainability of economic and social development, they then determined the geographical distribution of the mortality rate per hundred million yuan of GDP. Although their results were similar to those of this study, the analytical procedures in this study are more comprehensive. We take into account not only the number of fatal accidents and the economic development level, but also the number of deaths and the number of employees (
Table 4).
In terms of the numbers of construction accidents and deaths, August ranks first, while February ranks last [
10], a finding that is in line with that of our study. The temperature in August generally exceeds what employees can endure; moreover, they have to work to complete construction activities. In February, cold weather and the Chinese Spring Festival hinder construction activities.
4.2. Implications
According to accident cause theory [
48], accidents can be attributed to the unsafe behaviors of employees and the unsafe status of objects. The unsafe behaviors of employees mainly include commands against rules, operations against rules, and artificial operational errors. The unsafe statuses of objects mainly include incomplete or defective safety protection equipment, construction activities under severe weather conditions, and construction activities under insufficient lighting conditions.
To prevent the unsafe behaviors of employees, the following measures can be adopted. For example, operating procedures can be formulated, and those who violate regulations can be punished. Regular health checks of employees can be conducted. Education and training for employees can be increased, The necessary safety equipment can be worn. The number of work site safety inspections can be increased, and violations can be addressed in a timely manner.
To prevent the unsafe statuses of objects, the following measures can be adopted. For example, safety protection equipment can be used based on operational requirements. The status of safety protection equipment can be regularly checked. Construction operation specifications for abnormal weather can be formulated, and construction activities under extreme weather can be prohibited. The environmental conditions required for on-site construction operations can be ensured.
In construction projects, most aerial work occurs at openings and edges. Therefore, most construction accidents were located along openings and edges (
Figure 7). Countermeasures should be adopted to protect construction employees working along openings and edges, for example, the installation of guard railings and safety nets, as well as the establishing of safety marks. Previous studies have focused on the bodily location of injuries from construction accidents. Dumrak et al. reported that the most commonly injured part of the body among South Australian construction workers was the trunk, accounting for 28.4% of all injuries, followed by hands and legs, accounting for 20.4% and 11.9% of all injuries, respectively [
44]. This study focuses on the site of construction accidents and complements the findings of previous studies, and it should help construction workers better understand construction accidents.
4.3. Limitations
To simplify the discussion, this study conducted only a statistical analysis of construction accidents. Further research should focus more on the evolutionary mechanism and prevention of construction accidents. First, the risk of construction accidents can be determined by the accident frequency and the severity of accident consequences. By dividing the number of construction accidents by the total number of construction enterprises, the accident frequency can be obtained. The severity of accident consequences can be determined by the number of deaths, the number of injured employees, or the amount of property losses in a construction accident. Second, the sensitivity of and inherent randomness in the evolution of construction accidents might be investigated using chaos theory [
49]. Third, with the help of a synthetic theory model [
50], the causes of construction accidents could be analyzed by taking into account the national policy and production environment. Fourth, to determine the basic events in construction accidents, fault tree analysis could be performed [
38]. Fifth, for the key basic event identified, the bow-tie model could be adopted to reduce the risk of construction accidents by implementing corresponding safety measures [
51].
In addition, the initial condition of the first value of the first-order accumulating generation operator (1-AGO) data was chosen as only the first value of the original data in the GM(1,1) model [
27]. Future research should focus on the influence of the initial condition on the prediction results.