1 Objective

In an era of heightened globalization, the growing interdependence among nations has intensified cross-border environmental and socio-economic interactions. This has blurred traditional geographical boundaries, leading to effects that extend beyond the national scope from which they originate, commonly referred to as “spillover effects” [1,2,3]. A holistic evaluation of policy measures, considering both direct and indirect impacts, is crucial for mitigating negative outcomes and amplifying positive effects [2, 4, 5].

Among those spillovers, the ones related to international trade flows (or Global Value Chains (GVC)) pose distinct challenges. First, they necessitate a rigorous and comprehensive network analysis for an accurate evaluation. Additionally, these flows generate a variety of externalities (social and environmental), affecting local to global scales. Multi-Regional Input–Output (MRIO) models are a powerful macroeconomic analytical tool to tackle these complex issues [6]. MRIOs provide insights into the direct and indirect externalities of producing, transforming, and supplying commodities to end-users [7, 8]. Recent accounting framework developments have unlocked the capacity to trace those MRIO models’ complex networks, precisely measuring nations’ domestic, imported, exported, and transported footprints [9, 10]. While MRIOs have progressed in quantifying environmental footprints [11,12,13,14,15], their capability for measuring social footprints remains limited. The methodology remains nascent [16] despite methodological frameworks dating back to 2009 [17] and some developments [18,19,20,21]. This gap in measuring social impacts affects all models measuring the different aspects of sustainability [22]. Thus, there is an urgent need to improve social indicators within these sustainability models.

This work enhances the MRIO model EXIOBASE by creating a broadly applicable database that incorporates updated, nation-specific fatality data, addressing a notable metric gap for work-related injuries and fatalities. As shown in the figure listed in Table 1 and detailed in Koundouri et al. [23], this dataset is useful to underline the social impacts associated within a particular GVC. It offers tools for evaluating global social effects in industrial ecology, conducting cost-analysis in policy economics, assessing the true impact of global value chains for trade economics, and applying theoretical frameworks like the “Decarbonization Divide” [24] in geographical and social studies. Moreover, the study pioneers in quantifying health losses resulting from diseases and injuries, including those arising from both national and international spillovers, at the granularity of the economic sector. Thus, offering a valuable tool to enhance health systems through targeted prevention strategies, comprehensive assessment, and rigorous evaluation.

2 Data description

Ensuring social data’s reliability is essential in accurately evaluating social and economic impacts across geographical locations, economic sectors and stakeholder categories. Yet, the MRIO model used in our research, EXIOBASE, faced challenges due to outdated and significant estimates in labor data, labor types, and vulnerable employment ([25], SI7), leading to potential inaccuracies in employment and consequently in work related fatality statistics. We have comprehensively revised EXIOBASE fatality data to address this shortcoming, incorporating detailed, nation-specific, and up-to-date data.

The update includes work-related fatal occupational injuries as well as fatalities associated with occupational exposure to a variety of 17 hazardous substances and conditions such as: asbestos, arsenic, benzene, beryllium, cadmium, chromium, diesel engine exhaust, formaldehyde, nickel, polycyclic aromatic hydrocarbons, silica, sulfuric acid, trichloroethylene, asthmagens, particulate matter, gases and fumes, noise, ergonomic factors. Our methodological process is built on three pillars: data acquisition, raw data processing, and computation of fatal injuries by country, gender, year, and EXIOBASE economic sector.

Data were sourced from the global monitoring report “WHO/ILO Joint Estimates of the Work-related Burden of Disease and Injury, 2000–2016” [26] and Eurostat databases [27]. The WHO/ILO database meticulously compiles national data on work-related injuries and fatalities, offering comprehensive details for the years 2010 and 2016. This compilation is thoroughly documented in the global monitoring report. To summarize, it primarily integrates data from two key sources: the Global Burden of Disease Study [28] and the Institute of Health Metrics and Evaluation. The estimation of occupational risk factors is conducted using the Comparative Risk Assessment framework [29, 30]. Last, the methodologies and data sources utilized in deriving these estimates are rigorously reported, adhering to the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) standards proposed by Stevens et al. [31], ensuring the reliability, transparency, and accuracy of the WHO/ILO database. Eurostat data provides more granular information: work-related fatalities classified by economic activities in the European Community (or NACE Rev.2 [32]) and detailed per year. The strategy for allocating those injuries and fatalities depended on the countries’ geographical location and level of income.

For European countries: Fluctuations in fatality numbers within the NACE Rev.2 sector mirrored the changes registered by Eurostat.

For non-European countries: Injury and fatality numbers were distributed proportionally by economic sectors based on the NACE Rev.2 classification, considering a country’s workforce size and each NACE Rev.2 sector’s accident susceptibility. The injury and fatality count for a specific non-European country in the WHO/ILO database, per year and NACE Rev.2 activity, were derived by applying these ratios to that country’s total injuries and fatalities.

Geographical data gap: Due to incomplete geographical coverage in the WHO/ILO datasets from countries in Asia, America, and Africa, our study employed a combined regional and income-based methodology. This involved calculating fatality ratios across various NACE Rev.2 categories, aligned with each country’s World Bank income level and geographical classification. We achieved this by integrating existing data from countries with available information and estimating the missing data based on the income level of each absent country. A key example of this method is illustrated with American Samoa (ASM). The fatality data for ASM is neither included in the overall data for the USA nor in the WHO/ILO datasets. To address this gap, we identified ASM’s classification according to the World Bank, which is ‘High income’ in the ‘East Asia & Pacific’ region. Based on this classification, we calculated the average numbers of injuries and fatalities for countries in the ‘High income/East Asia & Pacific’ category, using the available WHO/ILO data, weighted by the population of this group. This calculation provided us with a representative figure, that we attribute to ASM (proportionally to its population), aligning it with similar countries in its income and geographical group.

Temporal data coverage: The WHO/ILO database covers exhaustively only 2010 and 2016. Given the closeness of these years and minor changes in national workforces annually, we assumed a linear trend in fatalities and injuries between these years. We have tailored the extension of data coverage for the years 2008, 2009, 2017, 2018, and 2019 based on the geographical location of the countries involved. For European countries, we directly utilized figures and trends from the WHO/ILO data, matching them with the comprehensive data available in the Eurostat database. This approach ensures that the European data is robust and directly reflects the trends observed in the region. For countries outside Europe, we adopted a different strategy. Here, the trends for each economic sector were aligned with the overall population growth of the country. This method is based on the assumption that within a relatively short time frame (less than 3 years), the proportion of a country’s population working in each economic sector remains relatively stable. It's important to note that this assumption might not hold for longer periods, where more significant shifts in sectoral employment could occur).

The result is a comprehensive database that includes the number of fatalities (expressed in the number of deaths for work-related fatal occupational injuries and Disability-adjusted life years (DALYs), for fatalities associated with occupational exposure to a specific risk factor), detailed at the country, gender, and NACE Rev.2 sector levels from 2008 to 2019, providing insights into work-related fatal injuries across different health effects and geographical regions (cf. Table 1).

Table 1 Overview of data files/data sets
Full size table

3 Limitations

In contrast to existing datasets on the subject, our dataset boasts comprehensive coverage across all countries and economic sectors, identified by the NACE Rev.2 classifications. However, it is essential to note certain limitations. The WHO data that served as a foundation for our research lacks exhaustive geographic representation, necessitating estimation techniques for regions within Asia, America, and Africa. To address this, we implemented a regional approach by calculating fatality ratios for each NACE Rev.2 category, using data from countries available for a reference year. These calculations were further refined through stratification by both geographic location and developmental stage, based on the different World Bank levels of incomes. While preliminary tests suggest that these estimates closely approximate actual figures, there remains a pressing need to address this data gap in future iterations of the dataset.

The Global Burden of Disease (GBD) studies conducted by the Institute for Health Metrics and Evaluation (IHME) are pivotal in providing detailed health analysis. They offer a comprehensive view of mortality and disability, categorizing data by country, time, age, and sex. This also explains why the IHME Global Burden of Disease database is one of the main sources of information for the WHO/ILO database. It would be relevant to integrate further data from the IHME GBD database. However, currently, the lack of sector-specific information in this database restricts our ability to further use it in the MRIO context. Looking to the future, the inclusion of economic sector data in subsequent updates of the IHME GBD database would be a significant boon to our research. Should such data become available, integrating it will be a top priority for enhancing our work. This addition would provide a more complete and nuanced perspective on the health risks associated with various economic sectors.

Regarding temporal coverage, the WHO data provided discrete points for the years 2010 and 2016. Recognizing the significance of continuous data for MRIO analyses, we assumed a linear progression in fatality counts between these 2 years. This assumption is predicated on the relative stability of working populations within economic sectors over short periods. Nevertheless, for more nuanced insights, future refinements could employ the ILOEST model [33, 34] from the International Labour Organization (ILO).

In our analysis, we examined the “39 established pairs of occupational risk factors and health outcomes” from the WHO/ILO database [26]. We excluded two recently added pairs by WHO/ILO: Stroke and ischemic heart disease linked to prolonged working hours. This dataset’s quality and integrity is high, as validated by extensive reviews [35]. However, while substances like sulfuric acid are confined to specific industrial sectors in EXIOBASE with relatively uniform exposure rates, extended working hours affect nearly all of EXIOBASE’s 163 sectors, each with distinct characteristics and exposure rates. This variability poses significant challenges in accurately distributing fatalities across the different economic sectors. Thus, for long working hours, adopting a similar approach where fatalities are allocated based on the population of each economic sector would skew the results for stroke and ischemic heart. Therefore, to maintain the integrity of our analysis, we have chosen to omit these conditions. Consequently, parsing the DALYs for this risk demands intricate sociological, epidemiological, and statistical evaluations. This inclusion will be crucial for future database updates.