0% READ
FULL TEXT
INTRODUCTION
Occupational Health and Safety (OHS) disclosures play a vital role in shaping public perceptions of corporate working conditions. As organizations strive to create safer and healthier workplaces, the voluntary disclosure of OHS procedures has become an essential element of corporate transparency and responsibility (Rikhotso et al., 2022a). These disclosures not only ensure the well-being and safety of employees but also contribute to the overall sustainability of organizations. Despite its importance, the integration of OHS into sustainability reporting remains a relatively new and evolving practice (Kawashita et al., 2005). Many companies recognize OHS as both a legal obligation and a key component of their sustainability strategies, using it to demonstrate their commitment to fostering safe work environments and responsible business practices (Alves and Ramos, 2022a). Through comprehensive OHS disclosures, companies provide insights into their efforts to protect and promote the well-being of stakeholders (Olanrewaju et al., 2022). These disclosures may include a variety of measures, such as OHS training programs, safety protocols, risk assessments and open communication regarding safety issues (Abdul Hamid et al., 2020).
Despite the growing importance of OHS disclosures, there is still limited research on how firms systematically incorporate these practices into corporate reporting. While Hahn and Kühnen (2013) investigated sustainability reports and Monteiro (2021), focused on employee-related disclosures, applying advanced methodologies such as Structural Topic Modeling (STM) to OHS disclosure research remains underexplored. STM has been successfully applied in other fields to uncover hidden thematic patterns-for example, bai et al. (2021) used STM to analyze trends in marine transport research, while Ali et al. (2019) applied STM in their analysis of hospitality management literature. Chen and Xie (2020) used STM for bibliometric analysis in sentiment analysis research, demonstrating its versatility across different domains.
The present study builds on this growing body of research by applying Machine Learning (ML) techniques, specifically STM, to explore key themes in OHS disclosure. Using a keyword search in Scopus, 956 articles published between 1974 and 2024 were identified and analyzed through STM and bibliometric analysis. These methods aim to uncover research objectives, affiliations, focal areas and epistemological perspectives related to OHS. Significantly, this study extends traditional bibliometric techniques by applying STM to identify trends in OHS corporate reporting and disclosure practices.
Moreover, this research goes beyond conventional methods by utilizing the findthoughts() function. (Roberts et al., 2019) This allows for identifying documents that align closely with thematic constructs derived from the STM model. This innovation enables the researcher to discover specialized publications related to OHS disclosures, thus contributing to a deeper understanding of the current state and future direction of OHS corporate reporting. While prior studies have successfully applied STM in fields like marine transport, hospitality management and sentiment analysis, there remains a significant gap in the literature regarding the systematic use of ML and STM to analyze OHS-specific trends. This study addresses this gap by introducing a novel approach to analyzing OHS disclosure trends, offering theoretical and practical implications for corporate reporting and sustainability practices.
MATERIALS AND METHODS
Research outline
The study builds upon (Bai et al., 2021) by modifying its research process. The first phase involves collecting articles, including selecting the database and keywords to retrieve relevant scholarly articles. The 2nd phase focuses on exploring the dataset and the third phase, STM, encompasses text processing, topic selection, knowledge discovery and identifying research articles related to specific topics derived from the STM model (see Figure 1).
Search protocol
The Scopus database was chosen to collect scholarly articles. A keyword search was performed using the following keywords: TITLE-ABS-KEY (occupation* AND health AND safety OR ohs AND disclos* OR report*.) AND (LIMIT-TO (DOCTYPE, “ar”) OR LIMIT-TO (DOCTYPE, “re”) OR LIMIT-TO (DOCTYPE, “ch”) OR LIMIT-TO (DOCTYPE, “cp”)) AND (LIMIT-TO (LANGUAGE, “English”)) A total of 956 scientific articles were collected from the period spanning from 1974 to 2024, covering 50 years of research papers.
Structural Topic Modelling
Topic modeling is an unsupervised algorithm that extracts latent themes from textual documents. These themes are learned based on the patterns and relationships within the text data without explicit supervision or predefined labels (Bai et al., 2021). Two commonly used methodologies for topic modeling are Latent Dirichlet allocation (LDA) and STM (Jelodar et al., 2019; Li and Lei, 2021).
STM offers several advantages over LDA (Bai et al., 2021). Firstly, STM allows for incorporating document-level covariates, which provide additional context and can improve the accuracy of topic modeling. For example, by including metadata about the author, publication date, or source of a document, STM can account for topic prevalence or word use variations based on these covariates. STM facilitates a better understanding of the relationship between topics and the underlying factors influencing them. Secondly, STM models give correlations between topics, providing deeper insights into how topics relate.
Additionally, STM allows for modeling variations in word distributions within a topic based on covariates and this means that the words associated with a particular topic may vary depending on the context provided by the covariates. This flexibility allows for more precise and context-specific topic modeling, leading to a better understanding of the data. Furthermore, STM provides the advantage of directly assessing the impact of covariates on topic prevalence or word use within a topic (Bai et al., 2021) and it provides more insights into how different factors impact the distribution of topics and the usage of words within those topics (Slade, 2016). (Figure 1) shows the 4 computational steps for STM modeling: text processing, selection of topics, knowledge discovery (Bai et al., 2021) and results.
Step1: Text processing
Inbuild text processor () function in the STM model, which will automatically create a corpus, convert to “Lower Case,” “remove punctuation,” “remove stop words,” “remove numbers,” “stemming,” and Create output for the following procedure (Küsters and Garrido, 2020).
Step 2: Topic selection
It is critical to find the appropriate number of topics. However, no single, generally acknowledged method exists for determining the ideal quantity of topics for STM estimations. (Ravenda et al., 2022). The search K () function determines the ideal number of topics for the model. It combines statistical methods with human judgment to find the optimal balance. By integrating coherence scores and expert input, this approach ensures that the model accurately reflects the complexity of the data, enhancing its interpretability and usefulness (Roberts et al., 2019). We begin by applying rigorous statistical diagnostic tools to examine dynamic fluctuations in semantic coherence and other relevant metrics to assess the goodness of fit of STM estimates. These evaluations are conducted throughout a spectrum of topic quantities ranging from 3 to 25 (Figure 2).
The held-out likelihood score increases as the number of topics increases. The scores rise steadily from k=3 to k=11, but after k=11, they fluctuate. The score improves again at k=16, with k=19 providing the best model fit. However, semantic coherence tends to decrease as the number of topics increases, with fewer topics generally having higher coherence (Roberts et al., 2019)This does not fully capture the model’s complexity. A better goodness of fit correlates with lower residuals and higher lower bounds. We chose k=11 for further analysis, as it best fits the first phase. Beyond this, the model risks overfitting and generating redundant or meaningless topics. Thus, k=11 balances fit and semantic coherence, outperforming k=19 in topic quality. While semantic coherence is important, it only measures internal consistency (Mimno et al., 2011) We followed Roberts (2014, 2019) and Wang (2022) to finalize the optimal number of topics, use labelTopics() and plot() to ensure meaningful topic labeling.
After selecting k=11, we fit the model accordingly. The plot Models () function averages all topics from each model run and plots them numerically. The model with favorable qualities in both semantic coherence and exclusivity will be chosen, with models scoring in the upper right corner of the plot considered for further analysis. (Roberts et al., 2019)The plot Models () function also displays each topic’s quality in the model (Figure 3).
Step 3: Knowledge discovery
Various methodologies can be used comprehensively to investigate the output of topic models, including exploring word associations with topics and examining the relationships between metadata and topics. (Roberts et al., 2019). Researchers frequently employ several key functions from the Structural Topic Model (STM) toolkit to obtain and visualize results, such as label Topics (), cloud (), topic Corr () and estimate Effect ().
The label Topics () function provides a detailed overview of the characteristics of each topic, including the highest probability words, FREX (Frequency×Exclusivity) words, lift words and score words (see Table 1). While the highest occurrence probability for words conditional on the topic offers insights into word association, certain common terms such as “study,” “research,” “management,” and “data” may lack specificity and provide limited value for topic interpretation (Bai et al., 2021)The FREX metric is crucial to address this as it identifies more distinctive and meaningful associations between terms and topics. (Bai et al., 2021; Bischof and Airoldi, 2012; Roberts et al., 2019) By applying the harmonic mean, FREX balances the influence of frequency and exclusivity, ensuring a more refined and relevant assessment of a word’s contribution to a specific topic domain.(Bischof and Airoldi, 2012). Equation 1 illustrates the FREX formula, as referenced by Bai et al. (2021) and W. Wang et al. (2023).
| Topic | Topic label | Representing words | mean |
|---|---|---|---|
| Topic 1 | OHS Auditing in Construction Supply Chains | Highest Prob: use, study, research, manag, data, construct, industri FREX: construct, chain, suppli, implement, audit, techniqu, data. | 10% |
| Topic 2 | Sustainable Business Practices | Highest Prob: social, sustain, csr, corpor, respons, compani, develop FREX: csr, corpor, sustain, social, respons, mine, stakehold. | 12% |
| Topic 3 | Healthcare Accessibility and Support for Mental Health and Disabilities | Highest Prob: health, care, profession, peopl, mental, support, communiti FREX: sex, care, privaci, disabl, mental, peopl, access. | 8% |
| Topic 4 | Pharmaceutical Hazard Management | Highest Prob: food, research, product, consum, medic, health, receiv FREX: food, grant, pfizer, honoraria, speaker, nanoparticl, hia. | 2% |
| Topic 5 | Regulatory Frameworks and Global Rights | Highest Prob: right, book, global, human, regul, state, reserv FREX: actor, law, cybersecur, privat, regulatori, reform, state. | 10% |
| Topic 6 | Health Studies and Safety Measures | Highest Prob: studi, use, health, result, exposur, safeti, patient FREX: ipv, self-report, preval, smoke, vaccin, injuri, stress. | 7% |
| Topic 7 | Environment Risk Assessment | Highest Prob: environment, chang, risk, system, assess, climat, sector FREX: lean, pollut, climat, wast, energi, ocean, plastic. | 7% |
| Topic 8 | Occupational Safety and Health Regulations | Highest Prob: safeti, health, occup, worker, risk, hazard, chemic FREX: chemic, hazard, osha, avail, abstract, accid, safeti. | 14% |
| Topic 9 | Employee Well-being and Workplace Dynamics | Highest Prob: employe, work, studi, relationship, effect, workplac, organiz FREX: mediat, organiz, employe, compens, negat, creativ, employees. | 8% |
| Topic 10 | Promoting Occupational Health and Safety Education | Highest Prob: book, research, student, provid, social, chapter, ethic FREX: edit, student, handbook, tourism, sociolog, reader, chapter | 13% |
| Topic 11 | Corporate Reporting and Disclosure Practices | Highest Prob: report, disclosur, inform, compani, studi, paper, indic FREX: annual, sdgs, report, gri, sdg, disclosur, legitimacy. | 8% |
Equation 1: FREX metrics
In addition to examining word associations, the cloud () function generates word clouds for each topic, providing a visual representation of the most frequent terms (Figure 6). The word cloud highlights the theme classified within each topic, with words sized according to their probability of occurrence. This visualization can help quickly identify the dominant themes across topics.
Moreover, the topic Corr () function allows for the exploration of topic correlations, shedding light on how topics are related to each other within the model. Such correlations can reveal underlying thematic linkages that may not be apparent through isolated topic analysis.
The estimate Effect () function enables researchers to examine the impact of metadata (such as publication year, region, or author affiliation) on topic prevalence. By doing so, the function provides insights into how the prominence of certain topics evolves over time or varies across different contexts. This study also leverages bibliometric analysis to assess publication trends, such as the most frequently cited journals and the most significant contributing journals within the domain of OHS disclosure. This analysis complements the STM results, providing a comprehensive understanding of the scholarly landscape related to OHS reporting.
RESULTS AND DISCUSSION
Most Contributed Scholarly Journals in Occupational Health and Safety Disclosure
Cumulative scholarly article production is a key metric for identifying the most influential journals in the field. Appendix 1 highlights the top 20 journals contributing significantly to OHS reporting and disclosure research. The Journal of Cleaner Production leads with 23 articles, representing 3% of the total output. Safety Science and Sustainability (Switzerland) follow, each with 19 articles, making up 2.12% of the total. Notably, out of over 704 journals identified, most have contributed only one article, while 34 journals have published more than three articles each
Overall Publication Trends
Figure 4 shows the trend in annual scientific production on occupational health and safety from 1974 to 2024, measured by the number of documents published each year. From 1975 to 1990, production was low and stable, with fewer than 20 documents annually. The early 1990s marked the start of an upward trend, peaking around 1995, followed by a slight decline and a subsequent rise in the late 1990s. The 2000s saw a gradual increase in document production, with fluctuations but a general upward trajectory. This trend accelerated sharply around 2010, leading to a significant rise in annual publications. The most notable surge occurred in the early 2020s, peaking at over 100 documents in 2022 before a decline. Overall, the data reflect substantial growth in occupational health and safety research, particularly from the 2000s onward, with a peak in the early 2020s.
Most Cited Paper
Appendix 2 highlights the most cited articles, including “Corporate Social and Environmental Reporting: A Review of Literature and a Longitudinal Study of the UK” by Gray et al. (1995), published in the Accounting, Auditing and Accountability Journal. The article explores integrating OHS concerns into corporate social and environmental reporting, focusing on the UK. It emphasizes the importance of OHS reporting for meeting stakeholder expectations and regulatory requirements. It examines trends in OHS reporting over time, its impact on corporate reputation and performance and challenges like data collection and standardization. The next most cited work is Patton’s (2016) paper, “Our Future: A Lancet Commission on Adolescent Health and Well-Being.”
Most Frequent Word
The analysis of scholarly articles highlights the prevalence of specific key terms in their abstracts. The most frequently occurring words include “health,” “social,” “safety,” “research,” “study,” “CSR” (Corporate Social Responsibility), “management,” “work,” “environmental,” “corporate,” “information,” “companies,” “occupational,” “sustainability,” and “rights.”
To visually represent the frequency distribution of these terms within the corpus, the Tree map () function was used. This visualization provides a detailed depiction of the relative occurrence of each term, offering valuable insights into the predominant themes and focal points within the scholarly discourse on occupational health and safety disclosure (Figure 5).
Topic Solution, Interpretation and Topic Prevalences
Topic labeling can often be simplified by selecting words with the highest probabilistic association with a given topic. However, this approach may limit the quality of the labels. To address this issue, Bischof and Airoldi (2012), introduced the FREX metric, which combines word frequency within a topic and topic exclusivity. FREX is calculated as the harmonic mean of a word’s rank by probability within the topic and its rank by the distribution across topics (Bai et al., 2021). Previous literature reviews using Structural Topic Modeling (STM) have applied both metrics to enhance the accuracy of topic labeling (Bai et al., 2021; Das et al., 2017; Kuhn, 2018).
A word cloud for each topic is generated using the cloud () function (Figure 6), which visualizes words based on their probability of occurrence within the topic. Words with a higher probability appear in larger font sizes, indicating their prominence in the topic. Researchers often use label Topics () and cloud () functions to determine topic labels through expert judgment.
Table show that most articles (14%) focus on occupational safety and health regulation. Promoting occupational health and safety education is closely followed, accounting for 13% of the research. In contrast, the least researched area is Pharmaceutical Hazard Management, with only 2% of articles dedicated to this topic.
Topic, one addresses the OHS Auditing in Construction Supply Chains; the highest frequency words are “construction” and “industry,” and FREX words are “chain,” “supplies,” “auditing,” and “techniques.” Topic two and Topic eleven discuss the corporate role in OHS; topic two addresses issues related to sustainable business practices with the high-frequency words “social”, “sustainability”, “CSR”, “corporate”, “response”, “companies” and the FREX words are “CSR”, “corporate”, “sustain”, “social”, “response”, “mine”, “stakeholder” furthermore, topic 11 related to Corporate Reporting and Disclosure Practices (8%). The highest frequency words are “report”, “disclosure”, “inform”, “company”, “studio”, “paper”, “indic” FREX: “annual”, “sdgs”, “report”, “gri”, “sdg”, “disclosure”, “legitimacy”. Topic three consists of the highest frequency words such as “health”, “care”, “profession”, “people”, “mental”, “support” and “communities” FREX words are “sex”, “privacy,”, “disable”, “access”, all together we can label the topic three as healthcare accessibility and support for mental health and disabilities.
Topic four is related to Pharmaceutical Hazard Management with the highest frequency of “food”, “research”, “product”, “consumer” and “medical,” and the FREX words are “grant”, “Pfizer”, “nanoparticle,”, “HIA”, this is least researched area in the model with 2% of research has done on this field. Topics five and eight address the common area of regulation frame related to OHS; the topic addresses research related to the regulatory framework and global rights. The highest frequency words are “right,” “global”, “human”, “regular”, “state”, “reserve,” and FREX words are “actor”, “law”, “cyber secure”, “privacy,”, “regulation”, “reform”, “state” and topic 8 is about Occupational Safety and Health Regulations with high-frequency words are “safety”, “health”, “occup”, “worker”, “risk”, “hazard”, “chemic” FREX: “chemic”, “hazard, “OSHA”, “avail”, “accid”, “safety”.
Topic six is about health studies and safety measures; the highest frequency words are “health”, “exposur”, “safeti, “patient,” and FREX: “ipv”, “self-report”, “preval”, “smoke”, “vaccin”, “injuri”, “stress”. Topic seven is related to environmental risk assessment with higher frequency word such as “environment”, “chang”, “risk”, “system”, “assess”, “climate”, “sector”, FREX: “lean”, “pollut”, “climate”, “waste”, “energi”, “ocean”, “plastic”. Topic Nine addresses Employee Well-being and Workplace Dynamics hight highest frequency words are “employe”, “work”, “studi”, “relationship”, “effect”, “workplace”, “organiz” FREX: “mediat”, “organiz”, “employe”, “compens”, “negat”, “creativ”, “employees”.
Topic ten is about Promoting Occupational Health and Safety Education; the highest frequency words are “book”, “research”, “student”, “provide”, “social”, “chapter” and “ethic” FREX: “edit”, “student”, “handbook”, “tourism”, “sociology”, “reader”, “chapter”. A key learning objective for sustainability that contributes to creating a safe environment (Moyo et al., 2022).
Topic correlation
Figure 7 shows the correlation between topics. Topic 1, “OHS Auditing in Construction Supply Chains,” is closely linked with Topic 2, “Sustainable Business Practices,” and Topic 11, “Corporate Reporting and Disclosure Practices.” This suggests a connection between auditing in construction supply chains and broader sustainability efforts, highlighting the importance of transparent reporting in promoting sustainable practices. Additionally, Topic 5, “Regulatory Frameworks and Global Rights,” is strongly related to Topic 10, “Promoting Occupational Health and Safety Education,” emphasizing the link between regulatory frameworks and educational initiatives to enhance awareness and compliance, especially in a global context. Similarly, Topic 4, “Pharmaceutical Hazard Management,” and Topic 7, “Environmental Risk Assessment,” are correlated with Topic 5, underscoring the broad scope of regulatory frameworks that cover areas like pharmaceutical hazard management and environmental risk assessment, all tied to global rights and standards.
Topic Trend Analysis
The impact of publication year on topic prominence is a key aspect of using STM in academic research. STM enables the analysis of interactions between variables like publication year and topics of interest. By including publication year as a covariate, researchers can track and compare the prevalence of topics over time, offering insights into their temporal dynamics.
Figure 8 illustrates these dynamics, showing trends in various topics over the years. Topics like “OHS Auditing in Construction Supply Chains,” “Sustainable Business Practices,” “Environmental Risk Assessment,” “Regulatory Frameworks and Global Rights,” “Employee Well-being and Workplace Dynamics,” “Promoting Occupational Health and Safety Education,” and “OHS Corporate Reporting and Disclosure Practices” have shown an upward trend.
In contrast, research focus on “Occupational Safety and Health Regulations” and “Health Studies and Safety Measures” has declined over the analyzed period. This shift suggests that earlier OHS research focused on regulations and safety measures, but as stricter OHS laws were implemented, interest has shifted toward newer topics within the field.
Exploring papers related to topics derived from the STM process
Roberts (2019) used the find thoughts () function to examine documents highly associated with topics and the researcher used the same function to find papers highly associated with the topics, resulting in a niche result. Topic 11 in the topic solution is associated with OHS-corporate reporting and disclosure practices and the results in (Appendix 3) show the scientific articles related to this topic. Research using this function to explore the relevant paper related to OHS-corporate reporting and disclosure is shown in (Table 2) though2<- find Thoughts (select Model, texts=data$Title, n=20, thresh=.25, topics=11) though2.
| Citation | Summary of Article | Sample | Theory | Measure | Major Findings |
|---|---|---|---|---|---|
| (Chan, 1979a) | This is the first scientific study to examine OHS disclosure in annual reports of Fortune 500 companies through content analysis. | Fortune 500 | N/A | Extent and nature of OHS disclosure. | The reporting of workplace injuries and illnesses needed to be more uniform. Existing social disclosure in OHS needs more data to evaluate enterprise performance. |
| (Kawashita et al., 2005) | The study aims to analyze corporate management’s viewpoint on occupational safety and health in Japanese companies through CSR-related reports. | Japanese companies | N/A | Extent of OHS disclosure. | Disclosure levels vary by industry; companies with more employees disclose more. OHS is a minor disclosure item compared to other segments like the environment and society. |
| (Montero et al., 2009) | Explores the presence of OHS concerns in CSR and the role of CSR in fostering a safety and health culture through content analysis of various CSR initiatives in Europe. | N/A | Stakeholder theory | N/A | OHS is key in CSR reporting, but its development varies across instruments. Integration of CSR frameworks with OHS can enhance practices. |
| (Koskela, 2014) | Explores OHS reporting in CSR reports through content analysis of three companies from different industries in Finland. | Three companies-Finland | N/A | Extent of OHS disclosure | Occupational safety was the most reported item in OHS disclosures, with variations among companies. |
| (Searcy et al., 2016) | Analyzes work environment disclosure indicators in CSR using content analysis in Canadian companies, including top CSR leaders and randomly selected companies from the Toronto Stock Exchange. | 100 Canadian companies | N/A | N/A | Top companies report 657 work environment indicators in CSR; safety indicators emphasized, less focus on psychosocial issues like work-related stress. |
| (Cahaya et al., 2017) | Examines factors influencing voluntary OHSD in the annual reports of 223 listed Indonesian companies using a GRI-aligned index. | 223 Indonesia-listed companies. | Coercive Isomorphism | Extent of OHS disclosure | Low disclosure levels, industry profile and MNC operations significantly impact OHSD. ROA has a positive association, while board members have a negative association. |
| (Nagata et al., 2017) | Explores OHS trends in CSR by industry sector and company size in Japanese listed companies. | Japanese listed companies. | N/A | Descriptive | OHS reporting in CSR is increasing, with larger companies reporting more extensively. |
| (Evangelinos et al., 2018a) | Examines voluntary OHSD reporting in top corporations across oil and gas, chemical, aviation and construction industries. | 40 sample companies from Forbes World’s Big Companies. | N/A | Extent and quality of OHS disclosure. | Comprehensive management approach to OHS, but reporting needs more depth beyond basic metrics. Disclosure levels vary across industries. |
| (Kumar et al., 2018) | Critically examines the sustainability reporting practices of the top 10 Indian banks using a GRI-G4-based persuasive communication framework. | Top 10 Indian banks. | Stakeholder theory | Extent and quality of SR. | Inadequate disclosure and transparency in SR among India’s top banks. Weak focus on stakeholder involvement. |
| (Tsalis et al., 2018a) | Examines firms’ accountability for OHS and the influence of institutional factors on OHS disclosure using content analysis and a benchmark scoring framework. | Companies selected from the GRI database | Accountability and Institution theory | Extent and quality of OHS disclosure | Quality and extent of OHS disclosure are superficial, with many firms providing injury rate info to protect social reputation. OHSAS certification enhances disclosure. |
| (Rahman et al., 2019) | Examines trade union views on OHS disclosure and whether it meets their expectations through a mixed-method approach, including 472 samples from trade unions and 40 annual reports for content analysis. | 472 Indonesia’s sample companies. | Stakeholder theory. | Extent of OHS disclosure. | Trade unions see OHS disclosure as crucial, highlighting an information gap between trade unions and companies. |
| (Fonseca and Carvalho, 2019) | Examines sustainability practices of 235 Portuguese firms with QEOHS certificates. | 235 Portuguese firms. | N/A | Disclosure on SDG. | Sustainable practices disclosure influenced by business volume, UN Global Compact Network Portugal membership and corporate website disclosure levels. |
| (Păun et al., 2020) | Examines the influence of the recent legislative framework (EU Directive 2014/95/EU) on non-financial disclosure in Romania, focusing on ownership structure, market share and field of activity. | 35 Romanian companies. | N/A | Extent of OHS disclosure. | Company size shows no significant relationship with OHS disclosure, but turnover does. Media exposure and market share positively influence disclosure levels. |
| (Roland et al., 2020a) | Investigates the determinants of OHS disclosure quality in Romanian companies. | Romanian companies | N/A | Descriptive | OHS disclosure is influenced by corporate debt, age and turnover. |
| (Mariappanadar et al., 2021) | Using a mixed research design, examining the quality of OHS and well-being disclosure in sustainability reports across industries and countries. | Forbes list companies | Institutional theory | Quality of OHS disclosure. | Limited focus on GRI reporting for OHS and well-being. Reporting driven more by legal requirements and social legitimacy than industry or market type. |
| (F. H. Ali et al., 2021) | Evaluates OHS practices in Pakistani manufacturing companies using an exploratory mixed design, with content analysis, benchmarking and statistical techniques applied to 181 firms’ annual reports. | 181 Pakistani companies | Legitimacy theory | Extent and quality of OHS disclosure. | Inadequate OHSD accountability index scores show significant differences in company-specific disclosures, with only 35% disclosing any OHS info. |
| (Mavroulidis et al., 2022) | Explores OHS reporting in multinational construction companies’ CSR reports. | Multinational construction companies listed on the GRI database. | Stakeholder theory | Extent of OHS disclosure | A limited range of OHS issues disclosed; larger companies disclose more and ISO 18001 certification boosts transparency. |
| (Alves and Ramos, 2022b) | Provides insights into the quality of OHS disclosures among EU companies, focusing on key determinants such as regional differences, certification of OHS management systems and external assurance. | European Union companies. | Institutional stakeholder theory | Extent and quality of OHS disclosure. | The average score was 12/30 points for OHS disclosure quality, which was significantly influenced by OHS management system certification. |
| (Shimizu et al., 2022a) | Studies OHS disclosure in CSR and integrated reports, comparing changes between 2012 and 2022 by industry and company size in Japanese companies. | Japanese companies | N/A | Descriptive | A positive relationship between company size and OHS disclosure; the secondary industry has higher disclosure scores than tertiary sectors. |
| (Gardiner, 2022) | Explores work health and safety in the top 100 listed companies on the Australian Stock Exchange, assessing the impact of mandatory reporting on OHS reporting quality. | Top 100 ASX-listed companies. | Stakeholder and legitimacy theory. | Quality of OHS disclosure. | Few companies report WHS metrics, with hazard-prone industries showing higher disclosure levels. Voluntary disclosure was found inadequate for protecting workers. |
| (Y. Wang et al., 2022) | Examines OHS performance in listed Chinese companies through a mixed-method approach, including quantitative data from questionnaires and qualitative data from CSR, SR and employee-related reports. | 69 listed energy companies in China. | N/A | Extent of OHS disclosure. | Low OHS management level in the Chinese energy industry; growing trends in reporting observed. |
| (Guo et al., 2022) | Using a game theory framework, examining how Chinese family firms perform in employee health and safety through information disclosure. | Chinese A-share listed companies. | Game theory | N/A | Family firms are less likely to prioritize employee health and safety, showing inconsistency in reporting. Legislative changes needed to improve reporting. |
| (Rikhotso et al., 2022a) | Examines how South African manufacturing and utility companies use CSR disclosures to report occupational health hazards without a national disease surveillance system. | South African companies. | Institution theory | Descriptive | With company variations, OHS is the least reported social aspect in CSR disclosures. Noise-induced hearing loss is the most reported occupational disease. |
| (Fotiadis et al., 2023a) | Explores OHS disclosure in the UK construction industry, assessing trends in reporting. | Top UK construction companies | Stakeholder, legitimacy and institutional theories. | Extent and quality of OHS disclosure. | Inadequate OHS disclosure in UK construction industry reports, with no standardized practices. Some key issues, like young workers in supply chains, are underreported. |
| (Zhang et al., 2023a) | Examines the impact of work safety information disclosure on corporate performance using a panel data approach, including 144 companies from high-risk industries. | 144 high-risk industries companies (China). | Stakeholder theory. | Extent and quality of OHS disclosure. | OHS disclosure positively influences financial performance and social reputation but is negatively influenced by institutional investor shareholding. |
| (Park and Kim, 2023) | Explores the relationship between corporate characteristics and workplace injury rates across different industries in Korea, applying quantitative analysis to HR data and annual reports. | Korean companies | Human capital theory | N/A | More full-time employees and skilled workers reduce injury rates. Wage increases have varying effects depending on industry, with hazardous industries showing more significant impacts. |
| (Oakman et al., 2024) | The quality and extent of OHS reporting in Australian companies are evaluated through a policy scorecard applied to OHS and non-financial reports. | Top 50 ASX-listed companies | Stakeholder theory | Quality of OHS disclosure | High-risk sectors score higher on OHS reporting; stakeholder pressure drives improved OHS reporting practices. Reporting gaps exist in non-high-risk sectors. |
Thematic evaluation of occupational health and safety disclosure
OHS disclosure has evolved, with early studies like Chan (1979b) highlighting the need for uniformity in reporting within Fortune 500 companies. Initial research focused on basic metrics such as workplace injuries but lacked depth. Geographical and industry-specific studies, such as those by Kawashita et al. (2005) and Nagata et al. (2017), demonstrated that OHS disclosure varies by industry and company size, with larger firms and high-risk industries like manufacturing disclosing more. In contrast, Japanese firms were found to report less OHS data compared to other CSR elements. Research by Rahman et al. (2019) on Indonesian trade unions and Evangelinos et al. (2018b) on multinational corporations revealed significant gaps between the union and corporate OHS communication, with reporting often lacking depth despite detailed management approaches. OHS is increasingly integrated into broader CSR frameworks, as studies by Montero et al. (2009), Searcy et al. (2016) and Kumar et al. (2018) indicate. However, it is often secondary to other dimensions like environmental sustainability. Legal requirements, rather than voluntary sustainability initiatives, frequently drive OHS reporting, as Mariappanadar et al. (2021) pointed out. The table showcases the evolution of OHS disclosure across industries, countries and frameworks, reflecting a trend towards integrating OHS into CSR and sustainability practices. Larger firms, certified companies and those in high-risk industries tend to disclose more OHS-related information, driven by both stakeholder pressure and legal requirements. However, gaps in standardization, voluntary reporting and sector-specific challenges continue to hinder uniform reporting despite its positive impact on financial performance and corporate reputation. Various factors influence OHS disclosure, including industry profile, company size, MNC operations and financial performance, as noted by Cahaya et al. (2017) and Zhang et al. (2023b) while board involvement and corporate debt may negatively impact transparency. High-risk sectors like construction show improved transparency due to stakeholder pressure, according to Fotiadis et al. (2023b) and Oakman et al. (2024). Stakeholder and institutional influences also play a role in disclosure practices, with studies by Tsalis et al. (2018) and Roland et al. (2020) suggesting that certifications like OHSAS 18001 enhance transparency. Alves and Ramos (2022a) further explored how certifications and external assurance improve disclosure quality in European companies. The distinction between voluntary and mandatory OHS reporting is critical, as studies like Gardiner (2022) and Oakman et al. (2024) highlight, with mandatory standards generally driving more comprehensive disclosures. Health-related OHS issues are also industry-specific, with studies like Park and Kim (2023) and Rikhotso et al. (2022b) focusing on noise-induced hearing loss and injury rates in sectors such as mining and manufacturing. OHS disclosure positively impacts financial performance and corporate reputation, as Zhang et al. (2023b) found, though factors like institutional investor shareholding can negatively influence it. Despite these developments, challenges remain, including inconsistent reporting practices and focus on injury rates rather than broader OHS indicators, as discussed by Shimizu et al. (2022b) and Ali et al. (2021).
CONCLUSION AND AREA OF FURTHER RESEARCH
An intensive study on OHS reporting and disclosure has been conducted. The sheer number and diversity of scholarly articles in the OHS disclosure make it time-consuming and challenging to identify the significant research themes and trends accurately. This review paper focuses on OHS reporting and disclosure of scholarly published articles from 1974 to 2024 in the journal indexed on the Scopus database. This study found 956 articles focused on OHS disclosure and reporting, most published between 2020 and 2022.
This review paper delves into OHS reporting and disclosure, employing a methodologically rigorous approach to analyze a vast array of literature spanning nearly five decades. Leveraging advanced computational techniques, particularly STM, the study uncovers latent themes and trends within the OHS landscape. “OHS regulation” is the predominant theme and noteworthy relationships have been identified. Firstly, a significant connection exists between “OHS auditing in the construction supply chain,” sustainable business practices and OHS corporate reporting and disclosure practices. This relationship highlights the integration of safety audits with broader corporate sustainability and transparency efforts. Secondly, promoting OHS education is closely linked to the regulatory framework and global rights, underscoring the necessity of educational initiatives to support compliance and protect workers’ rights. Thirdly, pharmaceutical hazard management is tied to regulatory frameworks, global rights and environmental risk assessment, indicating a comprehensive approach to managing health risks within the pharmaceutical industry.
The emerging topics in OHS include “OHS corporate reporting and disclosures,” “employee well-being and workplace,” “sustainable business practices,” and “OHS auditing in the construction supply chain.” Conversely, there is a noticeable decline in interest in topics such as “health studies and safety measures,” “OHS regulation,” “pharmaceutical hazard management,” and “healthcare accessibility and support for mental health and disabilities.” This shift reflects evolving priorities and focal points within the OHS field. The findthought function of the STM package assists researchers in identifying articles highly associated with specific topics, such as OHS corporate reporting and disclosure. Through this process, the researcher has pinpointed 27 highly relevant articles to OHS in corporate reporting.
