Testing of a Safety Leadership Model

Abstract

Fatal and serious injury rates remain unacceptably high in the construction industry. Leadership plays a critical role in safety management and serious and fatal injury prevention. However, limited research has examined industry practitioners’ perceptions of leadership and how it influences safety outcomes, particularly in the prevention of serious and fatal injuries in the construction industry. Therefore, a theoretical model for capturing perceptions of safety leadership was developed from a systematic literature review. To ensure that the labels and language used in the model can be understood by industry practitioners, a Delphi study was conducted involving twelve experts. Over three iterative rounds, the model was refined to include five leadership styles, seventeen elements, and eighty-five descriptive statements spanning the range from laissez-faire to transformational leadership. The refined model provides a comprehensive framework for understanding safety leadership and serves as a foundation for future empirical testing with frontline construction workers.

1. Introduction

Fatalities and serious injuries in the construction industry have remained unacceptably high worldwide over the past decade (Figure 1, Figure 2 and Figure 3). Leadership is vital for risk management and the prevention of serious and fatal injuries. However, previous research has predominantly focused on how leadership influences organisational safety culture, workers’ safety behaviour, and minor workplace injuries [1].

Figure 1. Worker fatalities: number of fatalities and fatality rate, 2014 to 2023 in Australia Construction Industry [2].

Figure 2. Worker fatalities: number of fatalities and fatality rate, 2011 to 2022 in USA Construction Industry [3].

Figure 3. Worker fatalities: number of fatalities and fatality rate, 2012 to 2022 in UK Construction Industry [4].

It is hypothesised that organisations with stronger leadership styles achieve better safety outcomes and lower rates of injuries and fatalities. Nevertheless, the systematic literature review revealed no articles or prior research that explored frontline workers’ perspectives on how different leadership styles influence safety. Therefore, further research is required to answer the following question: What leadership styles are perceived by frontline workers to contribute to better safety outcomes and less serious and fatal injuries in Australian construction industry?

In order to answer the question, information from the literature was transposed into a leadership model that could be used to survey frontline workers in a similar way to the other safety models [5,6,7,8]. The model described in this paper especially focuses on safety leadership in the construction industry. It consists of three parts comprising leadership styles, elements, and descriptive statements for each element across the different leadership styles. The styles of leadership described in this paper were adapted from Bass’s Full Range of Leadership Theory [9], which is widely accepted and recognised. The elements are very important components of construction industry safety and risk management systems. Then the descriptive statements were developed for each element in relation to each of the five leadership styles. As the model was grounded in academic sources, it required validating and testing to ensure it was understandable for industry practitioners. This article describes the research conducted to test and refine the model with industry experts using the Delphi technique.

Following three iterative rounds of the Delphi study, consensus was achieved across all components of the model. Specifically, all leadership styles and elements reached consensus in Round 1, eighty-eight statements in round 2, and thirty-three statements in round 3. The refined model comprises five leadership styles, seventeen elements, and eighty-five descriptive statements. It advances existing Full Range of Leadership Theory [9] by integrating these leadership styles, elements, and descriptive statements into a coherent and structured framework. Moreover, the model provides a practical self-assessment tool for organisations to evaluate leadership styles across seventeen elements, identify strengths and weaknesses in their safety management practices, and strategically allocate resources to improve leadership effectiveness and prevent serious and fatal injuries in the construction industry.

2. Delphi Study Method

The principle behind Delphi technique is the old proverb “Two heads are better than one” [10,11,12]. This premise is supported by other researchers; they assume that group opinion is more valid and reliable than individual opinion [11,13,14,15]. This assumption was explained by Oh [16] in his paper “Even when we know the best expert, his background knowledge is limited and accordingly his prediction may be less accurate than that from a group of experts”. Furthermore, de Villiers [17] remarked “pooled intelligence enhances individual judgement and captures the collective opinion of experts without being physically assembled.”

There are two approaches of developing questionnaires for Delphi studies. One is conventional Delphi, which use unstructured questions to extract opinions from the participants through an open-ended format; in essence, this is a brainstorming session [18]. The other one is modified Delphi, which devises structured questions, through a review of the literature, to pool opinions of individual expertise of the panellists [18]. Martino [12] found the participants are confused by the unstructured questions associated with conventional Delphi studies, while Uhl [19] indicated that a Delphi study with structured questions has significantly lower participant attrition compared to those with unstructured formats. In addition, Ludwig [20] disclosed that structured questionnaires minimise the expense and time of conducting a Delphi study compared with unstructured ones. In this Delphi study, a modified Delphi approach was adopted.

The modified Delphi technique employed involved using a panel of experts to review and refine the newly developed leadership model so that the language used in the model was understandable and credible from an industry practitioners’ perspective. The entire process of this Delphi study is portrayed in Figure 4. In general, for a successful outcome, this research was designed as a four-step process:

Figure 4. Process of this Delphi study.

• Develop a survey instrument to test the leadership model, which includes conducting a pilot test to identify ambiguities in wording and check the layout of the instruments.
• Conduct a literature review to confirm that the Delphi technique is an appropriate method for the research and to decide sample size, analysis process, and number of rounds required to develop a refined model.
• Recruit experts using certain criteria through a panel selection committee.
• Implement three iterative rounds to review and refine the leadership model.

In order to accomplish this Delphi study, the researchers recruited members of an expert panel from April 2022, and completed the third round in October 2022. Each step of the process will be discussed in detail in the following paragraphs.

2.1. Survey Instrument Development

There were two instruments developed for this Delphi study. The first instrument covered 5 different leadership styles and 18 elements of the model. Clear definitions of leadership styles were provided to eliminate ambiguity and prevent the use of emotive language [15,21]. The questions were designed to be easy to read, using simple and concise language while avoiding lengthy or complex wording [15,22]. These design features aimed to enhance response quality [12] and participation rates [23]. Furthermore, ample space was provided for participants to provide comments or further justification for their responses [12,15,18]. This instrument was divided into the following four parts:

• The first part is an instruction providing context information and an explanation on how to complete the survey.
• The second part is a demographic question which collected experience, occupation, education, gender, and age data to allow the researchers to better understand and analyse the data.
• The third part is the context information, which included a brief overview of the leadership model, providing fundamental information in regard to the model and the background information; this part also explained the purpose of this research.
• The fourth part is questions about leadership styles and elements of the model, which originally derived from the systematic literature review [1]. Participants were asked to record their level of agreement with the labels given to the leadership styles and elements of safety management system using a four-point Likert scale where 1 meant Strongly Disagree, 2 meant Disagree, 3 meant Agree, and 4 meant Strongly Agree. The participants also had the option to select “Don’t know” if they do not have enough knowledge to rate particular leadership styles or elements given in the instrument. The neutral point was removed from the Likert scales intentionally because a neutral option provides very little information about the issues discussed and we prefer that the participants take a particular option. In other words, the nonneutral stance can promote a debate which aligns with the main objective of developing agreement and disagreement [24].

The second instrument reviewed the 90 descriptive statements that were contained within the model and were also derived from a literature review [1]. Similarly to the first instrument, this instrument also contained Delphi study instructions and a brief overview of the leadership model. This instrument then asked participants to rate each statement across 5 leadership styles for each element.

Before the commencement of the Delphi study, a pilot test was conducted to identify ambiguities in wording, check the layout of the instruments, and ensure the language used was understandable by the industry workforces. The pilot test served as the basis for the Delphi study; for this reason, it is desirable to do this test.

A couple of experts from high-risk industries that were not part of the panel of experts were invited to test the survey instrument. On average, the two experts have 23 years of experiences in their industry.

On the basis of feedback obtained from the pilot test, the labels of Element 3, Element 6, Element 7, Element 9, Element 10, and Element 15 were renamed.

• Element 3—“Balance when making a decision” was renamed to “Balance priorities when making a decision”.
• Element 6—“Risk management perception” was renamed to “Perception of risk”.
• Element 7—“Hierarchy of risk controls” was renamed to “Risk controls”.
• Element 9—“Procurement” was renamed to “Contractor procurement”.
• Element 10—“Contractor management” was renamed to “Contractor management on the job”.
• Element 15—“Events management” was renamed to “Actioning of events”.

Some minor wording changes were made to 15 descriptive statements, and the layout of 18 elements and 5 leadership styles was turned into bulleted lists to make them easier to read in the instruments on the advice of participants.

2.2. Selection of Experts

The selection of expert panellists is a critical step in the Delphi study, as it directly influences the quality of the research [25,26]. While there are numerous debates regarding the optimal number of experts required, existing studies indicate limited empirical evidence that panel size significantly affects the reliability [14,27] or validity [14] of consensus processes. Rather, the representativeness of a Delphi panel is determined by the qualities of its members rather than its number [28]. The Delphi technique does not aim for statistical representativeness but instead seeks to synthesise the informed opinions of the particular group [29]. Johnson [27] demonstrated that panels of around 15 experts produce reliable outcomes, with little chance that a different panel would yield significantly different results. He further suggested that even a panel of approximately 10 experts can generate reliable results. Consequently, this study targeted a panel size of 10 to 15 experts. The appropriate sample size ultimately depends on the purposes of the research, the expertise required, and the complexity of the research [19,30]. In line with Powell [28] and Keeney [15], greater emphasis was placed on achieving diversity in participants’ backgrounds and experiences rather than merely increasing the number of experts.

To date, no Delphi technique research has addressed the issue of what criteria should be utilised to recruit their experts. Oh [16] indicated that choosing appropriate experts is generally based on the judgement and discretion of the principal investigator. Later, this statement was supported by Sumsion [31], who stated that “there is no ready answer and it becomes the responsibility of each researcher to choose the most appropriate group of experts and defend that choice”. For the purpose of this research, the inclusion criteria used for the selection of experts was those who were knowledgeable and/or experienced in safety and/or risk management and/or leadership in the construction industry. Specifically, the participants recruited had to meet at least one of the three criteria established for panel selection:

• The participants have at least 5 years of engagement in safety and/or risk management in the construction industry. The senior management of the organisations in the industry are identified in a list, subject to meeting the aforementioned criteria, and they are invited to participate in Delphi survey;
• Leastwise the participants have 5 years of construction and/or project management experience in the construction industry;
• The professionals have a minimum of 5 years of experience in safety and/or risk management in the construction industry, and are recognised by a government, institute, or association.

The researcher recruited experts directly via snowball sampling [32], where selected experts could nominate other experts that met the abovementioned criteria.

Once the individual had been selected as an expert in the specific field, this expert received a formal email invitation to participate in the Delphi survey. The participants were informed about the research purpose, process, benefit, and the reason for being selected.

• The purpose of the Delphi study was explained to the participants along with the reasons for conducting this research, the significant meaning of the academia, and the implications for the construction industry.
• The process of the Delphi study stated how the survey works, how many rounds there were, the duration of the iterative process, and what efforts will be required from the participants. For this research, the prediction was that the Delphi study would involve up to 4 rounds over a 4-month period, with each round taking approximately 30–45 min to complete.
• The benefit of the Delphi study was illustrated to participants, including what they will gain from this survey, the benefit of having an opportunity to learn about the Delphi method, and an opportunity to hear the thoughts and gain information from other experts.
• The reason for being selected was described to the participants, namely why the individual was selected to participate in the survey.

A total of 21 experts were invited based on the established selection criteria; 15 were purposefully selected, and 6 were nominated through the Snowballing method. Eventually, 12 experts accepted the invitation, forming the final Delphi study panel. First, all panel members possessed substantial knowledge and experience in safety and risk management within the construction industry. This aligns with Klee [33], who emphasised that the primary criterion for expert selection is knowledge; however, this alone is insufficient unless the expert can apply it effectively to the problem. Second, all experts signed consent forms to confirm their commitment to the study. This was supported by Johnson [27] who advised that the success of a Delphi study depends on the commitment of participation and continued engagement throughout the process.

It should be further understood that this Delphi study adheres to the Guidelines of the ethical review process of The University of Queensland and the National Statement on Ethical Conduct in Human Research, and all the experts were informed that the participation is entirely voluntary and all information provided will be anonymous in compliance with the ethical research process approved by The University of Queensland. All the experts invited were asked to sign a generic consent form prior to conducting the Delphi study. These experts were informed that they have the rights to withdraw at any time without penalty, and the comments provided are anonymous; this information was presented on the invitation letter of the Delphi study.

2.3. Consensus of Delphi Study

According to the literature review, to achieve consensus of opinions or judgements of a group of experts is the primary objective of the Delphi study [15,34]. Nevertheless, there is no universal agreement to declare consensus, and no definition of consensus or methods to yield consensus. Usually, this decision rests on the shoulder of researchers to choose the proper method to constitute consensus based on the demand of research. Consideration also needs to be given to the fact that consensus may not be reached and in some situations may not be required; this ultimately depends on the purpose of Delphi study [35,36].

The researchers need realise that the decided consensus depends on the purpose of the research, and the reality that the higher the standards, the harder it is to reach consensus between the experts [37]. Keeney [15] recommended that the researchers need to declare the agreed consensus before the commencement of a Delphi study. In this Delphi study, the measure of central tendency (median) was employed to reflect the convergence of collective opinions, while the interquartile range (IQR) was utilised to assess the dispersion level of collective opinions. Consensus was considered to have been achieved when the median of expert panel responses reached 3, and the IQR was less than or equal to 2. Correspondingly, the iterative process stopped once the agreed consensus had been achieved [18]. Any items ranking below the threshold were updated based on the feedback from the expert panel, and then these updated items went to the next round of review by all experts.

In theory, the Delphi study can iterate as many rounds as are required until the consensus is reached. Through literature review, the number of rounds demanded is a contentious issue; this can vary from two to five rounds [36,38,39]. The number of iterative rounds to some extent depends on a few factors: these are degree of agreement, new information to gain, time available, and fatigue. The researcher has the responsibility of balancing between the number of rounds and the pragmatic factors. Also, when successive rounds tend to have very little new information gained, the excessive repetition is unacceptable to the group experts [12,40]. Another influencing factor that needs to be considered is the time available when deciding the number of rounds [15,29]. Furthermore, the experts may become fatigued after three iterative rounds, and this will further decrease the response rate and increase the attrition [41]. Although there is not a standard for the exact number of rounds, Delphi researchers pointed out, in most instances, three rounds proved sufficient to collect necessary information and achieve consensus [20,35,40,42].

In this Delphi study, the researchers employed a three-round interactive process. Even if there are some items that cannot achieve consensus after three rounds, the researcher decided to stop at this point, because sometimes dissensus is as important as consensus from a research-findings perspective. In round 1, panellists were asked to refine and rate eighteen elements and five leadership styles of this model. In round 2, panellists were asked to refine and rate ninety descriptive statements in the grid of this model. In round 3, panellists were asked to reconsider and rerate these updated statements based on the comments of experts from the second round. In each round, the responses of panellists were aggregated and analysed by researchers through IBM SPSS Statistics 26 software. A summary of controlled feedback and results of the previous round were included in the cover letter of the new round. This three-round Delphi study took about 4 months to complete. To illustrate how the Delphi study works, the three-round procedure is described in Figure 5.

Figure 5. Three-round procedure of Delphi study.

3. Results

3.1. Round 1

The results from the demographic information revealed that on average, the expert panel has 20 years of experiences in the construction industry. Six of them hold safety professional roles, five of them construction and project management roles, and one of them an academic role. All of them have senior management positions in their own organisation. In total, 64% of the panellists have bachelor’s or master’s education, and the others have advanced diplomas or equivalent education. In the expert group, there are two female participants, and the rest of the participants are male. In total, 64% of the experts are over 55 years old.

The results from the round 1 review of element and leadership style labels were quite good. The response rate was 92%, with 11 experts completing and returning the round 1 survey, which included a total of 75 comments from experts. These were reviewed by researchers and used to update the elements, leadership styles, and statements in the model. The data analysis highlighted that all of the five styles of leadership and 10 of the 18 elements reached a median score of three, which means agree. The other eight elements reached a median of four, which means strongly agree. One element reached an IQR of two; all of the other 17 elements reached IQRs less than or equal to 1. At the conclusion of round 1, all the labels given to leadership styles and elements reached consensus as defined in Section 2.3. However, the leadership style “Management-by-exception Active” was renamed to “Proactive Leadership” based on the feedback from the expert panel, but the core meaning of the content was not changed. Two comments from the panellists are below:

One expert advised “Call this Proactive Leadership”.

Another expert commented “Title does not reflect the active nature of the work. This is not “Management by Exception” but proactively identifying and managing risk through multiple risk management processes.”

Four elements were also renamed based on the feedback from the expert panel, but the core meaning of these four elements remained unchanged. Element 1—Commitment has been renamed to “Commitment (e.g., attend safety walk and chair safety committee)”, because one expert advised “It’s important for leaders to talk about safety, attend safety walk and chair Health and Safety committees.” Element 5—People development has been renamed to “People Development (e.g., training and education)”, because one expert commented “Too generic in relation to risk management.”, Element 6—Perception of risk has been renamed to “Risk Assessment”, because one expert suggested to change to Risk Assessment. Element 17—Performance Measurement has been renamed to “Safety Performance Measurement”. Some of the comments regarding the labels were considered in the different descriptive statements of leadership model, which was reviewed in round 2. The results summary of round 1 is illustrated in Table A1 and Table A2.

3.2. Round 2

Round 2 secured a response rate of 91% with only one expert not returning the survey questionnaire. A total of 153 comments were received from panellists. At the conclusion of the second round, 73 statements reached a median of three, which means agree, and 16 statements reached a median of four, which means strongly agree. One statement of Management-by-exception Passive for Element 18 reached a median of two, which means disagree. One statement of Laissez-faire leadership for Element 6 reached a median of three, but an IQR of three, which means disagree; all the other 89 statements reached IQRs less than or equal to 2. The Likert scale ratings, median, and IQR of the distribution are illustrated in Table A3. However, following a thorough review and discussion of participant feedback within the research team, one element was renamed, another element was removed, and two statements were revised accordingly. The specific changes are outlined as follows:

• Element 11—“Safety Integrated in Supply Chain” was changed to “Safety in Design”.
• Element 15—“Actioning of Events” was removed from the model because this element has been covered in Element 14—“Event Report” and Element 16—“Event Investigation”.
• The statement of Management-by-exception Passive for element 18 “The leader punishes workers when safety target (e.g., injury rate) is not achieved.” was changed to “The leader does not react on changes in safety targets (e.g., injury rate)” because consensus was not reached.
• The statement of Laissez-faire Leadership for Element 6 statement “The leader allows fatal risks in the workplace to persist and often to become acceptable” was changed to “The leader does not consider safety risks or risk assessment information thereby allowing serious and fatal risks in the workplace to persist and often to become acceptable” because consensus was not reached.

In addition, 31 descriptive statements underwent minor revisions based on the comments of participants, despite already meeting the consensus criteria. These refinements were made to improve clarity and alignment with industry language. Consequently, the 33 updated statements were used to design the questionnaire of round 3.

3.3. Round 3

In round 3, the survey questionnaire consisted of 33 statements across 10 elements. All five statements of each element were presented in the grid of the model so that the panel members can consider the statements together with the others as a whole when they provide any comments. To avoid forcing consensus, the researchers intentionally did not provide feedback on individuals’ previous ratings and the collective median of the second round. There were two reasons for this. One is that the experts could be biassed by their responses from the previous round. The other is “when an expert sees himself in a minority and does not feel strongly about his position, that alone may be enough for him to change his answer” [29]. They were provided with a summary of the results of round 2, which included the consensus and dissensus of statements, elements renamed and removed, and the updated statements. The third round of the Delphi study was conducted in October 2022.

After completion of the third round, the response rate of round 3 was 90%, with one expert not responding to the survey. A total of 65 comments were received from panel members, and all the comments received were reviewed and analysed by researchers. At the conclusion of the third round, 33 statements achieved consensus based on the criteria of consensus defined in previous section, 26 statements reached a median of three, the other statements reached a median of four, and all the statements reached IQRs less than or equal to two (refer to Table A4).

Although all 33 statements reached consensus in round 3, one statement under Proactive Leadership of element 3 “The leader prioritises safety over other construction goals as long as the other construction goals are being met. If schedule and costs are under pressure, these will be prioritized ahead of safety.” was revised to “The leader prioritises safety over other construction goals as long as the other construction goals are being met. If schedule and costs are under pressure, these will be considered together with safety.” based on the feedback received. In addition, several minor edits, such as grammatical adjustments, were made without changing the core meaning of the statements. Ultimately, a refined and tested safety leadership model (presented in Table A5) was formed as the outcome of this Delphi study.

4. Discussion

On the whole, the response rate of every round is quite high in this study even though Delphi studies have been criticised for their low response rates [43] and high dropout rates [17,18,44]. According to Uhl [19], 83% would be considered an adequate response rate. Further to this, the researchers recommend that a 70% minimum response rate should be obtained for each round to maintain rigour when using the Delphi study [31,35,45]. However, there was no universal guidance regarding what level of response rate should be achieved according to the literature review. The response rates can vary from a very inadequate 10% [19] to an excellent 100% [46]. In order to maintain a high and stable response rate and low drop-off rate during the Delphi study, the researcher employed a variety of strategies to motivate and engage members of the panel in the whole process.

Firstly, the research process maintained the anonymity of experts’ responses, which, according to Rauch [47], can considerably increase the response rate. Furthermore, every expert signed off and returned a consent form to demonstrate their commitment before the commencement of the Delphi study. Johnson [27] stated the experts with commitment were interested and highly committed to participating in the study.

Secondly, the participants were kept informed with summaries of previous round results that highlighted the updated statements that were constructed based on their comments. This was done to keep the participants motivated and up to date with the Delphi study process. As stated by Sandrey [48], when motivating participants properly, the Delphi study enables experts to assume ownership of the study. Moreover, Sandrey [48] stressed that ongoing communication can be taken throughout the research process to retain panel participants. For this research a maximum of two reminders was used for each non-respondent in every round; even though Dillman [49] advocated three follow-up contacts, Beretta [50] suggested the ethics should be considered with this strategy because it seems likely that the experts would feel forced to continue, even though they wish to drop out, and Keeney [15] indicated that reminders could be considered coercion by the researcher to continue participation.

Thirdly, the motivation of participants can be influenced by the length of the study; participants may be fatigued after multiple rounds, which could lead to a reduction in expert responses [41]. Further, the participants sometimes become fatigued after two or three rounds [37], which may result in a low response rate. The researcher, in this iterative process, did not ask panellists to reconsider and rerate the agreed labels and statements without new information gained from comments of the previous round. Doing this would have consequently increased the workload and time spent for the experts due to the extra new rounds created. As stated by Keeney [15], the numbers of statements will not decrease as the rounds go on and the risk of expert’s dropout becomes greater.

In round 3, the participants were invited to reconsider and rerate the 33 updated statements, but they were not provided with individual scores of the statement in round 2 and the group median of round 2, because the experts can be biassed based on the statistical analysis, and there is a risk that “extreme opinions will be masked by the statistical analysis” [51]. When an expert realises they are in a minority and does not feel strongly about their position, that is enough for this expert to change their opinion due to strong group pressure to conform [27]. On the contrary, experts within a majority are reluctant to change their opinion, because they both mistakenly believe that the majority view is right. To make matters worse, experts with extreme views are more likely involved in dropouts than those experts within the majority [52]. Consequently, the consensus achieved is an artificial product because of the dropouts of participants with extreme views [53]. Additionally, Gordon [29] argued that the extreme views also represent a useful product.

To sum up, a high response rate and low attrition rate is achievable through applying various strategies across the different stages of the Delphi study. Moreover, the consistently high response rate demonstrates the reliability of the consensus achieved. The strong level of agreement among experts also indicates that the language and constructs used in the model were understandable, meaningful, and credible from the perspective of industry practitioners. Consequently, the newly developed leadership model has gained recognition and acceptance from industry experts. These outcomes provide the researcher with confidence to further test the model on a larger sample of frontline construction workers.

The researcher decided to stop this iterative process at round 3 because of three reasons. First, compared to the previous rounds, there was no extreme response “Strongly Disagree” in round 3. For most of the statements (31 out of 33), the responses have tended towards stability through two rounds iteration. As Rowe [54] stated, less variance is understood to mean greater consensus. Additionally, if either the consensus and/or stability of responses has been reached, the Delphi process completes [18]. Second, there were not new opinions generated from the comments of experts [27]. Third, participants may be fatigued after multiple rounds, which could lead to a reduction in expert responses. [41]. Most studies therefore limit the number of rounds to two or three [35]. Therefore, the researchers made decisions to stop this iterative process after the third round based on the needs and purpose of this Delphi study. In conclusion, a finalised leadership model was achieved after review and refinement by a group of experts through a Delphi study.

The nature of the construction industry presents significant challenges to safety leadership. This research addressed a critical gap: the absence of an integrated leadership model tailored for the construction industry. The proposed model extends Bass’s Full Range of Leadership Theory [9] by embedding leadership styles, elements, and statements into a structured framework. It enables the evaluation of safety leadership across seventeen organisational elements, offering construction companies a diagnostic tool to identify gaps between current and desired practices and to target improvements. Unlike previous models such as that of Stemn [8], which have been applied in the mining sector, this framework is developed for construction. In terms of practical contribution, the leadership model refined in this Delphi study provides a useful tool for construction organisations to assess their leadership styles across multiple elements. This self-assessment tool allows companies to systematically identify strengths and weaknesses in their leadership and safety management practices, evaluate their performance across projects and locations, and develop targeted strategies to enhance leadership effectiveness and safety outcomes.

The Delphi study was limited to Australian-focused experts with construction experience, but not frontline workers; the reliance on Australian experts may limit the transferability of the findings to other cultural or regulatory contexts. The panel also included only two female participants, leading to an under-representation of women’s perspectives. Moreover, the panel was predominantly composed of senior managers, which may have introduced bias and influenced the perspectives on leadership, thereby affecting the generalisability of the findings. Despite these constraints, the study generated valuable insights and contributes to leadership and safety practice. The inferences drawn are representative only of the populations studied. These limitations highlight the need for more diverse and broader studies to confirm its generalisability and strengthen its reliability. This can be achieved by collecting data across broader demographics, including international contexts such as the USA, UK, and other regions. The next stage of this research will involve applying the model to a broader sample of frontline construction workers to evaluate their perceptions of leadership. Understanding these perspectives is expected to yield new insights into improving leadership effectiveness and the management of serious and fatal risks in the construction industry.

5. Conclusions

This newly developed leadership model addresses a significant gap in the academic literature, as no prior research has established a framework specifically designed to capture frontline workers’ perceptions of leadership styles and their relationship to the prevention of serious and fatal injuries. The refined model extends Bass’s Full Range of Leadership Theory [9] by integrating leadership styles, elements, and statements into a structured framework. Importantly, this model offers practical value as a self-assessment tool for construction organisations, enabling them to evaluate leadership styles across multiple elements and identify both strengths and weaknesses in leadership and safety management practices. This diagnostic capability provides a starting point for organisations to prioritise resources and implement targeted strategies aimed at improving leadership effectiveness and preventing serious and fatal injuries in the construction industry.

Nevertheless, several limitations should be acknowledged. The Delphi study was limited to Australian-focused experts, with only two female participants, and the panel was predominantly composed of senior managers. These constraints may influence the generalisability of the findings. Therefore, further in-depth empirical research is recommended to systematically examine the reliability, applicability, and broader usefulness of this leadership model. To address this issue, this leadership model will be tested by frontline workers within the construction industry in future studies, and the feedback data of participants will be analysed accordingly.