Prioritizing Critical Factors Affecting Occupational Safety in High-Rise Construction: A Hybrid EFA-AHP Approach

Abstract

High-rise construction presents heightened safety risks due to vertical complexity, spatial constraints, and workforce variability. Conventional safety management often proves insufficient, especially in rapidly urbanizing or resource-limited settings. This study proposes a hybrid methodological framework to systematically identify and prioritize the critical factors influencing occupational safety in Vietnamese high-rise construction projects. Based on 181 valid survey responses from construction professionals, 23 observed variables were developed through extensive literature review and expert consultation. Exploratory Factor Analysis (EFA) was employed to empirically group 23 validated indicators into five key latent dimensions: (1) Safety Training and Inspection, (2) Employer’s Knowledge and Responsibility, (3) Worker’s Competence and Compliance, (4) Working Conditions and Environment, and (5) Safety Equipment and Signage. These dimensions were then structured into an Analytic Hierarchy Process (AHP) model, with pairwise comparisons conducted by industry experts to calculate consistency ratios and derive factor weights across three high-rise project case studies. The findings provide actionable insights for construction managers, safety professionals, and policymakers in developing and underdeveloped countries, supporting data-driven decision-making for safer and more sustainable urban development.

1. Introduction

The construction industry persists as one of the most hazardous occupational sectors globally, with high-rise construction presenting exceptionally complex safety challenges due to its inherent characteristics: extreme verticality creating severe fall hazards, confined working spaces intensifying worker proximity risks, intricate mechanical systems introducing multifaceted failure modes, and amplified environmental exposures to wind forces and lightning strikes [1,2,3,4]. This alarming reality manifests in sobering statistics worldwide, with South Korea’s construction sector accounting for over half of all workplace fatalities [5]. The United States Bureau of Labor Statistics reports a 12-year peak of 1075 construction fatalities in 2023, with falls from heights constituting 39.2% of these incidents [6]. European construction sites mirror this concerning pattern through consistently higher-than-average nonfatal injury rates compared to other industries [7].

Studies show that safety frameworks designed for low- to mid-rise buildings are often inadequate for high-rise projects without major adjustments [8]. For instance, Ansari et al. [9] indicated that falling from heights was the most critical safety risk in high-rise construction, with safety training and monitoring being the most influential criteria, accounting for approximately 35% of the total weight in risk assessment. High-rise projects involve intricate logistics, including hoisting and lifting operations, which are prone to accidents due to their scale and complexity [10]. Given these heightened risks, there is a clear and urgent need to systematically prioritize the critical factors affecting occupational safety in high-rise construction. While existing research has effectively documented the severity of specific hazards, particularly fall-related incidents and mechanical failures, few studies have operationalized this knowledge into structured, decision-oriented frameworks that support practical intervention [11]. In particular, the field lacks integrative approaches that combine empirical identification of latent safety dimensions with expert-informed evaluation of their relative importance.

Methodological approaches to safety factor identification and prioritization have advanced considerably, with researchers increasingly employing sophisticated multivariate techniques to move beyond descriptive case studies toward empirically validated factor structures [12,13,14]. Contemporary safety research frequently utilizes latent variable modeling approaches, particularly Exploratory Factor Analysis (EFA) and Structural Equation Modeling (SEM), to identify underlying constructs that explain patterns among observable safety indicators [15,16,17]. Mosly et al. [18] demonstrated EFA’s effectiveness in construction safety research in Saudi Arabia by reducing 37 observable variables to 10 manageable latent factors, with minimal information loss. Bayesian networks have emerged as another powerful tool for modeling probabilistic relationships between identified safety factors, with Wang et al. [19] developing dynamic Bayesian networks that capture causal relationships between safety management practices and accident occurrence in high-rise construction. However, most existing studies either focus on general construction safety without considering the unique complexities of high-rise projects or examine high-rise safety challenges without providing systematic prioritization tools for resource allocation. Furthermore, there is limited research that translates identified safety factors into actionable, context-specific management recommendations for rapidly developing urban environments like those found in Southeast Asia.

This study addresses these research gaps through three specific, measurable objectives. First, it aims to empirically identify the latent factor structure underlying occupational safety in high-rise construction in Vietnam. The approach allows the factor structure to emerge naturally from the data, while the large and diverse sample strengthens its validity and generalizability. Second, the research seeks to determine the relative importance of identified factors. This systematic prioritization creates a hierarchical framework that quantifies the relative importance of each safety factor and facilitates evidence-based resource allocation decisions. Third, the study endeavors to develop specific, actionable recommendations for enhanced safety management practices tailored to high-rise construction in rapidly degrowing urban environments, emphasizing the highest-priority factors identified through the integrated EFA-AHP methodology. These objectives collectively address the identified research gaps by combining statistical rigor in factor identification with systematic prioritization and translational focus, potentially contributing to quantifiable reductions in occupational injuries and fatalities as vertical development continues to characterize urban growth throughout Southeast Asia and globally.

The remainder of this paper is organized as follows. Section 2 details the research methodology, describing the sequential integration of EFA and AHP to identify and prioritize critical safety factors. Section 3 presents the results of the empirical analyses, including the extraction of latent safety dimensions and the hierarchical ranking of their relative importance across multiple high-rise construction projects. Section 4 discusses the key findings in the context of the existing literature, highlighting practical implications for safety management and outlining limitations and policy development in complex urban construction environments. Finally, Section 5 concludes the paper by summarizing the main contributions, and suggesting avenues for future research.

2. Method

This study employs a three-phase research framework to systematically identify, evaluate, and prioritize the critical factors influencing occupational safety in high-rise construction projects. The methodology integrates both data-driven statistical analysis and expert-based decision-making, combining EFA and the AHP within a sequential design, as illustrated in Figure 1.

Phase I focuses on context analysis and observed variable development. A comprehensive literature review was conducted to examine previous research on occupational safety in high-rise construction, with particular attention to accident causes, risk factors, and regulatory frameworks. This review facilitated the identification of key safety dimensions and observed indicators commonly used in international and local studies. These variables were then reviewed and refined through expert consultation to ensure their contextual relevance to construction practices in Vietnam. The resulting set of validated observed variables provided the foundation for the structured questionnaire used in Phase II.

Phase II comprises two sub-processes: factor extraction through EFA and factor prioritization using AHP. In the EFA sub-phase, survey data were collected from construction professionals with experience in managing on-site safety. The collected data underwent reliability assessment using Cronbach’s Alpha and validity testing through the Kaiser–Meyer–Olkin (KMO) measure and Bartlett’s Test of Sphericity. Latent safety factors were extracted using Principal Axis Factoring with Varimax rotation. These factors represent underlying dimensions of occupational safety and were interpreted based on the pattern of factor loadings. Subsequently, in the AHP sub-phase, the extracted factors served as evaluation criteria in a hierarchical decision model. Expert panels from three high-rise construction projects were engaged to conduct pairwise comparisons, assessing the relative importance of each factor. Consistency ratios were calculated to verify the coherence of expert judgments, and normalized weights were synthesized to produce a prioritized ranking of safety factors for each project.

Phase III centers on information delivery and practical insight generation. The results of the AHP analysis were compared across the three case study projects to identify both common and context-specific safety priorities. These findings were then interpreted to draw actionable implications for project managers, safety officers, and policymakers. Strategic recommendations are provided to support more targeted safety planning, resource allocation, and risk mitigation strategies in high-rise construction environments. By integrating empirical factor analysis with expert prioritization, this study offers a comprehensive framework for enhancing occupational safety management in complex urban construction settings.

2.1. Data Collection

The identification of observed variables in this study was conducted through a structured, multi-phase approach to ensure both theoretical rigor and contextual relevance to occupational safety in high-rise construction projects in Vietnam. Initially, 38 observed variables were compiled through a comprehensive review of 16 domestic and international studies on construction safety. This preliminary list was then evaluated by a panel of experienced professionals, including engineers, site supervisors, and safety managers, through targeted expert consultation sessions. Based on their feedback, eight variables were either redundant, unclear, or insufficiently relevant to the local construction context, and were therefore removed. At the same time, the experts proposed four additional variables that captured context-specific risks, such as limited working space and the use of stimulants during work. This refinement process resulted in a final set of 34 observed variables, which were then incorporated into the main questionnaire, as listed in

Table 1. Observed variables after reviewing related studies.

No.	Code	Observed Variable Name	Reference
1	OV1	Lack of safety equipment and tools	[11,20,21,22]
2	OV2	Outdated or damaged safety equipment and tools	[20,22]
3	OV3	Improper use of equipment and tools	[20,21,22]
4	OV4	Presence of safety warning signs	[20,22]
5	OV5	Established safety regulations and guidelines	[23]
6	OV6	Hot or rainy weather conditions	[11,20,21,22]
7	OV7	Dust and noise	[11,20]
8	OV8	Lack of protective gear for workers	[22]
9	OV9	Heavy workload and overtime	[22]
10	OV10	Electrical leakage	[20,21]
11	OV11	Working space	Expert recommendation
12	OV12	Managers’ knowledge of occupational safety	[11,22]
13	OV13	Work experience	[22]
14	OV14	Ability to plan reasonable construction site layout	[24]
15	OV15	Ability to schedule construction time reasonably	[25]
16	OV16	Ability to organize construction process reasonably	[24,26]
17	OV17	Ability to allocate tasks appropriately	[24]
18	OV18	Willingness to listen to subordinates’ feedback and suggestions	[22]
19	OV19	Ability to communicate with subordinates	[22]
20	OV20	Low-skilled workers	[20,22]
21	OV21	Workers’ compliance with safety rules on site	[21,27]
22	OV22	Workers lacking occupational safety certificates	[11]
23	OV23	Workers with poor health or underlying conditions	[22]
24	OV24	Workers using stimulants while working	Expert recommendation
25	OV25	Monitoring the use of personal protective equipment by workers	[11,21,22]
26	OV26	Regular inspection of safety and risks on site	[11]
27	OV27	Commitment to safety management on site	[21]
28	OV28	Training/education	[11,22,28]
29	OV29	Periodic assessment of workers’ safety knowledge	[11,20]
30	OV30	Ineffective safety training methods for workers	[29,30]
31	OV31	Periodic health check-ups for workers	[31]
32	OV32	Regular inspection and maintenance of equipment/machinery	[22]
33	OV33	Periodic labor management meetings	Expert recommendation
34	OV34	Inspections by related authorities on safety compliance	Expert recommendation

.

Based on the finalized list of 34 observed variables, a structured questionnaire was developed to evaluate the perceived impact of each factor on occupational safety in high-rise construction. A 5-point Likert scale was used, where respondents rated each item from 1 (very low impact) to 5 (very high impact). This scale allowed for the quantification of expert perceptions while maintaining clarity and consistency in responses. The questionnaire was distributed online via Google Forms and shared through professional networks, construction-related social media platforms, and organizational channels, to reach practitioners with relevant experience in high-rise projects in Vietnam.

2.2. Data Analysis

2.2.1. Exploratory Factor Analysis

To empirically identify the latent dimensions influencing occupational safety in high-rise construction, this study applied EFA to a dataset consisting of 23 observed variables. These variables were developed based on a comprehensive literature review and validated through expert consultation to ensure their contextual relevance to the construction industry in Vietnam. EFA was deemed appropriate to uncover latent constructs and organize the observed variables into coherent dimensions. This method is based on the assumption that correlations among observed variables are driven by a smaller number of unobserved latent factors that represent common sources of variance. Mathematically, the EFA model is expressed in matrix form as

$$\mathit{X}=\mathsf{\Lambda }\mathit{F}+\epsilon$$

(1)

where

• $\mathit{X}={[{\mathit{X}}_1,{\mathit{X}}_2,\dots {\mathit{X}}_{\mathit{p}}]}^{\mathit{\top }}$ is a $p \times 1$ vector of standardized observed variables,
• $\mathit{F}={[{\mathit{F}}_1,{\mathit{F}}_2,\dots {\mathit{F}}_{\mathit{m}}]}^{\mathit{\top }}$ is an $m \times 1$ vector of common (latent) factor,
• $\mathsf{\Lambda }=\left[{\lambda }_{ij}\right]$ is a $p \times m$ factor loading matrix, where each element ${\lambda }_{ij}$ represents the degree to which factor $F_j$ influences observed variable $X_i$,
• $\epsilon ={[{\mathit{\epsilon }}_1,{\mathit{\epsilon }}_2,\dots {\mathit{\epsilon }}_{\mathit{p}}]}^{\mathit{\top }}$ is a $p \times 1$ vector of unique variances or measurement errors.
• In this formulation, the subscript $i=\mathrm{1,2},\dots ,p$ indexes the observed variables, and $j=\mathrm{1,2},\dots ,m$ indexes the latent factors. Each observed variable $X_i$ is modeled as a linear combination of the mmm underlying factors $F_j$, weighted by their corresponding loadings ${\lambda }_{ij}$, plus a unique error term ${\epsilon }_i$. A large value of ${\lambda }_{ij}$ implies a strong contribution of factor $F_j$ to the variance of observed variable $X_i$, while the residual ${\epsilon }_i$ represents idiosyncratic variance, including measurement error.

The EFA procedure followed in this study consisted of five key steps. First, the adequacy of the dataset for factor analysis was evaluated using the Kaiser–Meyer–Olkin (KMO) measure and Bartlett’s Test of Sphericity. A KMO value exceeding 0.6 and a significant Bartlett’s test (p < 0.05) confirmed the dataset was suitable for EFA. Second, internal consistency was assessed using Cronbach’s Alpha, and items with corrected item-total correlations below 0.3 were considered for removal. Third, latent components were extracted using Principal Component Analysis (PCA), which aims to reduce dimensionality by transforming the original variables into a new set of orthogonal components that successively maximize explained variance. The number of factors to retain was determined using the Kaiser criterion (eigenvalues > 1) and validated by examining the scree plot. Fourth, to improve the clarity of the factor structure, a Varimax orthogonal rotation was applied, aiming to ensure each observed variable load, primarily on one factor. Items with factor loadings below 0.5 or with cross-loadings on multiple factors were subject to review and potential removal. Finally, the retained factors were labeled, based on thematic consistency among their associated observed variables. These extracted safety dimensions served as the structured input for the AHP conducted in the following phase of the study.

2.2.2. Analytic Hierarchy Process

In this study, AHP serves as a prioritization tool to quantify expert assessments regarding the significance of each factor previously extracted through EFA. This integration ensures that factor identification is data-driven, while the prioritization process captures context-specific expert insight, enhancing both the empirical grounding and practical relevance of the findings.

AHP structures decision problems into a hierarchical model that consists of three levels: (i) the overall goal at the top (i.e., prioritizing factors), (ii) the criteria layer composed of factors identified via EFA, and (iii) the alternatives or decision scenarios at the bottom (if applicable). The prioritization process begins by conducting pairwise comparisons of all criteria at the same level with respect to their contribution to the goal. Experts are asked to express their judgments using Saaty’s fundamental 1–9 scale, where a score of 1 indicates equal importance, and a score of 9 reflects extreme preference of one factor over another.

The pairwise comparison matrix is denoted as $\mathit{A}=\left[a_{kl}\right]$, where $k,l=\mathrm{1,2},\dots ,m$, and $m$ is the number of criteria (i.e., factors extracted via EFA). Each entry ${\mathit{a}}_{\mathit{k},\mathit{l}}$ reflects the relative importance of criterion ${\mathit{C}}_{\mathit{k}}$ over criterion ${\mathit{C}}_{\mathit{l}}$ based on expert judgment. The matrix is reciprocal by definition, satisfying $a_{kl}=1/a_{kl}$, and a $a_{kk}=1$ for all $k$.

$$\mathit{A}=\left[\begin{array}{ccc}1& \dots & a_{1m}\\ ⋮& \ddots & ⋮\\ {\displaystyle \frac{1}{a_{1m}}}& \dots & 1\end{array}\right]$$

(2)

The priority vector $\mathit{w}={[{\mathit{w}}_1,\mathit{w},\dots {\mathit{w}}_{\mathit{m}}]}^{\mathit{\top }}$, which contains the normalized weights of each criterion, is obtained by solving the eigenvalue equation:

$$\mathit{A}\mathit{w}= {\lambda }_{max}\mathit{w}$$

(3)

where ${\lambda }_{max}$ is the maximum eigenvalue of matrix $A$. It is important to note that ${\lambda }_{max}$ is conceptually distinct from the factor loadings ${\lambda }_{ij}$ used in EFA. While ${\lambda }_{ij}$ represents the strength of the relationship between observed variable $i$ and latent factor $j$, ${\lambda }_{max}$ in AHP is a scalar measure used solely to evaluate the internal consistency of expert judgments.

To determine the consistency of the pairwise comparisons, the Consistency Index (CI) and Consistency Ratio (CR) are calculated as follows:

$$CI= {\displaystyle \frac{{\lambda }_{max}-m}{m-1}},CR= {\displaystyle \frac{CI}{RI}}$$

(4)

where $RI$ is the Random Index, a standard value depending on matrix size mmm. A $CR$ ≤ 0.10 is considered acceptable. If $CR$ exceeds this threshold, expert judgments are subject to revision, for improved consistency.

Once consistency is verified, the normalized weights $w_k$ are synthesized to yield a priority ranking of the safety factor groups. This process was independently applied to expert panels from three real-world high-rise construction projects in Vietnam, enabling a comparative analysis of safety priorities across project contexts. The integration of EFA and AHP provides a robust analytical foundation for identifying, structuring, and ranking the critical dimensions of occupational safety, ultimately supporting data-informed decision-making in construction safety management.

3. Results

The survey targeted individuals with practical experience in high-rise construction projects in Vietnam, including engineers, site supervisors, project managers, and contractor and investor organizations. A total of 186 responses were collected through the online questionnaire distribution. After applying data screening procedures, including removing incomplete entries, duplicate submissions, and responses with uniform scoring patterns, 181 valid responses were retained for analysis. As illustrated in Figure 2, most respondents were site supervisors and engineers with more than three years of professional experience, ensuring that the collected data reflected informed and practical insights from the field. A preliminary screening based on mean values was conducted to ensure the analytical validity of the dataset. Since the questionnaire used a 5-point Likert scale where a score of 3 represented neutrality, only variables with a mean score of 3 or higher were considered relevant and retained. As a result, 11 variables with mean values below three were excluded, leaving a refined set of 23 observed variables, as listed in

Table 2. Observed Variables for EFA.

No.	Code	Observed Variable Name	Mean
1	OV2	Outdated or damaged safety equipment and tools	4.1713
2	OV3	Improper use of equipment and tools	4.1105
3	OV5	Established safety regulations and guidelines	4.105
4	OV22	Workers lacking occupational safety certificates	4.0608
5	OV21	Workers’ compliance with safety rules on site	4.0442
6	OV4	Presence of safety warning signs	4.0166
7	OV30	Ineffective safety training methods for workers	3.9448
8	OV31	Periodic health check-ups for workers	3.9392
9	OV32	Regular inspection and maintenance of equipment/machinery	3.9392
10	OV20	Low-skilled workers	3.9116
11	OV29	Periodic assessment of workers’ safety knowledge	3.9116
12	OV28	Training/education	3.9006
13	OV9	Heavy workload and overtime	3.8398
14	OV1	Lack of safety equipment and tools	3.8287
15	OV7	Dust and noise	3.8122
16	OV14	Ability to plan reasonable construction site layout	3.8122
17	OV8	Lack of protective gear for workers	3.7735
18	OV15	Ability to schedule construction time reasonably	3.6906
19	OV16	Ability to organize construction process reasonably	3.6906
20	OV12	Managers’ knowledge of occupational safety	3.6796
21	OV13	Work experience	3.6685
22	OV10	Electrical leakage	3.5801
23	OV27	Commitment to safety management on site	3.3702

. These variables were then analyzed using EFA to identify latent factor structures. The resulting factors formed a validated basis for the subsequent prioritization using the AHP.

3.1. EFA Results

Prior to factor extraction, the dataset’s suitability for EFA was confirmed by the Kaiser–Meyer–Olkin (KMO) measure of 0.750 and a highly significant Bartlett’s Test of Sphericity (Chi-square = 2261.683, df = 253, p < 0.001), demonstrating sampling adequacy and sufficient inter-variable correlation, as listed in

Table 3. KMO and Bartlett’s test of sphericity—analysis based on 23 observed variables.

KMO and Bartlett’s Test
Kaiser–Meyer–Olkin Measure of Sampling Adequacy.		0.750
Bartlett’s Test of Sphericity	Approx. Chi-Square	2261.683
	df	253
	Sig.	0.000

.

In this study, EFA was performed using the Principal Component Analysis (PCA) method as the extraction technique. PCA assumes that observed variables share a common variance structure, and aims to explain the maximum variance through a reduced set of components. Components were retained based on the Kaiser criterion (eigenvalues > 1.0) and further supported by the scree plot. As shown

Table 4. Principal Component Analysis.

Total Variance Explained
Component	Initial Eigenvalues			Extraction Sums of Squared Loadings			Rotation Sums of Squared Loadings
	Total	% of Variance	Cumulative %	Total	% of Variance	Cumulative %	Total	% of Variance	Cumulative %
1	4.182	18.183	18.183	4.182	18.183	18.183	3.926	17.071	17.071
2	3.524	15.323	33.506	3.524	15.323	33.506	3.358	14.601	31.673
3	3.114	13.540	47.046	3.114	13.540	47.046	3.111	13.525	45.197
4	2.626	11.419	58.465	2.626	11.419	58.465	2.992	13.010	58.207
5	1.955	8.502	66.967	1.955	8.502	66.967	2.015	8.760	66.967
6	0.977	4.246	71.213
7	0.813	3.536	74.749
8	0.750	3.260	78.009
9	0.611	2.655	80.664
10	0.588	2.557	83.222
11	0.553	2.405	85.627
12	0.481	2.093	87.720
13	0.445	1.933	89.653
14	0.394	1.713	91.366
15	0.332	1.445	92.811
16	0.315	1.371	94.182
17	0.277	1.202	95.384
18	0.214	0.932	96.317
19	0.213	0.927	97.244
20	0.191	0.832	98.076
21	0.168	0.729	98.805
22	0.150	0.651	99.457
23	0.125	0.543	100.000

Extraction Method: Principal Component Analysis.

, five factors were extracted, explaining a cumulative variance of 66.97%, which exceeds the conventional threshold of 50% for social and management sciences. The first factor accounted for 17.07% of the variance after rotation, while the second to fifth factors explained 14.60%, 13.52%, 13.01%, and 8.76%, respectively.

To enhance interpretability, the extracted factors were rotated using the Varimax orthogonal rotation method. This approach aims to simplify the factor structure by maximizing the variance of squared loadings across variables, thereby ensuring each variable loads highly on only one factor. As presented in

Table 5. Varimax orthogonal rotation.

Rotated Component Matrix a
	Component
	1	2	3	4	5
OV28	0.847
OV30	0.834
OV29	0.829
OV31	0.826
OV32	0.821
OV27	0.632
OV15		0.851
OV14		0.822
OV13		0.802
OV12		0.796
OV16		0.793
OV9			0.915
OV7			0.883
OV8			0.872
OV10			0.783
OV4				0.862
OV5				0.828
OV2				0.816
OV3				0.758
OV1				0.511
OV21					0.887
OV20					0.759
OV22					0.753

Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. ^a Rotation converged in five iterations.

, 23 observed variables had factor loadings greater than 0.5 on their respective components, indicating statistically significant contributions to the extracted structure.

Based on the pattern of loadings and conceptual coherence, the five factors were interpreted and labeled as follows:

• Factor 1 (TB): Safety Equipment, Tools, and Signage (OV1, OV2, OV3, OV4, OV5).
• Factor 2 (MTLV): Working Conditions and Environment (OV7, OV9, OV10, OV8).
• Factor 3 (NSDLD): Employer’s Knowledge, Skills, and Responsibility (OV12, OV13, OV14, OV15, OV16).
• Factor 4 (KTNLD): Worker’s Skills and Compliance (OV20, OV21, OV22).
• Factor 5 (ATLD): Safety Training and Inspection (OV28, OV29, OV30, OV31, OV32, OV27).

To verify the internal consistency of each factor, Cronbach’s Alpha was computed using the group of observed variables associated with each factor. The results are summarized in

Table 6. Cronbach’s Alpha Coefficients for each factor.

Factor Name	Cronbach’s Alpha
Factor 1: Safety Equipment, Tools, and Signage	0.813
Factor 2: Working Conditions and Environment	0.891
Factor 3: Employer’s Knowledge, Skills, and Responsibility	0.873
Factor 4: Worker’s Knowledge and Compliance	0.729
Factor 5: Safety Training and Inspection	0.888

. All five factors demonstrated high reliability, with Cronbach’s Alpha coefficients ranging from 0.729 to 0.891, exceeding the commonly accepted threshold of 0.6. This confirms that the items within each factor measure coherent latent constructs with acceptable internal consistency.

3.2. AHP Results

The results of AHP produced a clear prioritization of the five safety-related factors, reflecting expert consensus on their relative importance to occupational safety in high-rise construction. The aggregated pairwise comparison matrix yielded a CR of 0.05, indicating high internal coherence and methodological reliability. As listed in

Table 7. Pairwise comparison matrices of factor groups provided by the experts.

	ATLD	NSDLD	MTLV	TB	KTLND
ATLD		2.352	3.898	5.378	3.776
NSDLD			4.129	3.366	1.059
MTLV				1.644	4.183
TB					4.043
KTLND

, the matrix incorporates nine levels of comparative intensity, and the direction of preference is visualized through color-coded entries: black text denotes the more critical factor in each comparison, while red text indicates the less preferred counterpart. This allows for immediate interpretation of the dominance relationships between factors. For instance, the comparison between Safety Training and Inspection (ATLD) and Employer’s Knowledge, Skills, and Responsibility (NSDLD) yielded a preference ratio of 2.35216, suggesting that experts consider ATLD moderately more influential than NSDLD. Similar patterns are observed in comparisons with other factors, where ATLD consistently receives higher ratings, particularly over Safety Equipment and Signage (TB) (5.378:1) and Working Conditions and Environment (MTLV) (3.898:1).

The final normalized weights, presented in

Table 8. CI values of factor.

	CI Values
ATLD	0.447
NSDLD	0.216
MTLV	0.074
TB	0.058
KTLND	0.205
Inconsistency = 0.05 with 0 missing judgments

, reaffirm these preferences: ATLD received the highest priority score (0.447), followed by NSDLD (0.216) and Worker’s Knowledge and Compliance (KTNLD), at 0.205. The remaining two factors, MTLV and TB, were assigned considerably lower weights of 0.074 and 0.058, respectively. These results highlight a statistically validated hierarchy of safety factors, anchored in expert judgment and supported by consistent comparative reasoning.

4. Discussion

Based on a dual application of EFA and AHP methodologies, this study provides an understanding of critical safety drivers in high-rise construction. The five-factor structure derived from EFA encapsulates a holistic view of occupational safety risks. These include (1) Safety Training and Inspection, (2) Employer’s Knowledge, Skills, and Responsibility, (3) Worker’s Knowledge and Compliance, (4) Working Conditions and Environment, and (5) Safety Equipment, Tools, and Signage. These factors together reflect the procedural, behavioral, organizational, environmental, and technical domains that shape safety outcomes in high-rise construction. Notably, the factor structure was not only statistically valid, explaining over 66% of the total variance, but also conceptually robust, aligning with prevailing theories of safety systems that emphasize a balance between structural controls and human agency.

The AHP results provide an empirically grounded prioritization of these factors, offering new insights into how safety professionals in the field perceive their relative impact. The dominance of ATLD in the AHP results is consistent with the inherent complexity and vertical risk profile of high-rise construction projects. In these environments, the physical scale and multi-tiered workflows expose workers to a wide range of hazards, from falls at height to crane-related incidents and confined-space accidents. Consequently, formal safety training, routine inspection regimes, and structured knowledge reinforcement become essential mechanisms to mitigate the unpredictability and severity of potential incidents. This result reinforces global safety trends that emphasize proactive risk control through human systems, rather than relying solely on post-incident interventions or passive safeguards.

The strong prioritization of NSDLD and KTNLD further highlights the critical role of human capital in safety outcomes. These two factors, together with ATLD, account for nearly 87% of the total priority weight, signaling a strong consensus that organizational behavior and human decision-making are central to effective safety performance. This prioritization likely stems from the hierarchical nature of construction operations, where managerial oversight determines the consistency of enforcement and where workers’ situational awareness directly influences hazard identification and response. In high-rise projects, where construction sequencing is highly interdependent and tolerance for procedural errors is low, the technical competence of managers and the behavioral compliance of laborers become even more consequential.

In contrast, the lower relative weight assigned to MTLV and TB does not suggest these factors are unimportant, but reflects a recognition that their effectiveness is contingent on how they are implemented and monitored. In many cases, safety equipment may be available but improperly used, or warning signage may be present but ignored, without sufficient training and oversight. This underscores a crucial point: technical and environmental controls serve as enablers rather than guarantees of safety, and their impact is ultimately shaped by the procedural systems and human behaviors that surround them. Figure 3 illustrates a set of strategic research directions derived from the prioritized safety factors identified in this study. Four cross-cutting enablers, cultural adaptation, regulatory alignment, digital transformation, and longitudinal evaluation represent critical pathways for integrating these strategies into diverse and evolving construction contexts. Each domain is paired with forward-looking research themes that reflect evolving industry needs and technological advancements. For instance, procedural dimensions emphasize real-time safety monitoring, digital-twin simulation, and adaptive training models [32]. In contrast, organizational and behavioral dimensions point toward leadership development, multicultural communication, and behavior-based safety (BBS) systems. Environmental and equipment-related domains highlight the need for resilience planning, ergonomic design, and innovative safety technologies. Compared to other hybrid decision-support models such as Fuzzy AHP (F-AHP) or Bayesian networks, the EFA-AHP framework adopted in this study offers a practical and accessible approach for early-stage decision-making. F-AHP enhances the ability to handle uncertainty in expert judgments through fuzzy logic, and Bayesian networks are robust in modeling probabilistic dependencies and causal relationships among safety factors. However, these approaches often require extensive prior data, complex calibration, or specialized computational tools, which may be challenging to implement in resource-constrained environments. In contrast, the EFA–AHP framework allows for empirical factor discovery and structured prioritization with greater interpretability and lower implementation barriers. It is well-suited to the Vietnamese construction industry context.

While this study offers meaningful insights and methodological contributions, several limitations should be acknowledged. (1) The AHP results rely on expert judgments, which, despite high internal consistency, may still reflect individual bias influenced by personal experience, organizational roles, or project-specific contexts. (2) The expert panel was limited to professionals operating in high-rise construction projects within Vietnam, which may constrain the generalizability of the findings to other geographic regions or construction typologies. (3) The cross-sectional nature of the data prevents analysis of how safety perceptions and priorities evolve over the life cycle of a project. (4) The factor structure and prioritization outcomes are shaped by Vietnam’s regulatory and industry context, which may not reflect safety dynamics elsewhere. Future studies should apply the framework in other regions to evaluate the consistency of safety dimensions across diverse construction environments.

5. Conclusions

This study aims to advance construction safety research by identifying and prioritizing the critical factors influencing occupational safety in high-rise building projects. It uses a hybrid methodological framework that integrates EFA and the AHP. By integrating empirical factor identification with expert-informed prioritization, the study bridges the gap between theoretical understanding and actionable insight. The resulting framework provides a structured basis for strategic safety planning, training design, and resource allocation in high-rise construction environments. Since then, the framework has provided valuable lessons learned for developing and underdeveloped countries to improve safety practices toward construction sustainability. Future research should consider expanding the expert sample to include a broader and more diverse range of stakeholders, such as regulatory authorities, subcontractors, and labor representatives. Moreover, future research should incorporate Confirmatory Factor Analysis (CFA), to validate the extracted factor structure. This additional step will enhance the model’s statistical robustness and allow for hypothesis testing regarding interrelationships among safety dimensions, especially when applied to larger and more diverse samples. Longitudinal studies could also examine the temporal dynamics of safety interventions and evaluate how the relative importance of safety factors may shift across different construction phases.