Karst cave geotourism in Indonesia faces serious threats due to uncontrolled tourism pressure on vulnerable ecosystems. The lack of a data-driven management system leads to environmental degradation, tourist safety risks, and a decline in destination quality. This research aims to design a digital information system to measure tourism performance sustainably, with a case study of Jomblang Cave in Yogyakarta, Indonesia. Its contribution is to present a digital solution to monitor aspects of safety, enjoyment, and support decision-making in sustainable tourism governance. This research uses mixed methods with quantitative and qualitative approaches, and develops an information system based on the Waterfall-type System Development Life Cycle (SDLC) model through the stages of analysis, design, and implementation. Analysis was conducted using Use Case Diagram (UCD), Entity Relationship Diagram (ERD), and Data Flow Diagram (DFD) to design a system that meets the needs. The system can automatically calculate geosite performance values using the Modified Geosite Assessment Model (M-GAM) method, including two additional variables, namely safety and enjoyment. The web-based Jomblang Cave tourist attraction performance measurement information system consists of a home page, a questionnaire, and an assessment matrix that allows users to provide assessments, while the admin and owner monitor the results based on automatically processed inputs. The implementation phase maps the data of 32 tourists and six experts into the M-GAM Matrix, and calculates the MV and AV values to determine the Z-index position. The system automatically generates evaluative recommendations based on the lowest scores and has performed as expected as an evaluation and decision-making tool in tourism management. This research presents a real-time and easily accessible system for measuring tourist attraction performance. The system generates MV 6.37 and AV 16.09, placing the Z-index at a medium level (Z22). The automated system detected major weaknesses in the Education (VSE = 1.82) and functionality (VFn = 2.87) factors, especially regarding cave protection and tourist safety. Automated recommendations allow managers to focus on improving critical areas and support strategic decisions for geotourism development.

Background Artificial intelligence (AI) is rapidly transforming medical education and clinical practice. AI-driven clinical decision support systems, diagnostic tools, and smart tutoring systems are helping teach and guide medical students on developing clinical reasoning skills and making better-informed patient care decisions. AI literacy initiatives have grown in recent years to increase understanding of both how AI works and how to utilize it; however, medical educators receive minimal guidance regarding how to instruct their learners to appropriately question or override AI recommendations. Thus, the educational gap created by a lack of guidance places learners at risk of automation bias,the tendency to over-rely on computer-based recommendations, regardless of whether they conflict with clinical judgement or individual patient situation. Applied Insights The Applied Insights presented in this article are organized around commonly encountered educational contexts where learners interact with AI-assisted decision-making. They offer actionable strategies for helping learners recognize when AI recommendations should be questioned, contextualized, or overridden. For example, mismatches between patients and training populations, incomplete or inaccurate input data, and misalignment between the system’s priorities and the patient’s values. The Applied Insights presented are based on well-established literature on automation bias, patient safety, and clinical decision-making, and were written to be non-technology-specific, useful across multiple specialties and resources, and adaptable to current curricula without requiring AI-specific knowledge. Conclusion Medical educators have a responsibility to prepare learners to use AI safely in clinical practice. By providing strategies for teaching when and how to question AI recommendations, this article supports the development of professional judgment, patient-centered decision-making, and safe integration of AI in health professions education.

Nursing handovers in intensive care units (ICUs) are critical for ensuring continuity of care, patient safety and clinical decision-making. Beyond information transfer, handovers are key moments in which nursing expertise can influence treatment trajectories. However, power relations, communication practices and the visibility of nursing knowledge during handovers remain underexplored, particularly through the lens of critical medical anthropology and the concept of heterotopia. To explore how experienced ICU nurses participate in and influence clinical decision-making through handover practices. This ethnographic study, grounded in a critical medical anthropology approach, was conducted in a general ICU in Greece. Data collection involved overt participant observation and ad hoc ethnographic interviews with ICU nurses and physicians, as part of a larger PhD study. Nurses who participated in interviews subsequently consented to observation while working. Fieldwork spanned 2 years (2012-2014) and concluded in 2020. Multiple nursing handovers were observed, and thematic analysis was applied to fieldnotes using Atlas.ti. Nursing handovers functioned as heterotopias of nursing expertise, offering moments of access to clinical decision-making. Four sub-themes emerged: (a) narrating the patient's illness trajectory; (b) showcasing expertise and mentoring less experienced colleagues; (c) verbalising and consolidating clinical decisions; and (d) temporarily reversing hierarchical power dynamics. Through these practices, experienced nurses rendered their clinical reasoning visible and exerted influence over patient management. Nursing handovers represent strategic opportunities for nurses to navigate and renegotiate professional boundaries within the ICU. The findings highlight the potential of handovers to strengthen team collaboration, support clinical decision-making and enhance patient care. Recognising handovers as protected spaces for nursing expertise is essential for promoting professional autonomy within critical care settings. Protecting nursing handovers as spaces of professional dialogue can enhance communication, clinical reasoning and patient safety. Combining structured and narrative-based handover training may empower nurses, preserve mentoring and role-modelling functions and sustain professional visibility-particularly relevant in the post-COVID-19 era, where digital and hybrid communication models increasingly shape critical care practice.

Game-based learning in pediatric nursing: Escape room effects on safety and clinical decision-making

Enhanced safety and emergency decision-making in NPPs: Multi-step parameter prediction for complex accident scenarios

Intelligent defensive driving for autonomous vehicles: Framework, strategy and verification.

Background: Adverse drug reactions (ADR) pose substantial risks to patient safety and challenge clinical decision-making. However, traditional predictive approaches frequently fail to deliver interpretable insights into the complex interplay between pharmaceuticals and biological systems. Methods: We propose the KRDQN (Knowledge Graph Reinforced Deep Q-Network) predictive framework. First, a knowledge graph (KG) that encompasses five entity types—drug, target, pathway, gene, and adverse drug reaction (ADR)—is constructed, and each node is enriched with intrinsic attribute features. A Deep Q-Network (DQN) is subsequently deployed within a reinforcement learning paradigm to generate interpretable ADR predictions. Model performance is evaluated by five-fold cross-validation, with accuracy and AUC reported. Finally, the Spearman correlation coefficients between drug–drug similarity and path–path similarity are computed, and case studies are conducted to further assess the predictive capability of KRDQN. Results: We evaluated KRDQN on a comprehensive data set encompassing both drug–drug interactions and ADR records. Experimental results demonstrate that KRDQN surpasses state-of-the-art baselines, attaining a recall of 0.8171 and an AUC of 0.8327. Furthermore, to demonstrate the practical value of the KRDQN prediction framework, we applied it to predict potential ADRs and their mechanism pathways for the drugs sunitinib and indomethacin. The results indicated that the KRDQN framework could identify biological mechanism pathways consistent with clinical evidence. Conclusions: In this study, we developed the reinforcement learning-based KRDQN predictive framework, which outperforms existing baselines in predictive performance and yields interpretable adverse drug reaction (ADR) predictions, thereby serving as a powerful tool for pharmacovigilance and clinical decision-making.

A safe physical human-robot interaction (pHRI) in rehabilitation requires reliable perception and low-latency decision making under heterogeneous and unreliable sensor inputs. This paper presents a multimodal sensor-fusion-based safety framework that integrates physical state estimation, semantic information fusion, and an edge-deployed large language model (LLM) for real-time pHRI safety control. A dynamics-based virtual sensing method is introduced to estimate internal joint torques from external force-torque measurements, achieving a normalized mean absolute error of 18.5% in real-world experiments. An asynchronous semantic state pool with a time-to-live mechanism is designed to fuse visual, force, posture, and human semantic cues while maintaining robustness to sensor delays and dropouts. Based on structured multimodal tokens, an instruction-tuned edge LLM outputs discrete safety decisions that are further mapped to continuous compliant control parameters. The framework is trained using a hybrid dataset consisting of limited real-world samples and LLM-augmented synthetic data, and evaluated on unseen real and mixed-condition scenarios. Experimental results show reliable detection of safety-critical events with a low emergency misdetection rate, while maintaining an end-to-end decision latency of approximately 223 ms on edge hardware. Real-world experiments on a rehabilitation robot demonstrate effective responses to impacts, user instability, and visual occlusions, indicating the practical applicability of the proposed approach for real-time pHRI safety monitoring.

This study investigates strategic decision-making for integrating artificial intelligence–based predictive safety systems into occupational health and safety (OHS) management. The aim is to develop and apply a rigorous, transparent multi-criteria decision framework that helps organizations select among competing AI-driven safety solutions under uncertainty. The core research question is: Which AI-based predictive safety alternative offers the best balance of safety improvement, organizational feasibility, and strategic fit for OHS management? An integrated fuzzy MCDM approach combines Fuzzy Best–Worst Method (FBWM) to elicit criterion weights with MARCOS to prioritize alternatives evaluated by domain experts across technical performance, human factors, legal/regulatory fit, cost, and implementation readiness. The analysis highlights the dominant influence of safety impact and technological readiness on final rankings, while cost and legal compliance act as moderating considerations. Sensitivity tests across weighting schemes indicate stable priority orders without critical rank reversals, supporting managerial robustness. The findings provide actionable guidance for investment and OHS committees, demonstrate the practicality of a hybrid fuzzy model for high-risk settings, and clarify both the study’s aim and its central research question for future replications.

Drug-induced upper gastrointestinal bleeding (UGIB) is a serious adverse event that deserves close attention. This study conducted a real-world pharmacovigilance research, aiming to enhance the understanding of drug safety and more effectively identify and prevent potential risks. This study extracted data related to UGIB reported in the Food and Drug Administration Adverse Event Reporting System (FAERS) and Japanese Adverse Drug Event Report (JADER) databases from the first quarter of 2004 to the second quarter of 2024. We selected the top 50 drugs with higher frequency and conducted safety analyses using four signal detection methods: Reporting Odds Ratio, Proportional Reporting Ratio, Empirical Bayes Geometric Mean, and Bayesian Confidence Propagation Neural Network. Through data mining analysis, we found that the number of patients with UGIB reported was 62,941, including 57,414 in the FAERS and 5,527 in the JADER. It is particularly noteworthy that aspirin frequency and signal strength were among the top five in both databases. Rivaroxaban, warfarin, and pradaxa not only had the highest number of reports in the FAERS database but also showed highly in terms of their signal values. In the JADER database, clopidogrel, loxoprofen, apixaban, and bevacizumab had a higher number of reports, and it was also observed that esflurbiprofen/mentha oil and lornoxicam exhibited extremely high signal values. Meanwhile, meloxicam and prasugrel also had relatively high signal values. This study conducted a pharmacovigilance analysis of drug-related UGIB by integrating and analyzing multiple adverse drug reaction databases. In both the FAERS and JADER databases, we not only identified some common risk-signaling drugs but also discovered database-specific risk-associated medications. In particular, this study performed a systematic quantitative risk assessment of the selected high-reporting-frequency drugs. This analysis not only deepened our understanding of drug risk profiles but also provided important reference evidence for clinical medication safety decision-making.

Background/Objectives: Artificial intelligence (AI) is increasingly applied to drug safety evaluation, yet evidence is dispersed across lifecycle stages and tasks. This scoping review aimed to (1) map how AI supports safety- and treatment-related decision types across the drug lifecycle, and (2) examine evaluation strategies used to assess model reliability for clinical or regulatory use. Methods: Using Arksey and O'Malley's framework, we searched a major database for studies published in the past decade that applied AI or machine learning to drug safety or medication-related decisions. After screening, we extracted data on lifecycle stage, decision type, AI methods, data sources, and evaluation strategies. A lifecycle-decision matrix was constructed to characterize application patterns. Results: AI applications were concentrated in real-world clinical care × patient-level safety prediction and post-marketing × safety surveillance, using EHRs, spontaneous reporting systems, and clinical text. Common methods included gradient boosting, deep neural networks, graph neural networks, and natural language processing models. This concentration reflects structural incentives favoring safety-oriented applications with readily available data and lower decision liability. Evidence for treatment optimization, regulatory decision modeling, and evidence synthesis was limited. Most studies used internal validation; external validation and real-world deployment were uncommon, indicating early methodological maturity and limited translational readiness. Conclusions: AI demonstrates strong potential to enhance drug safety-particularly in risk prediction and pharmacovigilance-but its use remains uneven across the lifecycle. By situating AI applications within explicit lifecycle stages and decision contexts, this review clarifies where progress has advanced, where translation has stalled, and why these gaps persist. Limited external validation and minimal real-world testing constrain clinical and regulatory adoption. These findings suggest that external validation and real-world testing may contribute to further advances in AI for drug safety.

Abstract Background: Immune checkpoint inhibitors (ICIs) in combination with chemotherapy, have markedly transformed the therapeutic alogorithm of high risk early-stage triple-negative breast cancer (eTNBC). Although the safety and efficacy of ICIs have been extensively evaluated in randomized clinical trials (RCTs), their toxicity and effectiveness profiles may differ in real-world clinical practice. This systematic review and meta-analysis aimed to compare the incidence of all-grade adverse events (AEs), grade ≥3 AEs, treatment related discontinuation, and pathological complete response (pCR) rates across phase I-III RCTs and real-world evidence (RWE) in patients with eTNBC receiving neoadjuvant ICIs. Methods: Fourteen RCTs and ten RWE studies were included in the analysis. Incidence of all-grade AEs, grade ≥3 AEs, treatment related discontinuation and pCR rate were collected. Proportions were logit-transformed to stabilize variances and pooled using a random-effects model with Hartung-Knapp adjustment. Subgroup differences between RWE and RCTs were evaluated using the Q-test for subgroup analysis. Separate analyses were conducted for pCR, overall AEs, grade ≥3 AEs and treatment discontinuation to explore consistency of findings across different outcomes. Results: The pooled incidence of all-grade AEs was higher in RCTs compared to RWE, with estimates of 93% (95% CI: 78-98%) and 80% (95% CI: 27-98%), respectively. However, the difference between subgroups did not reach statistical significance in the random-effects model (Q = 1.18, p = 0.28). Similarly, the incidence of grade ≥3 AEs was 49% in RCTs (95% CI: 27-71%) versus 30% in RWE (95% CI: 14-52%), with no significant difference observed between groups (Q = 1.85, p = 0.17). Treatment discontinuation rates were consistent across study types, with both RCTs and RWE showing a pooled incidence of 19%, with no difference between subgroups (Q = 0.00, p = 0.97). Regarding pCR, the pooled rate was 58% (95% CI: 53-62%) in RCTs and 54% (95% CI: 43-65%) in RWD, with no statistically significant subgroup difference (Q = 0.46, p = 0.50). Conclusions: The incidence and severity of adverse events and pCR rates, were comparable between randomized clinical trials and real-world evidence, indicating that immunotherapy maintains consistent safety and effectiveness profiles across both contexts. These results reinforce the external validity of trial-derived data for application in routine clinical practice. Ongoing post-marketing surveillance and standardized reporting remain crucial to ensure patient safety, optimize outcomes, and inform evidence-based treatment decisions. Citation Format: A. Longobardi, R. Buonaiuto, C. Calderaio, A. Caltavituro, V. Cantile, G. Crimaldi, F. Puglisi, L. Del Mastro, C. De Angelis, M. De Laurentiis, M. Giuliano, G. Arpino. Safety and Efficacy of Immune Checkpoint Inhibitors in Early Triple Negative Breast Cancer: A Systematic Review and Meta-Analysis of Real-World Evidence vs. Clinical Trial Data [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS4-07-13.

Focusing on the construction of a 58-m-diameter double-layer steel space frame dome at the Han Culture Museum Assembly Hall, this study addresses scheme selection and safety control challenges in staggered jacking of large-span spatial structures. A three-dimensional finite element model in MIDAS Gen simulated the three-stage jacking process to compare three temporary support layouts. Numerical evaluation metrics included maximum vertical displacements, peak internal forces, the proportion of members undergoing stress state transitions, and spatio-temporal evolution of stress concentrations. Scheme B demonstrated superior performance, reducing peak vertical displacement by 44% under critical conditions, lowering peak stresses, and enabling more uniform internal force redistribution—effectively mitigating tension–compression cycling and buckling risks. Crucially, only nodal displacements and support elevations were monitored in situ using a 3D system based on magnetic prisms and total stations; no strain or force measurements were conducted due to practical constraints during construction. Monitoring data show good agreement with simulated displacements and support elevations under Scheme B, validating the model’s deformation response. However, localized deviations—including a 29 mm deflection discrepancy and elevation errors up to 28 mm—reveal the influence of uneven boundary conditions, with potential implications for long-term structural behavior. The findings confirm that numerical predictions of deformation are reliable, while internal forces remain unvalidated by field data. The integrated approach of “scheme comparison–construction simulation–full-process displacement monitoring” proves effective for safety control and decision-making in complex jacking operations, offering a transferable framework for similar large-span double-layer space frame projects.

Immune-mediated haemolysis caused by red blood cell (RBC) auto or alloantibodies depends on several factors, including antibody subclass. Immunoglobulin (Ig) IgG1 and IgG3 are more efficient at triggering phagocytosis and complement activation. This study evaluated whether IgG subclass determination can help predict the clinical relevance of RBC antibodies in different immunohaematological contexts. Blood donors and patients with IgG-positive direct antiglobulin tests (DATs) were included. IgG subclasses were determined using monospecific gel cards for IgG1/IgG3. The monocyte monolayer assay (MMA) assessed in vitro biological relevance. Antibody specificity was established by standard immunohaematological techniques. Statistical comparisons were performed using Chi-square, Fisher's exact and Mann-Whitney U tests. Among patients with IgG autoantibodies (n = 29), 51.7% had IgG1 or IgG3, versus 3.7% of donors (n = 27; p < 0.001). The presence of IgG1/IgG3 autoantibodies showed 93% positive predictive value (PPV) and 96.3% specificity for distinguishing patients from donors. IgG1/IgG3 autoantibodies were more frequently associated with positive MMA results (83.3% vs. 33.3%). Among RBC alloantibodies (n = 17), 64% were IgG1/IgG3, correlating with MMA positivity (sensitivity 78%; PPV 77%). Antibodies traditionally considered benign were often IgG1/IgG3 and MMA-positive. IgG subclass determination provides diagnostic value beyond IgG quantification alone. In DAT-positive donors, not detecting IgG1/IgG3 is compatible with a low haemolysis risk and often obviates follow-up; when IgG1/IgG3 are detected, selective evaluation may be appropriate. For alloantibodies, subclass identification may help predict clinical relevance, especially in urgent transfusion settings when MMA is unavailable, supporting transfusion safety decisions and efficient resource use.

Child welfare caseworkers have significant decision-making authority, ultimately determining if allegations are substantiated, if a case should be opened for ongoing services, and if removal from the parental home is required. This research considers the role of caseworker turnover on these decisions, considering decisions for 372,968 unique screened-in reports assessed by 2,128 unique child welfare caseworkers in Wisconsin. Specifically, this study uses logistic regressions to estimate the likelihood of maltreatment substantiation, case opening, child removal, and the timeliness of the assessment as a function of the caseworker's timeline to departure from the public child welfare field, net of case-related characteristics, caseworker demographics, and county/year fixed effects. Results indicate that the odds of substantiation, case opening, and timely assessment are lower as caseworkers near departure. Findings expand our understanding of the consequences of turnover, highlighting how it influences case decisions and providing insight into effects on child and family well-being.

Autonomous Vehicles (AVs) have the potential to reduce road accidents and improve transportation efficiency significantly. However, safety concerns and the lack of transparency in decision-making remain major barriers to their widespread adoption. Modern AV systems rely heavily on complex Artificial Intelligence (AI) and Deep Learning models, which often function as black boxes, making it difficult to understand or justify their actions. This paper explores the critical role of safety mechanisms and decision justification in autonomous driving systems. We discuss the AV decision pipeline, identify safety challenges, and highlight the importance of Explainable Artificial Intelligence (XAI) techniques in improving trust, accountability, and regulatory compliance. The paper concludes by outlining open research challenges and future directions for safer and more transparent autonomous driving systems.

Road traffic crashes are a leading cause of death across all age groups, particularly among children and young adults. This study evaluates the reliability of the empirical Bayes (EB) method for predicting crash frequencies to improve road safety decision-making. The EB method, initially introduced in an earlier road safety study, has been widely adopted and promoted by the Highway Safety Manual (HSM) as a standard approach for estimating crash frequencies. Despite the application of the EB method and the calibration procedures recommended by the HSM, uncertainties may still arise because of the inclusion of diverse crash-related variables in the prediction model. Reliability analysis, employing limit state functions and the Monte Carlo sampling method, was used to assess the probability of noncompliance P nc ) between expected and observed crash frequencies, and to identify and rank the sources of uncertainty in the model. The case study included 64 segments of two-lane, two-way rural highways in Iran, with 34 tangents and 30 horizontal curves. Based on actual geometric and traffic data, all EB model variables were computed. The results revealed segment-specific P nc , and a ranking of model variables was established using importance vectors to identify the most influential sources of uncertainty.

This retrospective study aimed to systematically evaluate the impact of catheter-related thrombosis (CRT) on the safety, timing, and clinical decision-making for femoral venous catheter removal in poisoned patients undergoing short-term extracorporeal treatments. This study characterized the incidence of CRT associated demographic, laboratory, and clinical factors, thrombosis-related complications, and the role of anticoagulation therapy in management, with the goal of establishing evidence-based guidelines for optimizing catheter removal strategies in this population. Seventy-one poisoned patients with femoral venous catheters for extracorporeal treatments were retrospectively classified into CRT and non-CRT groups based on femoral venous ultrasound findings. Group differences in thrombosis incidence, baseline demographics, laboratory parameters (fibrinogen, D-dimer), clinical characteristics, thromboembolic complications, and anticoagulation effects were evaluated to inform evidence-based catheter removal practices. CRT occurred in 38% of patients that received a duplex ultrasound. Local insertion-site symptoms (51.9% vs 9.1%, P < .001) and progressive deep vein thrombosis (11.1% vs 0%, P = .051) were more common in the thrombosis group, while no pulmonary embolism or major bleeding occurred in either group. Those with thrombosis had significantly higher fibrinogen (3.23 vs 2.79 g/L, P = .019) and D-dimer (1.16 vs 0.36 mg/L, P < .001) concentration, along with a higher D-dimer positivity rate (74.1% vs 27.3%, P < .001). Multivariate analysis identified positive D-dimer (adjusted odds ratio = 7.83, P = .002) and local symptoms (adjusted odds ratio = 10.90, P = .002) as independent predictors. Post-removal ultrasound showed complete recanalization in 81.5% of thrombosis patients, with thrombus progression in 3.7%; anticoagulation duration correlated with partial recanalization (thrombus reduction or stability) in 14.8% of cases. In this small cohort of poisoned patients with short-term extracorporeal treatments, CRT dose not pose immediate life-threatening risks for femoral venous catheter removal, justifying safe and timely removal. Elevated fibrinogen, positive D-dimer, and local puncture-site symptoms indicate increased thrombosis risk. Prospective studies are needed to validate these findings and refine management strategies.

End-tidal carbon dioxide (ETCO₂) monitoring is a vital, noninvasive technique for assessing ventilation, circulatory status, and predicting adverse events in pediatric emergency departments (EDs). This review aims to synthesize current evidence, examine barriers, and highlight strategies to optimize ETCO₂ monitoring in pediatric emergency settings. A narrative review of the literature was conducted, encompassing epidemiological data, clinical guidelines, expert consensus statements, and recent studies targeting ETCO₂ monitoring in pediatric emergencies. Key topics evaluated include physiological principles, airway management, prognostic value in cardiac arrest, procedural sedation safety, sepsis triage, limitations, and future directions. Data from both high- and low-resource settings were included. ETCO₂ monitoring demonstrates high sensitivity and specificity for confirming endotracheal tube placement and early detection of respiratory compromise-identifying hypoventilation, apnea, and airway obstruction minutes before pulse oximetry. During CPR, persistently low ETCO₂ values correlate with poor outcomes, while sudden increases signal return of spontaneous circulation. In procedural sedation, routine capnography reduces hypoxic episodes and adverse events. In sepsis, ETCO₂ inversely correlates with lactate levels, offering a rapid, non-invasive marker of perfusion, though its reliability diminishes in multisystemic shock. Challenges include equipment limitations, provider training gaps, lack of universal protocols, and cost barriers-especially in low-resource settings. ETCO₂ monitoring is an essential tool in pediatric emergency care, enhancing safety and clinical decision-making across multiple scenarios. Addressing implementation barriers through education, standardized protocols, and accessible technology is crucial to ensure widespread adoption and improved outcomes for critically ill children.

ABSTRACT Assessing and managing risks in offshore helideck operations is critical due to dynamic and hazardous conditions. This study introduces a hybrid risk assessment framework that combines the Fine–Kinney method with an intuitionistic fuzzy extension of the TODIM approach, representing the first integration of these two techniques for helideck safety evaluation. The Fine–Kinney method was applied to compute risk scores based on probability, exposure, and consequence, while the TODIM technique facilitated risk prioritization by reflecting decision-maker preferences. To handle uncertainty in expert evaluations, intuitionistic fuzzy sets (IFSs) were employed. The results revealed that turbulence and high wind speeds were the most critical factors, with dominance scores of 0.86 and 0.79, followed by human-related issues. In contrast, structural and inspection-related risks were perceived as less severe. The proposed model enhances the reliability of helideck risk prioritization and can support more effective safety decisions.