Early Signal Detection Research Articles

From regional signal to alert in addictovigilance

France has several monitoring systems that form the foundation of its health safety surveillance. This system is designed for the early detection of signals and their swift into an action system that enables timely, appropriate, and effective interventions to protect public health. These signals are considered alerts when sufficiently validated after an initial risk assessment and if they represent a potential threat to public health, necessitating an appropriate response. The French Addictovigilance Network (FAN) was established specifically to address complications associated with the use of psychoactive substances (PAS), both medicinal and non-medicinal, excluding alcohol and tobacco. This system, unique in Europe, provides information on one of the most complex vigilance systems due to its scope, and the risks associated with it are becoming an increasingly significant public health issue. FAN is at the forefront of identifying potential signals. It has been working closely with regional partners since 1990, particularly through a multi-source approach that combines data from various sources, in collaboration with local professionals. After outlining the missions of the addictovigilance centres and the general principles of signal detection in addictovigilance, this article will present recent examples involving similar products but different identifications or modes of consumption: synthetic cannabinoids, nitazene, pregabalin, methylphenidate. These examples demonstrate that while there are regional disparities, the risk of occurrence in other French regions should not be overlooked, and information should be rapidly communicated at the national level for preventive action.

AI as a Medical Device Adverse Event Reporting in Regulatory Databases: Protocol for a Systematic Review.

The reporting of adverse events (AEs) relating to medical devices is a long-standing area of concern, with suboptimal reporting due to a range of factors including a failure to recognize the association of AEs with medical devices, lack of knowledge of how to report AEs, and a general culture of nonreporting. The introduction of artificial intelligence as a medical device (AIaMD) requires a robust safety monitoring environment that recognizes both generic risks of a medical device and some of the increasingly recognized risks of AIaMD (such as algorithmic bias). There is an urgent need to understand the limitations of current AE reporting systems and explore potential mechanisms for how AEs could be detected, attributed, and reported with a view to improving the early detection of safety signals. The systematic review outlined in this protocol aims to yield insights into the frequency and severity of AEs while characterizing the events using existing regulatory guidance. Publicly accessible AE databases will be searched to identify AE reports for AIaMD. Scoping searches have identified 3 regulatory territories for which public access to AE reports is provided: the United States, the United Kingdom, and Australia. AEs will be included for analysis if an artificial intelligence (AI) medical device is involved. Software as a medical device without AI is not within the scope of this review. Data extraction will be conducted using a data extraction tool designed for this review and will be done independently by AUK and a second reviewer. Descriptive analysis will be conducted to identify the types of AEs being reported, and their frequency, for different types of AIaMD. AEs will be analyzed and characterized according to existing regulatory guidance. Scoping searches are being conducted with screening to begin in April 2024. Data extraction and synthesis will commence in May 2024, with planned completion by August 2024. The review will highlight the types of AEs being reported for different types of AI medical devices and where the gaps are. It is anticipated that there will be particularly low rates of reporting for indirect harms associated with AIaMD. To our knowledge, this will be the first systematic review of 3 different regulatory sources reporting AEs associated with AIaMD. The review will focus on real-world evidence, which brings certain limitations, compounded by the opacity of regulatory databases generally. The review will outline the characteristics and frequency of AEs reported for AIaMD and help regulators and policy makers to continue developing robust safety monitoring processes. PRR1-10.2196/48156.

Anomaly detection for geological carbon sequestration monitoring

Geological carbon sequestration (GCS) is a method to reduce the emissions of CO2 into the atmosphere. During GCS operations CO2 is captured from the atmosphere or industrial activities and stored in geological formations for permanent storage. Monitoring is an important element of GCS because it ensures that the stored CO2 remains safely contained in the intended formation during the long term. Additionally, monitoring wells can help to detect CO2 leaks, prompt remediation actions, and provide valuable information to optimize storage by monitoring the behavior of the CO2 over time. In this work, we propose a method for GCS anomaly detection based on an LSTM Autoencoder Neural Network and Isolation Forest. The LSTM-Autoencoder uses the monitor Bottomhole Pressure (BHP) response while CO2 is being injected into a geological structure. To account for the subsurface uncertainty, multiple subsurface model realizations are created, and using reservoir simulation, the multiple monitor BHP are generated to capture the subsurface uncertainty. Anomaly BHP points are detected using the residuals of the LSTM-Autoencoder and Isolation Forest. Additionally, an anomaly score based on the subsurface uncertainty is proposed. Finally, the method robustness is evaluated using point outliers, level shift outliers, and transient shift outliers as anomaly BHP signals. Early detection of abnormal BHP pressure signals can indicate the presence of subsurface fractures, faults, or leaks. Consequently, the correct detection of anomaly points in the pressure signals is of great importance.

Identification of Pregnancy Adverse Drug Reactions in Pharmacovigilance Reporting Systems: A Novel Algorithm Developed in EudraVigilance.

There is a need to strengthen the evidence base regarding medication use during pregnancy and to facilitate the early detection of safety signals. EudraVigilance (EV) serves as the primary system for managing and analysing information concerning suspected adverse drug reactions (ADRs) within the European Economic Area. Despite its various functionalities, the current format for electronic submissions of safety reports lacks a specific data element indicating medicine exposure during pregnancy. This paper aims to address the limitations of existing approaches by developing a rule-based algorithm in EV that more reliably identifies cases that are truly representative of an ADR during pregnancy. The study utilised the standardised MedDRA query (SMQ) 'Pregnancy and neonatal topics' (PNT) as a benchmark for comparison. Recognising that the SMQ PNT also retrieves healthy pregnancy outcomes, contraceptive failure, failed abortifacients as well as ADRs not associated with pregnancy, a novel algorithm was tailored to improve the accuracy of identifying suspected ADRs occurring during pregnancy. Upon testing, the algorithm demonstrated superior performance, correctly predicting 90% of cases reporting an ADR during pregnancy, compared to 54% achieved by the SMQ PNT. The implementation of the algorithm in EV led to the retrieval of 202,426 cases. The development and successful testing of the novel algorithm represents a step forward in pregnancy-specific signal detection in EV. Because signals associated with pregnancy may be diluted in a large database such as EV, this study lays the groundwork for future research to evaluate the effectiveness of disproportionality methods on a more refined subset of pregnancy-related ADR reports.

Modern models for predicting bankruptcy to detect early signals of business failure: Evidence from Montenegro.

This paper explores predicting early signals of business failure using modern models for bankruptcy prediction. It reviews how continuous operations enhance market value, strengthening competitiveness and reputation among stakeholders. The study involves medium and large companies in the Montenegrin market from 2015 to 2020, comprising 30 bankrupt and 70 financially stable firms. Logistic regression is also employed to create a logit model for early detection of bankruptcy signals in companies. This research establishes the empirical validity of modern models in predicting business failure in the Montenegrin market, particularly through logistic regression. Significant indicators, such as the Degree of Indebtedness (DI) and turnover ratio of business assets (TR), exhibit strong predictive power with a p-value less than 0.001 according to Likelihood ratio tests. The paper underscores the potential benefits of bankruptcy prediction for both internal and external stakeholders, especially investors, in enhancing the competitiveness of Montenegro's large and medium-sized companies. Notably, the research contributes by bridging the gap between theory and practice in Montenegro, as bankruptcy prediction models have not been extensively applied in the market. The authors suggest the possible applicability of the created logit model to neighboring countries with similar economic development levels. In that sense, the concept of predicting bankruptcy is positioned as integral to corporate strategy, impacting the overall reduction of bankruptcies. The paper concludes by highlighting its role as a foundation for future research, addressing the literature gap in the application of bankruptcy prediction models in Montenegro. The created logit model, tailored to the specific needs of Montenegrin companies, is presented as an original contribution, emphasizing its potential to strengthen the competitiveness of companies in the market.

Controversial issues in crisis management. Bridging public policy and crisis management to better understand and address crises

AbstractThe current body of multidisciplinary literature on crisis management still has some unresolved problems. This paper focuses on the following four “controversial issues” in dealing with crises: the usefulness of emergency plans; early signal detection; decision‐making amid high uncertainty; and the centralization/decentralization dilemma. The paper first presents the various, contradictory dimensions of these controversial issues, drawing on different strands of organization research, public policy theory, and crisis management studies. Next, these controversial issues are analyzed through the lens of public policy research, drawing specifically on the literature on policy robustness and policy capacities. This theoretical application shows how controversial issues can be framed differently and thus overcome—at least from an analytical and theoretical perspective—confirming that a bridge between crisis management and public policy can be very fruitful in improving our understanding of how crises can be addressed.

Unknown PD distinction in HVAC/HVDC by antenna-sensor with pulse sequence analysis

Ultra-high frequency (UHF) antenna-sensors are becoming popular for non-invasive detection of unwanted electric discharge i.e., partial discharge (PD) in high-voltage (HV) systems. Early PD signals are weak for detection and classification by UHF antenna-sensors. Early PD detection and distinction are crucial to prevent equipment failure. Typically, PD signals are distinguished in HVAC by phase resolved PD (PRPD) patterns. Whereas in HVDC, some unconventional methods are applied. However, a blind distinction of PDs under unknown HVAC/HVDC conditions still remains a challenge. In this article, we address this issue by using a lotus-shaped UHF antenna-sensor for early PD detection. Unlike cut-and-try technique-based conventionally designed bio-inspired antennas, our sensor is designed by a precisely derived equivalent circuit model to systematically optimize antenna gain over size to detect early PDs. The fabricated sensor has a size of 18×11 cm2, an average realized gain of 3.05 dBi in 740–1600 MHz frequencies, and a sensitivity index of 154.04 dBi/m2. The sensor prototype is applied to wirelessly distinguish PD signals from unknown sources under HVAC and HVDC by pulse sequence analysis. Early detection, characterization, and distinction of unknown PD signals are ensured by the proposed sensor and interpretation technique. This work offers a distinctive PD sensing method for HVAC/HVDC converter stations.

Identification of Signal of Clindamycin Associated Renal Failure Acute: A Disproportionality Analysis.

Early signal detection from post-marketing studies play a vital role in the safety of patients. Recent case reports have indicated the association of clindamycin with renal failure acute. However, renal failure acute is not known to be associated with clindamycin. The aim of the study was the identification of the potential signal of clindamycin-associated renal failure acute. The OpenVigil 2.1-MedDRA-v24 was used to query the FAERS database. The Reporting Odds Ratio (ROR) with a 95% confidence interval, was used to assess the association between clindamycin and renal failure acute. A total of 22326 cases associated with renal failure acute were found in OpenVigil 2.1- MedDRA-v24. Among these 125 were associated with clindamycin. The ROR was found to be 3.2 (2.7, 3.8) which indicates a statistically significant association between clindamycin and renal failure acute reporting. The signal strength was decreased after removing the cases of concomitantly administered drugs, however, the association of clindamycin with renal failure acute reporting remains statistically significant. The subgroup analysis results have also shown a positive signal of clindamycin with renal failure acute. The current study identified a positive signal of clindamycin as renal failure acute. However, further validation and causality assessment are required.

Conducting and interpreting disproportionality analyses derived from spontaneous reporting systems

Spontaneous reporting systems remain pivotal for post-marketing surveillance and disproportionality analysis (DA) represents a recognized approach for early signal detection. Although DAs cannot be used per se as a standalone approach to assess a drug-related risk and cannot replace clinical judgment in the individual patient, their role remain irreplaceable for rapid detection of rare and unpredictable adverse drug reactions with strong drug-attributable component (e.g., designated medical events), especially when developed by a multidisciplinary team and combined with a careful case-by-case analysis (individual inspection of reports for causality assessment or to uncover reporting patterns and clinical features). In the recent past, a remarkable increase in publications of pharmacovigilance studies using DAs was observed, albeit the quality was debated: several publications contained “spin”, namely, misinterpretation of results to infer causality, calculate incidence, or provide risk stratification, which may ultimately result in unjustified alarm. The development of dedicated Guidelines by the international READUS-PV project (https://readus-statement.org/) will allow reproducible and transparent publication of accurate DAs, thus supporting their real transferability and exploitation by regulators and clinicians. This review offered a perspective on methodological aspects (and understanding) of DAs, their rationale, design, reporting, and interpretation.

Analyzing the role of spectrum sensing for security enhancement in beyond 5G waveforms using energy detection

Spectrum sensing (SS) plays a pivotal role in bolstering the security of 5G waveforms. In this abstract, we explore how it contributes to safeguarding 5G networks. By continuously monitoring the radio environment, spectrum sensing allows for the early detection of unauthorized or malicious signals. Cognitive radio (CR) technology is a cutting-edge innovation that can alleviate the need for additional spectrum. It has the potential to influence the course of technology if applied effectively. A CR network allows secondary, unlicensed users to make opportunistic use of primary, licensed spectrum without negatively impacting primary, licensed transmission or customers. Spectrum sensing is the most crucial step in implementing opportunistic spectrum access. In the recent decade, many different spectrum sensing strategies have been proposed in the scientific literature. In this paper, the performance of energy detection-based cooperative spectrum sensing has been measured and visualized using MATLAB simulations.

A Case Study of Familiar Gastrointestinal Pathologies at Risk of Cancer. Early Detection of Degenerative Signals by Oxidative Stress Biomarkers

A Case Study of Familiar Gastrointestinal Pathologies at Risk of Cancer. Early Detection of Degenerative Signals by Oxidative Stress Biomarkers

Early FID Signal Detection Technique With a Gradient Coil for Magnetic Resonance Sounding

Early FID Signal Detection Technique With a Gradient Coil for Magnetic Resonance Sounding

Building Hyper-Awareness: How to Amplify Weak External Signals for Improved Strategic Agility

Early detection of weak external signals is increasingly critical for strategic agility. While many organizations scan for weak signals, most dismiss them as anomalies, principally due to poor amplification strategies. Several challenges hinder the necessary amplification and sensemaking of weak signals for organizational awareness. This article analyzes 139 proof-of-concept projects with startups and 15 interviews with executives involved in the projects at a leading German mobility corporation, and it reveals four actions to amplify weak external signals, thereby enhancing organizational hyper-awareness. It illustrates the actions with examples and presents the implications for both weak signals and strategic agility management.

Monitoring the safety of drugs and COVID-19 vaccines by the French Pharmacovigilance Centers during the pandemic: A win-win bet with Health Authorities!

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV2) pandemic virus was a "health crisis" and a significant burden also for the French pharmacovigilance system. It took its toll in 2 phases, the first being in early 2020 when very little was known, and during which the missions of the 31 Regional Pharmacovigilance Centers (RPVCs) from university hospitals were to detect adverse reactions of drugs used in the context of the disease. Whether as a possible aggravating role on COVID-19, or displaying a different safety profile during its course, or to assess safety of curative treatment, this phase preceded that of the arrival of dedicated vaccines. Then the RPVCs' missions were to detect, as early as possible, any new serious adverse effect leading to a potential signal that would modify the benefit/risk ratio of a vaccine and require the implementation of health safety measures. During these two distinct periods, signal detection remained the core business of the RPVCs. Each RPVC had to organize itself to handle an unprecedented surge of declarations and requests for advice, from health care professionals and patients alike. "Leading" RPVCs, who were in charge of monitoring vaccines, had to deal with an extraordinary workload (still going on to this date), to generate in real-time and on a weekly basis, a summary of all the adverse drug reaction (ADR) reports as well as an extended analysis of the different safety signals. The organization put in place at the beginning of the health crisis, adapted to the context of the vaccines, allowed to meet the challenge of real-time pharmacovigilance monitoring, and to identify many safety signals. Efficient "short-circuits exchanges" with the French Regional Pharmacovigilance Centers Network (RPVCN) were paramount to the National Agency for the Safety of Medicines and Health Products (ANSM) to develop an optimal collaborative partnership. The French RPVCN has shown at this occasion both agility and flexibility, swiftly adapting to vaccine- and media-related unrest, and demonstrated its effectiveness in the early detection of safety signals. This crisis also confirmed the superiority of manual/human signal detection over automated ones, as the most effective and powerful tool to date to rapidly detect and validate a new ADR and enable to elaborate rapid risk reduction measures. To maintain the performance of French RPVCN in signal detection and to monitor all drugs as they should, and as expected by our fellow citizens, a new funding model should be considered.

Abstract Number ‐ 140: Transcranial Doppler Emboli Monitoring for Stroke Prevention after Flow Diversion: A Single Center Experience

Introduction Following the favorable clinical and radiographical results of flow diversion, its use has continually expanded for the treatment of complex intracranial aneurysms. The high metal surface of flow diverters (FD) increases the risk of thromboembolic complications. Current emboli monitoring data after FD placement is limited. Transcranial doppler (TCD) offers a non‐invasive evaluation of microembolic signals and might be a valuable tool to prevent ischemic complications. We evaluated the rate of embolic signals and their management after FD implantation. Methods We performed a retrospective, single‐center evaluation of adult patients who underwent a FD procedure with the Pipeline Embolization Device (PED) for unruptured intracranial aneurysms between 2012 and 2016. Covariates included patient demographics, comorbidities, aneurysm characteristics, procedure characteristics, and post‐procedure TCD emboli monitoring (immediate and >12 hours post‐operative). TCD emboli signals were classified as positive (< 15 microembolic signals) and strongly positive (>15 signals). The primary outcome was the rate of symptomatic stroke at 2 and 12 weeks in these patients. Results 105 patients with 132 aneurysms were analyzed. Mean age was 59.7 years, and 78% were females. Mean aneurysm size was 7.8 ± 6.4 mm. Most of them were located in the paraclinoid region (40%). Two PEDs were implanted in 12 (11%) cases and adjunctive coiling was performed in 15 (14%) cases. Microemboli were detected in 11.4% of cases (12/105 patients) after PED deployment. In 5/12 patients, the emboli resolved upon repeat testing with no intervention. In 6/12 patients, further heparinization, additional clopidogrel administration, or change in antiplatelet agents were needed. These treatments reduced the microembolic events. In 1 patient who was clopidogrel‐resistant and a ticagrelor super‐responder, mechanical thrombectomy was necessary due to acute in‐stent thrombosis. There were no symptomatic strokes at 2 and 12 weeks. Conclusions TCD emboli monitoring can be a powerful non‐invasive tool for the early identification and prevention of thromboembolic events after FD implantation. Early detection of microembolic signals offers the possibility to modify postoperative management, avoid silent emboli, and potentially avoid symptomatic strokes.

Spiritual Care in the Intensive Care Unit: Experiences of Dutch Intensive Care Unit Patients and Relatives.

To gain insight into both patients' and relatives' experiences with spiritual care (SC) in the intensive care unit (ICU). Method used was qualitative interviewing. This was a thematic, topic-centered, biographical, and narrative approach, using semistructured interviews with thematic analysis. A purposive sampling method was used to select a sample of ICU patients and ICU patients' relatives. An interview guide facilitated individual, semistructured interviews. The interview data were recorded by means of note-taking and audio-recording. Verbatim transcripts were compiled for analysis and interpretation. All 12 participants-7 ICU patients and 5 family members of 5 other ICU patients-experienced ICU admission as an existential crisis. Participants would appreciate the signaling of their spiritual needs by ICU health care professionals (HCPs) at an early stage of ICU admission and subsequent SC provision by a spiritual caregiver. They regarded the spiritual caregiver as the preferred professional to address spiritual needs, navigate during their search for meaning and understanding, and provide SC training in signaling spiritual needs to ICU HCPs. Early detection of existential crisis signals with ICU patients and relatives contributes to the mapping of spiritual and religious needs. Spiritual care training of ICU HCPs in signaling spiritual needs by ICU patients and relatives is recommended. Effective SC contributes to creating room for processing emotions, spiritual well-being, and satisfaction with integrated SC as part of daily ICU care.

Early Detection of Adverse Drug Reaction Signals by Association Rule Mining Using Large-Scale Administrative Claims Data.

Adverse drug reactions (ADRs) are a leading cause of mortality worldwide and should be detected promptly to reduce health risks to patients. A data-mining approach using large-scale medical records might be a useful method for the early detection of ADRs. Many studies have analyzed medical records to detect ADRs; however, most of them have focused on a narrow range of ADRs, limiting their usefulness. This study aimed to identify methods for the early detection of a wide range of ADR signals. First, to evaluate the performance in signal detection of ADRs by data-mining, we attempted to create a gold standard based on clinical evidence. Second, association rule mining (ARM) was applied to patient symptoms and medications registered in claims data, followed by evaluating ADR signal detection performance. We created a new gold standard consisting of 92 positive and 88 negative controls. In the assessment of ARM using claims data, the areas under the receiver-operating characteristic curve and the precision-recall curve were 0.80 and 0.83, respectively. If the detection criteria were defined as lift > 1, conviction > 1, and p-value < 0.05, ARM could identify 156 signals, of which 90 were true positive controls (sensitivity: 0.98, specificity: 0.25). Evaluation of the capability of ARM with short periods of data revealed that ARM could detect a greater number of positive controls than the conventional analysis method. ARM of claims data may be effective in the early detection of a wide range of ADR signals.

Pharmacovigilance des médicaments et des vaccins contre la COVID-19 durant la pandémie : comment le Réseau français des centres régionaux de pharmacovigilance a relevé le défi ?

The pandemic subsequent to the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus resulted, for the French institutional pharmacovigilance, in a "health crisis" in 2 phases: the coronavirus disease 2019-"COVID-19" phase during which the missions of the Regional Pharmacovigilance Centres (RPVC) were to detect a possible impact of drugs on this disease, as whether existed a possible aggravating role of certain drugs, or the safety profile of drugs used for the management of COVID-19 could evolve. The second phase followed the availability of COVID-19 vaccines, during which the RPVCs' missions were to detect as early as possible any new serious adverse effect, source of a potential signal that would modify the benefit/risk ratio of a vaccine and require the implementation of health safety measures. During these two periods, signal detection remained the core business of the RPVCs. The RPVCs had to organize themselves to handle an historical surge of declarations and requests for advice, whereas the RPVCs in charge of monitoring vaccines had to deal with an extraordinary dense activity over a long period of time, in order to produce in real time and on a weekly basis, a summary of all the declarations and an analysis of safety signals. The national organization put in place made it possible to meet the challenge of real-time pharmacovigilance monitoring of 4 vaccines with conditional marketing authorizations. Short-circuit efficient exchanges with the French Regional Pharmacovigilance Centres Network was paramount for the French National Agency for medicines and health products (Agence nationale de sécurité du médicament et des produits de santé) to develop an optimal collaborative partnership. The RPVC network has shown agility and flexibility, has been able to adapt swiftly and demonstrated its effectiveness in the early detection of safety signals. This crisis confirmed the superiority of manual/human signal detection as the most effective and powerful tool to date, to rapidly detect a new adverse drug reaction and enable to elaborate rapid measures of risk reduction. In order to maintain the performance of French RPVCs in signal detection and to monitor all drugs as they should and as expected by our fellow citizens, a new funding model correcting the inadequacy of RPVCs' expertise resources in relation to the volume of reports should be considered.

Early detection of anthropogenic climate change signals in the ocean interior

Robust detection of anthropogenic climate change is crucial to: (i) improve our understanding of Earth system responses to external forcing, (ii) reduce uncertainty in future climate projections, and (iii) develop efficient mitigation and adaptation plans. Here, we use Earth system model projections to establish the detection timescales of anthropogenic signals in the global ocean through analyzing temperature, salinity, oxygen, and pH evolution from surface to 2000 m depths. For most variables, anthropogenic changes emerge earlier in the interior ocean than at the surface, due to the lower background variability at depth. Acidification is detectable earliest, followed by warming and oxygen changes in the subsurface tropical Atlantic. Temperature and salinity changes in the subsurface tropical and subtropical North Atlantic are shown to be early indicators for a slowdown of the Atlantic Meridional Overturning Circulation. Even under mitigated scenarios, inner ocean anthropogenic signals are projected to emerge within the next few decades. This is because they originate from existing surface changes that are now propagating into the interior. In addition to the tropical Atlantic, our study calls for establishment of long-term interior monitoring systems in the Southern Ocean and North Atlantic in order to elucidate how spatially heterogeneous anthropogenic signals propagate into the interior and impact marine ecosystems and biogeochemistry.

Solid oxide fuel cells for shipping: A machine learning model for early detection of hazardous system deviations

With rising concerns about the amount of pollutant emissions generated by shipping and the consequent pressure to curb the environmental impact of shipping activities, fuel cells are expected to take an important role in ship propulsion. In particular, Solid Oxide Fuel Cells (SOFCs) are envisaged to provide high electrical efficiency and offer the opportunity of combining heat and power production. This work deals with the safety issues related to the safety implications of the use of Fuel Cells in maritime applications. A machine-learning model for identifying and intercepting critical events, based on the early detection of the system weak signals, is developed and applied to a Solid Oxide Fuel Cell (SOFC) system. The model relies on a hybrid approach: a data-driven model based on gradient-boosted decision trees and a computational model of the SOFC system are integrated to enhance the data-driven approach by implementing physics-based knowledge to boost the resulting predictive capabilities. The outlined approach even if it requires further validation at the full scale may be considered a step forward in enabling the prediction of the conditions that may lead to an accident with remarkable accuracy.