EKElizaveta Kopacheva

Elizaveta Kopacheva

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Elizaveta Kopacheva has a total of 11 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Empirical evidence of direct and indirect relations between environmental pressure and conflict

    by Elizaveta Kopacheva, Giangiacomo Bravo, Davide Natalini

    The link between natural resources and conflict has led to a heated debate in the literature at the crossroad between social and environmental sciences.1 A recent metaanalysis concluded that, based...


  • Open Access

    Identifying Adverse Drug Events in Clinical Text Using Fine-Tuned Clinical Language Models: Machine Learning Study

    by Elizaveta Kopacheva, Aron Henriksson, Hercules Dalianis, Tora Hammar, Alisa Lincke

    Background: Medications are essential for health care but can cause adverse drug events (ADEs), which are harmful and sometimes fatal. Detecting ADEs is a challenging task because they are often not documented in the structured data of electronic health records (EHRs). There is a need for automatically extracting ADE-related information from clinical notes, as manual review is labor-intensive and time-consuming. Objective: This study aims to fine-tune the pretrained clinical language model, Swedish Deidentified Clinical Bidirectional Encoder Representations from Transformers (SweDeClin-BERT), for medical named entity recognition (NER) and relation extraction (RE) tasks, and to implement an integrated NER-RE approach to more effectively identify ADEs in clinical notes from clinical units in Sweden. The performance of this approach is compared with our previous machine learning method, which used conditional random fields (CRFs) and random forest (RF). Methods: A subset of clinical notes from the Stockholm EPR (Electronic Patient Record) Corpus, dated 2009‐2010, containing suspected ADEs based on International Classification of Diseases, 10th Revision ( ICD-10 ) codes in the A.1 and A.2 categories was randomly sampled. These notes were annotated by a physician with ADE-related entities and relations following the ADE annotation guidelines. We fine-tuned the SweDeClin-BERT model for the NER and RE tasks and implemented an integrated NER-RE pipeline to extract entities and relationships from clinical notes. The models were evaluated using 395 clinical notes from clinical units in Sweden. The NER-RE pipeline was then applied to classify the clinical notes as containing or not containing ADEs. In addition, we conducted an error analysis to better understand the model’s behavior and to identify potential areas for improvement. Results: In total, 62% of notes contained an explicit description of an ADE, indicating that an ADE-related ICD-10 code alone does not ensure detailed event documentation. The fine-tuned SweDeClin-BERT model achieved an F 1 -score of 0.845 for NER and 0.81 for RE task, outperforming the baseline models (CRFs for NER and random forests for RE). In particular, the RE task showed a 53% improvement in macro-average F 1 -score compared to the baseline. The integrated NER-RE pipeline achieved an overall F 1 -score of 0.81. Conclusions: Using a domain-specific language model like SweDeClin-BERT for detecting ADEs in clinical notes demonstrates improved classification performance (0.77 in strict and 0.81 in relaxed mode) compared to conventional machine learning models like CRFs and RF. The proposed fine-tuned ADE model requires further refinement and evaluation on annotated clinical notes from another hospital to evaluate the model’s generalizability. In addition, the annotation guidelines should be revised, as there is an overlap of words between the Finding and Disorder entity categories, which were not consistently distinguished by the annotators. Furthermore, future work should address the handling of compound words and split entities to better capture context in the Swedish language.


  • Open Access

    Detecting Adverse Drug Events in Clinical Notes Using Large Language Models

    by Elizaveta Kopacheva, Alisa Lincke, Olof Björneld, Tora Hammar

    Monitoring adverse drug events (ADEs) is critical for pharmacovigilance and patient safety. However, identifying ADEs remains challenging, as suspected or confirmed side effects are often documented solely in the unstructured text of electronic health records (EHRs). Manually reviewing clinical notes to detect ADEs is labor-intensive and time-consuming, highlighting the need for automated methods capable of analyzing and extracting ADE-related information from clinical documentation. In this short communication, we describe our ongoing research on fine-tuning and evaluating a large language model (LLM) for the detection of ADEs in clinical notes. Preliminary descriptive result of this study indicates that ADEs are poorly documented in discharge notes, with less than 15% explicitly linking ADEs to specific drugs, which highlights the need for improved reporting practices.


  • Open Access

    Fine-tuning Clinical Language Models to Identify Adverse Drug Events in Clinical Text (Preprint)

    by Elizaveta Kopacheva, Alisa Lincke, Aron Henriksson, Hercules Dalianis, Tora Hammar

    BACKGROUND Medications are essential for health care but can cause adverse drug events (ADEs), which are harmful and sometimes fatal. Detecting ADEs is a challenging task because they are often not documented in the structured data of electronic health records (EHRs) or explicitly written in clinical notes. OBJECTIVE This study aims to fine-tune the pre-trained clinical language model, SweDeClin-BERT, for medical named entity recognition (NER) and relation extraction (RE) tasks, and to implement an integrated NER-RE approach to more effectively identify ADEs in clinical notes from clinical units in Sweden. The performance of this approach will be compared to our previous machine learning method, which utilized conditional random fields (CRFs) and Random Forest (RF). METHODS We fine-tuned the SweDeClin-BERT model for the NER and RE tasks and implemented an integrated NER-RE pipeline to extract entities and relationships from clinical notes. The models were evaluated using 400 clinical notes from clinical units in Sweden. The NER-RE pipeline was then applied to classify the clinical notes as containing or not containing ADEs. Additionally, we conducted an error analysis to better understand the model’s behavior and to identify potential areas for improvement. RESULTS The fine-tuned SweDeClin-BERT model slightly outperformed the baseline model (CRFs) in the NER task and substantially outperformed it in the RE task (RF), achieving a 53% increase in macro-average F1-score. The F1-score of the integrated NER-RE pipeline is 0.815. CONCLUSIONS Utilizing a domain-specific language model like SweDeClin-BERT for detecting ADEs in clinical notes demonstrates improved classification performance compared to conventional machine learning models like CRFs and RF. However, the proposed fine-tuned ADE model needs to be further fine-tuned and evaluated on annotated clinical notes from another hospital to evaluate the model’s generalizability. CLINICALTRIAL This research has been approved by the Regional Ethical Review Board (Etikprövningsnämnden), permission number 2012/834-31/5 and permission number 2023-06920-01.

  • Open Access

    Explaining Protest Participation in Semi-authoritarian Regimes: The Power of Social Networks

    by Elizaveta Kopacheva

    This study uses the case of ecological protests in the Russian Federation and the social network of 903 263 VKontakte users to investigate the unique characteristics of protesters’ social networks and assess the effects of network topology on the likelihood of participation. By applying Bayesian structural equation modeling to the data, the study finds that social network topology plays a more nuanced and complex role in predicting protest participation beyond simple exposure to information. Notably, individuals situated as brokers in lower-density and higher-closure networks exhibit a higher likelihood of participation, while central nodes participate less. The results highlight the importance of strategic positioning as brokers for increasing the likelihood of participation. The paper discusses these and other findings, contributing to a deeper understanding of the interplay between social network structures and protest participation.

  • Open Access

    Online mobilisation strategies: Increasing political participation in semi-authoritarian regimes

    by Elizaveta Kopacheva

    Although the importance of weaker and no ties for political-information diffusion was established, previous research did not focus on their primary role in mobilization into low-cost political activities. This study examines online mobilization strategies that increase e-petition signing in semi-authoritarian regimes. The research question is approached by assessing mobilization carried out on Twitter, where communication takes place via weaker ties or no ties, and the Russian social networking service VKontakte, characterized by comparatively stronger ties. To examine how the benefits of participation affect the effectiveness of different mobilization steps, participation on the two biggest e-petitioning platforms in Russia, the Russian public initiative and Change.org, is compared. The significance of the information, persuasion and support effects on the petition signing is tested through linear regression and Bayesian network analyses. The results show that irrespective of the benefits of participation, information-sharing is the most important step for online recruitment into low-cost political activities. Meanwhile, persuasion and social support have no direct impact on e-petition signing.


  • Open Access

    The resource model of political participation 2.0 Protesting in semi-authoritarian regimes – A privilege of the privileged

    by Elizaveta Kopacheva

    Transitioning to and sustaining democracy cannot be possible without active citizens joining protests, distributing political information, or representing fellow citizens in legislative institutions. Due to this, political-science research for several decades has tried to investigate why some citizens are involved in political decision-making while others prefer to refrain from it. Many scholars have suggested that citizens’ political participation is, at large, explained by their interest in politics and political knowledge. However, in the time of shifting towards the digital era, social media has substantially increased the speed and scope of information sharing and overall political knowledge. Additionally, attention seekers populating social networking sites promote mindfulness, consciousness, pro-activeness, and altruism, popularising online activism, boycotting, buycotting, and protesting. Yet, the scale of protest participation in semi-authoritarian regimes, which have a high potential to democratise, remains limited. If political interest or knowledge cannot really explain why this is the case, what can? In this dissertation, I tested hypotheses grounded in political-participation, social-capital, political-mobilisation, and rational-choice research traditions, as well as new hypotheses generated by studying the patterns in original data. In this fashion, I sought to find the underlying factors behind limited protest participation in semi-authoritarian regimes. By studying what is traditionally referred to as unconventional participation (e.g., online activism, petition-signing, and protesting) in democratic and semi-authoritarian regimes and participation in the Russian Federation as a representative case, I have developed an explanatory model of contemporary political participation. In the Russian context, the model proved to be 96% accurate at predicting protest participation. Based on the results of this study and those reported by other scholars, I concluded that socioeconomic status (SES) is at the root of inequalities in political participation. While high-SES individuals acquire advantageous social networks that give them access to political information, low-SES individuals are often excluded from political processes altogether. This dissertation demonstrated that individual social networks—and not time, money, or civic skills—are the most critical resource for contemporary participation.

  • Open Access

    Age and Political Participation in Africa’s Electoral Regimes

    by Elvis Bisong Tambe, Elizaveta Kopacheva

    This article examines the relationship between age and four types of political participation. Previous research has examined the age-participation nexus in the context of established democracies. In contrast, few studies have been devoted to assessing age differences in other contexts, particularly those of Africa, where the meaning of age and age groups could be different than in industrialised democracies. We argued in Africa that age's effect on political participation would vary depending on the country and across different forms of political participation. Analysing Afrobarometer survey data for 34 African countries, we find that the relationship between age and three forms of participation (voting, contacting and collective action) is curvilinear, with younger and older people less likely to participate. While for protest participation, the relationship between age and participation is linear, with protest decreasing with age. Next, we uncover that the countries’ policy formulation, implementation and electoral integrity affect the relationship between age and political participation. In contrast, breaking down the analysis into regional subsamples (West, Central, East, Southern, and North Africa), we observe no patterns of regional differences concerning political participation.


  • Open Access

    Users’ polarisation in dynamic discussion networks: The case of refugee crisis in Sweden

    by Elizaveta Kopacheva, Victoria Yantseva

    This paper presents a study on the dynamics of sentiment polarisation in the active online discussion communities formed around a controversial topic—immigration. Using a collection of tweets in the Swedish language from 2012 to 2019, we track the development of the communities and their sentiment polarisation trajectories over time and in the context of an exogenous shock represented by the European refugee crisis in 2015. To achieve the goal of the study, we apply methods of network and sentiment analysis to map users’ interactions in the network communities and quantify users’ sentiment polarities. The results of the analysis give little evidence for users’ polarisation in the network and its communities, as well as suggest that the crisis had a limited effect on the polarisation dynamics on this social media platform. Yet, we notice a shift towards more negative tonality of users’ sentiments after the crisis and discuss possible explanations for the above-mentioned observations.