Scholarlink

Healthcare systems under increasing pressure require innovative strategies to optimize care delivery and reduce waiting times. This study presents a dual-method approach combining process mining and text mining to enhance insight into clinical workflows and support value-based healthcare. Process mining techniques are used to map diagnostic and treatment pathways, identifying inefficiencies, bottlenecks, and process variations across patient trajectories. By analyzing event logs extracted from electronic health records, patterns and deviations are uncovered, enabling targeted improvements in care coordination and resource planning. Text mining, particularly with domain-specific natural language processing models, is leveraged to extract unstructured information from clinical narratives. A deep learning classification model trained on medical texts was refined using techniques such as data filtering, class balancing, early stopping, and decision threshold optimization to identify specific complications from reports with high accuracy. Integrating structured event data with insights drawn from clinical notes creates a richer understanding of why certain process deviations occur, supporting the development of predictive models and personalized care strategies. Together, these methodologies illustrate the power of data-driven analysis in healthcare, providing the foundation for operational improvements and better patient outcomes. The study also demonstrates how combining structured and unstructured data enables a more complete representation of clinical workflows and informs future deployment of machine learning models in clinical environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimization of Care Pathways with the Use of Process Mining and Text Mining Techniques: Combining PM2 and CRISP-ML Methodologies

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Cheng Qi Wang,
  • Vera Bulsink,
  • Reini Bretveld,
  • Rogier Krom,
  • Shiva Faeghi,
  • Faiza Bukhsh

摘要

Healthcare systems under increasing pressure require innovative strategies to optimize care delivery and reduce waiting times. This study presents a dual-method approach combining process mining and text mining to enhance insight into clinical workflows and support value-based healthcare. Process mining techniques are used to map diagnostic and treatment pathways, identifying inefficiencies, bottlenecks, and process variations across patient trajectories. By analyzing event logs extracted from electronic health records, patterns and deviations are uncovered, enabling targeted improvements in care coordination and resource planning. Text mining, particularly with domain-specific natural language processing models, is leveraged to extract unstructured information from clinical narratives. A deep learning classification model trained on medical texts was refined using techniques such as data filtering, class balancing, early stopping, and decision threshold optimization to identify specific complications from reports with high accuracy. Integrating structured event data with insights drawn from clinical notes creates a richer understanding of why certain process deviations occur, supporting the development of predictive models and personalized care strategies. Together, these methodologies illustrate the power of data-driven analysis in healthcare, providing the foundation for operational improvements and better patient outcomes. The study also demonstrates how combining structured and unstructured data enables a more complete representation of clinical workflows and informs future deployment of machine learning models in clinical environments.