Development of an Intelligent AI System for Early Warning of Cardiogenic Shock Mortality in Critical Care: Against the backdrop of rapid advancements in modern medical technology, cardiovascular disease remains one of the leading causes of death among hospitalized patients. Cardiogenic shock, a severe clinical condition, carries a high mortality rate of 25-50%, even with the development of advanced cardiac interventional therapies and mechanical circulatory support. Facing such a daunting medical challenge, our research team is dedicated to developing an innovative artificial intelligence early warning system. The goal is to identify high-risk patients early, thereby securing precious golden hours for medical intervention. By integrating multi-faceted information including clinical diagnoses, physiological indicators, and laboratory data, the research team has constructed an intelligent predictive model aimed at effectively reducing the mortality risk for patients with cardiogenic shock.
The study employs the Long Short-Term Memory (LSTM) model as its core technology, utilizing clinical data from the patient's most recent day to predict the probability of mortality within the next week. To address the common issue of imbalanced data in healthcare, the research team innovatively processed the data into observations every two hours and used the Last Observation Carried Forward (LOCF) method to handle missing values. Crucially, the study introduced the SHapley Additive exPlanations (SHAP) interpreter. This not only enables prediction but also helps clinicians understand the model's decision-making logic, providing a more transparent basis for medical decisions.
The model demonstrated remarkable performance during both training and testing phases, achieving Accuracy scores of 0.98 and 0.97, respectively, and Area Under the Curve (AUC) values of 0.99 and 0.97. Notably, the model identified key variables such as fluid input volume, APACHE II score, and urine output as significantly important for predicting mortality risk in cardiogenic shock patients. This implies that medical teams can, in the future, more accurately assess patient criticality based on these key indicators and adjust treatment strategies promptly. This breakthrough represents not only an advancement in medical technology but also underscores the immense potential of artificial intelligence in the field of precision medicine.
Through this intelligent AI early warning system, medical teams are, for the first time, able to proactively, rather than reactively, address the severe threat of cardiogenic shock using a data-driven approach. The research findings provide not only a technically feasible solution but also bring about a revolutionary shift in thinking for clinical medical practice. In the future, this system holds the potential for widespread adoption in medical institutions globally, offering a beacon of hope for thousands of patients facing the threat of cardiovascular disease and truly realizing the grand vision of artificial intelligence technology serving human health.
Implementing an OR Nurse Scheduling System Based on E-Governance Concept: This research successfully developed an intelligent scheduling system based on the Z3 solver, specifically designed for the multiple constraints and dynamic requirements of operating room (OR) nurse scheduling. The system can rapidly formalize complex scheduling rules into logical constraints and automatically generate optimal scheduling solutions that satisfy multi-level conditions, including rational allocation of professional skills, comprehensive adherence to working hour limits, and effective implementation of fairness principles. Equipped with an intuitive and fully functional web interface, the system allows users to flexibly customize rules, manually adjust schedules, and make real-time modifications in response to emergencies. These features not only significantly enhance scheduling efficiency but also effectively reduce the burden on managers. Furthermore, through a fairer and more flexible scheduling model, it improves nurse job satisfaction and overall team morale.
This project was a collaborative effort involving the Nursing Department, Operating Room, Research and Innovation Center, Personnel Office, and Information Management Office. Data was collected and analyzed over four years (before, during, and after implementation) using creative, patient- and nurse-centric design thinking. We pioneered three innovative national firsts: "Intelligent Monthly Scheduling," "Intelligent Weekly Scheduling," "Intelligent Daily Scheduling," and "Intelligent Patient Safety Scheduling." Through dedicated effort and the implementation of these strategies, we enhanced scheduling and surgical care quality, reducing indirect losses from surgical delays or medical errors caused by personnel mismatches. In terms of costs, over five years, this accumulated to savings of over TWD$3.5 million in personnel and paper expenses for the hospital, along with a significant carbon footprint benefit of 46.816 Kg CO2e/38000.
This project has been invited for technology transfer and knowledge sharing with Pingtung Veterans General Hospital by iMedtac Co., Ltd. It has also been granted a Republic of China Invention Patent (I753522) and findings have been presented at domestic and international nursing conferences and shared with global critical care experts. Guided by our core values—"Cherish Life, Holistic Care, Cutting-Edge Medicine, and Innovative Quality"—this project utilized quality improvement methodologies to refine departmental processes. Through innovative strategies, we enhanced surgical scheduling and care quality, achieving positive clinical outcomes. Upholding the principles of continuous quality improvement, we strive to provide a bastion of high-quality surgical care for the public.
In summary, this intelligent scheduling system successfully overcomes many limitations of traditional scheduling methods in terms of efficiency and accuracy, offering an innovative and highly effective solution for complex and variable scheduling challenges. Future research will focus on two directions: optimizing the user interface to enhance system usability, and exploring the integration and application of more soft constraint parameters to improve the system's adaptability and flexibility across different scenarios.
Pharmaceutical Intelligent Dispensing Quality Management System (PIDQMS): This project is dedicated to the deep application of artificial intelligence technology in pharmacy dispensing management, aiming to elevate the hospital's performance in medication safety and quality control. As the medication environment in medical centers is often influenced by internal decisions and procurement changes, alterations in pharmaceuticals can lead to Look-Alike Sound-Alike (LASA) issues, thereby increasing the risk of dispensing errors. Therefore, utilizing smart technology and data management, we established a comprehensive digital dispensing assistance system covering critical steps such as medication verification, dispensing tracking, and optimization of medication bag labeling. This innovation not only reduces errors made by pharmacists during high-frequency dispensing tasks but also ensures the accuracy of each dispensing decision through AI analysis, providing a safer medication environment for patients. Concurrently, the system's implementation significantly boosts pharmacy operational efficiency, making medication management more precise and controllable. With AI assistance, pharmacists can more effectively track medication flow, reduce errors, and ensure patients receive the correct treatment. Furthermore, the application of this intelligent system enhances the pharmacy's ability to respond to unexpected situations, further strengthening the hospital's pharmaceutical management capabilities.
In the application of the intelligent dispensing system, we successfully developed and implemented several nationally pioneering technologies, including the "Customized Medication Bag Labeling Error Prevention System" and the "Red-Orange-Green Visual Anomaly Identification System." These effectively enhance pharmacists' identification capabilities during dispensing, reducing the error rate for LASA medications. Additionally, we introduced the "Real-time Medication Dispensing Flow Tracking System" and the "Dedicated Dispensing Counter Printing System" to ensure every step from dispensing to issuance is traceable and verifiable. Through these technologies, we drastically reduced the dispensing error rate from a pre-improvement level of 8.29 ppm to 3.3 ppm, not only meeting but exceeding preset standards and demonstrating the immense potential and value of AI in medication management. Simultaneously, the implementation of these technologies enhances internal operational accuracy and improves communication efficiency among the medical team, making patient medication use safer and medical quality more robust. These innovations serve not only our hospital but also act as exemplary cases for hospitals nationwide, driving the overall healthcare system towards a more efficient and intelligent future.
Our Pharmacy Department, upholding the core belief that "Patient Safety is Our Responsibility," actively utilizes AI technology and data analysis to continuously drive innovation and optimization in dispensing processes. We introduced the "Multi-dimensional Drug Information Zoning Design" into the medication bag labeling system, enabling patients to clearly identify their medication information, further enhancing medication adherence and safety. Furthermore, this project not only reduces the workload for pharmacists and increases dispensing efficiency but also promotes the modernization of hospital pharmaceutical management through technological innovation, creating a smarter, more precise, and safer medication environment. This successful case of AI technology application showcases the limitless possibilities of artificial intelligence in the medical field, provides an excellent example for the Smart Hospital evaluation, and lays a solid foundation for the future development of smart healthcare. Our goal is, through continuous technological innovation and process optimization, to enable healthcare institutions to monitor every medication step with greater precision, ensuring patients receive the highest standard of care services and further propelling the healthcare industry into a fully intelligent era.