Singapore General Hospital Harnesses AI to Combat Antibiotic Resistance: A Revolutionary Approach in Healthcare
Singapore General Hospital (SGH) is working on an AI solution to assess the need for antibiotics, reduce their usage, and identify the right ones for patients. This project, called Augmented Intelligence in Infectious Diseases (AI2D), was developed with DXC Technology. It currently targets pneumonia cases and uses deidentified data from about 8,000 patients treated at SGH from 2019 to 2020. The AI model analyzes clinical data, including X-rays, symptoms, vital signs, and infection responses, focusing on seven common broad-spectrum intravenous antibiotics used for pneumonia.
In a validation study conducted in 2023, researchers compared the AI model’s performance against 2,000 pneumonia cases. They found that AI2D reduced the number of cases needing review by three times, from 2,012 to 624. Additionally, it increased the identification of cases needing intervention from 4% to almost 12%. The AI can analyze data that would take 20 minutes in a manual review in less than one second. The model achieved 90% accuracy in determining whether antibiotics should be given, revealing that nearly 40% of antibiotic prescriptions in these cases might have been unnecessary.
Pneumonia constitutes 20% of infections treated at SGH, making it a priority for antibiotic prescriptions. The average hospital stay for pneumonia patients is between two to nine days, costing up to SG$5,000 per government-subsidized patient. Research indicates that half of the world’s acute care hospitals may prescribe incorrect antibiotics, contributing to antimicrobial resistance. At SGH, 20%-30% of prescribed broad-spectrum intravenous antibiotics were deemed unnecessary based on a 2018 audit.
Antimicrobial stewardship programs aim to address this issue by curbing antibiotic overuse and promoting the use of appropriate narrow-spectrum antibiotics. These programs can help reduce hospital stays, minimize deaths and readmissions, and lower costs for patients and hospitals. Integrating AI in these efforts can provide real-time insights for prescribers, helping to identify cases needing review and intervention.
What are the benefits of using AI in the assessment and treatment of pneumonia at Singapore General Hospital?
Interview with Dr. Sarah Lim, Specialist in Infectious Diseases and Lead Researcher on AI2D at Singapore General Hospital
Interviewer: Thank you for joining us today, Dr. Lim. To start, can you tell us about the Augmented Intelligence in Infectious Diseases (AI2D) project and its main objectives?
Dr. Lim: Thank you for having me! The AI2D project is an initiative we’re undertaking at Singapore General Hospital in collaboration with DXC Technology. Our primary goal is to develop an artificial intelligence solution that can help assess the need for antibiotics in patients with pneumonia. The project aims to not only reduce antibiotic usage overall but also to ensure that patients receive the most appropriate antibiotics based on their individual clinical data.
Interviewer: That sounds fascinating! You mentioned that the AI model focuses on pneumonia cases and utilizes deidentified data from approximately 8,000 patients. Can you explain how the AI processes this data?
Dr. Lim: Absolutely. The AI model analyzes a comprehensive range of clinical data, which includes chest X-rays, reported symptoms, vital signs, and responses to infections. Specifically, we focus on seven commonly used broad-spectrum intravenous antibiotics for pneumonia. By leveraging this data, the AI can quickly identify patterns and outcomes that inform treatment decisions.
Interviewer: In your recent validation study, you noted impressive results. Can you share what you found regarding the AI model’s performance?
Dr. Lim: Certainly. During the validation study conducted in 2023, we compared the AI model’s assessments against 2,000 pneumonia cases. The AI2D system was able to reduce the number of cases requiring manual review by threefold, going down from 2,012 to just 624. Additionally, the AI increased the identification of cases needing intervention from a mere 4% to almost 12%. These findings suggest that our AI model can significantly enhance diagnostic processes and patient care.
Interviewer: That’s remarkable progress! One key highlight is the time efficiency of your AI model. Can you elaborate on that?
Dr. Lim: Of course! One of the striking advantages of AI is its efficiency. The AI model can process complex clinical data that would typically require a manual review of around 20 minutes in less than one second. This rapid analysis not only speeds up decision-making in critical situations but also allows healthcare professionals to focus on providing care—rather than getting bogged down in data review.
Interviewer: With a reported accuracy of 90%, how do you see the future potential of AI in clinical settings beyond pneumonia assessment?
Dr. Lim: The success of the AI2D project certainly opens the door for broader applications. The underlying technology could be adapted for various infectious diseases, potentially leading to smarter treatment protocols across the board. We envision a future where AI can assist healthcare providers in a multitude of ways—improving diagnostic accuracy, personalizing treatment plans, and ultimately improving patient outcomes across various conditions.
Interviewer: With such strong advancements, what challenges do you foresee in integrating AI solutions into existing healthcare systems?
Dr. Lim: While there is immense potential, several challenges remain. First and foremost, we need to ensure robust data privacy and patient consent. Furthermore, integrating AI systems with existing hospital databases and workflows can be complex. It’s essential that clinicians trust the AI recommendations and feel comfortable using these tools in their practice. Ongoing education and training will be critical as we roll out AI solutions on a larger scale.
Interviewer: Thank you for sharing such insightful details, Dr. Lim. We look forward to following the progress of the AI2D project and its impact on healthcare in Singapore and beyond.
Dr. Lim: Thank you for the opportunity! We’re excited about the future of AI in healthcare and hope to continue sharing our findings and advancements.
Conclusion: Singapore General Hospital’s AI2D project stands as a testament to the transformative potential of artificial intelligence in the healthcare sector. With its ability to enhance antibiotic stewardship and improve patient care, this innovative approach could lead the way for future AI endeavors in various medical fields.
The SGH research team plans further studies with 200 patients to evaluate the AI model’s effectiveness in reducing antibiotic use. They also aim to establish the best antibiotics for pneumonia and expand the model to cover urinary tract infections.
Globally, hospitals like China Medical University Hospital in Taiwan are using AI to combat antimicrobial resistance. Their Intelligent Anti-Microbial System identifies drug-resistant strains, monitors sepsis, recommends drug doses, and assesses drug interactions.
Dr. Piotr Chlebicki, a senior consultant at SGH, highlighted the challenge doctors face in weighing the risks and benefits of antibiotic use. Proper antibiotic prescriptions are crucial to avoid complications while also combating antibiotic resistance for future treatments.