Singapore General Hospital (SGH), in partnership with DXC Technologies, has developed an artificial intelligence solution called "Augmented Intelligent Infectious Diseases" (AI2D), which aims to optimize antibiotic prescriptions, reduce antibiotic abuse, and select the most appropriate antibiotics for each patient. . The AI2D model was trained using de-identified data from 8,000 patients, covering multiple infection types such as pneumonia and seven commonly used antibiotics. Downcodes editors will give you an in-depth understanding of this innovation aimed at combating the global antibiotic resistance crisis.
The construction of the AI2D model is based on de-identified clinical data of approximately 8,000 SGH patients from 2019 to 2020, including X-rays, clinical symptoms, vital signs and infection response trends, covering seven commonly used broad-spectrum intravenous antibiotics. The research team conducted a preliminary validation study of the AI model in 2023, comparing it with 2,000 pneumonia cases.
In the study, SGH and DXC noted that AI2D was able to reduce the number of cases requiring review by a third (from 2012 to 624). The AI model also increased the likelihood of identifying cases requiring intervention by nearly 12% of cases reviewed, compared with only 4% for traditional manual review. In addition, the analysis time for a certain case was shortened from 20 minutes for manual review to "less than one second."
Research shows that the AI model is 90% accurate in determining whether antibiotics are needed in pneumonia cases. The study also revealed that in nearly 40% of these cases, antibiotic prescriptions may be unnecessary.
SGH said pneumonia accounts for 20% of all infections in its hospitals and is the type of infection for which antibiotics are most frequently prescribed. The average length of stay for patients ranges from two to nine days, and the government costs up to S$5,000 (approximately US$3,500) per subsidized hospital stay. According to a 2018 antibiotic use audit, SGH Hospital found that 20% to 30% of broad-spectrum intravenous antibiotics were redundant, while about 30% of hospital-acquired infections in Singapore are thought to be resistant to broad-spectrum antibiotics .
In response to this global problem, hospitals are establishing antimicrobial stewardship programs to prevent the overuse of antibiotics and identify when more appropriate narrow-spectrum antibiotics are recommended. The use of automation and artificial intelligence can better provide real-time insights at the time of prescribing, helping to identify and prioritize cases that require review.
The research team is currently conducting a comparative study on 200 SGH hospitalized patients to test the effectiveness of the AI model in reducing antibiotic use, and will develop similar models for urinary tract infections in the future.
The success of the AI2D project provides a new idea for solving the global problem of antibiotic resistance. Using artificial intelligence technology to accurately determine the necessity of antibiotic medication can not only effectively control antibiotic abuse, but also reduce medical costs and improve medical efficiency. The editor of Downcodes hopes that AI2D can be applied to more disease fields in the future and benefit more patients.