In the relentless battle against sepsis, a deadly blood infection claiming hundreds of lives annually in Colorado, physicians grapple with a challenging dilemma. Waiting until symptoms manifest presents a precarious 50-50 chance of saving the patient, complicated by the fact that early signs of sepsis mimic various other medical conditions. Dr. CT Lin, Chief Medical Information Officer at UCHealth, acknowledged the difficulty of spotting sepsis during a panel discussion, describing it as a "terrible disease."
The gravity of the situation is magnified by sepsis silently emerging as a major killer in the United States. According to estimates from the CDC, 350,000 adults each year who develop sepsis either succumb in the hospital or are discharged to hospice care. Strikingly, one in three individuals who die in hospitals had developed sepsis during their stay.
In response to this critical issue, Dr. Lin and his colleagues at UCHealth embarked on a groundbreaking initiative: leveraging artificial intelligence (AI) to predict and identify sepsis cases early on. The idea was to utilise the wealth of data accumulated over the years and employ machine-learning models to generate predictions. However, the journey into AI-driven sepsis detection was far from a seamless one.
UCHealth faced initial disappointment in a three-month pilot program, prompting a recalibration of the approach. Seeking insights from frontline healthcare providers, the team realised the challenges of shifting focus from immediate patient care to anticipating issues in advance.
This insight played a pivotal role in reshaping the approach to deploying AI tools, emphasising the need for continuous refinement and careful implementation in the integration of AI and healthcare practices.
Click here to read the original news story.