Stanford Health uses AI to reduce clinical deterioration events
Early detection of clinical deterioration in patients holds the promise of reducing mortality and improving outcomes. However, it remains a challenge, both in hospitals and outpatient settings.
Stanford Health Care has addressed this challenge by integrating validated artificial intelligence and machine learning models into clinical decision support systems. They also integrated AI into clinical workflows and improved the patient experience, including reducing wait times, improving quality of care and facilitating critical conversations.
Dr. Shreya Shah is a practicing academic internist, board-certified physician in clinical informatics, and expert in artificial intelligence healthcare integration at Stanford Health Care.
She will speak about the healthcare system’s AI efforts at the 2023 HIMSS AI in Healthcare Forumscheduled for December 14-15 in San Diego, with a case study titled: “How Stanford Health Transformed Patient Care by Combining Compassion with AI-Driven Innovations.”
We spoke with Shah to get a preview of her session and to gain more insight into how Stanford Health Care is using AI and ML.
Q. Why does detecting clinical deterioration in patients remain a challenge?
A. In patients in hospitals, the disease is becoming increasingly complex and serious, while less acute care is moving to home care, outpatient care or the subacute management level. Within an academic medical center, this is even more severe in patients at high risk for clinical decline.
The first signs can be subtle and vary greatly between patients. Identifying which patients need the most attention is like a needle in a haystack. Additionally, these patients are cared for by multi-person care teams and require assessments of large amounts of data that change over time.
Teams may experience communication gaps, information overload, and cognitive biases that can lead to unexpected clinical deterioration with major consequences, such as emergency resuscitation and unplanned transfers to ICU care. There may also be varying degrees of agreement among team members regarding the perception of risk.
Standardized care coordination workflows that empower all members of the care team to make patient care decisions can help overcome these challenges.
Q. How did you decide that AI and ML were the right choice to tackle this challenge?
A. We needed to identify high-risk patients and align the care team around a shared, standardized clinical response. We have found that an ML model can identify patients with a high probability of future clinical decline without additional tasks for our working physicians.
The predictions should be made early enough to give the healthcare team sufficient time to respond. Accuracy is always a concern, and doctors often think the AI system won’t tell them something they don’t already know.
In our implementations, the focus was not on whether the model predictions were correct. Rather, for each patient flagged by the model, healthcare team members of physicians and non-physicians were required to conduct a structured collaborative workflow to assess risk and response. So a probabilistic model creates a team-based trigger.
Our implementation efforts focused on these priority areas: 1) Designing a system that would integrate the ML model into a complex healthcare system, 2) Building effective teams and processes to enable the collaborative workflows necessary for successful implementation, and 3) Deploying these AI-integrated systems in a way that is both sustainable and scalable for the healthcare enterprise.
The focus was on creating a holistic system that not only encompasses cutting-edge technology, but also meets clinical, operational and strategic needs.
Question: What is an example of how integrating validated models of AI and ML into clinical decision support systems has helped Stanford with the challenge of clinical decline?
A. Our clinical deterioration model was validated against our data to ensure model performance; the signals were then integrated into our EHR with full transparency, including contributing factors, and supplemented with a mobile alert for the care team.
The ML model can update predictions about hospitalized patients every 15 minutes and was used to act as an objective risk assessor and helped facilitate alignment and coordination in patient care. AI integrated system.
The model underwent site-specific validation to ensure its effectiveness in predicting clinical deterioration such as unplanned ICU transfers within a 6- to 18-hour window. This workflow led to a significant increase in multidisciplinary standardized patient assessments and a resulting 20% reduction in clinical deterioration events.
Qualitative evaluation results showed that the model was useful in aligning mental models and driving necessary workflows for patients highlighted by the model with consensus among multidisciplinary team members. By using a reliable and continuously updated risk signal, we aligned the physicians with the rest of the care team to create a consistent workflow.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.