Early detection of clinical deterioration in patients can be expected to reduce mortality and improve outcomes. However, challenges remain in both hospitals and outpatient settings.
Stanford Healthcare addressed this challenge by incorporating validated models from artificial intelligence and machine learning into its clinical decision support system. We also integrated AI into clinical workflows to improve the patient experience by reducing wait times, improving quality of care, and facilitating important conversations.
Dr. Shreya Shah is a practicing academic internist at Stanford Health Care, board certified in clinical informatics, and an expert in healthcare integration of artificial intelligence.
She will speak about the health system’s AI efforts. The 2023 HIMSS AI in Healthcare Forum, scheduled for Dec. 14-15 in San Diego, will feature a case titled “How Stanford Health Transformed Patient Care by Combining Compassion with AI-Driven Innovation.” Study provided.
We spoke with Shah to preview his session and learn more about how Stanford Healthcare is using AI and ML.
Q. Why is it still a challenge to detect clinical deterioration in patients?
A. The illnesses of patients in the hospital are becoming increasingly complex and severe, and non-urgent care is moving to home, outpatient treatment, or subacute levels of management. Within academic medical centers, this problem is even more acute for patients who are at high risk for clinical deterioration.
Early signs can be subtle and vary widely from patient to patient. Identifying which patients require the most attention is like sticking a needle in a haystack. Additionally, these patients are cared for by multiple care teams, requiring the evaluation of large amounts of data that change over time.
Teams can experience communication gaps, information overload, and cognitive biases, leading to unexpected clinical deterioration and serious consequences such as emergency resuscitation efforts and unplanned transfers to ICU care. There is a possibility. There may also be varying degrees of agreement among team members regarding risk perception.
Standardized workflows for care coordination that empower all care team members to make patient care decisions can help overcome these challenges.
Q. How did you determine that AI and ML were best suited to solve this challenge?
A. We needed to identify patients at increased risk and coordinate care teams around a collaborative, standardized clinical response. We determined that ML models can be used to identify patients with a high likelihood of future clinical worsening events without any additional work by practicing clinicians.
Predictions must be made early enough to allow the medical team sufficient time to respond. Accuracy is always a concern, and clinicians often believe that AI systems can’t tell them anything they don’t already know.
In our implementation, the focus was not on whether the model’s predictions were correct. Rather, for a given patient flagged by the model, physicians and non-physician care team members were required to perform a structured, collaborative workflow to assess risk and response. Therefore, probabilistic models create team-based triggers.
Our implementation efforts focused on the following priority areas: 1) designing systems that integrate ML models into complex healthcare systems; 2) building effective teams and processes to enable the collaborative workflows necessary for successful implementation; and 3) deploying these AI- healthcare Integrate systems in a way that is sustainable and scalable for your enterprise.
The focus was on building a comprehensive system that not only incorporates advanced technology but also aligns with clinical, operational, and strategic needs.
Q. What is one example of how Stanford University has contributed to the clinical deterioration challenge by incorporating validated AI and ML models into clinical decision support systems?
A. Our clinical deterioration model was validated based on our data to ensure model performance. The signals were then integrated into the EHR with full transparency including factors and enhanced with mobile alerts to the medical team.
The ML model was able to update predictions about hospitalized patients every 15 minutes, serving as an objective assessment of risk and helping to facilitate coordination and coordination in patient care. AI integrated system.
The model was site-specific validated to confirm its effectiveness in predicting clinical deterioration events, such as unplanned ICU transfer within 6 to 18 hours. This workflow significantly increased multidisciplinary standardized patient assessment, resulting in a 20% reduction in clinical deterioration.
Qualitative evaluation indicates that this model is useful for adjusting mental models and driving workflows needed for patients flagged by the model with consensus among interdisciplinary team members. Confirmed. By using reliable and continuously updated risk signals, we aligned physicians and other medical teams and instituted consistent workflows.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email: bsiwicki@himss.org
Healthcare IT News is a publication of HIMSS Media.