Tech & Innovation in Healthcare

Sing the praises of a multicenter program.

Artificial intelligence (AI) has shown promise in many areas of healthcare, including revenue cycle management, coding, and aiding providers to diagnose a patient’s condition. One area where the technology could put its power to the test is in the intensive care unit (ICU).

Learn how providers can augment their critical care decision making with AI.

Analyze Data to Identify Common Issues

“ICU clinicians continue to have difficulties anticipating deterioration, handling high heterogeneity, and providing early interventions before patients decompensate,” says Ramzy Rimawi, MD, senior physician, critical care medicine, Emory University.

Developers are building AI systems that have the computing power to handle complex patient cases. As AI and machine learning (ML) continue to advance and grow, providers can improve their decision-making abilities to impact their patients. “Predictive AI models use patient care data to aid clinical decision-making by informing ICU resource allocation and recognizing morbidity and mortality outcomes earlier,” Dr. Rimawi says.

AI large learning models (LLMs) — like ChatGPT — deliver the technological power to quickly and easily scour electronic health record (EHR) data to supply providers with relevant data that’s concise and necessary to making a clinical decision. In turn, the AI programs’ multidimensional and multidomain data patterns could act as welcome alternatives to traditional monitoring systems and risk assessment tools in the future.

As a result, “ICU clinicians may soon be able to quickly tackle common ICU issues, such as disorders of volume status and acid-base, sepsis, acute respiratory distress syndrome, cardiorespiratory instability, and surgical complications,” Dr. Rimawi adds.

One hurdle that continues to remain regarding AI in healthcare is accessing a diverse dataset that accurately represents the patient population. Luckily, the National Institutes of Health (NIH) aims to reduce or eliminate that hurdle entirely.

Build the Data Repository

A multicenter program, named A Patient-Focused Collaborative Hospital Repository Uniting Standards for Equitable AI (CHoRUS), will generate and expand biomedical data for use in the monitoring, diagnosis, and treatment of critically ill patients. Plus, physicians can use the data to augment their decision making.

CHoRUS is a multicenter network of university health systems and is funded by the National Institutes of Health’s Bridge to Artificial Intelligence (Bridge2AI) program. Other health systems involved in the network include:

• Massachusetts General Hospital at Harvard
• Emory University
• Duke University
• University of Florida
• University of California, Los Angeles (UCLA)
• Nationwide Children’s Hospital
• Columbia University
• Mayo Clinic

“We define these sites primarily for their access to EHR data, waveform data, and diversity, We made sure that we spread as much as we could across the United States to really sample our patient care diversity in addition to our patient diversity,” said Eric Rosenthal, MD, medical director of the Neurosciences Intensive Care Unit at Massachusetts General Hospital in Boston, Massachusetts, in an October 2022 presentation at the Neurocritical Care Society annual meeting.

CHoRUS has six modules, which it’s using to develop critical care AI systems as well as the next generation of care teams. The modules include:

• Teaming: Build a national AI infrastructure in acute and critical care.
• Ethical and trustworthy AI: Establish ethics of equity, privacy, and consent with community input from vulnerable patients and policy leaders.
• Standards development: Broaden standards for AI use in critical care for multimodal, high-resolution EHR, imaging, waveform, and social determinants of health (SDOH).
• Tool development and optimization: Create tools for storage, visualization, labeling, interpreting, and blocking re-identification of AI-ready data.
• Data acquisition: Gather and standardize dataset of more than 100,000 patients.
• Skills and workforce development: Use curriculum development, community engagement, and workshops to train the next generation of critical care AI scientists.

One large issue standing in the way of robust AI development is data silos. The network is geared toward building an extensive data repository for AI research, and the data will come from more than 100,000 anonymous, critically ill patients. “We hold onto our data in silos. We cite institutional ‘IP loss’ and ‘patient privacy risks.’ But, what about the risk of not collaborating? We treat data like it is our personal property, but many create it and many steward it,” Dr. Rosenthal added.

But simply sharing the data isn’t enough. The information must be standardized to ensure clinicians and providers at healthcare centers with different systems can use the data. “We want a common data model, so we can do research that’s scalable, transparent, and reproducible. Also, if we all use the same standards, we can crowdsource and borrow and learn from each other,” Dr. Rosenthal said.

AI is Still a Work in Progress

There is great potential for AI adoption in the ICU to augment the clinician in their everyday work, and consortiums continue to form to ensure the safe sharing of de-identified and diverse patient data, so healthcare providers can build robust and knowledgeable systems. However, we’re still in the infancy of the technology.

“Streamlining the medical research pipeline using AI will largely benefit ICUs that seek to pinpoint inefficiencies, errors, and potential areas for improvement. There is still work to be done to overcome obstacles to AI before it will become a core component of an ICU clinician’s workflow,” Dr. Rimawi states.

Stay tuned to Tech & Innovation in Healthcare for more information as AI continues to evolve in healthcare.