Apr 19, 2024

Responsible AI in healthcare: Best practices from industry experts

The healthcare artificial intelligence space continues to experience significant growth.

AI applications have demonstrated tremendous potential to transform care delivery in many areas, from initiating early interventions for strokes and other acute medical conditions to predicting patient deterioration, promoting better medication adherence, supporting hospital-at-home programs and much more.

Hospitals and health systems are recognizing, however, that a commitment to “responsible AI” is essential for ensuring patient well-being.

To learn more about AI’s potential in healthcare and the need for responsible use of this technology, Becker’s Healthcare recently spoke with two healthcare AI experts who shared how to successfully select and deploy AI solutions:

  • Snehal Gandhi, MD, Chief Medical Information Officer and Medical Director of Hospital Medicine at Camden, N.J.-based Cooper University Health Care
  • A corporate director of business continuity and informatics at a large health system in the Southeast

Excitement about AI in healthcare must be tempered with due diligence

While buzz about healthcare AI applications began well before COVID-19, the pandemic increased interest in how these tools can change care delivery models. Many high-performing healthcare institutions are now either using AI or are actively looking to leverage the technology.

At the same time, healthcare organizations and other key stakeholders have expressed significant concerns about issues like bias, information security, lack of regulation and the lack of transparency of AI models. Finding “responsible AI” solutions is a top priority.

“I think these concerns are real and valid,” said the Southeastern health system’s corporate director of business continuity and informatics. “Anytime we introduce artificial intelligence into personalized healthcare, people will have concerns. What we’ve seen recently is regulation of the data sources and technology behind AI, which is great. The Office of the National Coordinator for Health Information Technology has an effort underway to regulate security within the healthcare AI space.”

As hospitals and health systems assess AI-based solutions, teams must determine whether these tools are built on a foundation of responsible AI. With so many AI vendors promoting their solutions, selecting the right one for your workplace is a strategic decision, and one that must not be taken lightly.

“Evaluating AI vendors is an ever-evolving process,” Dr. Gandhi said. “We firmly believe that the partners we work with must have a track record of accuracy. Other issues are bias and fairness. When it comes to AI systems, biases can be inadvertently perpetuated through training data that’s presented to the model. One of the things we always ask about is whether the solution was tested across a diverse demographic group, incorporating multiple scenarios.”

Cooper University Health Care also looks for AI vendors with an ethical framework in place, which includes a commitment to principles like maintaining confidentiality of patient data, ensuring informed consent, and providing transparency into the data and the model.

Data security is another essential element of responsible AI solutions. “When we select an AI vendor, we focus on three pillars: cybersecurity, evidence-based practice and scalability,” said the corporate director of business continuity and informatics at the large health system in the Southeast. “The ability to secure and protect data is a mandatory requirement.”

The right AI tools enable efficient patient care and deliver better outcomes

Cooper University Health Care has partnered with to transform patient care. With VizTM Neuro, the organization analyzes imaging and uses the resulting intelligence to improve patient outcomes.

“As soon as images come off our CT scanners, Viz Neuro can detect if a patient has a large vessel obstruction like a stroke and whether an intervention can be done,” Dr. Gandhi said. “The system alerts our interventional neurologists, so they can take action. With strokes and other conditions, time is of the essence. Rapid intervention can help patients regain full function or reduce the amount of deficit that occurs. That’s tremendous.”

Prior to the pandemic, the large health system in the Southwest also partnered with and implemented the radiology suite of products. “One of our goals for implementing AI in the radiology and imaging space was to decrease the amount of time needed to diagnose and intervene when patients come in with acute stroke,” said the health system’s corporate director of business continuity and informatics. “AI has played a key role because the algorithm is set up to work with images from our PACS system and enable faster decision-making among the stroke team.”

Members of the stroke team at the health system use the technology to ensure they are all on the same page regarding patient care. Smart, secure checks and communications on mobile devices alert the entire team about patient progress, so they know where they need to be and when.

“We’re very pleased with the results and hope to decrease the door-to-intervention time even more,” said the health system’s corporate director of business continuity and informatics.

Based on the positive results that Cooper University Health Care and the Southeastern health system have seen from Viz Neuro, both organizations have decided to adopt other modules. “We’ve implemented’s modules for pulmonary embolisms, aortic conditions and more,” Dr. Gandhi said. “The data collection is ongoing, but we are seeing positive outcomes. That in itself has been a game changer.”

Given the success of AI technology in the acute inpatient setting, the large health system in the Southeast and Cooper University Health Care would like to shift the application of AI to the outpatient arena. “By analyzing EKGs and echocardiograms, can we predict cardiomyopathy? We’d like to move into a space where primary care physicians are less burdened with referrals to cardiology and can leverage AI to save time and possibly lives by identifying conditions with a high degree of predictability,” said the Southeastern health system’s corporate director of business continuity and informatics.

Dr. Gandhi noted how the interoperability of EHRs has markedly improved access to comprehensive medical records, which is especially important for patients who receive care across multiple healthcare facilities. Additionally, more patients are using wearable health devices that capture valuable health data and insights. “We are looking at leveraging AI technology to efficiently summarize both internal and external clinical data into a concise, easily accessible summary for the clinicians,” Dr. Gandhi said. “By doing so, we decrease the amount of time clinicians spend navigating through the chart, allowing them to devote more attention to direct patient interaction. This allows for a much-needed move toward more patient-centric healthcare.”

When implementing AI, a focus on outcomes and stakeholder engagement is essential

Based on their experiences deploying healthcare AI tools, Dr. Gandhi and the Southeastern health system’s corporate director of business continuity and informatics offered several best practices. A top priority is understanding how new technologies will support people and processes.

“AI technology often looks really cool,” Dr. Gandhi said. “But we go back to the fact that it’s always people, process, technology and culture change. You don’t bring in AI technology just for the sake of implementing cool technology.”

The team at Cooper University Health Care considers how AI could help providers, operational staff or patients. From a process perspective, the organization evaluates whether technology could reduce defects or improve efficiencies. “Once we have the current state mapped out, we define the KPIs and metrics related to the goals and outcomes we desire. Only then do we bring in the technology and determine how it can help,” Dr. Gandhi said.

Another key to success when implementing healthcare AI solutions is early engagement with clinicians and financial decision-makers to generate buy-in. “When we consider changes to care delivery, physicians want to see evidence-based research on the AI algorithms,” said the corporate director of business continuity and informatics at the large health system in the Southeast. “Peer-reviewed studies showing that AI made a difference in terms of improved efficiency helped us with these internal conversations.”

Dr. Gandhi added that it’s important to work synergistically with clinicians. “It’s not about replacing physicians and advanced practice providers but creating a synergistic relationship, so they have the tools needed to work effectively when providing patient care,” he said.

In conversations with the leadership team, the Southeastern health system also focuses on the economies of scale associated with healthcare AI solutions. “With post-pandemic labor costs, many hospitals are in the red,” said the health system’s corporate director of business continuity and informatics. “Analyzing where to spend hard dollars is a tough question for any healthcare leader. If we have the evidence that an AI tool will scale and protect data, the conversations go more smoothly.”

Future state: AI will expand its reach into many aspects of health and well-being

In regard to healthcare quality, the corporate director of business continuity and informatics at the large health system in the Southeast said there are opportunities for AI to help improve patient safety.

“All organizations have publicly reported CMS ratings, Leapfrog Survey results and more,” he said. “Everyone wants to reap the benefits of AI algorithms and use predictability to prevent things like hospital-acquired infections or sepsis.”

The ability of AI tools to analyze huge volumes of data is also opening new possibilities for different approaches to patient care. Many believe AI tools will play a significant role in hospital-at-home initiatives by supporting remote patient monitoring, data aggregation and algorithm tracking. Other promising applications of AI include personalized medicine, genomics and accelerated drug discovery.

“An interesting area is how to use AI to analyze vast amounts of data across different sectors like healthcare, environmental health and social agencies so we can take a more comprehensive approach to health and well-being,” Dr. Gandhi said.

For organizations to realize the full value of AI in healthcare, leaders at the board and executive levels must be educated about the potential of this technology. Many myths and misconceptions about AI exist. To combat them, education and open dialogue are necessary.

“Mainstream medicine has joined in on AI,” said the Southeastern health system’s corporate director of business continuity and informatics. “Now, population health providers are interested, and infection prevention teams are jumping with joy, saying ‘show me the data.’ It’s an exciting time in healthcare and in AI. To move forward, we need a combination of vendors, education and organizations working together.” One™ is an AI solution that extends across your entire enterprise, integrating multiple clinical disciplines to expedite diagnosis and treatment, improving patient outcomes. To learn more, watch our recent webinar.