There is a lot of talk about the potential of artificial intelligence to transform healthcare delivery. New technologies offer advanced capabilities to ensure the right patients get the right treatment at the right time, while reducing the administrative burden on clinicians.
But what do doctors and clinicians think today about the power of AI to transform medical and health outcomes, and what are the risks associated with the rapid adoption of AI systems?
We interviewed Dr. Carrie Nelson, Chief Medical Officer of Amwell, a telemedicine technology and services company, to discuss how AI is currently transforming healthcare (patients and clinicians) and how AI in healthcare is changing. Hear insight into the risks of accelerated adoption of And the key to taking a values-based approach.
Q. What do you think is the most exciting AI tool coming up for clinicians?
A. Sure, the excitement has good reason.
for example, A recent survey of nurses found that documentation accounted for 15% of every 12-hour nursing shift. Analysis shows that technology-enabled activities, including the use of AI tools, have great potential to reduce that burden by 35%.
By leveraging AI to help clinicians make the most of their skill sets, we can reduce burnout and enable clinicians to focus more on the care we provide. In conversations I have had with doctors, many are very excited about the potential for AI to help improve work-life balance.
We are also seeing a wave of AI innovations that will enable healthcare systems to deliver hybrid care at scale to enhance access and improve outcomes. Seeing more than 200,000 emergency room patients annually, Spectrum Health’s automated chat-based check-in with patients post-discharge provides early awareness of changes in condition so medical teams can intervene quickly .
The program saved $1 million, reduced ED visits by 5%, and achieved a 90% patient satisfaction rate. Nurses, too, find their work rewarding because they are able to respond to the right patient at the right time.
At the St. Luke’s University Health Network, 71% of participants achieved clinically significant improvement using an automated digital behavioral health tool for employees with anxiety and depression.
Additionally, AI-powered referral management, pre-approval requests, and other tasks increase efficiency and reduce the administrative burden on clinicians and staff. Automating these tasks frees clinicians to devote more time to patient care, allowing them to focus on what matters most and reducing burnout.
Q. What are the possibilities for AI to improve the quality of care?
A. As the healthcare community gains experience with AI-powered automated chats, new use cases are being identified and leveraged to improve the quality of care, especially for vulnerable populations.
Consider maternal health. Women in our country today are twice as likely as their mothers to die from complications of pregnancy, especially those on low incomes or living in rural areas. At Northwell Health, a virtual automated companion for a pregnant woman helped identify high-risk patients in her 16% of interactions.
Many of these chats were unexpected based on direct correspondence. This has enabled the organization to escalate cases to clinician attention between visits, providing timely and professional support to these women and their families.
Meanwhile, Northwell experienced 69% of interactions with AI to fill gaps in care across more than 35 specialties. These successes have been demonstrated across a wide variety of populations and conditions, facilitating rapid adoption. The more we learn about how automation can improve quality of care and health outcomes, the more we realize the value of this approach for both common and complex conditions.
There are also certain aspects of care that can be automated within the context of real patient visits for better short- and long-term outcomes. For example, a doctor doesn’t need to know that a woman over the age of 50 who has an average risk of breast cancer needs to order a mammogram.
Things like this can be automated. Even in more complex care scenarios, automation can be used to gather information about a patient’s complex family medical history and other risk factors for illness. AI could potentially synthesize relevant information from patient medical records to help doctors fill gaps and ensure accurate medical documentation.
QWsAre there risks associated with the accelerated adoption of AI tools in healthcare?
A. AI-centric innovation is happening so fast, and going too fast carries absolute risk. For example, there is talk that ChatGPT may be used to enable doctors to respond to messages from patients received via patient portals. But is that the right use of AI in healthcare? The inbox is tough, but we need to stop and assess the risks before we move forward.
We also know that longstanding inequalities and biases in the healthcare system can be incorporated into AI algorithms and potentially magnified. This bias, including what data is collected and documented in medical records and what is not, limits AI’s potential to improve the quality of care today, especially for vulnerable populations. .
Identifying these gaps and working to enhance datasets is essential if AI is to unlock its potential to help healthcare workers improve the quality of care.
More experience with AI-assisted models of care will be needed to discover what is possible, what is not, and how to establish appropriate guardrails. Any errors in medical care are unacceptable. I’m optimistic, but recent data shows we still have a long way to go.
In fact, one study found that 60% of consumers said they would be uncomfortable if their healthcare provider relied on AI for care. A recent Pew Research Trust survey.
Q. What are the keys to taking a value-based approach to AI adoption in healthcare?
A. In the same way that a hospital in Boston is hiring an AI chat engineer to design and develop AI prompts for large language models like ChatGPT, how can AI innovation in healthcare deliver value? , will require a combination of intelligent discovery and human guardrails.
Advances in medical knowledge have far outstripped the ability of clinicians to do it all in the context of patient care. AI can help extract knowledge from vast literature to inform complex diagnoses and create treatment plans. We hope that doctors will reach a stage where they can use AI in efficient ways to improve the accuracy of their diagnoses.
What is diagnostic error? Serious patient safety issue. Clinicians can leverage AI tools that are applied in the context of patient knowledge and patient preferences to best and precisely tailor care to the individual.
To realize our vision of what is possible, we need to take a structured approach to generative AI, one that doesn’t involve as much free text messaging. We’ve seen the funky answers AI can provide when we ask it to have free-flowing conversations with bots and deduce meaning from those encounters.
A better approach at this point would be to ask a specific question where the chatbot would generate a yes or no answer, or prompt the patient to answer on a discrete data point such as blood sugar level or glucose level. , to establish AI-supported guardrails in managing complex states. weight. AI algorithms detect when to prompt clinicians to take action based on input data.
At Amwell, we are applying this approach to many medical conditions and patient populations, transforming the quality of care and health of populations.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.
