By 2030, the healthcare AI market is expected to be worth almost that much $188 billion.
The Institute of Electrical and Electronics Engineers, the world’s largest nonprofit technical organization dedicated to advancing technology for humanity, is keeping a close eye on artificial intelligence – both the benefits and challenges of the technology that is exploding in health care.
That’s why I recently met with IEEE member Chenyang Lu. We asked him how healthcare professionals are using AI to improve patient outcomes, the challenges of implementing AI in healthcare and how to overcome these challenges, and what he thinks the future of AI in healthcare looks like. He gave some very insightful answers.
Q. What is your view on how healthcare professionals can utilize AI to improve patient outcomes?
AND. Artificial intelligence will effectively become the co-pilot of our doctors and enable timely, precise and effective treatment of each individual patient. AI models can create personalized predictions about a patient’s clinical outcomes, risk factors, and response to various treatments. Here are three examples of artificial intelligence in healthcare that have the potential to improve patient outcomes.
First, depression screening. According to WHO, more than 280 million people suffer from depression. Among them, over 50% are not diagnosed or treated. The problem of underdiagnosis results from the significant time and costs incurred by psychiatrists for diagnosis. A recent study showed this Deep learning models can detect depression and anxiety disorders based on data collected via wearable devicesopening a fresh way to discreetly detect depression.
This AI-powered screening tool will enable physicians to deliver individual selective prevention programs in a targeted and timely manner, including: a critical evidence gap for depression prevention identified by the United States Preventive Services Task Force.
Second, cancer care. Cancer patients are at high risk of clinical deterioration: 6.4% of cancer patients are transferred to at least one intensive care unit, and 2.7% of them die in hospital wards, according to a study recent research. Machine learning models can generate early warnings of clinical deterioration in cancer patients by integrating heterogeneous data in electronic health records.
AI-generated early warnings, along with risk factors linked to predictions, allow clinicians to identify at-risk patients in advance and provide early interventions to prevent deterioration. Clinicians also face challenges when deciding whether to discharge patients from oncology units. A longer stay reduces hospital availability for cancer patients. Machine learning models can be used to determine when a patient hospitalized with cancer is clinically stable for hospital discharge, thereby improving the efficiency of cancer care while ensuring patient safety.
And third, perioperative care. Surgery carries significant risks and costs for patients. Early identification of risk factors may be crucial for early intervention and improved treatment outcomes. For example, pancreatic resection is the only treatment for pancreatic cancer, but it is often associated with a high rate of sedate complications. Using data collected from certain fitness bands, Machine learning models can predict a patient’s risk of serious complications before surgery.
If a patient is high risk, he or she may be enrolled in prehabilitation programs to augment his or her readiness for surgery. Machine learning models were also developed using EHR data identify risks during surgery and to predict complications after surgeryto improve the safety and outcomes of patients receiving perioperative care.
Q. What do you see as the main challenges in implementing AI in healthcare, and how can hospitals and healthcare systems overcome these challenges?
AND. Integrating AI models into EHRs and clinical workflows is indispensable to implementing AI in healthcare. However, there are significant challenges in implementing AI models on current EHR platforms, as opposed to commercial cloud platforms that have made it much easier to build and deploy AI.
We currently have many AI models in pilot stages, but few of them have been implemented in EHRs. We are still in the early stages of applying artificial intelligence in healthcare. Looking ahead, it is imperative that we reduce the barriers to deploying AI models in our infrastructure.
Additionally, we need to change our workflows and protocols so that clinicians and AI can collaborate effectively. The experience of recent years has shown that artificial intelligence and doctors provide complementary capabilities. Artificial intelligence will be the co-pilot, working with clinicians to make the best decisions and treatments together. Significant research is needed to develop effective human-in-the-loop AI in clinical settings.
Q. What do you see as the future of artificial intelligence in healthcare? What’s next and where is it heading in the coming years?
AND. We are seeing the early adoption of generative AI to improve operational efficiency by automating clinical documentation and patient communication. Despite implementation challenges, we will see increasing adoption of AI-based clinical decision support, driven by enormous potential to improve patient outcomes and healthcare efficiency.
Importantly, we need to build evidence on the effectiveness and benefits of AI in healthcare in terms of patient outcomes and cost-effectiveness, so that we can progressively expand AI capabilities in our healthcare systems. In the meantime, we must ensure fairness, safety, security, privacy and access to AI in healthcare through both policy and technology.
This is another area where significant research is needed to enable the sustainable development of artificial intelligence in healthcare.