Eric Topol sees a bright future for tech in healthcare – after some false starts

Despite vast growth in medical research and knowledge, clinicians struggle to apply it to care – AI and other tech holds much promise in improving care.


Eric Topol, MD

While medical knowledge has advanced rapidly in recent years, there’s a lot of room for improvement in applying it to healthcare delivery and medical treatment. 

That’s where a variety of technologies can be brought to assist clinicians in making care decisions, but the industry is still in the early phases of doing it effectively, says Eric Topol, MD, who sees a bright future in melding digital technology with genomics to improve the state of the art in healthcare. 

Topol is the chair of innovative medicine at Scripps Research and a director and founder of the Scripps Research Translational Institute, in addition to other roles there. He is a recognized futurist who frequently is called on to detail where technology is headed to improve care, also is a go-to source for making medical trends understandable to the masses. For example, he took on the role of disseminating credible medical information daily during the months of the COVID-19 pandemic. 

But his passion is to improve medical care, and he sees an urgent need to funnel all of a patient’s medical information into the clinician’s hands to make a fully informed care decision. 

Strong science, weak implementation 

“I’ve had a lifelong interest in genomics; in college in 1975, I wrote a thesis on the prospects for gene therapy, so that was about four decades ahead of when it actually started to click,” he says. 

Despite the advances of medical research and widespread use of technology in delivering care and keeping records, healthcare continues to have a poor track record in applying knowledge to care and preventive treatment. 

“The science is really strong, but the implementation is really weak,” Topol contends. “You know, we have these polygenic risk scores that can help people know if their risk is high for things like coronary artery disease, breast cancer, colon cancer, prostate cancer, atrial fibrillation, a long list of things – and we don’t use them. We have 200 drugs where we have pharmacogenomic interactions that we can predict if a person should not get the drug because of the risk of side effects or if the dose should be adjusted. We’re not using that.” 

Voluminous medical research also is not impacting clinical decision making, he says. “We have a number of high-quality peer-reviewed papers, but have we changed medicine? Not substantially. I don’t know what it’s going to take to implement a lot of these things we’re talking about today, but we’re so far behind where we could be.” 

Electronic health records hold potential to do more than they do today, he says, but they need to be re-imagined to hold more data of consequence to making treatment decisions. “Electronic health records … don’t have genomes, they don’t have sensor data, they don’t have a lot of environmental data – they don’t have a lot of things that make us unique,” he says. “But what we’re getting now is a kind of ability for a person to track a chronic condition, like diabetes or high blood pressure, through their sensors and algorithms. So we’re just starting to see that beginning now.” 

Patients given more control over their medical information and tools to self-assess broadens the ability of the system to provide pre-emptive care, he believes. Patients are gaining “the ability to screen for a diagnosis, whether it be atrial fibrillation through a smartwatch or a urinary track infection or ear infection or skin lesions without having to go to a doctor,” he notes. “But basically, we’re stuck in this lack of democratization of a healthcare system while its professionals are overwhelmed.” 

Improving the aim of care 

Topol eschews the term “precision medicine,” saying the profession needs to strive for accuracy in care. “What we want is accuracy; we don’t have that now,” he contends. Artificial intelligence holds potential, particularly in imaging analysis because it enables “far better accuracy when you combine trained deep learning algorithms with an expert clinician. 

“I do think this is terrific, because we do have serious accuracy issues; we have at least 20 million major diagnostic errors in the United States each year,” he adds. “So we can clean that up; we’re not going to get to zero with AI, but I think it’s clear that all the data supports that we will get to a higher degree of accuracy.” 

AI can offload routine, voluminous imaging studies from radiologists, for example, and give them more opportunities to interact with peers and patients, Topol suggests, and that can improve both performance and reduce burnout risks. “A lot of radiologists I interact with would be pleased to have more patient interactions and being able … to not just be in the dark basement looking at scans all day.” 

Smartphones and other related technologies can enable care democratization, provide more meaningful datapoints for clinicians and lighten routine assessment burdens on clinicians, and Topol has been a vocal advocate of adopting digital technologies for the past 15 years. That technology is getting to a point that it obviates the need for some routine – and overused – diagnostic procedures. 

As a cardiologist, Topol sees widespread use of routine echocardiograms and the impact of interpretation burdens on clinicians. AI can be applied to assess simpler smartphone-acquired images and heartbeat patterns to sharpen the focus on patients who are truly at risk. “We’re moving toward a time when acquisition of certain images and interpretations with algorithms is going to be used as a screening, as an initial tool, it’s going to be very useful,” he believes. 

Gathering and sifting patient data 

Current technology holds promise for gathering more relevant patient data and making sense of it, Topol says. AI, in particular, will be crucial in this large-scale, patient-centric analysis. 

“The next big thing is to take all the data of a given patient that is available … and to quickly process all that data, including all the labs and all the scans and everything that’s possible, to help get that teed up for the clinician,” he predicts. “These days, that’s very hard with the time that’s allotted (to physicians), but AI can help that. And it’s getting better at unstructured text. So we’re going to see, more and more, the streamlining of data … and I think that’s going to be a big help. 

“One of the things I’m really excited about in the imminent phase of AI in healthcare is keyboard liberation, so that we don’t have any doctor or nurse or clinician having to interact with keyboards, because the voice from the visit or at the bedside can be made into synthetic notes. That will get us back to the face-to-face, the true visit, the true bond of a patient to their clinician.” 

The eventual promise of advanced computing is what Topol calls multimodal AI, taking various swaths of medical data and making it useful for clinicians, health officials and individuals. For example, that level of technology may have blunted the worst effects of the pandemic, better predicting where outbreaks were likely and factoring in a patient’s individual risk factors to better suggest individualized preventive measures. 

Gazing into the future, as Topol frequently does, he sees technology enabling the use of digital twins to better refine treatments, individualized health coaching for patients and support for remote monitoring that would enable patients weather illnesses at home – supported by an intelligent grouping of sensors and AI that can summon medical care home, or advise transport to a facility, if a patient’s condition turns south. 

The pandemic should advise the need to ramp up what technology can deliver to the nation and its healthcare system, he concludes. 

“I think technology could do much more. AI was thought that it would be transformative in a pandemic. And it did help get us a couple of drugs more quickly,” he says. “We've had a lot of false starts. But, you know, I think one thing to emphasize is we're still relatively early with AI applications. And we're still learning all the time.” 

More for you

Loading data for hdm_tax_topic #better-outcomes...