Today, many decisions made in medicine are based upon prior observations, training, memory, and flawed studies.
With artificial intelligence (AI)- based tools, doctors will have powerful tools to make better diagnoses and treatment decisions based upon analysis of real-world clinical data and use of strong science.
The promise of AI is that medicine becomes more of a science than an artform. Guidelines promulgated by specialty societies for physicians’ use in daily practice are often based upon flimsy evidence and flawed studies. In a 2009 study, Duke University researchers found that only 11 percent of the recommendations from the American College of Cardiology and American Heart Association were based upon evidence from multiple randomized trials or meta-analyses – the gold standards for study design . A more recent study in the Journal of Neurosurgery reviewed compliance with evidence-based guidelines in patients with traumatic brain injuries (TBI) and found that the overall compliance rate was 73%, with only 3 out of 11 Level I trauma centers achieving a compliance rate exceeding 80%. The study concluded that despite widespread dissemination of EBM guidelines, patients with severe TBI continue to receive inconsistent care.
It is important to acknowledge that medical decision making is not just science – indeed, the voice of the patient must guide treatment based upon personal preferences. That said, the starting point should be science not conjecture.
The Potential of AI
Already, industries are increasingly adopting AI and cognitive computing technologies because of their ability to analyze vast amounts of information, recognize patterns, automate processes, and learn and improve. In fact, the research firm, IDC, predicts that global spending on AI and cognitive computing technologies will increase from $12.5 billion this year to $46 billion by 2020.
AI development and adoption in healthcare is in its infancy. There is a mixture of excitement and understandable skepticism. According to Healthcare IT News and HIMSS Analytics report, HIT Market Indicator: Artificial Intelligence, 35 percent of healthcare organizations plan to deploy AI to some degree within two years, and more than half intend to do so within five years. This might be a result of “fear of missing out” amid market hype, or it might reflect a real belief that AI can help solve intractable problems in healthcare. We do believe that AI will prove very valuable – it will bring automation to what is a very labor-intensive profession and precision to anecdote based decision making.
The New Physician Tool for a Science-based Approach to Medicine
Applying AI can amplify and augment physicians, not replace them. Three prominent ways this unfolds is by making better sense of patient data, using patterns to better predict patient health and improving physician training.
Making sense of patient data:
One way to use AI is to summarize and unlock key insights out of patient data. The sheer volume of healthcare data is growing so large that it becomes unmanageable for people to curate. Research firm IDC likened storing healthcare industry data on a stack of tablet computers. In 2013, IDC calculated that the stack measured nearly 5,500 miles high. By 2020 that stack is predicted to grow to reach nearly a third of the way to the moon, measuring in at 82,000 miles high. This data is largely unstructured – notes, forms, surveys, and other care related documents.
The collection, storage, and analysis of this data is critical for improved care, yet it is a drain on physicians. After care providers and others enter data into electronic health records (EHRs), it becomes trapped in a silo of sorts. In fact, it is so cumbersome to use patient data within EHRs that working with them is cited as a leading cause of physician burnout, according to a study in the Mayo Clinic Proceedings.
AI has the potential to unlock data and connect the dots to help physicians better predict risks and select more precise treatments for their patients. Machines trained to decipher physician notes from hospital or clinic encounters, radiology and procedure reports, and prescriptions and lab result data can then construct a definitive patient problem list. In some cases, diagnoses can be made by the computer prior to a physician confirming the diagnosis. We have developed software to read medical records directly from EHRs as well as records which have been printed and scanned or transmitted by facsimile. This process of understanding patient conditions, known as electronic phenotyping, can become the key for better predictive and treatment decisions.
Using patterns to predict patient health:
Highly subjective and sometimes anecdotal applications of a physician’s medical knowledge are no match for AI and machine learning algorithms trained on massive data sets with billions of analytic events. In fact, a recent study found that intelligent, self-learning computers were able to better predict heart attacks than use of calculators using observed risk factors. Since a physician’s practice is based upon the patients they have treated and the studies they’ve read, it is not surprising that their pattern recognition is more limited than machines fed millions of patient records.
Google’s DeepMind project has partnered with the United Kingdom’s National Health Service (NHS) to augment ophthalmology through computer-based identification of early degenerative eye disease. Google is developing a machine learning system trained on one million eye scans to recognize sight-threatening conditions from a simple digital retinal scan. The application of intelligent machines can dramatically increase physicians’ scale, diagnosing more patients than previously possibly at a stage where disease modification and reversal is possible.
Machines can also better predict when medications could produce unintended harmful side effects. In one study, by digesting large data sets, Stanford researchers could detect adverse drug events for commonly used medications in a systematic and reproducible way (Banda, J. M., Callahan, A., Winnenburg, R., Strasberg, H. R., Cami, A., Reis, B. Y., Vilar, S., Hripcsak, G., Dumontier, M., Shah, N. H. 2016; Feasibility of Prioritizing Drug-Drug-Event Associations Found in Electronic Health Records). In another study with an initial dataset of 1.8 million individuals, researchers were able to identify that the Food and Drug Administration issuance of Blackbox warning on an important drug — flagged for possibly increasing cardiovascular mortality — was not supported by evidence, reducing the access and limiting the drug's use without just cause (Leeper NJ, Bauer-Mehren A, Iyer SV, LePendu P, Olson C, Shah NH (2013)- Practice-Based Evidence: Profiling the Safety of Cilostazol by Text-Mining of Clinical Notes) ].
Training physicians:
In the new era of AI-assisted care, doctors will focus on delivering what computers cannot: guidance, counseling, and advocacy for patients. In a medical era in which physicians spend more time staring at screens rather than at patients, the human touch is becoming a lost art form. Traditionally, medical schools have selected students based upon their ability to memorize a corpus of facts and perform well on multiple choice tests. It is important to understand the underlying mechanisms of disease and the determination of what works best for patients.
Certainly, the next generation of physicians and healthcare providers will need to understand the underlying mechanisms of health and disease and the basis for treatment, but medical professionals must also acquire a basic knowledge of how to effectively use AI to augment their skills.
We predict that the use of AI based technology in the exam room will create more time for meaningful patient interaction. This is very promising since surveys continually show that people most trust physicians and that the healing touch and guiding voice has great impact.
The balance between practicing medicine as both an art and a science dates as far back as the Hippocratic Oath where a section translates to: “I will remember that there is art to medicine as well as science.” With AI assistance, doctors can concentrate providing their patients guidance, support, and a healing touch, with a focus on “well care,” and spend less time on computer clerical work and treatment guess work. The “art” of medicine will be the human element rather than the process of diagnosis and treatment decision making.
The greatest potential of AI lies in helping reduce the cost of care while improving outcomes for more people. McKinsey notes that if U.S healthcare were to use big data creatively and effectively to drive efficiency and quality, the sector could create more than $300 billion in value every year[1] and improve the lives of countless patients in the process.
The views, opinions and positions expressed within these guest posts are those of the author alone and do not represent those of Becker's Hospital Review/Becker's Healthcare. The accuracy, completeness and validity of any statements made within this article are not guaranteed. We accept no liability for any errors, omissions or representations. The copyright of this content belongs to the author and any liability with regards to infringement of intellectual property rights remains with them.