by Elizabeth Hofheinz, M.P.H., M.Ed., September 3, 2019
While many “sexy” technologies have come and gone in the field of orthopedics, artificial intelligence (AI) is clearly here to stay. In fact, the AI health market is burgeoning and predicted to reach $6.6 billion by 20211.
A recent article explores which areas in orthopedics might experience the most value from AI. The paper, “Value-based Healthcare: Can Artificial Intelligence Provide Value in Orthopaedic Surgery?” appears in the July 17, 2019 edition of Clinical Orthpaedics and Related Research.
Co-author and orthopedic surgeon Prakash Jayakumar, M.D. Ph.D. is Assistant Professor of Surgery and Perioperative Care at Dell Medical School at The University of Texas at Austin. He commented to OSN, “It is important to recognize that there are a wide range of different technologies that fall under the banner of AI. For one, many current applications involving AI use machine learning algorithms that are different to deep learning. The latter involves more complex, brain-like, neural networks enabling elevated processing power and functions capable of understanding and organizing large unorganized data such that often found in electronic health records (EHRs).”
AI…boosting the value-add
Along with co-authors Meredith L. G. Moore, B.S. and Kevin J. Bozic, M.D., M.B.A., Dr. Jayakumar describes the range of technologies using machine and deep learning AI within three domains that promise to hold value in orthopedics: (1) Advanced data discovery and extraction, (2) improved diagnostics and prediction, and (3) enhanced clinical and decision support.
“Predictive models can use advanced algorithms to discover and extract patterns occurring in large, often unorganized and chaotic datasets,” says Dr. Jayakumar. “The algorithms are also able to create increasingly complex layers of information that can be processed on demand to reveal actionable insights. AI has already shown itself to be useful in the imaging realm, with studies finding remarkable accuracy for machine diagnoses of various conditions. This technology has at times surpassed human accuracy in mapping, for example, the progression of spine degeneration. AI may also enhance predictions of adverse events from raw EHR data in real time. The technology can further augment decision-making in the real-world, improve workflows and positively influence how we interact with patients. In general, all these functions to varying extents can improve outcomes benefiting patients relative to costs.”
AI + PROs = better engagement, experience and outcomes…
“One of the things we’re doing with AI/machine learning is to combine patient reported outcome (PRO) data with demographic and medical data to derive personalized metrics that can guide decision making around surgical procedures. For example, patients with knee osteoarthritis being considered for a total knee replacement complete general health and knee osteoarthritis-specific PROs prior to their clinic visit. These are applied to a machine learning algorithm developed from massive national data sources. The PRO data involving patient perceptions of their physical limitations as well as their mental health is combined with patient demographics, clinical comorbidities, and prior use of health care services to provide personalized information related to the surgery. It can be extremely powerful to say to a patient ‘Based on what you’ve told us about what you can do and how you feel, combined with your medical background, we are able to provide you with some extremely personalized information regarding surgery using a sophisticated computer program. Someone with your profile, physical ability, mood and social circumstances, could have X% potential benefit from the joint replacement, % risk or chance of a complication, % improvement in pain and stiffness and % improvement in quality of life.”‘
“As surgeons we’ve never been able to truly do this before and instead have been reliant on our personal experience, variable evidence-based, clinical culture and, in reality, ‘guesswork’. We can now walk patients through a shared decision-making process around a potentially life changing treatment and include accurate, patient-specific information to add richness and quality to our consultations.”
This also highlights how something that is relatively high-tech could also have a positive influence on the softer skills and doctor-patient relationship in medicine. Dr. Jayakumar: “AI can and should be humanized; we shouldn’t just use this technology to provide ‘cool data’ and perform ‘neat tricks’, but leverage it to enhance human interactions with patients and our clinical teams to deliver better care. Also, AI is really not a replacement for the surgeon. Although AI-driven programs may be delivering personalized metrics on demand, way beyond the capabilities of a human, and AI-driven robots are advancing the precision of techniques, the ‘human touch’ is still critical if we are to truly meet the preferences, needs and values of our patients and truly deliver high value, quality care.”
“When one takes a sky-high view of clinical pathways and looks at the full cycle of care from the point when the patient is seen in clinic through to surgery and afterwards; an AI layer can be applied to monitor and analyze the pathway itself, something that could be valuable in terms of enhancing workflow efficiency, resource utilization, scheduling, redesigning care teams and care pathways, which would ultimately provide better performance, outcomes and costs. Zoning in on the operative phase, data points, including time spent, could be provided to assess each step of a procedure—this could feed back data to look at adapting components to improve operative efficiency. The more data points we have, the more lenses we have trained on certain conditions and interventions. AI can accelerate different phases of these processes.”
At present, says Dr. Jayakumar, in the U.S., AI is primarily being developed and applied within academic centers. “There is growing worldwide interest and enthusiasm for this technology. However, to date there are only a relatively select band of entities that are actively engaged with AI/machine learning and fewer still with deep learning solutions. The vehicle to deliver AI to the front lines is the data itself. Most institutions still need to establish an effective EHR system that is interoperable with other technology platforms such as PRO collection and communication systems. We have made great strides at our institution with such efforts, but people should realize that this is an evolving situation. Ultimately, these core technologies should be simply built to meet the real needs of care teams and patients, enabling the collection of robust and reliable data that can then be fed into AI solutions. With AI more than ever, it’s clearly a case of quality in – quality out.”
As for cost, Dr. Jayakumar notes, “As proof of concepts are being achieved with AI solutions, clearly the discussions around commercialization are around the corner and even underway in some cases. As per Moore’s Law, there is an exponential growth of computational power, which is also aligned with a decline in technology costs. Time will tell how our profession will navigate discussions around costs (including resources and time) for the AI related technology itself as well as the services that wrap around them. We are likely to reach a point at which access to the technology itself comes at minimal or no cost with most of the costs borne around data, services and systems delivering AI functions.”
The more we know, the more questions arise…
“Right now, it is imperative that we keep pace with this technology and how it is being integrated into the care of our patients. For instance, patients will ask, if they are not doing so already, ‘Where is this information coming from? How is this calculated? How confident are you with what it says?’. After all, we are using their data and the data of populations we may not even have treated to inform future care. AI is here to stay and like any medical technology (including any treatment, drug or device) we need to understand the risks, benefits, limitations, alternatives and nuances in order to maintain our position at the forefront of the advances it can deliver. Development should be done in a controlled manner and regulated effectively without losing track of why it was developed in the first place—to improve the outcomes that matter to patients.”
References
