Artificial Intelligence Should Start With Artificial Joints
Artificial intelligence (A. I.) will inevitably converge with medicine, yet how and where it will yield the greatest impact remains to be seen. If wielded wisely, machine learning and other related forms of A.I. may prove to be the lifeline health care needs.
While the practice of medicine is an art and its history is storied, in today’s modern practice we permit ill-conceived electronic health records (EHRs) primarily designed for billing to enter patient rooms and detract from the doctor-patient relationship. Weighing available data stored in the EHR paints the picture of the patient, but reliance on the screen and the infinite clicks it demands creates nothing more than the illusion of patient-centered care at the expense of truly attentive, human-centered care.
By failing to tailor our plans by listening to the expectations, thoughts, and feelings of the patient in front of us, we practice cookie cutter medicine where one size fits no one. Continuing down this path will only lead to the further dehumanization of both doctor and patient. As Eric Topol describes in his new book, Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again, we are in the midst of a Fourth Industrial Age wherein Big Data, A.I., and robotics may be able to revolutionize health care inefficiencies, provide custom care, and maximize the latest evidence to guide treatment.
First described in 1956, A.I. has quickly become a reality with the ubiquity of advanced computing power and the newfound availability to collect and store vast quantities of data, colloquially known as Big Data. In studying this data, sophisticated algorithms may be created and improved upon to recognize patterns that aid in diagnosis or projection of value metrics.
The practice of medicine has simultaneously transformed into a value-based industry focused on the best possible patient experience at the lowest possible price point. Orthopaedic surgery, specifically the field of lower extremity arthroplasty wherein diseased joints are replaced with artificial ones made of metal and plastic, is ideally poised to evaluate the impact of A.I. on the rest of medicine.
First and foremost, joint replacement is usually elective surgery. The patient with end-stage arthritis on x-ray may be referred to a specialist for joint replacement, but the decision to operate requires shared decision-making that involves careful consideration tailored to the patient’s functional demands, medical status, quality of life, and expectations. A.I. boasts the ability to detect such nuance and anticipate the future with enough high quality patient data, which may lend to a sophisticated algorithm that predicts the risk of eventually undergoing a joint replacement, the cost and length of staying during admission, or even their post-operative recovery trajectory.
Of course, to develop such an algorithm requires broad commitment to collecting patient data on the order of hundreds of thousands of patients. Fortunately, arthroplasty represents one of the highest volume procedural subspecialties, with hip and knee replacement representing the most common procedures reimbursed by Medicare. Moreover, the patient experience after surgery is increasingly tied to reimbursement. As a result, hip and knee replacement surgeons for over a decade have serendipitously been stockpiling troves of data in patient-centered registries primed for machine learning analysis and algorithm development.
Joint replacement is also uniquely positioned to evaluate the impact of changing payment models in medicine. The Center for Medicare and Medicaid Services has recently instituted the “bundled payment” as an alternative payment model whereby a flat fee is paid to the hospital for all care up to 90 days after surgery, agnostic of preoperative patient complexity. With machine learning analyses, preoperative factors may be finally quantified to propose more equitable patient-specific payment models.
Radiology, robot-assisted surgery, and physical activity all closely affect the daily clinical workflow surrounding hip or knee replacement and similarly generate large quantities of data capable of being studied and characterized with A.I.-based algorithms. One company in particular, FocusMotion, works with orthopaedic surgeons to remotely monitor arthroplasty patients using a machine learning algorithm that accurately captures range of motion, gait, therapy compliance, opioid dependency, and activity – among thousands of other data points – from conventional smartphone sensors.
The excitement surrounding the eventual merger of human and artificial intelligence in medicine is great and likely substantiated by the prospect of increasing efficiencies while reinvigorating the doctor-patient relationship. In order to maximize clinically meaningful and appropriate application of A.I. – which presently remains unregulated and untested in medicine – further study is critical. Orthopaedics, particularly arthroplasty, embodies a high volume subspecialty with nuanced care plans, readily available clinical data, and far-reaching implications to payment models, policy, device manufacturing, radiology, surgical techniques, and everyday human movement. With this unique position as a generator of impactful big data, artificial joints may be the ideal starting point to evaluate the bandwidth and harness the power of artificial intelligence.
You can read the original article by Dr. Prem Ramkumar on Forbes Magazine online:
