IntelliMedicine
For many diseases, there are far fewer physicians than there are patients who need their care.
Take type 1 diabetes. In the United States, there are roughly 8,000 endocrinologists. There are close to 1.5 million type 1 patients. If you also include type 2 patients who require insulin therapy and would benefit from the specialized care of an endocrinologist, you add another 8 million patients to the pool! That translates to each physician needing to care for almost 1,200 patients.
If type 1 diabetes was the only disease that an endocrinologist treated, that would be one thing. But they also treat thyroid disorders, osteoporosis, infertility, hormonal imbalances, etc. It’s not hard to envision that quickly this can create an unmanageable dynamic. In fact, in many instances, these physicians can only spend 15-20 minutes with a diabetes patient because of their case volumes. Beyond that, they are typically seeing the most severe diabetes patients every three months. In order to ensure they are providing the right care and treatment recommendations, they need to review the patient’s glucose, insulin, medication and if available, nutrition and exercise data since the patients last visit. Now, put yourself in the endocrinologist’s shoes. When you have 20 minutes to spare, how can you review 90-days of all this data, analyze trends and make an effective therapy recommendation for your patient? It is challenging, to say the least. To alleviate some of the pressure, the physician needs more time. So, how does one free up time when seemingly no additional free time exists?
One answer: intelligent telemedicine, or what I call “Intellimedicine”.
Telemedicine, on its own, is good, and certainly COVID-19 has shown many the value of using video technology to enable healthcare. However, using what amounts to tele-conferencing as a simple substitute for an in-person office visit alone doesn’t solve everything. Physicians still have the challenge of trying to care for more patients than they can handle. However, when combined with the intelligence of decision support, virtual care has the potential to improve health while also unlocking efficiencies. Let’s look at a few ways as to how this is possible.
Digital Triage
Within the 1,200 patients a specialist must care for, not all of them require immediate assistance. Some may be urgent, some may be trending towards urgent, while others may be doing just fine. Through the use of technology, the health status of a patient could be determined in near-real time, remotely and proactively. With the data that exists surrounding a patient – including diagnostic measurements from medical devices, lab results, pharmacy data or even information from their smart watch – the capabilities exist for decision support algorithms to digitally triage these patients.
The triage doesn’t need to only assess a patient’s health status, it can also help match them to an appropriate physician. In other words, based on the level of patient risk, the system can ensure the most appropriate physician is assigned to a particular patient. For example, not all issues require intervention from a specialist and could be routed to a primary care physician, a nurse practitioner, etc.
This isn’t fantasy. There is a diabetes clinic in the Netherlands doing exactly this kind of triage on a daily basis. Using a patient’s blood glucose and insulin data, the clinicians in this office walk in every day to a prioritized list of patients. For the highest risk patients, they proactively reach out instead of waiting for them to come in on their next scheduled office visit.
There is another company here in the United States, where I serve as a member of the Board of Directors, that uses Artificial Intelligence to identify the most at-risk Cardiology patients. The company’s algorithm uses quantitative data from diagnostic imaging devices as well as qualitative data from the electronic medical record to assess a particular patient’s risk score. Once the algorithm has determined all the high-risk patients in a population, it then cross-references this population with the list of patients who are either scheduled to see a specialist or who have been scheduled for cardiac surgery. The proper care pathway for a high-risk cardiac patient is either a follow-up visit to a specialist for more tests / consultation, or surgery. What the company is finding is that approximately 30% of high-risk patients have neither been sent to see a specialist, nor scheduled for surgery. These patients have essentially fallen through the cracks. With this capability, a health system can easily identify their most at-risk population, reach out and ensure appropriate and timely care is provided.
Remote Intervention
With the explosion of sensor technology, the capability also exists to proactively provide care to a patient. Continuous glucose monitors (CGM) are sensors that measure a diabetes patient’s glucose levels every 5 minutes. Prior to CGM, a patient would have to prick their finger, collect their blood manually and check their glucose levels using a glucose meter. They typically did this 3-5 times a day, before a meal or a snack. With CGM, a patient gets up to 288 glucose measurements a day without having to prick their finger each time. From a statistical and clinical perspective, it’s hard to determine a trend with 3-5 data points. However, with 288 data points significantly more can be understood about the glycemic status of a patient. For instance, with all of this additional data, predictive algorithms can be developed. If glucose values for a patient have been rising these algorithms could trigger an alarm – depending on the slope of the glucose curve and the rate of change – to alert a physician that a patient is in danger of a potential hyperglycemic event. In parallel, the same alert could even be sent to loved ones or care partners who also provide oversight. Through a remote connection, or even through simple text message, the physician and/or care partner could then communicate with the patient and advise them on the best course of action to prevent the event. In much more sophisticated constructs, the text message sent to the patient with instructions on what to do needn’t be sent by a physician, it could be computer generated (see AI-based interpretation below).
This doesn’t only apply to diabetes. Multiple chronic disease states can be monitored with biometric data from digital scales, blood pressure cuffs, pulse oximeters, spirometers, pedometers and even activity trackers. The more real-time the data, the greater the ability to identify and even predict issues in a way where successful, proactive remote intervention can occur.
AI-based Interpretation
As mentioned earlier, it’s very difficult to assess 90-days of glucose, insulin, medication, exercise and meal data in order to understand how a patient has been managing their diabetes and then coming up with a therapy recommendation – especially in 20 minutes. This feat is not that hard for an AI-based algorithm to accomplish, though.
Artificial Intelligence algorithms have been around for a while in healthcare. They are virtually interpreting MR, CT and X-Ray images. They are screening patients for cancer. These algorithms are quite sophisticated. There is no reason that these algorithms could not take patient specific data, cross-reference it against accepted disease management protocols and clinical guidelines and provide a clinical recommendation to the physician. This recommendation could include the types of pharmaceutical, and the dose, to prescribe. It could also include the rationale for why the algorithm came to this conclusion, so the physician can understand the logic the algorithm applied. In addition, multiple possible recommendations could be presented along with a confidence level – a score that indicates how certain the algorithm feels that each prediction is the right recommendation. Of course, none of this is intended to replace the role of the physician. In fact, he or she may choose not to accept any of the systems recommendations. It is merely a tool that the physician can use to aid in the clinical decision-making process and help their overall productivity and efficiency.
The Change Required
In order to make Intellimedicine more mainstream a number of things must change:
Patient behavior … using digital devices to facilitate care is not todays standard or the norm. Therefore, “if you build it, they will come” principles have not broadly taken hold. This is especially true for the elderly who are the most afflicted by chronic conditions. However, in a world influenced by COVID-19, we could experience a sea-change in behaviors and attitudes towards digital health. My 72-year old mother has become quite proficient with Zoom and other remote technologies over the past 60-days. In addition, she is also now very hesitant about going into a care setting and potentially risking exposure to the Corona virus. For her, the collision of digital proficiency and fear of exposure have changed her attitudes towards the benefits of telehealth. She is likely not alone.
Physician-behavior … just as patient behavior needs to change, the attitudes of physicians must as well. In over 20 years in the healthcare space, I have seen more skepticism than enthusiasm from physicians about the value of digital tools. The skepticism stems from any number of reasons – from the fact that these tools represent a change to their existing processes and workflow, to a lack of trust in the ability of a digital algorithm to accurately assess, diagnose and provide therapy advice for a patient. Some of the skepticism is warranted, however, in order for Intellimedicine to work, physicians must be firmly on the side of helping to move the technology forward.
Approval, Coverage & Payment … our regulatory authorities need to prioritize this area. It has been encouraging to see the Center of Medicare and Medicaid Services (CMS) provide coverage and payment for telemedicine. It is an encouraging step forward, but more needs to be done, particularly in risk-sharing and value-based payment models. In addition, we need the FDA to more aggressively embrace technologies – particularly AI-based algorithms – that can facilitate this migration. Can a digital algorithm that provides a therapy recommendation make a mistake? Yes. Can a physician who has 20 minutes to interpret 90-days of data make a mistake with their assessment? Also, yes. We need our regulatory agencies to embrace this aspect of care while working more aggressively to ensure safe and effective digital tools enter the marketplace AND that the necessary and reasonable one’s are covered and adequately reimbursed in a timely fashion.
Access … chronic care afflicts not only the elderly disproportionately, but it also disproportionately impacts minority populations. Access to the healthcare coverage and to these technologies is therefore critical. With the right data and through the use of statistical methods, it is possible to assess an individual patient’s risk of a future ER visit or of future hospitalization. These are the patients that could potentially benefit the most from remote monitoring, in general, and Intellimedicine specifically. Since these digital tools allow a more efficient use of a healthcare system – certainly much more efficient than a hospital stay or a trip to the ER – making these technologies ubiquitous and more affordable, especially to the highest risk populations, must be a common goal.
In many areas of the global health system today, demand for care far exceeds the system’s ability to supply the care. This dynamic is particularly true for the most prevalent chronic diseases. Through a more aggressive push for the use of Intellimedicine, the healthcare system gives itself a chance to better manage this dynamic.
