It was 1999, and I was speaking at a prestigious academic center’s ‘Innovations in Dermatology’ symposium. I presented work we had done on a web application that would allow a non-dermatologist (primary care doctor or other front-line provider) to upload images of a patient’s skin as well as some history. Subsequently, a dermatologist could review both the history and the images and enter a diagnosis and recommendations on the same website. My talk was greeted with intense scorn by one audience member (“You are cheapening our specialty,” he told me) and general lack of enthusiasm by most of the others.
Fast-forward to 2014. Dermatologists share digital images via email and a number of social networks thousands of times a day. If the specialty is cheapened, I can safely say it is not due to this activity. So far, we’ve not created any ‘cutaneous radiologists’ which was another fear of nay sayers at the time. In fact, the American Academy of Dermatology has an officially sponsored software application that members can use to provide volunteer teledermatology services to underserved clinics, and is planning on promoting this concept in the coming year. What a difference 15 years makes!
I use this story to set context for a thoughtful discussion on one of the most controversial telemedicine questions of our current time. Is it safe and effective care for providers to evaluate and prescribe for patients that they’ve never met face-to-face?
Join me in thinking through this question.
1. Most health care requires authentic relationships.
You don’t have to go very far to find graphic stories about folks who form relationships online. These relationships become quite ‘close,’ then the parties meet face-to-face and find that one or both was being highly deceptive with their online persona. It seems safe to conclude that the possibility of fraud in online relationships is much higher than in person. Most clinicians feel that forming a relationship with their patients is a core part of providing quality health care. Until this authentication challenge is solved, it’s hard to imagine many health care interactions with new patients being conducted in an online environment. Some of you may be thinking that ubiquitous, embedded videoconferencing solves this and certainly adds a great deal of value beyond text-based interactions. But I don’t know if we can say applications like Skype and FaceTime are the complete answer.
2. How do we define quality care in this new environment?
When you go to a hotel, you expect clean sheets and towels, a warm shower, a bed you can sleep in, and comfortable room temperature. What are the same minimum quality requirements for an initial doctor’s visit? As doctors, we’re taught that this includes a complete medical history and physical exam relevant to our specialty. As a dermatologist, I am not expected to do a complete neurological exam (you should be thankful for that) but unless you decline, I am expected to do a complete skin exam, even if you come in for a 4 X 4 cm patch of poison ivy on your arm. This is something that I really cannot do online, even with current state-of-the-art technologies.
The idea of allowing patients to upload images of body moles or facial acne for evaluation by providers they have not ever met gives many of my colleagues great consternation. (Even more controversial are image analysis scientists who are creating applications that can analyze these images without human intervention. More on that phenomenon at a later date.) My colleagues fret that a patient will send them an image of a mole that is benign, but ignore a mole that is an incipient melanoma. To me, this seems less onerous than the authentication issue. Patients are capable of managing their own risks when these risks are spelled out. For example, it seems perfectly reasonable to alert an individual that she may be missing an important lesion if she chooses to submit an image over the Internet, rather than come in for a complete exam. A closely related fear or objection is one of physician liability. In this context, the doctor is not only afraid they will provide substandard care by not doing a complete physical exam, but that they may be held liable for that omission. Once again, spelling out that a patient is taking accountability for those aspects of his care not addressed in an online interaction seems reasonable to me. I think the liability concerns are overstated.
3. Is the technology up to the task at hand?
There is not a general yes/no answer to this question. It is medical problem specific. The answers are in the realm of clinical research. Taking you back again to the late ‘90s, we (and many others) did painstaking clinical studies to empirically test whether a set of digital images is of sufficient quality to be a diagnostic tool in lieu of an in person exam. There now exists a body of literature that demonstrates this equivalency, with the possible exception of some pigmented lesions. We also carefully examined the feasibility that patients could take their own, clinically accurate, facial images of acne. Are there other examples? Can a psychiatrist do an initial evaluation of a patient via interactive video? I am not up on this literature, but my guess is that it has been studied and the answer is yes. There are probably a few other examples as well.
Where does this leave us in my logical analysis? To provide a quality care experience online without having met the patient in person, I content that the following criteria would need to be met:
1. Identify a medical problem that has a diagnostic data set, easily and reliably acquired by a consumer/patient.
2. Assure that the patient is capable of understanding that the online interaction is problem specific and may carry risks, particularly for omission of care involving other health problems.
3. Assure that the treatment decisions for the specific condition at hand are algorithmic and do not require an authentic relationship (i.e., the problem is transactional or of low emotional value to both provider and patient).
They accept images of facial acne and if they are comfortable making a diagnosis, will prescribe a limited array of therapeutics for these patients. Both are gaining some traction, indicating that there is consumer demand. Interestingly, since acne images involve the face, it makes it much less likely that an individual can assume a fraudulent online personality. And even if they decided to have their friend send in facial images what would be the point — to clandestinely procure a prescription for a topical antibiotic? I’d say there is not much risk on the authentication side in this model.
Another interesting comparison is the rise of retail clinics. These were initially scorned by primary care providers, but consumers are drawn to the convenience. The repertoire of problems is limited, as are the therapeutic options. Patients are made aware of these limitations and the associated risks.
Perhaps Direct Dermatology and DermatologistOnCall are the vanguard of a new set of medical services that are like retail clinics but delivered in an online environment. I’ll watch their evolution with great interest. And I think the risk of cheapening our specialty is low.
What an amazing time to be a champion of new health care delivery models. Three megatrends are driving change at a pace that I only wished for 20 years ago. Back then, I first had the insight that care could be improved, in some instances, if we separate doctor and patient in time and space.
The first trend is provider reimbursement reform. For provider organizations going at risk with their payers, these are nerve-wracking times. What were once revenue streams are now cost burdens and what were once costs are now opportunities to improve efficiency and thus our bottom line.
Secondly, all of the talk and activity around consumer health insurance exchanges is resulting in sharpened consumer awareness of all aspects of health, including costs, provider options and ways to improve health. A more engaged consumer is almost always a healthier consumer. Though I won’t spend any time on it today, these exchanges are almost as upsetting to the insurers as risk contracting is to providers.
The third trend is the adoption of mobile technologies, which of course gets a lot of coverage in this blog. We see the rapidly growing adoption of smart devices (phones, tablets) among patients and consumers now as inexorable. Their familiarity with these technologies results in an openness to connected health that we haven’t seen before in history.
Three phenomena are at play here. The first is that it is getting harder to find new molecules that represent any true breakthrough for our current disease mix. By some estimates, 75% of costs are due to chronic illnesses that are mostly lifestyle related, so new pills for hypertension or diabetes are of less interest to those who pay the bills. Instead, payers are focusing more on prevention, rather than treatment. Secondly, generics are increasingly prevalent and offered at prices that leave only the smallest margins for manufacturers. It is not hyperbole to say that when Walmart began offering generics at $4/month in 2006, the world of the pharmaceutical industry was turned upside down. Related is the looming ‘patent cliff’ that some large pharma companies are experiencing as blockbuster branded molecules go off patent and the associate margins vanish. Finally, the age-old business model of driving pharma sales through influencing doctors has fallen on hard times. Due to the Sunshine Act and related publicity, many physicians are no longer interacting with pharmaceutical detailing representatives. No more pens, mugs, pads or sales meetings. More importantly no more dinners masquerading as education, lavish trips, etc. Of course this is a good thing for all parties involved, but when your product is ten to twenty times as expensive as a generic and not really differentiated, it makes it harder to get the doctor to write that prescription. These effects are synergistic. For instance, at Partners HealthCare’s hospitals, clinicians are not able to see drug detail reps and our electronic purchasing system reminds us constantly to prescribe generics whenever possible.
Another force-multiplier is that as we health care providers take on more risk, we are pushing our suppliers (e.g., the device and pharmaceutical manufacturers) to lower our costs. One of the first things we signed up for during the journey from fee-for-service to fee-for-value was targets around generic prescribing. We’ve been doing far less branded prescribing for the last several years.
At the same time, here at the Center for Connected Health, it’s been a privilege to have a steady stream of visitors from most of the high-profile pharmaceutical companies. Some come in for a ‘pick-your-brain’ session, and many have hired us as consultants to help them examine how they can respond to these market changes via connected health solutions. The phrase we keep hearing from our pharma colleagues is that they want to move ‘beyond the pill.’
We’ve shared insights into how to use sensors and devices to improve patient engagement and accountability. We’ve talked about our experience with various technologies to improve adherence. Lately, we have been sharing our insights into how important motivation is and how assessing motivational state and individualizing messaging around that is critical for success in patient engagement and outcomes. With a couple of notable exceptions (which I cannot share because of non-disclosure), we have largely gotten lots of blank stares or polite head nodding. I kept wondering whether we just don’t get their industry, aren’t articulating our view of the solution or they simply aren’t ready to embrace it yet.
I had an insight recently that helps me put it all into perspective. We’ve been thinking a lot about the power of data analytics and targeted messaging of late. We have several very promising studies in progress to demonstrate that we can develop personalized messaging programs and keep patients/consumers engaged in healthy behaviors for long periods of time. This work breaks down into three areas (I guess this is the blog post of threes). One is data collection: what new sensors and tools can we use to capture more finely-textured information about you that relates to your health state? The second is the analysis of those data – the realm of predictive analytics and machine learning. The third is the psychology of engagement, an area well known to marketing professionals.
The insight is that pharmaceutical firms know how to do marketing and messaging, having done direct-to-consumer marketing for years. When they had the chance, they were frighteningly good at changing physician behavior to write prescriptions for their drugs.
Health care providers don’t have these skills in the work force and I dare say payers do not either. If the goal is behavior change in chronic illness management, which equates to skillful use of engaging messaging, the pharma companies should have a leg up.
Their challenge, however, is that pharma companies think of that skill and knowledge as an expense to support the sales of molecules. For decades, molecules have been the source of revenue and all of the marketing/messaging is an expense item. In order to transform their businesses ‘beyond the pill,’ they will need to turn this thinking on its head.
The engagement becomes the product. The therapeutic is almost a give away or ‘marketing expense.’ There aren’t many better ways to develop a relationship with a patient than through a prescription for a medication to treat a chronic illness. I’ve taken Simvastatin for years now and no one yet has leveraged that Trojan horse to upsell me other products, get me to do other things to lower my cholesterol, bring me additional diet and exercise opportunities or anything like that. Once a drug becomes a generic, it becomes a forgotten step child. But at $4/month, that molecule should be viewed as an inexpensive tool to develop a relationship with a patient around which to sell other services.
Health care providers are scared of the consequences of taking financial risk. Some bright, strategic-thinking pharmaceutical executive is going to come up with a service package that includes engagement tools, perhaps some connected health devices or an app and is centered around a therapeutic area but not a brand. Gillette did it with razors and blades. Bill Gates did it with computer hardware and the operating system. I know the pharma industry will rise to the occasion.
What do you think?
Since I gave a keynote at the 2013 Connected Health Symposium called “Making Health Addictive,” I’ve been posting on this topic in order to explain some of the concepts in more detail and to get your collective feedback (always incredibly helpful). Previous posts include a framing post, and further detail on what I laid out as three strategies to achieve addiction to healthy behaviors, “Make it About Life,” “Make it Personal” and “Reinforce Social Connections.”
Since strategy is not much use without a tactical component, I turn my attention in this post to tactics.
My current working model includes three tactics: Employ subliminal messaging, Use unpredictable rewards and Use the sentinel effect. Today’s post focuses on the first of those three, subliminal messaging.
Making health addictive is really about harnessing the power of our fascination with mobile devices, particularly smartphones. We check these devices up to 150 times per day. What if we put a personalized, relevant, motivational and unobtrusive message in front of you some of those times? Could we induce permanent behavior change? I am searching for examples of these customized mobile, personalized messages and any resulting behavior change, so if you know of any, please let me know.
The term ‘subliminal messaging’ has its roots in the advertising industry. When I was a teenager, I remember stories about psychological experiments where advertisers would splice still frames of product images or messages into unrelated film clips. Watching the film, the story went, you’d never actually see those images or messages, but they would subliminally imprint on your mind and influence your behavior. Another common use of the term is to refer to subtle visualizations in various advertising imaging that allegedly speak to the subconscious. A Google search will show you many examples, most with some sort of sexual double entendre.
Let’s think about how we might use this tactic to motivate healthy behavior. While getting someone’s attention with a colorful, catchy, fun, attractive or useful message of some sort, you’d slip in a health-related message. This tactic is an implementation tool for the strategy Make it About Life. I have two examples to illustrate this tactic applied to health care.
The first is to review the impressive work of the American Legacy Foundation’s Truth campaign. The current home page is a terrific example. To convey the message that cigarette smoke contains harmful chemicals (in this case methane and urea), there are videos illustrating, in grand urban settings, these messages using large stuffed animals. There are accompanying Twitter hashtags and a tool at the bottom of the page to allow visitors to build and disseminate their own video.
The work is amusing, edgy and takes full advantage of viral marketing. Since the messaging is aimed at teenagers, the context of bathroom humor is completely relevant. This campaign (and others like it illustrated on the site) serves to educate teenagers about important aspects of smoking in a fun way (who would want to smoke if cigarettes contain chemicals that are found in animal excrement?), BUT never does the content lecture, talk down to or browbeat the viewer regarding whether they choose to smoke or not.
Powerful subliminal messaging!
The second example is from our own work at the Center for Connected Health. It involves a study we did a while back to test the impact of text messaging on sunscreen adherence. The headline that came out from the study was that a daily text message reminder was a powerful motivator of adherence (about 60% of the time compared to about 10% in a control group).
What is even more relevant to this post was the design of the messages. Each morning, our participants had a text message delivered sharing the weather report and secondarily, a reminder to apply sunscreen.
In exit interviews, the most interesting thing was that study participants told us the thing that kept them coming back was the weather report. They didn’t really pay attention to the sunscreen adherence message. They didn’t object to it, but it didn’t really strike them either. Pretty good adherence rate for a forgotten message.
I think you see where this is going now. By designing health-related messages so that they apear within something that is either funny, inspiring or just plain useful, it seems we can have a greater impact than messages that threaten, scold or embarrass an individual. I’m talking about the type of messaging that has, over the years, led so many of my dermatology patients to say to me, “Please don’t yell at me because I got a sunburn.” I confess, I never yell at patients, but the feedback is that we need to message better.
What do you think?
“Two roads diverged in a wood, and I—
I took the one less traveled by,
And that has made all the difference.” Robert Frost
I received two emails from organizational leadership in the past week that represent milestones to me. One was about our efforts to better manage populations in the context of risk-bearing (pay-for-value) contracts with our payers: “[Partners] is undergoing a planning process to determine what we can do over the next 3 years (2015 to 2017) to better prepare for increased risk in payer contracts….there is a segment of our Hypertension and Diabetes population (and CHF) that would benefit from telemonitoring.” The language here is not dramatic, but the recognition that patient-generated data can be a critical tool for population health management and improved efficiency of care delivery is something we have strived to demonstrate over the last 10 years or more. It feels good to hear it come from others.
The second email came from a senior IT architect: “[The data integration team] are working with the Center for Connected Health to get remote monitoring data into [a scalable, secure database]. From there, the plan is to have the data viewable from, but not necessarily resident in, Epic.” Again, the words are not dramatic in nature, but the recognition, at the highest levels of our IT organization, that it is critical to our future to incorporate patient-integrated data into the care delivery process is a milestone.
I blogged a few months back about how we created the platform referenced in the paragraph above, allowing the easy integration of patient-generated data into our core clinical systems. In fact, we are deliberately using this platform to integrate our Patient-Reported Outcome Measures into the system. This will enable us to bring in both patient self-reported data as well as objective, biometric data and eventually perform analytics on these two data sets.
Partners HealthCare believes that patient-generated data is critical to our future and we’re managing this integration proactively. That does not mean our work is finished. For example, there is currently a significant overhead in managing remote sensors (BP cuffs, scales, etc.) and home hubs (2Net hub, Alere’s Connect device), and getting them back from patients after the program is complete (all of our Connect programs are time-limited). The overhead of purchasing that gear and managing inventory adds cost to the programs and strains ROI calculations.
Why not use what the mobile industry calls BYOD (bring your own device) and software apps on smartphones/tablets? That way, patients who own their own health sensors (so many are available in the consumer space these days) and mobile devices can participate at a much reduced cost to the delivery system.
A recent article in Health Data Management (Providers, Payers Get Appy) sheds some light on what other marquee health care organizations (Kaiser, Geisinger, Children’s Hospital and Aetna were featured) are thinking about the issue of integrating patient-generated data. There are some parallels and some differences to the way we’re approaching it at Partners.
I should caveat the next section of this post by saying that I only read the piece in HDM and did not talk to any of these folks personally. I do know all of these organizations, and have, in the past, spoken with their leadership about their strategies and continue to learn from each other. So I think my analysis is solid. However, I could be wrong.
The biggest difference I see is their belief that patients are already hopping on the bandwagon of buying their own sensors and generating their own data in some sort of massive wave –and that they are all using smartphones and tablets.
As the mHealth field was emerging four to five years ago, I went to conferences and listened to all of the enthusiastic entrepreneurs talk about self-tracking, Quantified Self, consumer enthusiasm for buying health tracking devices, etc. For some time I had a rather unsettled feeling because what I heard in the field did not match our experience at CCH. If we plot degree of illness vs. smart device adoption the graph looks something like this.
At Partners, the patients we are targeting for connected health programs — to improve care, increase patient self-management and decrease cost — are those sickest five percent — and they are not smart device adopters. The quantified-selfers brimming with smart devices, measuring and monitoring their health and fitness tend not to be folks with chronic illness or patients who drive much utilization of health care dollars.
What the organizations featured in the Health Data Management piece have in common is that they are creating mobile apps/platforms to enable patients who already own smartphones and connected devices to share those data. This approach seems less proactive to me, meaning if you have a smart device and one or more connected sensors and want to share your data with these organizations, they’d be happy to have it. If you do not, they’ll manage you the old-fashioned way with office visits, hospitalizations, etc. The advantage of this strategy, as noted above, is that the cost of technology falls on the patient and the organization rides the adoption curve. However, the adoption of these technologies in sicker patients is still low.
At some point, the five percent of sickest patients will move from the top left quadrant to the top right quadrant in the graphic above. We’ve started to see this with patients asking us for apps, using their own tablets (thank God for Skyping with grandchildren!), etc. But there is almost no penetration of connected sensors in this population yet. We’ve made the commitment at Partners to be more proactive in deploying devices and hubs to these patients. We’re also thinking about how to facilitate adoption of smart devices among these patients — maybe offer coupons to facilitate the purchase of their own scales, BP cuffs, etc.
Then there is the matter of towing the regulatory line. Many of the devices available at Best Buy or the Apple Store are not FDA approved to be used in treating illness. Many of the smartphone apps out there are also not FDA approved. I’m sure this is one reason Kaiser, Geisinger, et al. are creating their own. At least one firm, iHealth, is going in this direction (their platform is FDA approved and available to consumers), and combining their consumer platform into a care delivery process seems like an intriguing option. Qualcomm is offering the 2Net ‘hub’ as a mobile app.
So what are the two roads? One is a proactive approach, using patient-generated data as a core part of population health management and making sure that the sickest patients have the devices to generate objective and self-reported data. The other is perhaps a less aggressive approach, opening the APIs and allowing those with smartphones and connected devices to bring data in. For now, we believe that approach will generate more data from healthy, low-utilization individuals and miss the sickest five percent responsible for 40 to 50 percent of health care costs.
At some point the second path will make the most sense as more, sicker individuals will have smart devices and connected sensors. We’re preparing for that world at Partners.
When do you think it will arrive?
Just four to five years ago we were beginning to understand the power of self-tracking for health. Prior to that, individuals interested in gathering measurable information about their lives resorted to paper journals and Excel spreadsheets and manually entered data from tracking devices themselves. Those that participated were really motivated.
We soon started to see the power of passive tracking. In about 2005, my friend Penny Ford-Carleton came up with the term “wear and forget” sensors to call attention to all of the new form factors coming on the market – clip-ons, bandaids, smart clothing, etc. The early versions of these trackers required you to plug them into a computer to offload their data. Being disciplined enough to regularly do the upload and study the data required motivation. Not quite wear and forget, but passive sensing drew us a step closer.
The next innovation was to move from wired to wireless data upload. Sensor manufacturers began to routinely put wireless chips into their sensors allowing easy migration of the data from the sensor to some sort of hub and then onto the cloud. The vision of continuous sensing is now upon us.
So, what’s next?
I’d highlight three things.
- Seamless connectivity through the mobile phone. This is happening quickly. The technology is ready (smartphones, apps, better battery life) and the patient adoption of these devices is increasing fast as well (for example, we saw 65% adoption in a recent ad hoc survey of one of our own underserved patient populations). Imagine not needing any hub device but having the wireless sensor pair effortlessly with your smartphone, uploading the data without extra effort on your part. Companies like iHealth are doing this already and many, including industry leader Qualcomm Life are moving in that direction.
- Increasingly innovative sensors. Right now at the Center for Connected Health, we’re testing two innovative sensors in preparation for the developers to file for FDA clearance. One is a wrist-worn device that measures, among other things, blood pressure continuously without inflating any cuff. The second is a shirt with sensors embedded in the fabric that enable passive tracking of heart rate, respiration, EKG and activity. There is tremendous innovation in this area. Look for more bandaids, pendants, etc. to come our way.
- Tracking beyond physiology. This may be the most exciting development of the three. Up until now we’ve thought about tracking mainly vital signs (heart rate, blood pressure, weight, blood glucose) and more recently activity, via the myriad of devices for this purpose. But remember, the vision for Connected Health is to make care a continuous function in your daily life – to move beyond the doctor’s office as the only place where care is accomplished. If we’re going to get there, we need to expand tracking beyond vital signs and start to collect data that provides further context for your state of health. It turns out the smartphone is a worthy device to accomplish this as well. Firms like Ginger.io are working on software that can monitor your mobile activities and infer your state of wellness. Their first effort centers around depression, but they will go beyond that. Just think of how much more powerful it will be for your primary care doctor to not only know your blood pressure and activity level, but your mood and state of motivation. This is the context for our excitement about mood trackers.
If you read the descriptions of the mood trackers we’re featuring at wellocracy.com, you’ll see that they are pretty primitive at the moment. If you’re motivated, and have the time to open an app on your smartphone, you can record your mood at any time, as well as key in associated places, people, activities, etc. One tracker in particular, Emotion Sense, allows you to enter your mood, but also tracks frequency of phone calls, texts, and other smartphone activity to correlate with your own data entry. It grows smarter as time goes on, predicting your mood from associate data. This sounds a lot like the dawn of physiologic trackers just a few years ago.
Emotion Sense and Ginger.io represent the exciting trend in mood tracking, moving as we did in physiologic tracking from active data input to passive ‘wear and forget’ tracking. This type of analysis, combined with physiologic tracking, should lead to very insightful snapshots of your health without you having to enter anything.
What’s left? Well, the mood trackers have a few iterations to go before they get really good. Then we have to develop the analytics to merge their data with physiologic data. Finally, it would be really handy to get an analysis of an individual’s state of motivation. This is slightly different than mood. We’ve found motivation or readiness to change to be really important in targeting messages to chronically ill individuals to help them improve their health. Right now all we can do is ask. It’d be great to infer motivational state from mobile phone data.
Do you know if anyone is working on this?