This is a fitting time of year to pause and take a moment to look back and to look forward in one’s industry. An evergreen issue for me is that healthcare delivery in the United States is well, complicated. It’s complicated due to the basics of medicine, biology, behavior, pharmacology, illness and the like. But it’s also complicated in its delivery, the systems of and access to care, opaque payment structures that disconnect consumer from provider, and so on. Add to that technology and innovation and startups and things can get very messy very fast. “Users” of medical technologies aren’t solely patients, but also physicians and other healthcare providers, perhaps clinical trials, insurers, and regulatory bodies (whose rules can vary from state-to-state), all of which conspire to make sustainability, iteration and scaling an even tougher challenge for digital companies in this space. It’s complicated.
Fool Me Once
Nevertheless, I have to say, I am continually amazed with the proverbial brave new world of tech and medicine and how it seems to have a Groundhog Day-like propensity to repeat over and over with just slight differences. But we still haven’t cracked the code for electronic medical/healthcare records to play well together or interoperate with other systems—practice management, outpatient with inpatient, registries, payers, etc.
A few years back, I had goosebumps with the promise of Hadoop in helping me make sense and connections with my large clinical outcomes datasets combined with unstructured clinician variables, patient comorbidities and medical history, and payer datasets. While I can do some things today that I could not do before, the easy-button promise of black-box-algorithm-magic has yet to happen.
Every time I think that this time it will be different when I hear about the latest gee-whiz announcement of something promising to be groundbreaking, paradigm shifting or actually disruptive, I soon come away with that disappointed feeling of here we go again. I do have to admit my excitement with Atul Gawande’s new position as CEO of the Amazon-Berkshire-JPMorgan Chase (or “ABC”) healthcare partnership, as I have long been a fan of his writing and perspectives on healthcare and medical service provision—so time will tell—but I do like this mix of smart folks and solid funding.
Show Me the Money, Part 1
Rock Health notes that funding in just US-based companies providing artificial intelligence (AI) and machine learning (ML) approaches to medical services have seen a 2010% increase in total investment between 2011 and 2017, to the tune of $98.4 million and via risk capital, aggregated investments are reaching $7 billion so far in 2018. found that healthcare AI has more than 300 first equity rounds since 2016. Indeed, even non-per se medical companies are also heavily wading into the depths of healthcare and technology; between 2013 and 2017, Apple had filed 54 healthcare related patents while Microsoft filed for 73, and Alphabet submitted a whopping 186 healthcare patents.
Not So Fast Unicorn
In spite of this level of growth in funding, not everything is butterflies and rainbows. In the world of digital health startups what leads to almost certain death of the company is falling in love with the tech and then looking for a healthcare issue it can address. As Paul Yock wrote in Fast Company “Entrepreneurs and investors from the tech world mistakenly assume that (the) ‘lean startup’ approach, which works well for products like photo-sharing tools and meal-delivery apps, should be equally successful for tackling any kind of problem. However, this strategy is ill-suited to healthcare, a much more complex and regulated industry…For example, many founders coming from tech are focused on building and marketing products to consumers. They don’t realize until well into their company’s development that doctors and insurers are actually the gatekeepers and customers to whom they should be selling their products. This is why 61% of digital health companies that start B2C end up pivoting to B2B and selling to insurance companies, employers, hospitals, or other healthcare providers.”
AI to the Rescue (?)
And this is not to say that there are no amazing and sustainable successes. Adam Bluestein recently ran the numbers in Fast Company magazine on AI in healthcare and found that Arterys’ FDA approved cardiac AI could diagnosis MRI scan data in 15 seconds versus 30 minutes for a human. The FDA has been busy vetting—and approving—companies using AI and machine learning in their services. Here is just a sample of approved companies this year:
aidoc: AI that detects and prioritizes acute cases in the radiology workflow.
MaxQ AI: delivers artificial intelligence based decision support to improve clinical
Neural Analytics: A robotically assisted ultrasound system for brain health assessment
Viz.Ai: Provides intelligent infrastructure through which time sensitive, life-threatening
diseases are diagnosed and managed, and new therapies discovered.
IDx: Autonomous AI that instantly detects eye disease.
Bay Labs: Combines deep learning with cardiovascular imaging to help in the diagnosis
and management of heart disease.
Zebra Medical Vision: Use CT scan data to calculate risk of cardiovascular disease
Densitas: Improve the processes and quality of breast screening and breast imaging
iCAD: Offers computer aided detection and workflow solutions to support detection of
breast, prostate and colorectal cancers.
Arterys: Clinical SaaS analytics platform revolutionizing medical imaging and healthcare
through ultra-fast cloud computing, advanced visualization, and deep learning.
Mirada Medical: Provide complex image analysis problems in the diagnosis and treatment
of cancer and other diseases
Imagen: Apply AI to medical image analysis.
STAT reported that hospitals are employing AI for mortality and readmission predictions, as well as operating room scheduling along with clinical uses in early diagnoses of sepsis and in clinical decision making. The Federal government has stated, via HealthIT.gov, that clinical decision support systems or approaches are meant to augment/improve patient care via “…computerized alerts and reminders to care providers and patients; clinical guidelines; condition-specific order sets; focused patient data reports and summaries; documentation templates; diagnostic support, and contextually relevant reference information, among other tools.” AI enhanced electronic medical/health records are a great place to have such tools reside. In addition to better care and thus improved clinical outcomes, there is also an expectation of improved efficiency vis-à-vis the delivery of care and that it be more timely and error-free.
In the National Institute of Medicine’s report, Optimizing Strategies for Clinical Decision Support that was published last year, they note what they consider to be key aspects to such support:
- “provide measurable value in addressing a recognized problem area or area for improvement;
- leverage multiple data types to bring the most current and relevant evidence and evidence-based practice recommendations to bear on clinical decisions;
- produce actionable insights from the abundant multiple data sources;
- deliver information to the user that allows the user to make final practice decisions, rather than being opaque and acting autonomously;
- demonstrate good usability principles, including clear displays and rapid action options;
- are testable in small settings with a clear path to larger scalability; and
- support successful participation in quality and value improvement initiatives.”
Patients are perhaps less enthusiastic about being early adopters of medical AI. A Salesforce “2017 Connected Patient Report,” conducted by Harris Poll, found that:
- “Among baby boomers, 74% are worried about AI providing an incorrect diagnosis vs. their doctor’s recommendation—something only 60% of millennials were worried about, indicating a trust gap between generations.
- When asked about currently available AI applications, baby boomers (28%) were significantly less likely than millennials (63%) to agree that they are interested in a digital assistant like Siri or Alexa recommending personalized healthy habits, similar to how online retailers recommend things to buy based on purchase history.”
Big pharma is using AI to digest and make sense of their big data in order to streamline their drug development pipeline and processes, and to better understand the biological mechanisms of illnesses in order to improve drug targeting and clinical trials. Pharmaceutical companies have an opportunity to engage with patients in new ways. The same Salesforce survey discovered:
- Nearly three in four (72%) respondents agree that they would choose drugs from pharmaceutical companies that are engaged in their outcomes versus those who are not.
- Three in five (60%) Americans are open to virtual support service options (e.g., video conference calls) with pharmaceutical companies to help them understand their medications. This is especially true with millennials, as 70% want to leverage these modern technologies to communicate with their drug providers.
- More than four in five Americans (83%)—and 88% of millennials—would share their experiences and direct feedback about medications with a pharmaceutical company to help improve their abilities to develop and support new medications.
Big Data and N=1
I have written about the Precision Medicine Initiative® (PMI) that former President Obama instituted during his tenure. Through advances in research, technology and policies that empower patients, the goal is to enable a new era of medicine in which researchers, providers, and patients work together to develop individualized care. It’s a bit “moonshot-ish,” which I like. It’s also very integrative, which I also like. Results may not be as immediate as anyone would prefer, but I think the sorting and sifting of the resulting big data will be aided by AI. I would like to see the combination of large patient outcome registries to be combined with the Federal findings in order to have the best of both worlds, or rather truly personalized medicine.
I’ve previously written in a LinkedIn Influencer post about how Google scooped the Centers for Disease Control and Prevention for flu prediction almost a decade ago. Many universities and professional guilds and even practice groups have perhaps become some of the best go-to entities to understand the real-world experiences of heterogeneous populations. For example, The University of Michigan plans to invest $100 million into a big data program. The University of Massachusetts Medical School developed the Function and Outcomes Research for Comparative Effectiveness in Total Joint Replacement and Quality Improvement (FORCE-TJR), a data system that guides total joint replacement practices. The North American Spine Society has a Spine Registry that is a “diagnosis-based clinical data registry that tracks patient care and outcomes.” The American Academy of Orthopedic Surgeons has developed the American Joint Replacement Registry. The American Association of Neurological Surgeons uses the Quality Outcomes Database to collect, analyze and report on nationwide clinical data from neurosurgical practices. And the list of registry developers is growing to include psychologists, physical therapists and other healthcare providers.
As I have discussed in a geek.ly article, I’ve established one of these registries in our company that focuses on outpatient orthopedic rehabilitation cases. Each registry application goes through an vetting process at ClinicalTrials.gov and if approved, is added to Agency for Healthcare Research and Quality’s Registry of Patient Registries. Then clinicians and researchers have access to and can benefit from the clinical trials performed by other groups, or they have visibility into outcomes of certain interventions conducted in more “real-world” clinical settings. This also allows for research to be leveraged much more broadly than ever before and for clinicians and researchers to test hypotheses without incurring the time and expense of conducting primary research or doing their own data collection.
While super-sized Ns make me giddy, the problem with big data is that it generalizes to no one since it’s predicated on everyone. The growing “N of One” movement falls short in the opposite way. So I’m hoping that we can finally be able to meaningfully impact public health via the PMI. (I also recommend that readers interested in this area take a look at the brief article in the March 2018 issue of the Harvard Business Review by Making Better Use of Health Care Data by Hsu and Griese on a real world example of success.)
As the information that feeds AI and ML and deep learning (DL) comes from a number of often diverse places the internet of things (IoT), smart devices, wearables, apps and the like, all can produce feeds. In a report by IQVIA Institute, it was found that there were more than 318,000 health and wellness apps on the market, with a growth rate averaging about 200 a day. A promising relatively new aspect is that medical apps are increasingly being vetted via clinical trials, and IQVIA noted that there were about 860 such trials currently underway.
But clinical trials are not easy, inexpensive, or fast. Joseph Smith writing in STAT News said “One of the big promises of digital health is the speed with which it can transform health care delivery. The FDA has recognized that a different regulatory process is required for digital health solutions, signaling that regulation may be willing to move at the pace of innovation. But the full impact of digital health innovation is seen only with broad adoption of truly valuable solutions, and that rightfully requires evidence. Yet the best way to gather that evidence—the carefully conducted, prospective, randomized controlled trial and its subsequent publication—is currently ill-suited for the typical digital health startup.”
The (New) Chain Gang
There is are two sacrosanct aspects of healthcare that lies in ethics codes and oaths as well as the state and federal laws, and they are privacy and confidentiality. The decades old Health Insurance Portability and Accountability Act came about as a way to “ensure appropriate protection of electronic protected health information.” Today, health records are at risk from various villains for numerous nefarious reasons, which often test, and frequently break, even well encrypted and thought to be secure systems. I interviewed the CIO of the Cleveland Clinic, Edward Marx for my podcast and also wrote a piece for LinkedIn Influencers about such concerns and the then nascent idea of applying Blockchain approaches as a solution to securing electronic medical records (EMRs).
Such theft is serious and big business. You won’t be surprised for me to tell you that EMR theft is more complicated than situations like credit card theft because Visa can freeze accounts or cancel cards in the case of breach, but healthcare organizations or medical practices have no corresponding strategy for addressing the negative repercussions their patients face in the event of fraud or theft.
For small practices, those with less than 75 employees, products and services such as HITRUST Cyberaid may be of help. Although it’s unclear how much of a threat hacking is to small practices, when the Health Information Trust Alliance performed test on small and medium-sized hospitals, they found malware at over half of them. People like Marx and Peter Nichol argue that Blockchain technology (which is most famously used as the Bitcoin ledger) could revolutionize healthcare data security. However, it’s not an easy-button fix either due to the selection of a protocol, creation of Blockchain regulation, unknown costs, and limitations of the technology itself.
Show Me the Money, Part 2
A dear friend and behavioral healthcare policy wonk, Monica Oss, recently opined about what seems to be the evolution of telehealth/telemedicine as “…now part of a much broader category of tech-enabled therapy services…(to be known as) ‘virtual health’ services ranging from virtual online provider networks; asynchronous ‘store-and-forward’ diagnostic services; and automated programs like eCBT and mindfulness apps.”
This caused her to wonder about how payers/insurers view this evolution. Her thinking is that if this is a harbinger of a true Holy Grail of whatever value-based care is, then guess what? Traditional, dare I say, orthodox psychotherapy service provision may actually be disrupted. It’s a hard argument make in support of the good old “face-to-face 50 minute hour” of psychotherapy when there are published, peer reviewed randomized control trials that some digital therapeutics and other virtual approaches have outcomes and satisfaction on par with the old school orthodoxy.
And believe you me, what gets reimbursed for is what will be provided.