Over the course of the last 18 months artificial intelligence (AI) has matured to the point where there are several viable vendor options for nearly every use case.
AI dominated every aspect of the annual gathering of the Radiological Society of North America (RSNA18) in Chicago. Self-described ‘machine learning’ vendors with a presence on the conference floor more than doubled from 49 in 2017 to over 100 in 2018, 25 of which were first-time presenters.
I moderated a panel hosted by Life Image on practical uses cases of imaging AI and was blown away by the conversation that ensued, particularly what I learned about how the veteran radiologists feel about being “replaced.” During the question period, a senior radiologist approached the microphone to address a comment made by a more junior radiologist on the panel which he interpreted to be too pessimistic about the potential for AI. To paraphrase the elder, “Listen here sonny, you are too young to fully appreciate what you don’t know, and you don’t know how many mistakes you are truly making on a day to day basis. 1-2 percent error rate due to fatigue alone. WE NEED AI to save us from ourselves.”
Not all old school radiologists are so optimistic: “When you’re going up the ride, you get excited,” noted University of Chicago radiologist Paul Chang said during his workshop on AI. “But then right at the top, before you are about to go down, you have that moment of clarity—‘What am I getting myself into?’—and that’s where we are now. We are upon that crest of magical hype and we are about to get the trench of disillusionment… It is worth the rollercoaster of hype. But I’m here to tell you that it’s going to take longer than you think.”
Last year, the major cloud vendors each had a significant footprint at RSNA, but this year the two largest, Amazon and Microsoft, were nowhere to be found. Only Google Cloud had a significant, if smaller than last year’s, presence. Donny Cheung, one of the Google Cloud team leaders, was on the panel I moderated and his message to the imaging community could be boiled down to two words: storage and compute. No dashboards or toolkits or tensorflowing, just storage and compute, a smart and refreshing strategy amidst the obvious feature creep many other vendors suffer from.
Over the course of the last 18 months artificial intelligence has matured to the point where there are several viable vendor options for nearly every use case.
While it was surprising that Amazon had no noticeable presence, it was even more surprising to find Facebook making news on the conference floor. Facebook AI Research (FAIR) has partnered with the Center for Advanced Imaging Innovation and Research (CAI2R) in the Department of Radiology at NYU School of Medicine and NYU Langone Health to release the fastMRI, an open source dataset for training and testing machine learning algorithms to reconstruct MRI images.
This offering is roughly equivalent to similar X-Ray and CT datasets released by NIH. Given that algorithms ALWAYS significantly outperform on all metrics against the data used to train them versus new data, the industry needs independent validation of AI claims so it is unlikely that Facebook moves the needle with this offering.
PACS vendors want to get in on the AI action by positioning their existing products as AI marketplaces or platforms (Philips HealthSuite Insights, PureWeb, LifeImage, GE Edison, FujiFilm REiLI, Nuance AI Marketplace, Blackford Analysis). Nuance has shown there is a viable market for these platforms, counting 40 startups and health systems among user groups for its marketplace. There is no shortage of startups taking this approach (MDW, Envoy.ai, Medimsight, Lify, Fovia). Imaging hardware vendors refused to be left out too, with many partnering with AI vendors to embed their algorithms on the “edge.”
International AI startups, particularly from Israel, China, and South Korea, stood out from the crowd in terms of their approach to product design, but only the companies from Israel have been able to break into the US market so far. One Korean company voiced frustration with the FDA, saying it couldn’t understand what was wrong with their application. I wonder if it underestimates the importance of using data from US patients to validate their algorithms?
Not everything we learned about AI at RSNA was positive. A paper presented at the conference showed that neural networks could be used to insert malignant features into mammograms giving a false positives, and then reverse the alterations without detection. Even scarier, it took about 680 images to train the algorithm that executed the adversarial attack. Cyberattacks have been increasing in healthcare over the last couple years, but mostly just for taking data hostage and demanding ransom to get it unencrypted. This type of attack would represent a frightening new paradigm in cyber-vulnerability, and it is certainly not difficult to imagine ways this could be exploited to make money. It could be used for a different sort of ransom, with every image appearing to show cancer until a ransom is paid and the adversarial attack is reversed. Another conceivable way this type of attack could be exploited would be falsifying data for clinical trials.
Matt Guldin · 2 years ago
Chilmark Team · 1 month ago
Chilmark Team · 2 months ago
Brian Edwards · 2 weeks ago
2019 Predictions: M&A, Big Tech, and the Fate of ACOs
The Meaningful Use gravy train finally came to an end in 2018. As the strongest EHR vendors struggle to define new revenue streams, weaker ones faded from view through acquisitions or leveraged buy-out. Meanwhile, funding for ‘digital health’ start-ups continued to increase, though it likely hit the high water mark in 2018. And lest we forget, Amazon, Apple and Google continue their forays into the healthcare sector as the market is simply too big to ignore.
So what’s in store for 2019?
We brought together our analysts’ brain trust and came up with the following baker’s dozen of 2019 predictions. Over the near decade of making these annual predictions, our batting average has consistently been well above .500, so don’t ever say we didn’t give you an advanced warning on the following:
Revenue cycle management M&A activity will continue to pick up with the most notable acquisition by Optum as it doubles down on its Optum 360 managed revenue cycle business and acquires Conifer Health Solutions from Tenet.
Despite the hype and media attention around alternative primary care clinics (e.g. Oak Street Health, Chen Med, One Medical), the actual number of physical locations serving patients will remain paltry at less than ten percent of the number of retail health clinic locations.
Walgreens will likely make the first move to acquire Humana in 2019, but Walmart will outbid Walgreens to win Humana over.
The number of FDA approvals for algorithms in 2018 was impressive and shows no signs of abating. Additionally, 2020 will see a further tripling of regulatory approvals for AI.
Consumers’ use of telehealth will continue to see rapid growth and rising competition leading to significant consolidation among the plethora of vendors. By year-end, a major non-healthcare-specific consumer brand will join the mix, and the market will be down to five direct-to-consumer (DTC) nationwide brands.
By the end of 2019, every major healthcare analytics vendor will provide a cloud-hosted offering with optional data science and report development services.
Cloud offerings have become far more robust, concurrent with HCOs’ struggles to recruit IT talent and control costs. Amazon’s AWS and Microsoft’s Azure will be clear winners while Google’s own cloud infrastructure services will remain a distant third in 2019.
Laws and regulations to-date have not compelled providers to freely share data with patients. ONC’s information blocking rule, which will be released before the end of 2018, will make it easier to transfer data to other organizations but will do little to open the data floodgates for patients, clinicians, and developers.
Despite loud protests, the vast majority of provider-led MSSP ACOs will take on downside-risk as CMS shows flexibility in waivers. However, hospital-led ACOs, who continue to struggle with standing up a profitable MSSP ACO, will exit the program in 2019.
Continued changes in post-acute care reimbursement, especially from CMS, combined with the migration to home-based services, puts further economic strain on these facilities. Nearly twenty percent of post-acute care facilities will shutter or merge in 2019.
The warning signs are there over the last couple of months that the stock market has become skittish. This will extend well into 2019 (if not lead to a mild recession). It will hardly be an ideal time to do an IPO, and those planned by Change Healthcare, Health Catalyst and others will wait another year.
Elon Musk will have a nervous breakdown leading him to reinvent the healthcare system from his bed during his two-week recovery at Cedars-Sinai.
Revisiting Our 2018 Predictions
As is our custom here, we like to look back on our predictions for the closing year and see just how well we did. Some years we do amazingly well, others we over-reach and miss on quite a few. For 2018, we got seven of our 13 predictions spot-on, two were mixed results and four predictions failed to materialize. If we were a batter in the MLB we would have gotten the MVP award with a .538 batting average. But we are not and have to accept that some years our prediction average may hover just above the midpoint as it did this year.
Stay tuned, 2019 predictions will be released in about one week and it is our hope that they will inspire both rumination and conversation.
(Note: the bigger and plain text are the original predictions we made in 2017, while the italic text is our review of 2018).
Major mergers and acquisitions that mark the end of 2017 (CVS-Aetna, Dignity Health-CHI and rumored Ascension-Providence) will spill over into 2018. Both Humana and Cigna will be in play, and one of them will be acquired or merged in 2018.
MISS – neither happened. However, Cigna did pick-up PBM service Express Scripts and rumors continue to swirl about a possible Humana-Walmart deal or more recently, even a Walgreens-Humana deal.
Hot on the health heels of CVS’ acquisition of Aetna, growth in retail health reignites, albeit off a low overall footprint. By end of 2018, retail health clinic locations will exceed 3,000 and account for ~5% of all primary care encounters; up from 1,800 and ~2%, respectively, in 2015.
MISS – Modest growth in 2018 for retail health clinics with an estimate of around ~2,100 by year’s end. Telehealth, which is seeing rapid growth and on-site clinics may be partially to blame.
In a bid to one-up Samsung’s partnership with American Well, and in a bid to establish itself as the first tech giant to disrupt healthcare delivery, Apple will acquire a DTC telehealth vendor in 2018.
MISS – Apple continues to work on the periphery of care with a focus on driving adoption of its Health Records service in the near-term with a long-term goal of patient-directed and curated longitudinal health records.
Despite investments in population health management (PHM) solutions, payers still struggle with legacy back-end systems that hinder timely delivery of actionable claims data to provider organizations. The best intentions for value-based care will flounder and 60% of ACOs will struggle to break even. ACO formation will continue to grow, albeit more slowly, to mid-single digits in 2018 to just under 1,100 nationwide (up from 923 as of March 2017).
HIT – MSSP performance data showed only 34% earned shared savings in 2017 (up from 31% in 2016) and by year’s end it is estimated there will be ~1,025 ACOs in operation.
While some of the major EHR vendors have announced support for write access sometime this year and will definitely deliver this support to their most sophisticated customers, broad-based use of write APIs will happen after 2018. HCOs will be wary about willy-nilly changes to the patient record until they see how the pioneers fare.
HIT – FHIR-based read APIs are available from all of the major EHR vendors. Write APIs are still hard to find. To be fair, HCOs as a group are not loudly demanding write APIs.
True cloud-based deployments from name brand vendors such as AWS and Azure are in the minority today. But their price-performance advantages are undeniable to HIT vendors. Cerner will begin to incent its HealtheIntent customers to cloud host on AWS. Even Epic will dip its toes in the public cloud sometime in 2018, probably with some combination of Healthy Planet, Caboodle, and/or Kit.
HIT – adoption of cloud computing platforms is accelerating quickly across the healthcare landscape for virtually all applications. Cloud-hosted analytics is seeing particularly robust growth.
Providers will continue to lag behind payers and self-insured employers in adopting condition management solutions. There are two key reasons why: In particular, CMS’s reluctance to reimburse virtual Diabetes Prevention Programs, and in general, the less than 5% uptake for the CMS chronic care management billing code. In doing so, providers risk further isolation from value-based efforts to improve outcomes while controlling costs.
HIT – Awareness of the CCM billing code (CPT code 99490) remains moderate among providers and adoption is still estimated at a paltry less than 15%.
Mobile accessibility is critical for dynamic care management, especially across the ambulatory sector. More than 75% of provider-focused care management vendors will have an integrated, proprietary mobile application for patients and caregivers by end of 2018. These mobile-enabled solutions will also facilitate collection of patient-reported outcome measures, with 50% of solutions offering this capability in 2018.
MIXED – While the majority of provider-focused care management vendors do have an integrated mobile application (proprietary or partnership), collecting PROMs is still a functionality that remains limited through an integrated mobile solution.
A wide range of engagement, PHM, EHR, and care management solutions will make progress on documenting social determinants of health, but no more than 15% of solutions in 2018 will be able to automatically alter care plan interventions based on SDoH in 2018.
HIT – despite all the hoopla in the market about the need to address SDoH in care delivery, little has been done to date to directly affect dynamic care plans.
The hard, iron core of this issue is uncertainty about its real impact. No one knows what percentage of patients or encounters are impacted when available data is rendered unavailable – intentionally or unintentionally. Data blocking definitely happens but most HCOs will rightly wonder about the federal government’s willingness to go after the blockers. The Office of the National Coordinator might actually make some rules, but there will be zero enforcement in 2018.
MIXED – Last December we said, “The hard iron core of this issue is uncertainty about its real impact.” Still true. Supposedly, rulemaking on information blocking is complete but held up in the OMB. The current administration does not believe in regulation. So “data blocking” may be defined but there was and will be no enforcement or fines this year.
Providers will pull back on aggressive plans to broadly adopt and deploy PHM solution suites, leading to lackluster growth in the PHM market of 5% to 7% in 2018. Instead, the focus will be on more narrow, specific, business-driven use cases, such as standing up an ACO. In response, provider-centric vendors will pivot to the payer market, which has a ready appetite for PHM solutions, especially those with robust clinical data management capabilities.
HIT – PHM remains a challenging market from both payment (at-risk value-based care still represents less than 5% of payments nationwide) and value (lack of clear metrics for return on investment) perspectives. All PHM vendors are now pursuing opportunities in the payer market, including EHR vendors.
This is a case where the threat of alert fatigue is preferable to the reality of report fatigue. Gaps are important, and most clinicians want to address them, but not at the cost of voluminous dashboards or reports. A single care gap that is obvious to the clinician opening a chart is worth a thousand reports or dashboards. By the end of 2018, reports and dashboards will no longer be delivered to front-line clinicians except upon request.
MISS – Reports and dashboards are alive and well across the industry and remain the primary way to inform front-line clinicians about care gaps.
Arterys, Quantitative Insights, Butterfly Network, Zebra Medical Vision, EnsoData, and iCAD all received FDA approval for their AI-based solutions in 2017. This is just the start of AI’s future impact in radiology. Pioneer approvals in 2017 — such as Quantitative Insights’ QuantX Advanced breast CADx software and Arterys’s medical imaging platform — will be joined by many more in 2018 as vendors look to leverage the powerful abilities of AI/ML to reduce labor costs and improve outcomes dependent on digital image analysis.
HIT – With about a month left in 2018 the count of FDA approved algorithms year to date is approaching 30 and could potentially hit three dozen by year end. This is a significant ramp up in the regulatory pipeline, but more is needed in the way of clear guidance on how they plan to review continuously learning systems and best practices for leveraging real-world evidence in algorithm training and validation.
What do you think of 2018 for health IT?
Unlocking Healthcare’s Big Data with NLP-powered Ambient and Augmented Intelligence
It wouldn’t be a radical statement to say NLP bridges the human-computer divide more than many technologies. ROI has been elusive, leaving prospective adopters reluctant to embrace it despite the numerous opportunities for NLP-driven solutions. NLP technologies have reached an inflection point with the emergence of advanced deep machine learning methods that are on-par with humans for an ever-increasing list of core natural language skills, such as speech recognition and responding to questions. In our newest report, Natural Language Processing: Enabling the Potential of a Digital Healthcare Era, we profile 12 vendors, all with a track record in text mining and speech recognition, including 3M, Artificial Intelligence in Medicine (Inspirata), Clinithink, Digital Reasoning Systems, Health Catalyst, Health Fidelity, IBM Watson Health, Linguamatics, M*Modal, Nuance, Optum and SyTrue. Each has a reputation for delivering solutions that serve a particular set of use cases or customer groups, distinctions we capture using heat maps for each company.
NLP is particularly well suited to address two huge problems in healthcare – easing the clinical documentation burden for clinicians and unlocking insights from unstructured data in EHRs. Documentation consumes an ever-increasing portion of clinician’s time. Recent research has shown physicians spend as much as half of their work day (6 hours of a 12 hour shift) in the EMR. Another recent study showed clinicians spend two hours on clinical documentation for each hour spent face-to-face with patients. Unsurprisingly it is often cited as a key factor contributing to physician burnout. Ambient Intelligence refers to passive digital environments that are sensitive to the presence of people, aware context-aware, and adaptive to the needs/routines of each end user. The familiar virtual personal assistants (VPAs), such as Amazon’s Alexa and Google’s Assistant, are familiar examples.
Speech recognition technology is approaching 99-percent accuracy, a milestone that some argue means that voice will become the primary way we interface with technology. I am skeptical of this prediction, at least when it comes to the broader utility of voice-based interfaces for consumers. The visual display, with its links and rich media, is an indispensable element of the modern digital experience.
Smart speakers, the input device for speech recognition, are the hottest technology trend of the moment, with an adoption curve that exceeds even the smartphone (see graphic below from Activate). We expect the smart speaker to rapidly become a fixture in both the home and office setting, following a similar path to maturity as the smartphone, offering applications for consumers and enterprises.
Interest and adoption in healthcare is already apparent. In September Nuance announced a smart speaker virtual assistant that uses conversational cloud-based AI (Microsoft Azure) to engage physicians during clinical documentation. In late November a post on the Google Research Blog described internal research and a pilot at Stanford investigating the potential to use a similar smart speaker interface and Automatic Speech Recognition (ASR) technology to create a virtual scribe.
Startups are taking on this problem too. Saykara, led by former executives at Nuance and Amazon, is developing a virtual assistant similar to Google’s. The company claims to have far more advanced speech recognition technology than its heavyweight competitors. Other are developing ambient scribes to passively document patient encounters, including Suki.ai , Robin Healthcare, and Notable Health.
EHR vendors are also making investments in ambient intelligence. Epic has partnered with Nuance and M*Modal to embed their ambient scribe technology directly into clinical workflows. Allscripts and athenahealth have partnered with startup NoteSwift. eClinicalWorks has launched a virtual assistant called Eva. Eva operates is initially intended to respond to queries for things like recent lab data or past clinical note content.
Barriers remain on the road to ubiquitous adoption of NLP technology by healthcare enterprises. NLP provides HCOs a low-risk opportunity to experiment with advanced machine learning and deep learning technologies, but its not the type of technology that can be implemented optimally by just any analyst in the IT department, but instead requires specialized expertise that is in short supply. While free text and mouse clicks will dominate the clinical documentation landscape in the near-term, healthcare enterprises will soon expect their users to talk their applications.
Signaling the Tipping Point of AI Value in Healthcare
Top minds across a broad spectrum of healthcare disciplines converged to discuss artificial intelligence in Boston April 23-25 at the 2018 World Medical Innovation Forum (WMIF), which drew 1700 registrants and 169 speakers. Researchers, clinicians and industry leadership paraded on stage for a collection of panels and “1:1 fireside chats” to discuss healthcare applications of AI and related technologies, their progress, roadblocks and the promise of future value. The majority of provider participants were from Harvard teaching hospitals as Partners Healthcare runs the conference.
The healthcare innovation community is always looking for applications to illustrate the potential of AI, but also to grasp a tangible that is closer to the now rather than solely as a futuristic vision. At this point, there are many diverse applications for AI and related technologies, as highlighted in the framing of the conference, which opened with a discussion of 19 “AI breakthroughs” and closed with the “disruptive dozen” (scroll to the end of the post for listing). It was encouraging that these innovative pursuits were not limited to any small facet of care or research, but instead cover a broad spectrum of healthcare challenges.
Across the rest of the conference were overlapping discussions of the impediments and implications of advances in AI and the many applications. These AI-centric discussions can be organized into 5 themes that broadly mirror key concerns of the healthcare industry as a whole, indicative of both the hype and potential for these technologies to fundamentally change our methods for care delivery:
Unsurprisingly, the importance of good sources of high-quality data was a core theme across all panels. In some cases, the data already exists but has not yet been analyzed. Drug manufacturers sitting on mountains of clinical trial data being a prime example. New data sources like that those coming from companies like Flatiron – recently acquired by Roche – provide a fresh resource for research. All of this data, however, exacerbates challenges of large data analysis. In order to run AI and machine learning systems, the data needs to be clean. Surveys have indicated that data scientists spend 60% of their time cleaning data. The challenge before everyone includes sourcing data (as clean as possible), cleaning and making data usable for tools.
In many cases, the limiting reagent to progress is our understanding of biology – as much as or even more so than the technology. Technology, in fact, is the means to understand the biology further. Life sciences companies like Novartis, Pfizer and others are actively using, in their words, “higher order data,” including ‘omic data sources. In doing so, they hope to reach breakthroughs faster by redefining how they consider new drug discovery.
William Lane, MD, PhD delivered one of the 19 breakthroughs that opened the conference, bloodTyper, which offers a new outlook for the long-standing method of categorizing patient blood types based on the presence or absence of 2 antibodies and 2 antigens. bloodTyper uses DNA-based categorization as opposed to serologic testing alone. Whole genome sequencing costs have declined dramatically in recent years and are expected to continue to decline making this solution viable for widespread clinical adoption. Technology enables this effort, but much research needs to be done in order to illustrate specific value and bring these new methods into practice. With continued advancement of technological capability, understanding the biology of health and/or disease will continue to be a primary obstacle.
One of the consistent themes across panel discussions was the notion that medical technology, AI in particular, is futuristic, while our care system is stone-aged. The challenge of adoption and change to the care paradigm is not a limitation of technology. Panelists repeatedly remarked about how the pairing of “Jetsons technology and a Flintstones care system” would take substantial time to evolve, because change requires evidence and trust that are not ascertained lightly.
In some cases, however, evidence-based change is already upon us. A group at MGH, represented at WMIF by Erica Shenoy, MD, PhD, is using Machine Learning to more quickly identify cases of hospital-acquired infection, identifying C Diff cases full days before the currently accepted 5-day standard process. We expect to see substantial growth in research publication volume illustrating the value of AI and Machine Learning technologies in areas like this.
On the other side of the curtain, AI is changing the way that industry operates. Drug development, in particular, is undergoing a revolution of sorts. Companies like Exscientia, with diverse data science capabilities, provide life sciences partners with a way to look across different data types including ‘omic data, research text and other sources. This allows manufacturers to potentially repurpose molecules and sub-organize disease for more precise targeting.
Not all areas of healthcare are ready environments, but the ubiquity of efforts to utilize AI technologies to accelerate processes, improve accuracy, increase access, increase bandwidth and offer precise care points to a tipping point.
Another recurring theme was the impact of AI on the workforce. As in other healthcare sub-verticals, there is a large and increasing demand for data experts. Radiologists and pathologists across the conference echoed the surprisingly optimistic resolutions of a panel dedicated to this topic both with respect to the increased demand for data science professionals in all organizations as well as the potential obviation of some roles.
The future will likely position these data scientists alongside healthcare professionals as part of multidisciplinary/cross-functional teams in care and non-care settings alike. There is also an expectation that data science literacy, at least at a high level, will become a core component of education for many healthcare professionals, not just IT specialists.
The fear of the obviation of certain roles was interestingly framed. At a very high level, the introduction of AI and related technologies will exacerbate the division between the highly educated and the less educated, as Glenn Cohen of Harvard Law School eloquently pointed out. It was also highlighted, however, that a lot of the lower skilled manual roles have already been dramatically reduced with the introduction of EMR systems. The natural targets often discussed beyond lesser skilled workforce are experts in clinical disciplines of radiology and pathology. In both contexts, AI technologies have been shown to approximate or even beat the accuracy of some clinical practice by human experts. There is also a global shortage of these types of experts. The prevailing opinion expressed was that new technologies will be used to make these professionals more efficient and effective, and it is unlikely that technology will be used to truly replace the human element.
The burden of AI technologies on regulators is substantial. They need to be able to evaluate, audit and assure quality of these new technical capabilities. Linguamatics is one example of the many companies that are making progress in crafting technology solutions for healthcare while also targeting the FDA as a customer. Providing a mechanism for auditability of otherwise “black box” AI systems is a great benefit to the regulators.
Standardization is a key to enable scalability and support industry-wide progress. Consider DICOM, for example, which offers a standard for many imaging modalities. The expectation has been set that at least some standardization must be a focus of innovators so as not to run into an even greater challenge of interoperability with more complex technology. Building gold standard training data sets was a solution highlighted that begins to address this challenge with AI.
Precision medicine was an end target for many of the applications and topics of discussion at the conference. The development of a clearer, richer phenotype and genotype (or other “’omic” and new forms of data) essentially yields the potential for a digital twin for each individual patient. To accompany this, the development of precision drugs, diagnostic capabilities and other therapies is the end promise of many of the applications of AI, machine learning and other technologies.
We will only be able to deliver on these promises if our data, technology and systems for the delivery of care are able to adapt. To get there though, we need to establish trust and confidence across the healthcare ecosystem and for patients with AI and its elevated role.
Currently, there are little islands of the right components to make great progress and work is well underway. In oncology, for example, there is data access, a willingness and need from physicians and patients, a motivated industry (opportunity for profit or strategic position), research capacity and funding, payment and regulatory feasibility. At WMIF, Atul Gawande described these environments as “ready environments,” in that they are capable and motivated. The historical example he gave to juxtapose “ready” and “not ready” was the spread of anesthetized surgery over weeks or months, vs. antisepsis, which he estimated to take 20 years to spread.
Not all areas of healthcare are ready environments, but the ubiquity of efforts to utilize AI technologies to increase accelerate processes, improve accuracy, increase access, increase bandwidth and offer precise care points to a tipping point. Within the next year, I expect to see a thicker volume of applications and more importantly more examples of impact.
Predictions in Healthcare: Unpacking the Complications
Interest in predictive analytics in healthcare is on the rise. How long will someone be in the hospital? What are their chances of being readmitted? Which treatment will work the best?
On the surface, these seem like straightforward questions, but it turns out that prediction is not a simple topic. Questions about what data and effort goes into the prediction, how accurate it is, who gets to see it, and how actionable it is can all make a huge difference in its value.
As we start to develop our major Market Trends Report for Advanced Analytics, scheduled for release later this year, I thought I would unpack these questions a bit.
To start, it makes sense that the more data and the better data you can include in making a prediction, the more accurate it will be. However, you don’t always know what factors will matter the most, or at all.
If you over-collect data to ensure you factor in most everything, you unnecessarily raise the cost, time, and complexity, as data warehouses are not inexpensive and interoperability is still a struggle. If you under-collect, you might miss something important, and clinicians will continue to complain about the burden of documentation.
The recent focus on social determinants of health is promising, but it requires much more data collection and management, along with more sophisticated analytic models. Here, we welcome the contributions of AI and Machine Learning (ML) to help determine which variables to focus on.
Prediction is not a simple topic. Questions about what data and effort goes into the prediction, how accurate it is, who gets to see it, and how actionable it is can all make a huge difference in its value.
Accuracy has all kinds of complications. You’ve likely come across the difference between precision and recall.
With precision, you measure how right you were that something was going to happen when it did, in relation to how often you thought it was but didn’t. With recall, you measure how often you thought something was going to happen, and you were right, but you missed some that did happen. In the first situation, false alarms could be costly, leading to additional tests and much worry. In the second situation, missing an important prediction (such as a cancer diagnosis) could also be costly.
Fortune tellers make many predictions and eventually get some stuff right, but most of us don’t take them seriously; Chicken Little got into trouble when he cried out too many times. Conversely, someone who makes too many conservative predictions might have a lot of credibility but also miss too many opportunities.
Consider who wants to know or gets to see the predictions. Most clinicians, healthcare workers, and leadership teams want do the best they can, but negative predictions about their approaches or performance may not be met with welcome arms. Exposing their problems can lead to potential unwanted oversight by others, reduced funding, etc.
In other words, expect pushback about the quality of the data, the accuracy of the prediction, the motivations of the prediction team, and who is really responsible for the bad trends. Think about how predictions are supported and presented. Also consider the flip side: Positive predictions, if not accurate, may lull an organization into a false sense of security.
This leads to one of the most important questions: How actionable is the prediction, and what can be done about it? (This also brings prescriptive analytics into the discussion.) If the answer is “not much,” there may still be value in knowing it – but you have to brace for impact, as it could also lead to helplessness and negativity; consider the effect of telling a patient they have an incurable disease.
Better in most circumstances to focus on those predictions where positive action can be taken. Some actions might be obvious: You’re running out of medication x, so you’d better reorder now. Others might be more complex: What will be the effect of changing our portfolio mix?
Complexity of action can arise when dealing with interdependent systems, multiple parties, and lots of variables. (Welcome to healthcare!) Like the old game Whack-a-Mole, addressing one trend may lead to unanticipated or unwanted changes in others. Image how well you could play if you knew what was going on under the table. To properly act on predictions, you have to employ systems thinking.
As our data sources grow, our ability to analyze them using AI and ML advances. Given regulatory and market pressure to provide higher-quality care while reducing care costs, healthcare’s need to manage to value also continues to increase.
In the coming years, we’ll see a lot more use of predictive analytics in healthcare, and a lot more tools and vendors to support it. As we analyze these predictive tools, we will focus on how healthcare will employ them to deal with the complexities of data selection, model accuracy, user perception, and actionability.
HIMSS18 – A Cacophony, Not Yet a Symphony
My post before HIMSS talked about how jazzed (not jaded) I was to be attending my 20th HIMSS conference. Now that HIMSS18 is in the bag – what did I learn?
On Monday I presented the results of our AI survey at the Machine Learning & AI for Healthcare preconference event at the Wynn. Although there were a few hundred seats, the event sold out weeks in advance. A number of folks I knew who planned to buy a ticket at the door were shut out. So that’s a lesson – when it comes to attending hot topics, don’t procrastinate!
Keynote speaker Lynda Chin from the University of Texas compared using of AI to having a paralegal on your team – it’s someone intelligent that could pull resources together to help you make better decisions. She summed it up simply: “Machines serving humans, not humans serving machines.”
Many other speakers came from large health systems and spoke to important use cases:
It’s become a given that these leaders and their vendors use AI and use it well. My favorite from the above examples was Srinivasan Suresh, CMIO at Children’s Hospital of Pittsburgh of UPMC. His slide highlighted that, although he had no impressive AI or ML credentials, he was still able to use these kinds of tools successfully to predict pediatric readmissions due to seizures, asthma, and pneumonia, which led to more effective interventions.
HIMSS and health IT may be more of a cacophony than a symphony, but I’m glad to be in the orchestra.
AI and the cloud were key themes this year and have become mainstream topics. For our views on Eric Schmidt’s keynote about data, analytics, and AI, see our earlier HIMSS18 recap blog.
Glad I had teammates that made it into Seema Verma’s CMS keynote the next day – her announcement about patient data access, open APIs, and Blue Button 2.0 was welcome. You may recall the previous year, given the change of administration, there was little that CMS or ONC could say about anything. Although we’re seeing some progress, it doesn’t seem substantial enough to move the needle on value-based care.
A big part of my week was meeting with NLP vendors. Chilmark Research is close to releasing our major report on this topic, and it was great to get insights from more than a dozen vendors. Some of the smaller ones are highly focused on specific use cases (Health Fidelity and Talix on risk stratification; Clinithink on matching patients to clinical trials). 3M and its partnership with Alphabet’s Verily are a powerful combination on determining the “dominoes” of costs and care. Also of note: M*Modal’s virtual provider assistant and use of ambient devices, as well as Nuance’s partnership with Epic to add more conversational AI functionality. We are seeing voice assistant success paving the way to virtual scribes – those that can “whisper” in the physician’s ear will be most valuable to ensure that decision support is not bypassed by passive systems.
As John Moore posted in his earlier HIMSS18 recap, it’s sad (well, infuriating) that we still have to address interoperability. I attended two events held by the Strategic Health Information Exchange Collaborative (SHIEC), which has been successful in providing a rallying point for 60 HIEs and 40 vendors to share knowledge and provide comments to ONC regarding TEFCA and data exchange. But it only represents a fraction of the hundreds of private and public HIEs in the country, so there is still a long road ahead. A payer committee was a welcome sign that convergence was part of their agenda.
At the opposite end of the interoperability spectrum, I attended a session by Houston Methodist on body sensors, where the distances are measured in inches and the signals are often so weak that temperature or motion (such as a kicking baby) are enough to throw them off. Sensor network fusion is the frontier – the more information you can capture from more places with more context, the better. For example, one of Methodist’s use cases was rapidly predicting a patient fall.
I also met with Somatix, a small vendor with a big idea we’ve been hearing about for years – using data from wearables to track more routine activities of daily living (some of which, like smoking, are harmful). The vendor is attempting to take this to the next level with more accurate gesture detection and predictive analytics so appropriate (and even real-time) interventions can be made using specific apps. As Brian Eastwood recently posted, we’re still waiting for wearables to provide insight. I didn’t sport a wearable at HIMSS18 (I broke two and lost another in 2017), but I’m on the lookout for a good, waterproof one.
Another key area of focus for us is the use of AI to interpret digital medical images. An impressive talk by University of Virginia and the National Institute of Health included use of speech recognition (using Carestream and Epic) to embed hyperlinks of AI-recognized areas of interest into reports for the EHR. The two-year effort showed productivity improvements of 3x over unassisted analysis and reporting.
A presentation by Entlitic claimed AI-enabled “superhuman” techniques able to detect lung cancer two years sooner than existing approaches. Their solution made it easy to compare an existing case to similar cases where timelines of data showed disease progressions. The company has 65 radiologists that label their training data, claiming only 1 in 4 that apply for the job pass their test. We’ll dive into detail about these kinds of advances in our Digital Medical Imaging Report scheduled for Q4’18.
I spent time with Ambra, a major provider of image exchange solutions (others include Nuance and lifeIMAGE). Aside from the challenge of the size of medical images, it always surprises me how difficult it is to move them around and make them available despite good standards (DICOM). It was only recently that Epic, for example, addressed image exchange, and it’s not part of many HIEs. I’m glad to see we’re moving beyond the vendor neural archiving discussion and toward a focus on the cloud and useful exchange of images in clinician workflow.
I also attended half a dozen receptions during the week. The biggest was sponsored by a large consulting firm. It was an evening of fun, but it reminded me of what was right and wrong about our industry and a conference in Las Vegas – who’s really paying that bill? My last reception was with BetterDoctor, which specializes in the quality of provider directory data. It always seemed ironic to me that the most regulated profession in the world has such a problem with accurate information (retirement, credentialing, locations, and so on).
To rework my “I’m Jazzed” comment from the top with a music metaphor, HIMSS is more like a blaring of thousands of different instruments with each of the “sections” competing to be louder than the other – and the sounds of Vegas don’t help. There are many great musicians and an increasing number of duets (e.g., partnerships, ACOs), but we’re still playing off too many different pages. Adding to the problem is the conductor (the government) changing every few years.
It may be more of a cacophony than a symphony, but I’m glad to be in the orchestra. I hope you are, too.
HIMSS18 – A Bipolar Experience
A decade of attending the annual HIMSS conference and I leave both excited and depressed. Excited and enthused by meeting so many people who are dedicating their lives to affect positive change by improving healthcare delivery through IT. Depressed as yet again I find a lack of real leadership and vision among many who repeat the years’ worn phrases of interoperability, patient-centered care, reducing physician burden, and the like.
“Oh please, can’t we just get on with it,” I scream to myself.
In keeping with the bipolar theme that is HIMSS, following are my takeaways, in an up-down fashion.
Up: Anthem goes public on its deal with Epic r/e HealthyPlanet. This partnership is an exciting step in enabling provider-payer convergence wherein Anthem will embed IP (risk, prior authorization, claims adjudication, etc.) into HealthyPlanet and take HealthyPlanet to market with wrap-around services.
Down: Head-in-the-sand vendors who are entrenched in FFS model. These vendors told me point blank that the market will revert back to FFS, that value based care is DOA. Gotta wonder what they’re smoking.
Up: Telehealth going mainstream. Saw loads of examples/demos of telehealth with direct or near-direct integration to the EHR. Been hearing about the coming of telehealth since I started this company in 2007. I believe we are finally there.
Down: Almost zero discussions on managing the costs of care/cost containment. There was some discussion on reducing clinical variability – but beyond that, HIMSS was devoid of any deep conversations on this critical variable in the value equation.
Up: Clear demonstrable, scalable use cases for AI. I was particularly impressed with the work 3M has done with Verily, leveraging Deep Mind technology for specific measures. Though just released, 3M has already landed 17 provider clients and 2 payers.
Down: The preponderance of AI vendors with little sense of scaling their solution. Many of the AI vendors I talked to have ongoing projects with “Big Brand” healthcare systems. That’s great – but disturbingly, few have taken the next step to address how they plan to scale their solution within an organization for widespread adoption and use.
Up: New solutions leveraging FHIR to insert actionable insights directly into clinical workflows. This is near nirvana for me, as it gets beyond the Herculean task of interoperability writ-large and tackles those points where significant friction and opportunity exists.
Down: One policy pundit after another talks yet again about the need for interoperability. Frankly, this is no longer a technical issue. Interoperability is a policy issue and really does not belong at an event such as HIMSS – where we should be talking about the future, not rehashing the past ad nauseum.
Clearly, a lot of work lays ahead for us in the health IT arena, which provides us all meaningful work going forward. And frankly, we are in but the top of the third inning – there is so much to do, it really is an amazing time to be in the healthcare IT market.
Thankfully, we are at last moving beyond the prescriptive use of IT via meaningful use, transitioning to meaningful insights from the data we are collecting and placing into clinical workflows. There is a near unfathomable opportunity to begin leveraging clinical, genomic, and other data sets that will lead us to dramatic improvements in care delivery – improvements that are likely beyond our comprehension at this time.
Despite some of my downer moments at HIMSS18, I could not be more excited for what the future holds for us as an industry – and, personally, in how even I and my care team will leverage new insights to more effectively and efficiently manage my own condition.