sign in sign up
<<
Home , Health care, Medicine

Stanford 2018 Trends Report The Democratization of

December 2018 Stanford Medicine Foreword from Dean Minor

Data and the democratization sophisticated medical knowledge is no of health care. longer confined to the . One day, perhaps soon, this expertise will live in our smart devices—readily accessible, In last year’s inaugural Health Trends whenever and wherever it’s needed. This Report, we discussed the explosion of kind of access could have an enormously data in medicine and what it meant for positive impact on , especially , researchers, and . This for patients who lack high quality care potential of data while ensuring year, we drill down into operation and close to home. its and security at every step will implementation, exploring how this wealth be one of the defining challenges of this of information is changing traditional Today, we’re starting to see walls come new era. health care roles, the experience of down in the health care industry, allowing patient care, and access to services in the data to flow more freely to where it can But we are undoubtedly on our way to digital age. What we’ve found has vast do the most good. Realms of historically a future of care that is more predictive, implications for medicine writ large, and siloed expertise are opening up to preventive, and personalized. This is our especially for patients: all early point more and more people. Patients can vision for Precision Health at Stanford toward the democratization of health care. now get access to their personal health Medicine, one in which we build a health information. And digital tools are giving system that goes beyond curing For our second annual report, we rise to new health platforms that are and moves toward preventing it before again interviewed industry experts and increasingly useful to physicians and it strikes. It’s a future that is good for conducted a comprehensive review of patients alike. patients, physicians, and the long-term publicly available data from a variety of viability of the American health care media, analyst, and academic sources. With all of this said, we haven’t yet reached system, and it’s one that we are earnestly We also spoke to our own to gain the promised land of digitally enabled working to achieve. insight into the trends influencing their health care and to data. work and the issues they see driving the Health care lags other industries when it This report is not intended to offer future of health care. comes to data sharing and interoperability. firm conclusions, but rather to begin a Case in point: We can manage our money conversation about how we can best In short, more individuals and from mobile devices anywhere in the realize the promise information gleaned organizations now hold more data than world, but we still can’t manage our health from data offers our patients and those ever before, and the pace at which data is records in this way. It’s clear that we have who care for them. I hope you find the moving between them is accelerating. At work to do when fewer than one in three report valuable, and I look forward to the same time, new tools are now available can electronically find, send, hearing your thoughts in the coming year. that can rapidly and accurately interpret receive, and integrate patient information medical data—from imaging from another provider. to —and push insights directly to the point of care, which is less and less We must also address the very real issues defined by physical location. around privacy and with regard Lloyd B. Minor, MD to patient data. The opportunities we Dean Together, these trends have the stand to realize are huge, but the risks Stanford School of Medicine potential to bring about a future where are similarly significant. Maximizing the Stanford Medicine Table of Contents

Introduction 1 1. Data is Growing and Flowing 1 2. New and Industry Players 1

Three Pillars of Health Care Democratization 3 Intelligent Computing 3 Sharing 8 Security, Privacy, and Safety 12

The Road Ahead 16 References 17 Stanford Medicine Introduction

In last year’s Health Trends Report, care concerns will put less on multiplicity of information-sharing we took a global look at the the health care system, lower costs, relationships—one-to-many, many- emergence of big data in the clinical and improve overall. to-one, and many-to-any. and experimental fields of medicine. The explosion of data sources, This transformation puts the data types, and analysis tools 1. Data is growing— patient at the center, encouraging foreshadowed a sea change coming and flowing—across the spread of medical knowledge for health care: in the laboratory, in in unprecedented ways. It is also the clinic, in the living room. our health care challenging the health care sector to adapt. The net result is that the This year we continue to track the system faster than public now has access to medical exponential growth in the volume ever before. data, both personal and general, and utility of that data but begin in ways it never has before: 93% of to take a deeper look at how data hospitals and health systems enable sharing and its use will transform Historically, health care has operated patients to access their , research, medical practice, and the as a closed ecosystem of siloed interact with health data, and obtain role patients play in their own health , with the research health services.2 care. We have seen this year a rapid as the hub. And within increase in the flow of data and this environment, physicians have information across an increasingly served as the primary gatekeepers 2. New technologies wide range of stakeholders. More of medical information. The flow and industry players people hold more data than ever of information has gone in one before. As a result, for the first time direction: from expert to patient. are taking medical in history, there is an opportunity to truly democratize health care. Largely due to the digitization knowledge from a of patient health records—rare scale to a This health care democratization a decade ago but now nearly is characterized by two major ubiquitous—data is flowing more digital scale. factors: the distribution of data freely. Patients are now engaged and the ability to generate and with the wider health care system in The growth and spread of data have apply insights at scale. It promises more complex forms of information generated more information than any a world in which patients—armed sharing. More people are using one person could possibly interpret. with data, , and access to digital devices, from smart watches However, recent breakthroughs in expertise—can take charge of their to internet-connected insulin data and artificial intelligence own well-being and manage their pumps, to monitor and manage their (AI) are quickly overcoming this own health. health. DNA testing is increasingly challenge. Medical experts and popular and convenient—one in 25 are now training Democratization will mean that American adults has access to their that can analyze vast quantities of providers focus less attention on personal genetic data.1 Patients no data and extract insights. routine tasks and more on the areas longer get all of their health care where they provide the most value information from the . This These developments, though recent, and find the most satisfaction. And one-to-one relationship between have already attracted tech firms to individuals managing their health expert and patient is giving way to a the health care sector in a big way.

3 Stanford Medicine

Tech companies see opportunity in the need to bring health care into the digital age, and they are investing heavily to address these Advances in Digital new markets. Health Taking Shape

Many roadblocks and inefficiencies at Stanford Medicine: that have encumbered this digital transition are gradually being phased out and eliminated. Although we are not yet in an ideal state of 1. A Stanford-developed AI 3. Stanford and Apple teamed up democratization, early signs of where for radiology can on the Apple Study, we’re headed are loud and clear. We reliably screen chest x-rays must now take steps to address the recruiting 400,000 participants challenges that stand in the way of for more than a dozen types in a year with the goal of that ideal state. of disease, and does so in less detecting atrial fibrillation in time than it takes to read Apple Watch wearers. Key Questions Going Forward this sentence.

• What are the key opportunities presented by democratization? 2. Stanford Medicine’s Second • How do we realize this vision? 4. Stanford Medicine’s Center • How do we ensure that the benefits Opinion platform, powered by for Clinical Excellence play out equitably for all people? Grand Rounds, allows patients (CERC) is collaborating with • What safeguards are necessary? to get a second opinion from Stanford’s Artificial Intelligence The 2018 Health Trends Report will a Stanford Medicine Lab via the Partnership examine these questions across without leaving home. in AI-Assisted Care, with three domains that represent some current pilot tests focusing on of most important trends taking computer vision technology. shape in health care today.

1. Intelligent Computing 2. Sharing 3. Security, Privacy, and Safety

4 Stanford Medicine Three Pillars of Democratization in Health Care

1. Intelligent Scope and Scale of a diagnosis in a matter of minutes. And as the AI Market they continue to do this, they will become Computing more accurate over time. In some cases, The AI and learning market is these algorithms are already capable of Algorithms are fueled by data; the more immense and fast-growing, especially providing diagnoses more accurately than they have, the further they can go—and in health care, where it promises steady specialists working alone. the more we can learn. An increase in data improvements in cost, access, and quality. is fueling better algorithms, creating a The size of the AI health market is expected The health care industry appears ready virtuous cycle for the industry. As analytic to reach $6.6 billion by 2021—that’s a to embrace this development. In a survey methods continue to improve, these compound annual growth rate of 40%. of senior health care decision-makers approaches will have a compounding That means that from 2014 to 2021, a regarding AI maturity levels by industry, effect on the industry, in both the period of just five years, the health AI health care ranks higher than retail and near- and long-term: increasing efficiency, financial services, with pharmaceuticals improving predictive capabilities, enabling and being considered the greater personalization, and democratizing Health AI Market Size 2014 - 2021 most AI-ready.4 access to this enhanced care. As health care leaders adapt to this technology, they $600 M $6.6 B Near-term and Long-term must take into account the practical and Impacts ethical implications it poses.

11x AI has significant implications for the health care sector, perhaps more so than other industries. The McKinsey Global Institute estimates that 15-20% of the [In] the next 2 to 5 2014 2021 health care market has the potential to be years there will be a lot impacted by AI, making it one of the most Acquisitions of AI startups are rapidly affected sectors. 5 of attention on data increasing while the health market is analytics and artificial set to register an explosive CAGR of intelligence that will 40% through 2021. Intelligent computing has the potential to add a significant amount of value by allow us to learn from improving quality of care, specifically by large observational Source: Accenture (December 2017). Artificial Intelligence in Healthcare. allowing for quicker and more accurate datasets. It will teach us diagnoses, higher quality treatment plans, what we do today which and new ways of managing processes. we don’t understand, market will have grown more than 10x.3 Some of the most anticipated advances: how varied our practices This surge of health data has led to the development of highly sophisticated Prediction: are, and what outcomes algorithms that are able to parse Intelligent algorithms have already shown we get with them. information at an unprecedented scale. promise in predicting and identifying - Dr. Paul Tang, VP, Chief Image recognition algorithms, for example, public health threats as well as outcomes Health Transformation Officer, can now process hundreds of thousands for at-risk patients in the hospital. As they IBM Watson Health of images, such as x-rays, photos of skin continue to improve, medical professionals conditions, or images of tumors and reach will have a powerful tool to provide patient

5 Stanford Medicine

care that is more precise, immediate, By ensuring that the right care takes place By one estimate, AI could help reduce and preventive. The Sloan Kettering at the right time, and in the right setting, health care costs by $150 billion by 2026 Institute estimates that physicians use advances in data science and AI have the in the U.S. alone.12 From a business only one-fifth of available trial-based potential to significantly slow the onset of perspective, intelligent computing creates information when diagnosing and treating chronic disease at a population level and an opportunity for industry players to patients.6 When effectively utilized, prevent some forms of disease altogether. optimize savings and profitability while algorithms could process vast amounts Even small improvements—powered by still taking advantage of growth. Various of data and provide a larger picture of intelligent computing—could result in large AI applications are expected to bring medical evidence to a doctor within economic savings for all stakeholders in billions of dollars in near-term value to seconds. the health care system. This incentive is the economy, the most valuable being likely to encourage partnerships and more robot-assisted ($40 billion), Personalization: value-based purchasing models in the virtual assistants ($20 billion), Physicians will soon be able to customize future, further bolstering the industry’s and administrative workflow assistance treatment plans, and even drugs, to shift to preventive care.11 ($18 billion).13 These applications will only individual patients based on a complex become more accurate and efficient over interplay of factors such as their time, implying even greater value down genetic makeup, habits, and digitally the line. monitored biometrics, to name a few. As this continues to evolve, standardized health care will increasingly give way to personalization. Research shows that tailored treatments could reduce health expenditures by up to 9% and add up to 1.3 years to average .7 Top Perceived Advantages of Using AI for Health Care Access: Virtual and mobile care now serve as a primary health resource for many patients. Health care would be easier and In fact, over the next five years the global quicker for more people to access 34% mobile health market is expected to have a compound annual growth rate of 29%.8 When asked to name the top advantage Faster and more of AI in health care, over a quarter of accurate diagnoses 31% consumers cite having their own health care specialist available at any time, Will make better treatment on any device.9 recommendations 27%

Cost and Health Outcomes: Like having your own health care Today, 90% of the U.S.’s $3.3-trillion annual specialist, available any time 27% health care spend is on people with and on any device chronic health conditions. 10 Source: PwC (November 2016). Survey: The new imperatives for health.

6 Stanford Faculty on the Benefits of Intelligent Computing across Stanford Medicine Areas:

You have these large genetic associations between a phenotype and a certain genotype, and it is very hard for the human to figure out which ones are real and Estimated Annual Benefits of AI Applications by 2026 (Billions) what they’re telling us. But there have been machine Robot-Assisted Surgery $40 learning approaches applied to genomic data sets that are Virtual Nursing Assistance $20 now suggesting candidate Administrative Workflow Assistance $18 disease teams. Fraud Detection $17 - Stanford Faculty Dosage Error Reduction $16 Connected $14 Clinical Trial Participant Identifer $13 Preliminary Diagnosis $5 Automated Image Diagnosis $3 Cybersecurity $2 On the more clinical side, there’s been a proliferation Source: Accenture (December 2017). Artificial Intelligence in Healthcare. recently of National Cancer Registry based studies that analyze data from tens or New advances in intelligent computing Specifically, AI can provide relief to hundreds of thousands of will reduce the cost of treatments and physicians’ already hectic work days, six patients. That also has let us procedures that were previously seen hours (on average) of which are currently ask some questions that we as inaccessible. The price of genome spent on (EHR) could never have asked with sequencing continues to fall while at the data entry—a task that can be easily prospective studies. So, I 15 same time the actionable information handled with automated methods. By think that big data has been available from combining the results of making clerical processes faster and more a huge impact. genomic tests with other key determinants efficient, AI stands to play a meaningful of health is growing. Treatments for role in reducing physician fatigue and - Stanford Faculty complex such as cancer are allowing health care professionals to focus being made increasingly more specific more time on their patients. as a result of information obtained from populations of patients with comparable characteristics. Greater specificity AI in the Clinic typically means less trial and error as well as better outcomes. Various specialty areas within health care have already experienced the When you look at cells or Additionally, AI holds the potential to benefits of intelligent computing. neurons under the , address a significant challenges in modern For example, AI can be used to assist there are a lot of features medicine: physician burnout. A recent radiologists in identifying abnormalities going on in those cells and the Stanford Medicine study found that a such as heart disease and cancer and in human eye and brain can only majority (55%) of physicians conducting image guided procedures really detect a few of them. report symptoms of burnout—or, and . The proper deployment There have been exciting exhaustion, cynicism, and feelings of of algorithms to such processes has the developments where machine reduced effectiveness. The study also potential to greatly improve the accuracy learning algorithms can detect notes that burnout influences quality of of screening and diagnostics. things and patterns that we care, , turnover rates, and can’t really see with our . patient satisfaction. 14 - Stanford Faculty 7 Stanford Medicine

International Competition Publications on AI Research The is not the only market looking to increase capabilities in AI and China 41,000 intelligent computing. China has become United States 25,471 progressively competitive with the U.S. 10,120 in terms of investment in AI applications and solutions. China leads the world in the 7,957 number of research publications on AI.16 6,221 The country’s intense focus on this area 5,227 has resulted in the rapid development of new health care diagnostic tools. 2,458 Singapore 2,432 The U.S. leads the global AI sector in Hong Kong 2,205 many other respects: in average years of 1,685 experience of data scientists (majority have more than 10 years), number of AI 1,537 patent applications (15,317), number of workers in AI positions (850,000), and Source: The Times Higher World University Rankings (May 2017). Which countries and percent of AI investment coming from are leading on AI research? private sources (66%).17

Ethics of Automation Top Perceived Disadvantages of Using AI for Health Care Despite the many perceived benefits of AI in health care, public opinion splits on some sensitive questions about its role in If something unexpected is found, don’t delivering care. In a recent global study of trust robots to make decisions on what to do 47% consumers, just over half of respondents indicate that they would be willing to let an People need the human touch when it AI “doctor” perform an eye exam, while the comes to their health care 38% majority say they would be opposed to brain surgery by an AI-driven machine.18 Only a human health care professional can make the right decisions 36% Less invasive, more routine procedures (e.g. look beyond the data) and processes seem to be more We don’t understand this kind of acceptable roles for AI in patients’ technology enuogh to know if it can benefit minds. But when it comes to patients’ or be dangerous in health care 30% biggest concerns about AI in health care, nearly half lack trust in robots to make the right decision when faced with an Source: PwC (November 2016). Survey: The new imperatives for health. unexpected result.19

AI causes concern in the job market regardless of industry, with increasing health care sector, with studies showing abilities or skills, or even losing control apprehension over labor displacement, nearly half of patients and caregivers over machines.20 worsened income inequality, and express concern over job losses caused by a potential rise in unemployment. AI. Similarly, about a third worry that AI will Despite the perceived ability of health Apprehension is significant throughout the result in losing certain

8 Stanford Medicine

Automation Improves Corporate Performance in Ways Beyond Simple Labor Substitution care roles to be automated, studies show that the health care industry itself Relative weight of performance gains vs. labor substitution is the fourth lowest when it comes to job automation potential, suggesting that Oil and gas (lobal, upstream, non-mature mega fields) 85 15 these fears are largely unfounded. In fact, Retail AI could have a net positive impact on (U.S. grocery stores) 32 68 the health care . Research shows that performance gains from AI are Health care (U.S. emergency departments) 30 70 sizable in comparison to the projected labor impacts.21 Aircraft maintenance 34 66 (Global commercial aircraft maintenance services) Intelligent computing and AI should not Mortgage origination 12 88 be viewed as a means of replacing doctors; (United States) the human element is an important Automotive keystone in -patient (Larger redesign or new development) 86 14 relationship. Research supports the Pharmaceutical fact that most patients are more (Research an development) 96 4 comfortable with humans managing Marketing their health; 56% cite lack of human (Consumer marketing ) 90 10 oversight as a concern about virtual nurse assistants. More than a third of patients Performance gains Labor substitution are concerned that without human Source: McKinsey Global Institute (December 2017). Jobs Lost, Jobs Gained: Workforce Transitions in a interaction, their Time of Automation. would not know them personally.22

Percent of Roles That Can Be Automated by Sector Using Current Technologies

Hotel/Food services

Manufacturing The potential for AI to Transport positively impact our practices and make Agriculture radiology professionals Retail even more valuable to Mining our patients and health Construction systems is enormous, Utilities

but to be effective the Finance/Insurance design and development Real Estate of these use cases must Health care have substantial input from radiologists. Information -Bibb Allen, MD & Keith Dreyer, Management DO, PhD, Journal of the Education American College of Radiology 02 04 06 08 0 Source: McKinsey Global Institute (December 2017). Jobs Lost, Jobs Gained: Workforce Transitions in a Time of Automation.

9 Stanford Medicine

they are not native English speakers and industry is unable (or unwilling) to are slower to respond to test prompts as share information and collaborate with When it comes to a result. others. Fundamentally, democratization collecting analysis of necessitates openness, which invites Quality and safety checks should be technical and cultural changes that must data, software can do built in well before data is used to inform be navigated. it better than a human health decisions. The Framingham Heart being, but I also still Study in Massachusetts Improving Data Sharing at strongly believe in the is one classic example of health care a Basic Level value of having a human data potentially being skewed by being examine the certain factors—in this case, most study For health care to be truly democratized, participants were white New Englanders.24 information must flow freely between and patient. The study remains one of the most famous among various participants in the system, AI can be incredibly and influential in including health care providers, patients, helpful to doctors to epidemiology and raises important technology providers, insurers, and others. really elaborate on this concerns about data objectivity. While this is happening to some degree, enormous amount of data the health care industry has a long way to that we are collecting, Issues with data objectivity affect more go in order to reach an ideal state of data than just cardiovascular epidemiology. sharing, and, ultimately, an ideal state of but I don’t think we will Data shows that many psychological democratization. replace our roles. and studies comprise - Stanford Faculty participants that are overwhelmingly The foremost hurdle, and perhaps the Western and educated, and from most trying, is interoperability: the industrialized, rich, democratic countries. capacity of of health information Further, 67% of American systems to work together within and studies are conducted at universities using across organizational boundaries in order AI can also have unintended college students, bringing age bias into the to advance the effective delivery of health consequences. Algorithmic bias is mix.25 As we transition into the age of AI- care data and information for patients, becoming a major concern in this critical assisted health care, software developers, practitioners, and providers. moment in the of intelligent practitioners, and patients must be on the computing. If biases built into health care lookout for subjectivity in the data being Health care lags behind other industries algorithms (or those that evolve within the used to make life-altering health care when it comes to data sharing and program naturally) go unchecked, they decisions. Without inclusivity, there cannot interoperability. We can send text could jeopardize quality of care, and even be a democratization of health. messages between different networks and patient safety. access our money from mobile devices all Intelligent computing and AI have the over the world but cannot access all our A seemingly simple test to screen for potential to be a major pillar in the own health data in a central location. Alzheimer’s disease may boast 90% democratization of health care, but accuracy but miss the mark for someone only if the industry is prepared to address Currently, there is no central, shared from a different country or region than the challenges. database to reference all, or even the the computer scientists who developed majority of, health care data around a the test. That was the case for Toronto- specific issue or patient. This limits the based Winterlight Labs, which develops 2. Sharing quality of care physicians can deliver auditory tests for neurological diseases. to their patients, as some of the most After publishing initial research on their Intelligent computing will allow health essential data points may be shielded technology, they found that their test only care practitioners to generate and behind virtual walls or slow-moving worked for English speakers of a certain apply insights at a scale that has never processes. For instance, fewer than one Canadian dialect.23 A participant’s results before been possible. Yet this potential in three hospitals can electronically might indicate Alzheimer’s simply because could be constrained if the health care find, send, receive, and integrate patient

10 Stanford Medicine

information from another provider. two organizations have the exact same health care. For example, Apple reports Instead, most rely on paper or fax when standards governing how data is entered that 142 health care institutions now share sending a care summary for patient into their EHRs, let alone individual medical records data with their Health discharges or referrals.26 departments and . The result app.31 These partnerships promote data is large disparities in data quality and interoperability in a way that could lead A recent study revealed that less than 30% consistency. This “messy” data makes to more comprehensive and accessible of U.S. hospitals were able to meet the four it extremely difficult to share usable patient records across the board. key metrics necessary for interoperability: information between organizations that data integration, reception, distribution, can be subsequently analyzed. Companies and funding.27 To realize the benefits of like Flatiron Health are working to resolve democratization, the industry must make this by aggregating, cleansing, and A Few of the Health Care a shift toward increased integration and anonymizing data from various health Systems Linking Medical Records practices that promote interoperability. care platforms to make it transferable for to Apple Health: research purposes.30 Application programming interface (API) • Stanford Medicine technology has the potential to advance interoperability in health care, and by • Geisinger extension, accelerate democratization. APIs, which are commonplace in • Johns Hopkins Medicine other sectors of the economy, help to standardize the way applications interact • Medstar Health and access data from one another. In health care, APIs have been slower to take • Cerner Health Clinic hold, but most agree that the greatest The biggest problem is opportunity to use APIs in the near term that our data are not • OhioHealth is to foster and enhance data sharing prepared in a way that between the various EHR systems that allows us to even make • Adventist Health System exist in the market today. sense of it. Once the data One such solution is Health Level 7 are readily analyzable, International’s (HL7’s) Fast Healthcare frankly, the majority Interoperability Resources (FHIR), a of the critical clinical standards framework for data formatting questions can From end-to-end, the experience of and an API for exchanging electronic be addressed. health care is being bolstered by new health records. The FHIR framework, and non-traditional partnerships. That which is available to the public, has - Amy Abernethy, Chief Medical includes patient transportation, where a grown increasingly popular.28 In practice, Officer/Chief Scientific Officer & flurry of recent activity has taken place. it encourages the transfer of data by SVP Oncology, Flatiron Health For example, Lyft has partnered with EHR reducing the time and effort it takes company Allscripts to provide patients application developers to connect to EHR with rides to appointments. Uber similarly systems and exchange health data. Tech gives doctors the option to order rides for companies like Microsoft have already patients rather than using taxi vouchers. begun integrating FHIR technology into Sharing with New Market Even the more traditional automotive their cloud-computing services.29 Entrants companies are entering the health care market. In April, Ford launched an on- EHR data preparation represents an Increasingly, tech partnerships are driving demand medical transportation additional hurdle for interoperability. No a significant portion of the innovation in for non-emergency medical purposes.32

11 Stanford Medicine

Hospitals were once the center of the Sharing in Discovery health care universe. Now, organizations that did not start in health care are The democratization of becoming increasingly prominent in could also bring advantages, exemplified the industry, with 42 of the Fortune by new kinds of organizations entering the 50 companies engaging in health care field as well as the emergence of “citizen Headlines from the past year to some extent (up from 38 in 2013).33 science,” enabled by crowdsourcing announce new collaborations Moreover, 20% of consumers in 2018 websites. These developments represent between health care providers reported having been treated at a health a cultural shift taking place in research and large corporations, with clinic in a retail setting.34 toward the notion of “discovered together, projects ranging from biotech to deployed everywhere”. wearables and transportation. The continued challenge for the once-cloistered field of medicine will Technology companies are increasingly be determining how best to share in the focused on fostering innovation in health responsibility of patient care with outside care. Alphabet, Microsoft, and Apple Comcast partners with organizations and how to collaborate collectively filed more than 300 health Independence Health in ways that actually lead to improved care patents between 2013 and 2017.35 to create outcomes for patients. This is still a Recent health care patents by these company relatively new concept, for all involved. and other tech companies focus most on data storage and processing-related developments as well as wearable Ford launches on-demand technologies.36 medical transportation service Crowdsourced health research is a powerful example of democratization in This can be good for the health care since anyone can contribute Lyft partners with Allscripts field and can be good for to crowdsourcing studies regardless of background or qualification. for patient rides the society. It will bring attention to health care, It starts with data—as data becomes Doctors can now call Uber the fact that new players readily available online, more information for patients, ditching taxi come in. It will create is being shared freely across groups vouchers interest for investors to and channels. Kaggle, for example, is a engage in the health care crowdsourcing platform for data science projects, and aggregates datasets for system and, hopefully, public competitions. The Heritage Provider Amazon, Berkshire, J.P. to engage in a business Morgan partner to cut U.S. Network hosted one such competition model that will lower health care costs on Kaggle in which participants used health costs and deliver de-identified hospital data to develop significant advancements an algorithm capable of predicting and preventing unnecessary hospitalizations. Apple and Stanford and breakthrough Medicine fire up Heart for patients. Medical researchers are also beginning to Study to spot irregular - Stanford Faculty call upon the public to help brainstorm beat, atrial fibrillation solutions to scientific and medical problems. Eterna, a citizen-science crowdsourcing platform that was co- developed at Stanford, is another example of such efforts. It challenges users to work

- 12 Stanford Medicine

through medical issues ranging from the development of a new diagnostic device Health Care Patents Filed by Tech Companies 2015-2017 to detect tuberculosis to improving the of CRISPR editing—all by Other playing visually based puzzles online. Technology Community biolabs, like other Battery-Based crowdsourcing platforms, bring members Technology for batteries and charging 13% Data-Related of the public into the world of medical Patents related to experimentation. These communal labs 5% data storage, 30% processing, focus on biotech applications and refer to Smart Home protection, customer their work as “do-it-yourself ” or Processes and 8% access, and analysis devices to connect “garage biology”. These efforts have grown the home or office into physical operations and facilities 10% across the globe in which members of the community can participate freely. Touchscreen Haptic medical % 23% Wearales devices and force 11 Body and eye sensing technology wearable devices, pocket sized devices, and personal health devices

Image Processing Technology and methods to capture, process, and analyze images

Source: U.S. Patent and Trademark Office (2018, 2017, 2016).

Prevalence of DIY Biology Groups

USA 50 Europe 40 Asia America 30 Australia and New ealand Africa 20

10

0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Source: Brookings (October 2017). Do-it-yourself biology shows safety risks of an open innovation movement.

13 Stanford Medicine

Prevalence of Recent Significant Security Incidents in Past 12 Months

Recent Don’t Recent security Know security incident incident 3. Security, Privacy, NO YS and Safety

In an increasingly democratized world, public and private payers, research more data is in motion—and when data institutions, developers, is in motion, it is at risk. Patient privacy software companies, and a diverse and 3% and safety are, and always have been, widespread patient population. With the cornerstones of health care. As health data addition of intelligent computing systems becomes more fluid, security practices to this dynamic landscape, it becomes 21% around that information should be held increasingly unclear which parties own to the highest standard. In addition to the which data, even if doctors are still the safety of patients, at stake is health care’s ones providing care. ability to benefit from the advantages that digitization has brought to virtually every Uncertainty around ownership brings other sector of the economy. to light questions of confidentiality. The conventional understanding of 76% confidentiality as it relates to EHRs, for example, requires that a physician withhold information from the medical Distribution of the Benchmark record in order to maintain doctor-patient Sample by Root Cause of the We all benefit [from] the confidentiality. Some of the greatest Data Breach robust sharing of our perceived benefits to be gained from AI in health data, however, imply the need Malicious System Human data, because insights or glitch error to compare the totality of a patient’s criminal that may benefit all of us against large volumes attack can be garnered. of other patients’ data. The concept of confidentiality and ensuring patient As we all know too well, privacy and safety will need to be re- there are also nefarious examined within this kind of environment. forces at play who aren’t Patient safety should remain at the thinking about how [to] forefront of industry priorities, for good use [data] for insights into reason: if protections erode, trust in these human health, but how to medical advances will dissolve, greatly use it for [manipulation]. limiting their benefits. - Stanford Faculty 24% 24% Data Breaches

Health information security professionals report a high number of breaches occurring each year, with more than three Ensuring Patient Protection quarters citing a security incident in the past year.37 Further, a majority of breaches 52% The health care industry is incredibly in health care are caused by a malicious complex, composed of large health or criminal attack (52%) as opposed to a Source (top): HIMSS (2018). HIMSS Cybersecurity systems, single physician practices, system glitch (24%) or human error (24%).38 Survey. Source (bottom): Ponemon, IBM (2017). Cost of a Data Breach Study: Global Overview

14 Stanford Medicine

Data breaches are expensive and tend Seven Potential Costs of a Health Care Breach to cost more in health care than in any other industry. Health care breach costs surpassed the next industry (finance) by 39 $135 billion in 2017. Breach costs have Forensics $610 k gone up over time as well, meaning they are likely to only get more expensive.40 Notification $560 k Data breach costs can be quantified by a Lawsuits $880 k variety of factors, including lost business, HIPAA fines, lawsuits, and forensics Lost Business 3.7 M spend. But this still doesn’t reflect the Brand Value $500 k reputational and trust impact of lapses in data security. HIPPA Fines $1.1 M Post-Breach Costs $440 k Health Care’s Current Security Posture—Bright Spots and 0 1,000,000 2,000,000 3,000,000 4,000,000 Current Gaps

Source: Ponemon, IBM (2017). Cost of a Data Breach Study: Global Overview The health care industry generally recognizes the need to invest in cybersecurity.41 In a majority of U.S. health care organizations, funds are being may not be investing enough. Times have Earning (and Maintaining) allocated specifically for cybersecurity clearly changed with respect to innovation, Public Trust purposes. That being said, more than a but security practices have not—in large quarter of organizations do not have a part due to spending priorities. Consumers tend to be wary of sharing specific portion of their budgets carved their personal data despite the out for cybersecurity. This suggests the growing desire for more personalized need for health care entities to become experiences.43 From a health care more intentional about their security standpoint, consumers are reluctant to practices and preparedness. share data with non-traditional health care actors; tech companies rank last among As technology companies delve deeper Concerns about data entities with which consumers are willing into health care, they, too, are expected to privacy are not being to share their health data.44 Despite being adapt to the additional security risks and taken seriously enough. the creators of such technologies, tech responsibilities that come with handling companies are less trusted to provide health data. Many companies already The safest health care virtual health care when compared to understand the significance of these risks: record was a file in a doctors and hospitals.45 , for example, recently expanded its doctor’s office, because list of HIPAA compliant products. no one was going to Although patient data is essential to Despite this focus, questions are being come in and steal it. Now, research and fueling the development raised about whether cybersecurity can with electronic medical of intelligent computing in health care, keep pace with emerging technology. In a the industry lacks standardized, explicit 2017 study, 36% of all U.S. organizations records, we know at permissions or anonymization, processes were breached—a marked increase from some point these could that would help to mitigate concerns for 26% in 2016.42 Although health care be hacked and hundreds consumers. Health care providers must organizations are investing in security, they of millions of people’s work with tech companies and other health care data will be partners to overcome these barriers released. and ensure patient trust remains at the

- Stanford Faculty 15 Stanford Medicine

forefront of their collaborations because in a highly regulated field like health care, Percent Willing to Share by Stakeholder without patient trust, little can be achieved. 79 Physican 76 Even though patients are wary of sharing 64 data with tech companies, the context of 53 Family 50 these interactions has an interesting effect 51 on trust. Studies show that more than half Insurance 37 of consumers rely on Google searches 45 Company 32 for health information, with smaller 31 subsets relying on other platforms like Research 36 Wikipedia and Facebook. 46 This points to a 25 29 divergence between the stated preference 35 of respondents in surveys, and the 18 13 revealed preference of how tech company Biopharma 17 platforms are already being used to inform 12 health decisions. 9 Government 13 9 What is clear is that the traditional Tech 8 12 physician-patient relationship remains Company 9 a vital component of health care. This existing foundation of trust should be cultivated in congruence with tech Health History Physical Activity Genetic Data partnerships to provide high-quality, high-trust patient care. Which Tech Company Would You Share Data With? Only shown to those who answered ‘yes’ to being willing to share their health data with a company at all.

Google 60

We want people to feel Microsoft 56 confident that when they disclose [data] for one 54 thing, it’s only used for Apple that thing, and the fact 50 that it is digital doesn’t Amazon 39 make it riskier to disclose. - Lucia Savage, Chief Privacy Facebook 39 & Regulatory Officer, Omada Health IBM 37

Source: Rock Health (2016). Consumer Survey Data.

16 Stanford Medicine

Trust in Organizations / Individuals to Provide a Virtual Nurse: Which of the following groups or organizations, if any, would you trust to provide a virtual nurse assistant? How trustworthy do you consider each of the following companies in providing a virtual nurse assistant? (Specific to the four technology companies provided as examples.)

EU U.S.

56 Online health care website, 15 Physican 57 like WebMD, Doctissimo 26 44 Family 37 Technology company such 14 as Google, Apple, Amazon, 20 Insurance 39 and Facebook 17 Company 12 33 None of the above 19 Research 37 19 Don’t know 8 Pharmacy 13 12

18 8 Biopharma 18 Drug manufacturer 8

Tech companies ranked by trustworthiness 1. Google 2. Apple 3. Amazon 4. Facebook

Source: Syneos Health (2018). Survey: Artificial Intelligence for Authentic Engagement: Patient perspectives on healthcare’s evolving AI conversation.

Preventing Discrimination government. We see this clearly with the and issuing guidance to modernize and Promoting Privacy European Union’s General Data Protection existing policies. (GDPR), a new regulation that Genetic testing offers many opportunities fundamentally calls into question previous State regulation is also a factor. for improving health. The information from notions about data ownership in health California, for example, recently passed these tests also has major implications care. In May of 2018, organizations with a comprehensive consumer privacy for health care ethics with imperatives business ties to the EU were required to (the California Consumer Privacy for protecting privacy and preventing comply to GDPR standards or face fines, Act, A.B. 375) that is likely to significantly discrimination. While individuals express prompting any organizations who treat affect industries relying on personal data, openness to their employer paying for patients from any of the 28 EU nations especially if that data is digital. The law genetic testing services, for example, to gain affirmative consent for any data gives California residents the right to the majority (65%) are concerned about collected from people who reside in the be informed about how companies whether the results of genetic testing will EU. Now more than ever, health care collect and use their personal data, be used to deny insurance coverage. 47 institutions will need to consider data allowing them to request the deletion flows, cross-border data transfer, privacy, of personal information, opt out of the and security to ensure their sale of personal information, and readily Guidelines and Guardrails policies are compliant with the law. access personal information that is made available to third parties. Regulators have enormous influence over In an attempt to keep pace with the the health care industry and its trajectory, digitization of health care, the FDA has What’s clear is that proactively engaging and this isn’t likely to change on the recently taken steps to regulate digital regulators and policymakers will be path to democratization. The regulatory health. The FDA’s Digital Health Innovation increasingly important to achieve the environment greatly affects investment Action Plan aims to “ensure patients trust and support needed to push the decisions: policy and regulatory factors have timely access to high-quality, safe boundaries of health care while ensuring rank third in importance when making a and effective digital health products” adequate standards are in place to protect health care-related investment decision.48 by committing to several goals, such as patient privacy and safety. The status of tech partnerships or increasing the number of digital health advances in heath, as a result, may be staff at the FDA, launching a digital health dependent on the decisions of those in software precertification pilot program,

17 Stanford Medicine The Road Ahead

Key Enablers of Democratization: As the trend of democratization continues to develop, traditional health care entities will need to make adjustments to account for new technologies, new ways in which patients will experience health care, 1. Taking advantage of innovations: 3. Building trust with patients: Adequate and new kinds of partnerships. Perhaps AI and machine learning have proven data sharing requires that the proper most of all, health care democratization benefits in health care; these and rules and guidelines be put into place other innovations will not only help the has the potential to recast the patient- to ensure the needs and privacy of the industry progress, but also ensure data is patient come first. doctor relationship, giving patients appropriately cleaned, managed, an opportunity to play a much more and shared. prominent role than they have before. If we get this right, it could be a very positive development, for both sides.

2. Collaboration: Working together with tech Over the past year, we’ve seen the future companies and other health care entities of a democratized health care ecosystem will preempt knowledge sharing and closer to reality. But there is a long strategic partnerships. road ahead. Here is a summary of the key issues that are likely to influence the course and speed at which we arrive at a more open, innovative, and equitable health care system: At the Same Time, the Obstacles to Data Sharing in Health Care Are Numerous and Include:

1. Accessibility: Consumers and health care 3. Physician burnout: Doctors are becoming providers are reluctant to share data with increasingly frustrated with inputting data into tech companies and non-traditional health EHR systems and find themselves with less care actors. time with their patients.

2. Data quality: Most data requires thorough 4. Privacy and ethics: Patients are largely cleaning and structure alignment in order to uncomfortable with their data being used be referenced and shared between systems; for research and other purposes; the there are not enough processes to ensure industry lacks any explicit permissions or cleaning gets done. anonymization processes.

If these issues can successfully be addressed, the road to democratization will achieve substantial benefits for patients, providers, and the system as a whole.

18 Stanford Medicine References

1 Technology Review (February 2018). 2017 was the universities are leading on AI research? https://www. 32 TechCrunch (April 2018). Ford Launches On-Demand year consumer DNA testing blew up. https://www. timeshighereducation.com/data-bites/which-coun- Medical Transportation Service. https://techcrunch. technologyreview.com/s/610233/2017-was-the-year- tries-and-universities-are-leading-ai-research. com/2018/04/18/ford-launches-on-demand-medi- consumer-dna-testing-blew-up/. cal-transportation-service/. 17 Science Magazine (February 2018). China’s 2 American Hospital Association (2016). Annual Survey massive investment in artificial intelligence has an 33 PwC (2018). The New Health Economy in the Age of Supplement. https://www. insidious downside. https://www.sciencemag.org/ Disruption. https://www.pwc.com/us/en/industries/ aha.org/system/files/2018-03/expanding-electron- news/2018/02/china-s-massive-investment-artifi- health-industries/health-research-institute/ ic-engagement.pdf. cial-intelligence-has-insidious-downside. new-health-combinations.html. 3 Frost & Sullivan (January 2016). From $600 M to $6 18 ARM, Northstar (June 2017). AI Today, AI Tomorrow: 34 PwC (2018). The New Health Economy in the Age of Billion, Artificial Intelligence Systems Poised for Awareness, Acceptance and Anticipation of AI: Disruption. https://www.pwc.com/us/en/industries/ Dramatic Market Expansion in Healthcare. https:// A Global Consumer Perspective. https://www. health-industries/health-research-institute/ ww2.frost.com/news/press-releases/600-m-6-bil- businesswire.com/news/home/20170627005941/en/ new-health-combinations.html. lion-artificial-intelligence-systems-poised-dramat- Robots-Enhance-Replace-Humans-Jobs-ARM-Survey. 35 Ernst & Young (2018). When the is the ic-market-expansion-healthcare. 19 PwC (November 2016). Survey: The new imperatives biggest data platform, who will capture value? 4 Infosys (2017). AI for Healthcare: Balancing Efficiency for health. https://www.pwc.com/gx/en/industries/ https://www.ey.com/Publication/vwLUAssets/ and Ethics – Survey of 1,600 Senior Business healthcare/publications/ai-robotics-new-health/ ey-when-the-human-body-is-the-biggest-data- Decision Makers. https://www.infosys.com/smart-au- survey-results.html. platform-who-will-capture-value/%24FILE/ey-when-

tomation/Documents/ai-healthcare.pdf. 20 the-human-body-is-the-biggest-data-platform-who- Syneos Health (2018). Artificial Intelligence for will-capture-value.pdf. 5 McKinsey Global Institute (January 2018). Artificial Authentic Engagement: Patient perspectives on Intelligence: The Time to Act is Now. https://www. healthcare’s evolving AI conversation. https://syneo- 36 U.S. Patent and Trademark Office (2018, 2017, 2016). mckinsey.com/industries/advanced-electronics/ shealthcommunications.com/uploads/articles/972/ https://www.uspto.gov/patents-application-process/ our-insights/artificial-intelligence-the-time-to-act- Syneos_Health_Communications_2018_Artificial_In- search-patents. is-now. telligence_for_Authentic_Engagement.pdf 37 HIMSS (2018). 2018 HIMSS Cybersecurity Survey. 6 Wired (February 2013). IBM’s Watson is better at diag- 21 McKinsey Global Institute (December 2017). What the https://www.himss.org/sites/himssorg/files/ nosing cancer than human doctors. https://www. future of work will mean for jobs skills and wages. u132196/2018_HIMSS_Cybersecurity_Survey_Fi- wired.co.uk/article/ibm-watson-medical-doctor. https://www.mckinsey.com/~/media/mckinsey/ nal_Report.pdf. featured insights/future of organizations/what the 7 38 McKinsey Global Institute (December 2016). The Age future of work will mean for jobs skills and wages/mgi Ponemon Institute, IBM (2018). Cost of a Data Breach of Analytics: Competing in a data-driven world. jobs lost-jobs gained_report_december 2017.ashx. Study. https://www.ibm.com/security/data-breach. https://www.mckinsey.com/business-functions/ 39 22 Ponemon Institute, IBM (2018). Cost of a Data Breach mckinsey-analytics/our-insights/the-age-of-analyt- Syneos Health (2018). Artificial Intelligence for Study. https://www.ibm.com/security/data-breach. ics-competing-in-a-data-driven-world. Authentic Engagement: Patient perspectives on healthcare’s evolving AI conversation. https://syneo- 40 Health IT Security (February 2018). How 8 Research and Markets (March 2018). Global Mobile shealthcommunications.com/uploads/articles/972/ Much Do Healthcare Data Breaches Cost Health (mHealth) Market - Segmented by Type of Syneos_Health_Communications_2018_Artificial_In- Organizations? https://healthitsecurity.com/news/ Service, Device, Stake Holder, and Geography - telligence_for_Authentic_Engagement.pdf how-much-do-healthcare-data-breaches-cost-orga- https:// Growth, Trends and Forecast (2018 – 2023). nizations. www.researchandmarkets.com/research/mv6s5d/ 23 Journal of Alzheimer’s Disease (August 2015). global_mobile?w=4.. Linguistic Features Identify Alzheimer’s Disease 41 HIMSS (2018). 2018 HIMSS Cybersecurity Survey. in Narrative Speech. http://www.cs.toronto. 9 https://www.himss.org/sites/himssorg/files/ PwC (November 2016). Survey: The new imperatives edu/~kfraser/Fraser15-JAD.pdf. u132196/2018_HIMSS_Cybersecurity_Survey_Fi- https://www.pwc.com/gx/en/industries/ for health. nal_Report.pdf. healthcare/publications/ai-robotics-new-health/ 24 National Institutes of Health (September 2014). The survey-results.html. Framingham Heart Study and the Epidemiology of 42 Thales (January 2018). 2018 Thales Data Threat Cardiovascular Diseases: A Historical Perspective. 10 Report. https://dtr.thalesesecurity.com/. Centers for Disease Control and Prevention (2018). https://www.ncbi.nlm.nih.gov/pmc/articles/ https://www.cdc.gov/chronicdisease/about/costs/ PMC4159698/. 43 Retail Systems (September 2018). Consumers wary of index.htm. data-sharing despite personalization trend. http:// 25 Slate (May 2013). Psychology is WEIRD. https://slate. 11 www.retail-systems.com/rs/Consumers_Wary_Of_ McKinsey Global Institute (June 2017). Artificial com/technology/2013/05/weird-psychology-so- Data_Sharing_Despite_Personalisation.php. Intelligence: The next digital frontier? https:// cial-science-researchers-rely-too-much-on-west- www.mckinsey.com/~/media/McKinsey/Industries/ ern-college-students.html. 44 Rock Health (2016). Digital Health Consumer Advanced%20Electronics/Our%20Insights/How%20 Adoption Survey. https://rockhealth.com/reports/ artificial%20intelligence%20can%20deliver%20 26 Healthcare Finance (October 2018). Interoperability healthcare-consumers-in-a-digital-transition/. remains limited, according to the real%20value%20to%20companies/MGI-Artificial-In- 45 telligence-Discussion-paper.ashx. of Medicine. https://www.healthcarefinancenews. Syneos Health (2018). Artificial Intelligence for com/news/interoperability-remains-limited-accord- Authentic Engagement: Patient perspectives on 12 Accenture (2018). AI: An Engine for Growth. https:// ing-national-academy-medicine. healthcare’s evolving AI conversation. https://syneo- www.accenture.com/us-en/insight-artificial-intelli- shealthcommunications.com/uploads/articles/972/ gence-healthcare. 27 HealthAffairs (October 2017). Progress in Interop- Syneos_Health_Communications_2018_Artificial_In- erability: Measuring US Hospitals’ Engagement In 13 telligence_for_Authentic_Engagement.pdf Accenture (2018). AI: An Engine for Growth. https:// Sharing Patient Data. https://www.healthaffairs.org/ www.accenture.com/us-en/insight-artificial-intelli- doi/pdf/10.1377/hlthaff.2017.0546. 46 Syneos Health (2018). Artificial Intelligence for gence-healthcare. Authentic Engagement: Patient perspectives on 28 Health Level Seven (February 2014). HL7 Fast 14 healthcare’s evolving AI conversation. https://syneo- Stanford University School of Medicine (July 2018). Healthcare Interoperability Resources Specification shealthcommunications.com/uploads/articles/972/ Medical errors may stem from more physician (aka FHIR®), Release 3 (STU). http://www.hl7.org/ Syneos_Health_Communications_2018_Artificial_In- https:// burnout than unsafe health care settings. implement/standards/product_brief.cfm?prod- telligence_for_Authentic_Engagement.pdf med.stanford.edu/news/all-news/2018/07/medical- uct_id=449. errors-may-stem-more-from-physician-burnout.html. 47 Wamberg Genomic Advisors (November 2017). 29 HealthIT Analytics (November 2018). Microsoft Azure 15 Wamberg Genomic Consumer Survey. https://www. Annals of (October 2017). Releases Open Source FHIR Server for Health Data. prnewswire.com/news-releases/65-of-employees- Tethered to the EHR: Physician https://healthitanalytics.com/news/microsoft-azure- would-pay-for-genetic-testing-from-their-employer- Workload Assessment Using EHR Event Log Data releases-open-source-fhir-server-for-health-data. sponsored-health-account-wamberg-genomic- http://www. and Time-Motion Observations. survey-300560945.html. annfammed.org/content/15/5/419.full. 30 Flatiron Health. https://flatiron.com/. 48 31 Brunswick Insight (January 2018). Global Investor 16 The Times World University Apple (2018). Institutions that support health records Views on Healthcare. https://www.brunswickgroup. Rankings (May 2017). Which countries and on iPhone. https://support.apple.com/en-us/ HT208647. com/global-investor-views-on-healthcare-i6917/. 19 med.stanford.edu/healthtrends