GAFAMS, STARTUPS & INNOVATION IN HEALTHCARE by PHARMAGEEK

One of the fastest-growing parts of the economy in the last ten years has been healthcare, and in light of the growing threats of pandemics like the coronavirus outbreak, the industry is set to rise once again. To stay ahead of the curve in demand for healthcare services and solutions, organizations worldwide are turning to advanced techniques like AI, machine learning, and Big Data.

 

AI is going to be huge in healthcare. According to Acumen Research and Consulting, the global market will hit $8 billion by 2026 and there is a huge overlap of skills in AI and big data—where the processing of information is optimized to help solve business and real-world problems. AI and big data provide numerous potential benefits for individuals and companies alike, including:

Empowering patient self-service with chatbots
Diagnosing patients with faster computer-aided design
Analyzing image data to examine the molecular structure in drug discovery, and by radiologists to analyze and diagnose patients
Personalizing treatments with more insightful clinical data
Let’s take a look at a few examples of AI and big data at work in the healthcare sector.

 

AI Combats Serious Illness with Better Predictions
AI and big data provide great value when they can boost the speed with which scientists and healthcare professionals can process and utilize data. One company that is at the forefront of change in the industry is Amgen, the world’s largest independent biotech company, as highlighted recently in Wired Magazine. By combining life sciences with big data, the company can combat dangerous illnesses such as cancer and cardiovascular disease. Various examples cited include:

 

Boosting the accuracy of osteoporosis risk predictions in women, thereby reducing the risk window from ten years to two
Creating machine learning algorithms and devices to predict the risk of cardiovascular disease before it strikes
Arming clinicians with AI-driven insights into patient responses to various therapies to improve overall patient satisfaction

 

In-Patient Mobility Monitoring
The clinical staff is busy people. Take intensive care unit (ICU) nurses, for example, who often have multiple patients in critical condition under their watch. Limited mobility and cognition during long-term treatments can adversely affect the patients’ overall recovery. Monitoring their activity is vital. To improve outcomes, researchers at Stanford University and Intermountain LDS Hospital installed depth sensors equipped with ML algorithms in patients’ rooms to keep track of their mobility. The technology accurately identified movements 87 percent of the time. Eventually, the researchers aim to provide ICU staff with notifications when patients are in trouble.

 

Clinical Trials for Drug Development
One of the biggest challenges in drug development is conducting successful clinical trials. As it stands now, it can take up to 15 years to bring a new – and potentially life-saving – a drug to market, according to a report published in Trends in Pharmacological Sciences. It can also cost between $1.5 and $2 billion. Around half of that time is spent in clinical trials, many of which fail. Using AI technology, however, researchers can identify the right patients to participate in the experiments. Further, they can monitor their medical responses more efficiently and accurately — saving time and money along the way.

 

Quality of Electronic Health Records (EHR)
Ask any healthcare professional what the bane of their existence is, and undoubtedly cumbersome EHR systems will come up. Traditionally, clinicians would manually write down or type observations and patient information, and no two did it the same. Often, they would do it after the patient visit, inviting human error. With AI- and deep learning-backed speech recognition technology, however, interactions with patients, clinical diagnoses, and potential treatments can be augmented and documented more accurately and in near real-time.

 

Physical Robots Use AI Too
Robots (the physical kind) are being used today in many types of businesses, such as in manufacturing and warehousing. But, robots are increasingly being used in hospitals as well, and many are designed to leverage AI. The National Center for Biotechnology Information (NCBI) reported that physical robots are becoming more collaborative with humans and can be trained to perform various tasks empowered by AI logic. And it’s not just delivering supplies in hospitals. Surgical robots can “provide ‘superpowers’ to surgeons, improving their ability to see and create precise and minimally invasive incisions, stitch wounds, and so forth.” With AI driving their decision-making processes, robots can improve the speed and quality of a wide range of medical services.

 

Improving Population Health
Population health studies patterns and conditions that affect the overall health of groups (unlike “public health,” which focuses on how society ensures more healthy people). Big data is a massive part of this effort. A recent article in BuiltIn highlighted various companies that are leveraging big data to help healthcare organizations and researchers read the trends to improve health conditions.
For example, one company called Linguamatics in Cambridge, MA uses natural language processing to mine through unstructured patient data to detect relevant lifestyle factors and predict which patients may be at higher risk for disease. Another company in Santa Clara, CA, called Hortonworks, helps organize and integrate billions of records so that pharmaceutical companies can do better research for clinical trials, raise the level of safety, and get products to market faster.

How Big Data Can Fight Cancer
Big data technologies are also being used in the battle against cancer. As reported in National Geographic, big data technologies can process clinical data to reveal hidden patterns that result in earlier diagnosis of cancer. The earlier it’s detected, the better the chances are for treating it. Big data technologies are adept at analyzing genome sequencing to identify biomarkers for cancer, and can also reveal groups that are at particular risk for cancer and find otherwise undiscovered treatments. The most progressive companies are using big data techniques to speed their analyses and create treatments faster and with more tangible results.

Americans on average will visit a care provider about 300 times over the course of their lives.

 

That’s hundreds of blood pressure readings, numerous diagnoses, and hundreds of entries into a patient’s medical record—and that’s potentially with dozens of different doctors.

 

So it’s understandable, inevitable even, that patients would struggle to keep every provider up-to-date on their medical history.

 

This issue is compounded by much of our healthcare information being fragmented among multiple, incompatible health systems’ electronic health records.

 

The majority of these systems store and exchange health information in unique, often proprietary ways—and thus don’t effectively talk with one another.

 

Fortunately, recent news from Apple points to a reprieve for patients struggling to keep all of their providers up-to-date. Apple has teamed with roughly a dozen hospitals across the country, including the likes of Geisinger Health, Johns Hopkins Medicine, and Cedars-Sinai Medical Center, to make patient’s medical history available to them on their phone.

 

Patients can bring their phone with them to participating health systems and provide caregivers with an up-to-date medical history.

Empowering patients with the ability to carry their health records on their phone is great, and will surely help them overcome the issue of fragmented healthcare records.

 

Yet the underlying standardization of how healthcare data is exchanged that has made this possible is the real feat. In fact, this standardization may potentially pave the way for innovation and rapid expansion of the health information technology (HIT) industry.

 

Growing agreement upon a standard way to store and exchange electronic healthcare information is what made Apple’s foray into health records possible in the first place.

 

Fast Healthcare Interoperability Resources (FHIR) emerged four years ago as an interoperability standard for electronic exchange of healthcare information.

 

It is a standard framework for the sharing, integration, and retrieval of clinical health data and other electronic health information. Enough agreement upon such a standard for health information exchange has promoted modularity.

 

How modularity fast-tracks innovation

 

A system is modular when all its components fit together in a standardized way, whether physically, mechanically, chemically or in this case digitally. This standardization enables people to design one component without having to know how everything else in the system works.

 

An everyday example of this is the USB port. It is a standard cable connection interface upon which any number of products can connect—whether it be a keyboard, a charger, external memory, or any other device that can meet the specification.

 

This differs from interdependent systems, in which the design of parts are customized, nuanced, and how they work together is not widely-known. Thus, a designer has to know how the whole system works to be able to design any part of it.

 

In the case of the FHIR standard, the manner in which digital healthcare information is exchanged is modularized—the rules of the road are established and easy to follow.

 

Adoption of this bit of digital standardization, by an influential group of healthcare providers, is what allowed the third-party giant, Apple, entry into the modular electronic health records game. Even though their experience in healthcare is limited, the standard lays out the rules well enough for them (and other third parties) to participate in the HIT market.

 

We’ve learned in the past that the creation of and agreement upon standards can expand industries by creating a new ecosystem in which third-party players can add value. In fact, the preeminent example of this type of ecosystem creation is Apple itself, and their AppStore.

 

Along with their AppStore, Apple created a set of standards that specified how third-parties (from companies to individual hobbyists) can more easily create applications that make use of the information on their phone and the Internet.

 

These apps were made available to Apple’s network of users and developers were paid according to the amount of revenue the app generated Apple (based on usage).

 

Over the span of 10 years Apple has paid AppStore developers $86.5 billion (paying out $26.5 billion in 2017). The rapid expansion of the market for creating substitutable apps in return gave everyday users the ability to harness information in any number of more convenient, simple, and potentially meaningful new ways.

 

What does this relatively recent and still unfolding story mean for HIT? It means that as opposed to merely viewing your health record, standardization may also allow for the creation of new tools that actually make use of your health record in new, meaningful ways.

 

In this way, not only does modularity stand to make healthcare data more accessible to providers, researchers, and public health organizations (current consumers of health data), but to a new market—the patient. Standardization mediated by the adoption of FHIR opens up the market for innovators outside of the traditional health IT industry.

 

These new players can then compete to reach everyday people (just as app creators did on Apple’s AppStore platform), with useful tools that empower them in their struggle for health.

As ambient technologies improve, additional use cases to leverage voice will emerge – that leaves us with the question of how patients and physicians are responding to voice-enabled tools in their healthcare encounters.

For a while now, we have been watching how voice-recognition-based artificial intelligence tools can improve physician productivity, reduce burnout and improve the quality of the patient experience.

In addition, health systems have looked at voice-enabled transcriptions to identify reimbursable conditions identified during the diagnosis while ensuring that the diagnosis doesn't miss any critical health indicators.

It is common knowledge that the most significant burden for many caregivers is documenting and annotating clinical encounters in electronic health record systems; Voice recognition is one of many tools that can alleviate the problem and reduce clinician workloads today.
Voice-enabled tools fall in the broad category of conversational AI, along with chatbots and other productivity and automation tools. However, the maturity of the tools, especially in a clinical context, is a long way off from the promise of the technology.

Users of the leading voice-recognition tools acknowledge that the technology delivers better caregiver productivity. However, they also point out that ambient artificial intelligence, or the underlying assumption about software that can make sense of a conversation and provide clinical decision support in real-time, is still very nascent.

According to Stephanie Lahr, CIO and CMIO of Monument Health, voice recognition in a clinical context is complex, and Doctor-Patient encounters are hard to capture in voice recognition software.

Dr. Lahr points out that even with leading technology providers for voice-based tools, a "person behind the curtain" often interprets the conversation and separates the clinical terminology from the overall conversation.
BJ Moore, CIO of Providence Health and a user of voice-recognition tools, states: How does an AI tool pluck out the necessary components from a doctor-patient encounter and add that to the EHR while ignoring the rest of the chitchat in the room?

Big tech and voice-recognition in healthcare
Big tech firms and startups alike are keen to expand voice recognition capabilities, considering the significant potential for voice-enabled tools to improve productivity and transform patient experiences.

Amazon, Google, and Apple have all invested in consumer-facing voice applications. Microsoft, whose Cortana platform has not made much of an impact in the marketplace, went ahead and acquired voice-technology software developer Nuance for nearly $20 billion in 2021. The move essentially implied that Microsoft was doubling down on its commitment to healthcare.
Amazon, the only other big tech firm with a voice-based offering for healthcare, has deployed Alexa services in several healthcare organizations. However, Alexa uses voice in a non-clinical (or quasi-clinical, depending on how you see it) context. Amazon's recent announcements point to using voice-enablement in senior living communities and patients in hospitals to stay connected, informed, and entertained, much as consumers use Alexa today for general information.

While these solutions are not directly enabling clinicians and caregivers with diagnosis and treatment, they still have an essential role in care delivery. For example, voice assistants allow patients with routine, non-medical needs such as medication reminders, almost like having a healthcare attendant at home but using a voice assistant instead.
This brings us to Oracle, now a major new player in healthcare tech with its planned acquisition of Cerner. The news release made multiple mentions of voice-recognition software as a significant driver of productivity and reduced workloads for clinicians in the future.

While Oracle is not the first name that comes to mind when hospitals and health systems think of voice-recognition technology, its intent to bring voice-recognition technology to the Cerner platform to address clinical workloads is indicative of the perceived opportunity for voice technologies in healthcare. (Interestingly, Cerner is currently in collaboration with Nuance for its voice-enablement capabilities).
Ambient clinical computing is still in the early stages
Ambient computing using voice and other conversational interfaces is an exciting area, and several startups are getting into the field.

However, the progress towards more intelligent uses of voice recognition for clinical decision support has been slow. As mentioned earlier, separating the clinical terminology from other aspects of the conversation is a non-trivial challenge, implying that voice-recognition technology fits in well with some specialties but not others.
Regardless of the pace of adoption, most providers see a reduction in clinician burnout for those using it. Speech recognition software can transcribe encounters three times faster than a human typing into a clinical system, potentially freeing up a couple of hours a day for a typical caregiver who sees twenty to thirty patients a day.

We can only hope that we will see higher adoption rates as the technology gets better and needs less and less human involvement in reviewing the note. The entry of big tech into the voice space will hopefully result in significant new investments that will advance AI tools and intelligent automation of aspects such as coding and quality abstraction from encounter notes.

Today, a vital consideration for automation with voice-enablement and similar technologies is that it can help providers get through the high demand and low staffing levels across the board in healthcare – compounded by the "Great Resignation".
Allowing clinicians to perform the most demanding in-person work with the highest and most complex patients also means using technology to assist those patients who don't have high acuity needs – something we have seen work very effectively with telehealth and virtual consults in primary care and specialties such as behavioral health. As ambient technologies improve, additional use cases to leverage voice will emerge.

That leaves us with the question of how patients are responding to voice-enabled tools in their healthcare encounters. Early indications are that most patients accept ambient technologies because it provides an opportunity to regain the intimacy of their relationship with their provider, which was lost to the burdensome requirements of documentation in the EHR.
However, questions around data privacy and patient education about ambient technologies suggest that voice-enabled applications will need to tread carefully.

At a broad level, voice-recognition technology's true potential lies in going beyond documentation and becoming an intelligent decision support tool through effective listening for clinical indicators and proactively supporting clinical decisions.

The level of integration between emerging technology tools and core clinical platforms such as EHR is a significant factor in increasing adoption rates. Today's fundamental challenge for voice recognition in ambient computing is the same for AI applications in general in the healthcare context.

As with all new technologies, voice-enabled solutions will stand a better chance of broad adoption by addressing important and urgent problems in care delivery, which builds support from clinician owners and champions inside the organization.
There are many promising technologies emerging today that can impact healthcare. However, it is critical for clinicians and digital health leaders to recognize that no matter how good the tech, success can be elusive without organizational alignment and demonstrated performance.

New technologies also involve business process changes in addition to integration with core clinical platforms such as EHR and require effective change management approaches. Success requires alignment between the supplier of the technology and the healthcare organization's internal stakeholders and developing an end-to-end view of solving problems.

Often this means paying close attention to understanding stated and unstated needs. When all these elements come together, the digital transformation of the healthcare sector can make giant leaps forward.