Improving Healthcare with Technology Depends on Faster Adoption of Innovation
When Electronic Medical Records (EMRs) were introduced in the early 1970s the vision was clear: digitizing patient records would make healthcare more efficient, and better for the patient and provider. Fast forward to today—we’ve come a long way towards the goal of a better system, but we aren’t quite there. We’ve identified the industry’s pain points, and laid the groundwork for solutions. Widespread adoption of the three healthcare tech ideas below will be crucial to building a healthcare system that is better for patients and the clinicians alike.
Interoperability and SMART on FHIR
In 2018, the Centers for Medicare and Medicaid (CMS) renamed “Meaningful Use” to “Promoting Interoperability”, which is a good indicator that we are inching closer to greater efficiency in healthcare. Essentially, interoperability means the ability for health information systems and software apps to communicate, exchange data and use the data they’ve shared. “Digitizing healthcare” was initially driven by revenue optimization and then care standardization. Clinical efficiency seems to be the forgotten promise but will ultimately be one of the most meaningful outcomes. While EMRs have made huge strides in digitizing healthcare and allowing for data to lead the way, they haven’t always communicated well with each other or other technology. Enter Fast Healthcare Interoperability Resources, or ‘FHIR.’
FHIR is a standardized API that enables better discreet data exchange between different EHRs and health IT apps. FHIR created a shared language for health IT systems that all health IT players could use to exchange, interpret and process health data.
Substitutable Medical Applications and Reusable Technologies (SMART) added an additional layer of consideration to FHIR, addressing how users would engage with third-party applications inside of the Electronic Health Record (EHR), or other health IT platform workflow. The standard supports both single sign-on to third party applications as well as an integrated user experience. While it’s certainly still in its early days, SMART on FHIR is catching on across the industry.
Clinical Decision Support – CDS Hooks
The primary care physician checklist is ever-growing. Does the patient use tobacco? Do they wear a seatbelt? Do they floss? Do they eat enough kale? Instead of leveraging technology to empower the physician to know what is best for their patients, we’re adding more mandatory boxes to check that may not even pertain to that patient’s problem that they are presenting with. On top of that, physicians are tasked with keeping up with rapidly changing clinical guidelines and new therapeutics. Arming clinicians with the information they need to make the best decisions for their patients—while working within their workflow–is essential.
What if the software systems that clinicians use could be more dynamic? What if technology could prompt the clinician with the information they need when they need it in order to make a more personalized clinical decision? The CDS Hooks specification was created to support this idea. The vision for CDS Hooks is this: A ‘hook’ is programmed within the clinician’s workflow in the EHR, and can share information with third party applications, like the SMART on FHIR apps mentioned above. The app can then trigger a CDS, or Clinical Decision Support, notification and prompt the clinician with relevant information. For example, a new lab result may be available in the EHR indicating worsening kidney function. When the clinician opens that patient record, the EHR silently communicates with the third-party SMART on FHIR app, or launches the app automatically, to show guideline-based medication dosing recommendations for patients with compromised kidney function. Put simply, embedded clinical decision support tools will make it easy for physicians to do the right thing for their patient, and hard to do the wrong thing.
Remote Patient Monitoring and Personalized Medicine
The vison of ‘personalized medicine’ is entirely dependent on bringing in patient-generated health data from connected devices (think step counters with apps but measuring more clinical information, like blood pressure or blood-glucose levels). Clinicians need a way to know what’s going on with patients the 8,516 hours per year that they aren’t engaging with the healthcare system, versus the four hours they may be with their doctor. This data set has been largely ignored but as price points come down on cellular and Bluetooth-enabled devices, data transmission, and cloud storage, connected devices are finally be available to the masses. Interoperability allows us to pull this data into analytics platforms and connect it with patient record data from the EHR, giving clinicians a more complete view of their patients’ health—whether at the point-of-care or virtually.
Not only can technology like this arm clinicians with a better view of each patient’s unique needs, but it can enable remote monitoring to detect costly complications before they happen, and prompt clinicians to intervene with appropriate guideline-based management.
Make no mistake: The convergence of these three emerging healthcare technology trends can change healthcare—but providers must be ready to adopt them.
After working as a physicist, a venture capitalist, an academic clinician and scientist, Lucienne Ide, MD, PhD, decided to start Rimidi to drive innovations in the healthcare industry.
Viruses: Biological versus Computer
By Mark Webb-Johnson, Chief Technology Officer of Network Box
During this time of the COVID-19 pandemic, those of us working with computer viruses continue to be amazed at the similarities between the techniques used by the medical community to fight SARS-CoV-2 (the virus that causes the COVID-19 disease) and the processes involved in our electronic anti-virus systems. Let’s look at some of these similarities, and see how computer anti-virus researchers help protect.
By Mark Webb-Johnson, Chief Technology Officer of Network Box
What is the Virus?
SARS-CoV-2 is an RNA virus. Essentially, a strand of information wrapped up in a protective shell and a mechanism to infect cells – information, envelope, infection mechanism.
By comparison, a computer virus consists of the payload, a carrier, and an exploit mechanism. For example, the payload would be the malicious code, the carrier an email message, and the exploit mechanism something to take advantage of a vulnerability in a particular mail client. Another example would be script downloaded from a web page, taking advantage of a web browser exploit.
The virus’s primary purpose is to replicate – to make copies of itself. SARS-CoV-2 does this by infecting cells (in the lungs, stomach, and other areas of the human body), then using the cell’s own mechanisms to make copies of its RNA and make new virus particles.
Computer viruses have the same primary purpose of replication. They want to infect as many hosts as possible. Once in a computer system, they make multiple copies of themselves and transmit out to new hosts. It is this self-replication capability that defines this as a virus.
Identification and Testing
The gold standard for identifying viruses is the whole genome sequence. This maps the full sequence of the chemical components of the RNA (made up adenine, uracil, guanine, and cytosine; abbreviated as A, U, G, and C), and results in a very long string of these four letters. That is excellent for precise identification, but not so good for testing. Due to mutations, what we know of as SARS-CoV-2 is actually a collection of hundreds of different strains of the same fundamental virus, each with their own slightly different sequences (more on this later).
To test for the virus, researchers instead concentrate on relatively small portions of the full sequence. This way, they can extract samples from a potentially infected human, amplify the RNA in the sample to obtain enough to test with, and then compare that against the portion of the full sequence. That is the RT-PCR test.
Computer virus researchers work similarly. The payload of the virus itself is a sequence of computer code that can be expressed in binary, or more commonly in hexadecimal notation. Computer viruses are often intentionally self-encrypted and randomized (we call these polymorphic viruses) to avoid whole sequence detection. Nowadays, these are by far the most common form of computer virus seen.
Researchers extract portions of the sequence that don’t change and use pattern matching techniques to detect those partial sequences in suspicious samples. We call these ‘signatures,’ and they can be effectively used to detect known viruses in suspicious samples.
Immune Response and Vaccination
The human immune system has a component known as the ‘adaptive’ immune system. The system works by identifying portions of viruses already in the body, and creating antigen-specific cells designed to identify, remember, and attack that specific antigen. These cells protect against future infections of the same virus and can survive in the body for some time (months, or years, typically). This is why after you’ve had the measles once, for example, you usually don’t get it again. Vaccinations work by purposely injecting the body with antigens that will generate such an adaptive immune response, to protect you from future specific infections.
Computer anti-virus systems store databases of signatures of known viruses. When your computer receives a new file, it can scan it, look for a match against those signatures, and take action if a match is found (quarantine, etc.). Such signature-based systems are the computer equivalent to the body’s adaptive immune system.
Innate Immune Response
Another component in the human immune system is known as ‘innate.’ This system can detect what is not ‘you’ (what is ‘foreign’) and attack the invader. It relies on the antigen’s chemical properties and doesn’t need to have previously seen that specific antigen.
For computer anti-virus, this is extremely hard to achieve well. The capability to detect and block previously unknown viruses is what differentiates good anti-virus systems from the poor. Various techniques are used, but mostly revolve around a) decoding the virus code to the rawest form, b) detecting suspicious encoding or exploit behavior, and c) using emulation or sandboxing techniques to see what the virus does when executed. Looking at behavior, rather than code sequences.
Mutations and the Future
RNA viruses such as SARS-CoV-2 are very poor at accurate replication, and sometimes the copies made are not perfect. Base pairs get flipped. Portions of the sequence are lost. Parts of other viruses are incorporated. Before long, you are dealing with bad copies of bad copies of a bad copy. It is like the story of a million monkeys with a million typewriters, eventually producing the works of William Shakespeare. Sometimes these viral mutations are beneficial to the virus, but most often not. Whatever the outcome, these mutations are the way the virus adapts to further its goal of replication.
Thankfully, we do not see the same with computer viruses. A computer virus can and does make perfect copies of itself, 100% of the time. Sure, we have self-encrypting polymorphic computer viruses, and randomizing fragments are often introduced, but the core code of the virus is not changed, and certainly not randomly. Perhaps in time, we will see this, but with today’s non-forgiving computer CPU architectures, it is unlikely to be a successful approach.
Of course, given enough monkeys and enough typewriters, anything is possible. Perhaps they can even improve on Shakespeare. Before that day comes, however, make sure you are prepared. Subscribe to a Managed Security Services Provider (MSSP) that can adapt, and protect you from cyber threats.
AI to help kids struggling with ADHD, PDD-NOS
Stephane Bourles, CIO, Brain Balance
AI to help kids struggling with ADHD, Autism, Asperger Syndrome , PDD-NOS and other ASD’s
Left Brain or Right Brain?
In a properly functioning brain, both hemispheres communicate equally and at lightning speed, millions of times per minute. In a poorly functioning brain, the left and right sides of the brain only impart partial information, causing frequent miscommunication. This is called Functional Disconnection and is the root of many types of learning, behavioral and social problems found in children. The Brain Balance program puts the left and right brains back in sync using sensory motor exercises, academic skill building, and nutrition guidelines.
How does the Assessment Work?
The assessment consists of sensory, motor, and academic testing of more than 900 functions. The outcome of this assessment is a highly customized report providing parents with a complete understanding of their child’s behavioral, social, and academic skill levels.
We use AI to determine which brain hemisphere we believe to be stronger or weaker. The Machine Learning algorithm used for the assessment is not always accurate, which we know based on the feedback from our staff—yes we let them disagree with the system, which is intended as a tool to help them support their own assessment, but not to necessarily force them into a decision they don’t agree with.
But as valuable as an individual observation based on years of experience is, it still remains just the view of one individual. That is why we look at Artificial Intelligence as a new solution combining neural network architectures with massive computing power to enable our solution to learn a pattern from large datasets and make statistical predictions based on test results and feedback we already have for tens of thousands of students.
What is next?
Thanks to many product releases with different Machine Learning models we tested, we were able to improve our assessment accuracy and achieve precision, recall and F1 scores over 0.95. The limitation of this AI model is it is not 100% accurate and you don’t know for sure the source of truth. Since AI is a “black box” which can’t explain its prediction for most models, you have to trust your staff first.
Similar to clinical decision support systems helping healthcare practitioners, we believe this fast growing dataset about children, combined with new Artificial Intelligence models such as Explainable AI, will help our staff improve a child’s initial assessment, which will then improve our overall program’s results.
The importance of information to empower health-app users
Liz Ashall-Payne, CEO, ORCHA
Shining a light between the rock and the hard place: The importance of information to empower health-app users
Apps, and their use for the promotion of health and wellbeing, are the subject of increasing interest and enquiry; particularly in light of the NHS’ Long Term Plan, and ever-increasing pressures on scarce NHS resources. But finding and downloading effective, engaging, and most importantly, safe health-apps, is a significant challenge. While many have suggested that health-apps could be a ‘silver-bullet’, aimed at quashing the woes that stem from decreasing funding and increasing demand for incumbent NHS services, it is important to tread with caution.
Unlike pharmaceuticals, over the counter remedies, talking therapies, physiotherapy or surgery, health-apps can not only be sought, but readily obtained in absence of gatekeeping and safeguarding. The result is that you or I can download upwards of 200,000 health-apps today, with as little as a click of a button on the app-store. While this can be argued as the unique value-proposition of app-based health technologies, a lack of guarantees regarding the quality and content of such apps, means that this open-access feature of apps also represents the chief concern; and is likely limiting the enthusiasm with which healthcare professionals engage and promote their use.
The number of apps labelled as ‘pranks’ or for ‘illustration purposes only’ has been subject to year-on-year growth. But for those looking for convenient, easily accessible, and often cost-free support; or for those who are either vulnerable or impressionable; such technologies can pose significant health risks, beyond those that are apparent and conveyed (if at all) in the short description within the app store. An app that misleadingly claims to monitor blood pressure or blood sugar for example, may result in either (1) misleading information, (2) modifications to disease management not representative of true disease status (including the use of insulin or warfarin), or (3) serious adverse events including hypoglycaemia or stroke. One such app for monitoring blood pressure has been downloaded upwards of 1,000,000 times on the Android Google Play app store, while boasting a review rating of 4.3/5.
This raises questions regarding the utility of both the subjective user review score (which is inherently limited in only capturing both very positive or very negative views) and the number of prior downloads of an app, as potential surrogates for app quality. Recent research into apps targeted at chronic insomnia disorder conducted by the Organisation for the Review of Care and Health Applications (ORCHA) confirmed this. In this analysis, which is now available in the journal BMJ: Evidence-Based Mental Health, it was demonstrated that both user review scores, and the number of prior downloads of a health-app, were not only unreliable as proxies for quality, but in fact were inversely related; and as these metrics increased, the objectively measured quality of apps decreased (as assessed by presence of clinical evidence, data privacy policies etc.) This has left a considerable void regarding where reliable indicators of the safety and quality of app-based health technologies can be found, something which at least in part, has been tackled by the National Health Service (NHS).
Courtesy of the NHS’s digital assessment questionnaire (DAQ), and the NHS apps library (beta version), the number of ‘accredited’ and NHS-approved health-apps is increasing, albeit at a leisurely pace, with approximately 50 apps accredited so far. While a step in the right direction, the rigour with which such reviews are conducted comes at the expense of speed, and an estimated review time of 6-8 weeks, limits the possibility of such initiatives being conducted at scale. As such, the difference between the total number of health-apps available (~300,000), and those which have been quality-tested and approved by regulators or accreditation bodies (~50); is if anything, increasing; leaving little or no information regarding the validity, efficacy, and most importantly the safety, of 99.99% of health-apps available today; which to date have been downloaded by potential users upwards of 50million times.
This is where the Organisation for the Review of Care and Health Applications (ORCHA) aim to transform the process of health-app research and information provision. Positioned as an open-access, free of charge and publicly facing repository regarding the user experience, clinical efficacy, and data privacy of thousands of health-apps, ORCHA’s aim is simple; to empower and inform potential users of health-apps regarding potential risks and benefits, prior to committing to use. Through a structured, objective and peer-reviewed assessment, consisting of 160 question areas, answered on a ‘yes’ or ‘no’ basis, ORCHA provides information to the 99% of users of health-apps, which are yet to be formally assessed by health technology assessment bodies, and in doing so, ORCHA removes the information asymmetry that is currently clouding the health-app market in, distrust, disbelief, and uncertainty. Publishing a new review of a health-app every 30minutes on average, ORCHA pull in information for all 300,000+ apps available on both iOS and Android Google Play, organise these into medically approved sub-categories (including diabetes, mental health and diet and exercise), and then order these apps from the most to the least downloaded.