Komodo Health captures fingerprints to enable precision medicine
Precision medicine has attracted a lot of attention in recent years as two technological advances - inexpensive mapping of human genomes and the near ubiquitous use of electronic medical records - have unleashed a torrent of healthcare data. Making sense of this data is easier said than done, though one company is already making significant inroads.
Komodo Health sits on an exponentially-growing rich repository of hundreds of billions of data points aggregated from hundreds of sources and stitches together a high-quality tapestry of 300 million people profiles across hundreds of different therapeutic conditions. The beauty is not only in its grouping of macro populations but also its extraordinary detail down to micro-subsets of individuals.
“We’re building real-time feeds and lining them up so disparate organizations can have a more holistic view of a single patient as well as patient groups based on a constellation of different data points, collected over many years to the present,” said Dr. Arif Nathoo, Founder and CEO of Komodo Health. “We’ve essentially created a fingerprint of a person’s health.”
Komodo Health, based in both San Francisco and New York City, started in 2014 to pursue the ambitious goal of mapping individual health identities that, when combined over many individuals, can surface patterns to better detect diseases well before symptoms manifest. Today it has nearly 100 US-based employees, over half of whom are engineers trained in machine learning, artificial intelligence, and data visualization/interpretation.
“We started Komodo because the healthcare data ecosystem is fundamentally broken. When companies and providers make decisions based on bad data, the patient loses,” Dr. Nathoo said. “Today, we’re now seeing a much more 24/7 sense of how healthcare works on the ground.” In other words, how healthcare is working in practice vs in theory and in real-time vs in snapshots.
Mapping out the flow in real-time
Not even a generation ago, nine-out-of-ten doctors’ offices still had physicians holding a pen in one hand and a clipboard in the other, writing down observations, organizing documents in manila folders and categorizing them in filing cabinets. Today, that’s flipped around, with nine-out-of-ten offices digitizing patient records.
Yet this data still lives in silos within hospitals and clinics.
“There are all these incredible silos of fragmented data, and our healthcare map is the first time that most of these data sources are being linked together,” said Dr. Nathoo.
By seeing what’s happening across the different silos, Komodo is making great progress equipping healthcare providers with new insights for detection and treatment.
“Historically, this work is done on the ground by providers,” Dr. Nathoo explained. “Along with genetic counselors and a handful of specialists, these providers are diagnosing at the point of suspicion maybe a year into the disease, or worse yet - 20 years.”
This process works when there are brilliant, astute providers available, and even then it may still be too late for some patients. This is where capturing historical data into an identifiable image is helpful.
“Because we understand from a data standpoint how patients who developed these diseases evolve over many years, we can see the fingerprint they leave over time and train algorithms to learn these patterns. Therein lies the opportunity for precision medicine and AI to identify cues early enough in the process to predict patients experiencing symptoms for the first time before they actually present.”
Indeed, the power of precision medicine lies in the ability to zero in on patients even before the first onset of symptoms. Ideally, patient populations would be categorized into subpopulations and even narrower subpopulations not based on a disease they have, but their similar preclinical makeup and susceptibility to certain diseases.
“Every day, we’re able to screen millions and millions of interactions of people with the healthcare system, in real-time, for anyone with a disease,” said Dr. Nathoo.
Identifying patients across the spectrum of diseases
From the diagnostic side of precision medicine, Komodo has succeeded in detecting patients across the full spectrum of rare diseases to common chronic conditions. As a consequence of Komodo’s findings, the right patients are tested and treated earlier, reducing the need for unnecessary diagnostic testing and therapies.
One case study that Dr. Nathoo points to is TTR Amyloidosis, a rare, slow, progressive disease resulting from the buildup of a protein known as amyloid, which leads to nerve or heart failure and eventually death. Misdiagnosis (and underdiagnosis) is common. To understand this life-threatening disease one needs to assess a number of variables, such as multiple genetic mutations and the location of the amyloid deposits, and importantly discern the symptoms, which are similar to other diseases. For instance, someone with carpal tunnel syndrome may not be identified as someone to be tested for heart problems down the road, even though for some with the disease, hand symptoms are a precursor eight to 10 years before cardiac manifestations.
“This disease can manifest itself in different constellations of symptoms in different patients, and it's therefore not simple or cost-effective to test every person with carpal tunnel syndrome,” said Dr. Nathoo. “There's a ton of diversity in how patients experience symptoms and when they get neurological vs cardiac symptoms. By looking at various patterns of those symptoms and thousands of patient journeys with known disease, we successfully predicted which patients needed to be tested before they were ever diagnosed with TTR.”
One disease that affects one-out-of-ten American adults is diabetes. Up to a third of people with prediabetes will develop type 2 diabetes within five years absent weight loss and physical activity. Average medical expenses are almost $17,000 a year. Left unmanaged, diabetes can lead to heart failure and a heart attack can cost up to $1 million dollars, if you add in loss of productivity.
“The cost of diabetes, when linked with cardiovascular events later in life, is borne institutionally over the course of the disease,” said Dr. Nathoo. “The question is: can we predict in the early days who is most likely to advance to these more critical states?”
While epidemiologists do a great job identifying where diseases are prevalent, they fall short of recommending more targeted and personalized treatments because they don't take into account the variability in factors across geographies. Just because two towns have the same prevalence of diabetes doesn’t mean the patient populations in both towns can be treated in the same way.
“Prevalence can be the same in different areas, but the behaviors can be profoundly different,” said Dr. Nathoo. “Social determinants [work, upbringing, community] of a given area show how behaviors are locally networked and how clusters of patients across varied geographies have contrasting reactions to similar interventions. A certain intervention may work well for someone who lives in an urban area, like Manhattan, but may not for someone in a rural town, like Chickasaw, Alabama.”
Komodo takes these social determinants into account to offer up more precise interventions for a given population.
Spotting areas of underinvestment and unmet need
Importantly, by determining what’s working and what’s not, Komodo is also able to show where there are great areas of underinvestment and areas of need.
“There are these incredible hidden pockets of high disease burden that are underappreciated, overlooked, or just not visible, leading to massive amounts of areas of underinvestment,” said Dr. Nathoo. “Part of the value of our healthcare map is that it shows higher costs of maintenance, letting you see where you have to start investing or lowering costs.”
This visibility is useful for all constituents across healthcare, from payers, to providers, to pharmaceutical companies, and ultimately, to patients.
Pharmaceutical companies typically have gone into areas where they’ve had relationships but they don’t go to areas where they’re really needed, said Dr. Nathoo.
“The strongest predictors of where pharma will make investments are where they’ve done so in the past and where they have historical relationships. But late-stage clinical trials should be conducted where there’s more need, not where there’s a relationship. We can identify where needs are really high and patient outcomes are terrible.”
For insurance companies, investing in treatment coverage is often a function of how one patient population responds, but this thought process often doesn’t take into account the variances in patient population across geographies or payers. “They see the performance of a certain population and say, ‘We covered this particular product and it’s added two years of life to this subset so we’ll cover another product in other regions and we should see the same outcomes,’” said Dr. Nathoo, “We have cost and outcomes data on every zip code in the US and the divergences are enormous.”
Commitment to privacy
Komodo’s healthcare map consists of 300 million patient profiles, with a typical patient’s journey mapped across hundreds of healthcare encounters and treatments and linked to the patient’s diagnostics and outcomes over many years.”
“We’re building the broadest and deepest index of patient encounters in the US,” said Dr. Nathoo. “This enables us to more accurately address the needs of communities across the US.”
Using AI and machine learning, Komodo can create correlations across patients, facilitating real-time clinical trials that reach a broader population. While the intent is good, the process relies on trust by all parties liable for any misuse of their data.
“Building trust is so important to creating the foundation for real-world evidence. This begins with how we handle sensitive information. Providers send us anonymized data, which is done at a higher standard than what the law requires. We then link anonymous records together across large populations, which massively lowers the risk of re-identification,” said Dr. Nathoo. “As a Qualified Entity under CMS, our security environment is assessed by the agency and monitored to make sure we continue to enforce a high privacy standard.”
On the trust front, Komodo’s partners want to know that their data is not being resold to others and that the algorithms developed on their behalf are not being used to help competitors.
“They have to believe, and do, that Komodo will operate in the best interests of its clients while aligning to its mission to improve patient outcomes,” said Dr. Nathoo. “We do this by working only on problems that align with our mission: to reduce global disease burden through the most actionable healthcare map. We employ the right technologies and security controls to engender trust in our technologies and are radically transparent with our partners on how their data is being used.”
A new health future
Komodo is at the forefront of a data revolution transforming healthcare from one-size-fits-all to a more tailored approach. By bringing in billions of data points to create a fingerprint image of anonymized individuals, Komodo is becoming a critical source of truth for the major constituents of today, from providers to payers to pharmaceutical giants.
But even more exciting and impactful is Komodo’s value to the most important constituent of all: the patient.
(Image source: GEhealthcarepartners)