The
Coming Storm in Biotech
A storm is coming in Biotech. This storm is born of the convergence of years
of computer science, medical and societal forces. This storm will play out over a period of
decades but will fundamentally change the way we think about and deliver healthcare.
With
Big Data, comes big responsibilities
The first megatrend driving this is the generation and
management of data. This starts with the patient and the ever increasing
personal and corporate spend on health awareness, wellness and the mining of
traditional bio markers, advance genomics and proteomics in a metric driven, data
hungry world. People are paying ever
closer attention to diet, exercise, cholesterol and environmental factors. This has led to a generation of humans who
welcome being monitored, poked, prodded and sampled at a dizzying rate to
collect a personal corpus of actionable data.
When we think of Big Data in the traditional IT sense, it
is defined by such sources as: machine exhaust, web traffic, the Twitter “Fire
Hose” and “internet of things” e.g. smart-grid.
All of these IT and consumer generated data, although massive, are
dwarfed by the amount of data that is generated from human biology.
The societal question is what do we do with this data. What is the next generation of personalized,
predictive medicine? Who pays for it? Predictive
medicine is the ability to predict and assign probabilities to likely events in
your medical future. It flies in the
face and represents a radical departure from traditional medicine where the
medical community is only engaged after something goes wrong. Once something is determined to be wrong,
the physician deploys a hunt and peck method of seeing what combination of drug
therapies, formulae, dosage and delivery method gets you on the path to
wellness without messing something else up.
One of the fundamental problems is that the entire medical eco-system
and path to sustained wellness is predicated on the simple question “What’s
wrong with you today?” Imagine if the
answer was “Nothing, I’m just here to make sure I stay healthy”. How would they bill for the battery of test to
1) Give you a steady state of what your target health profile should look like
and 2) Based of the battery of predictive tests, what should you do to maintain
target health. The irony of course is
that in the long run, the economic benefit of preventative care far outweighs
the trillions spent on a traditional “break-fix” mentality.
Waves
of Change
The Biotech revolution will happen in three distinct
waves. The first wave was the wave of EMR adoption. This
started in the early 2000’s and continues today. This wave was about two major
vectors: 1) Digital records e.g. Basic Document Management displacing the paper
chart, handwritten notes, prescription pads, sticky notes. The promise was process efficiency, reduction
of errors, lower malpractice insurance costs etc. and 2) Enterprise and Supply Chain efficiency. By using an EMR, Doctor Groups, hospitals and
HMO’s had cross departmental and pan-facility visibility into patient care.
Before EMR, patient records were locked up in best of breed silos and profit
centers such as: radiology, pharmacy and emergency. Like manufacturing ERP in the 90’s, EMR’s
deliver an integrated flow of transactional related data on patient services,
billing, payments, and relationships with critical suppliers etc.
The second wave of change was the rise of unstructured,
non-relational data “big data” modalities. Like business intelligence and the
analytic capabilities that appeared post the ERP era, this tipping point for transforming
traditional EMR transactional efficiency to now include imaging, PACS, and
output from next generation sequencing data.
The first step and opportunity area will be around the simple storage
and access of this data and the potential blending of these unstructured
sources with transactional EMR data to enable the physician to simply view
these new modalities when making a clinical decision. The analysis and insights derived from these
data types will still come from experts in the supply chain e.g. radiologists,
geneticists. But the clinician will now
be able to draw upon these data sources for targeted action.
The third and arguably most significant wave of change
post the data collection and integration phase will be delivering actionable
insight at scale. In order to accomplish
this, we must be ready for predictive analytics augmenting the role of the
clinician. Clinical healthcare analytic
and decision platforms will need to be developed that perform the IT heavy
lifting described below and deliver clear decisive recommendations for the
clinician to then impart their expert opinion towards the optimal course of action.
Figure 1 – The three waves
of change and opportunity
Change
is often about convergence
In traditional IT, we see this explosion of data
particularity the rise in No-SQL stores such as Hadoop and Mongo enabling
massive horizontally scalable stores of unstructured data. Combining these
web/machine generated data sources with OLTP structured relational data from
traditional sources such as customer demographic data, buying patterns, we see
real-time analytic engines being able to curve fit a likely algorithm, run the
multivariate linear regression and fit a demand and profitability curve. This is useful in the case of an online
retail environment to custom tailor a promotion, A/B test it and launch the
promo. In Financial Services, we see
this as the domain of the quant trading, hedge fund gurus developing algorithms
that predict the minute changes in a commodity based on a myriad of related and
seemingly unrelated data sources.
The above examples are great when it comes to selling
more shoes or doing currency arbitrage, but the Holy Grail is still untapped.
The perfect storm of convergence is happening when you
combine modern medicines ability to generate data at an unprecedented rate with
IT’s ability to store, manage, analyze and combine this clinical data across a
huge number of disparate sources and modalities. The good news is that the processes are similar. A thriving eco-system of entrepreneurs, investors,
universities, incubators has already built the base platforms that can address
the market of predictive medicine.
Figure 2 is a sample marketscape of traditional IT
vendors that make up the various layers of the predictive analytics category.
Figure
2 - Sample Data Science Marketscape of Traditional IT Vendors
Moving upper left to upper right starting with the raw
data stores of relational and data stores optimized for analytic processing all
the way to predictive analytic engines, dashboards and KPI’s. Moving
lower left to lower right, you have the important layers of cleansing
normalization, MDM to visualization.
This horizontal technology mapping illustrates the
aggregation of 20+ years of traditional computer and data science
thinking. These are known market
segments that anyone having spent a few years in the field would be able to
grasp and apply to traditional data problems. This market landscape represents
about $65B in annual middleware spend in 2012.[1]
The same analysis/mapping needs to be done in the context
of healthcare. The differences will be subtle
such as: the source systems, modalities and density of the data across
modalities. The good news is that most of the concepts and many of the
horizontal layers will be the same.
The
Economics and Societal Readiness
The most important question that needs to be answered is
that in industries like Retail, Financial Services, or the consumer web, businesses
are willing to pay billions of dollars to track consumers, optimize spending or
increase on-line conversion by one percent.
Such is not the case of predictive medicine. The expensive advanced testing required in
building up a personal knowledge of one’s health profile remain by and large
not covered by health insurance and therefore out of reach of the common person. This requires a fundamental overhaul and
perspective on how we look at preventative, quantitatively driven healthcare.
The clinical readiness is the transition from a largely
human based intuitive decision making process to one that is on the sliding scale
of augmented to highly predictive. It is
this evolution towards the clinician taking full advantage and improving their
own judgment and reliability with predictive data based algorithms that will be
the tipping point of acceptance and adoption. As with most mega-trends, this
evolution will start locally and spread globally. In
early 2012, India instituted a nationwide program issuing a Unique
Identification Number (UID) to all of their 1.2 billion residents. Each of the
numbers will be tied to the biometric data of the recipient using three
different forms of information – fingerprints, iris scans, and pictures of the
face. All ten digits of the hand will be recorded, and both eyes will be
scanned. We’re only beginning to
comprehend the repercussion of such a societal orchestration.
During our
daily lives in the US, we are bombarded with predictive analytics looking to
monitor, optimize and recommend action, either opt-in, prescribed our
automated. Our lives are increasingly and subtlety controlled by such everyday
events like variable HOV toll changes based on real-time traffic patterns or
the perfectly timed coupon that magically appears in a banner add on
Amazon. These are but a few examples of
everyday human behavior being measured, optimized and guided by deep data analysis.
The tipping point in health care will come equally as
stealthy as our societal acceptance masked as the new era of customer service
evolves over time.
Copyright 2012: Kevin Chew
