The breakthrough platform for the automatic
discovery of disease signatures and scoring patients accurately for risk
Applying machine learning techniques to modern patient data, starting with electronic medical records (EMRs), has long promised better outcome predictions. But this has proven harder than expected, with useful results but no real breakthroughs. Until the Backdrop team developed its wholly novel and uniquely powerful way of applying machine learning to create a new way of computing on all kinds of patient data.
Backdrop's solution gets the maximum value from any set of patient data, including EMRs, claims data, clinical notes, genetic sequencing, and more. Backdrop enables pharma companies to bring new therapies to market more quickly and economically, health care providers to preemptively identify health issues and improve patient outcomes, and insurers to optimize patient health at lower cost.
Health care spending is almost $4 trillion per year in the US alone,
almost 20% of the US economy.
A breakthrough in the precision with which we can predict medical outcomes and disease trajectories for individual patients and populations will enhance the lives of millions.
Backdrop will deliver huge cost savings in three big markets that rely daily on electronic medical records: pharmaceutical companies, health care providers, and insurers, worldwide.
Clinical trials are expensive, risky, and take years. It's expensive to license medical records, and they're often hard to harness for use. Backdrop's software can materially accelerate clinical trials by enabling improved precision in the inclusion and exclusion criteria, by simplifying the creation and management of synthetic control arms for real-world evidence trials, and more.
All health care systems have a deep interest in optimizing care and improving patient outcomes. Backdrop makes it easier for providers to preemptively identify and proactively assess likely health issues and improve care. Backdrop can also greatly reduce the 'diagnostic odyssey' of patients, especially for rare diseases, and get them the treatments they need sooner.
All payers want to understand which diagnostics, therapies, and procedures work and how well. They care about quality of care and equality of care. They are interested in earlier interventions that enhance health and save money. Backdrop can help insurers in each of these areas, reducing costs while helping their customers stay healthier.
How can you truly understand and use your medical records?
A comorbidity is the existence of two or more medical conditions, such as diabetes and obesity, in a patient. Comorbidities have been used for decades, precisely quantified using proven statistical techniques. But they've usually been discovered only between pairs of diagnosed medical conditions. Most of the available clinical data has never been used completely, if at all.
Classic comorbidities only show a small part of a big picture. It's a field ripe for the application of new big data techniques. Backdrop Health was launched to deliver this breakthrough. Backdrop greatly extends the concept of comorbidity to a full set of comorbid factors that include all pairs of co-occurring clinical variables: diagnosed conditions, procedures, medications, and lab tests.
Get the maximum value from your available medical records.
CoDE is the software at the heart of our solution. CoDE analyzes any body of electronic medical records to discover and quantify how every factor -- diagnosed conditions, lab tests, procedures, medications, any genetic testing data, and more -- relates to every other factor. Equally important, CoDE also computes the temporal connections between factors, allowing additional insight into disease progression.
The output from CoDE is the backdrop for that body of medical records: a comprehensive probabilistic knowledge graph that makes visible every clinical variable's relationships to every other. While traditional comorbidities reveal some useful facts, a backdrop reveals everything, for every factor of interest. And, a backdrop is free of any PHI (Personal Health Information), rendering moot any privacy concerns.
It's expensive to acquire patient data, so get the most from it.
Previous best practices for comorbidity discovery rely on stratification, or subsetting the data. That's done to ensure uniformity of risk in the face of confounding variables such as age or ethnicity. But this results in diminished statistical power. Even a large starting data set can shrink dramatically when adjusted for many confounding variables. It's common, for example, for a drug company to license data for a trial and end up with just a handful of patients after stratification.
Backdrop's solution does not rely on stratification. CoDE calculates and uses each patient's individual probability for every condition, based on his or her full medical history. The full history and demographics of every patient in the sample is brought to bear on each outcome analysis. Backdrop amplifies the effective size of medical data sets by orders of magnitude. This is particularly valuable when dealing with rare diseases and with demographic factors under-represented in the data.
A backdrop enables an unlimited set of useful and actionable analyses.
Knowing all of the comorbid factors in a medical data set is not enough for accurate and useful outcome predictions. These factors do not exist as isolated pairs, but combine to create complex webs of influence on any given outcome. Useful results require the ability to calculate the joint contribution of multiple factors which in turn depend on one another.
HOPE, the other key software module in our solution, does just that when you query your backdrop about any outcome of interest. Backdrop includes a powerful means to understand the impact of all demographic factors on outcomes. The contribution of every clinical and demographic factor to every prediction is clear, so all conclusions are fully explainable.
Backdrop is commercializing the latest innovations from
The University of Utah's Prof Mark Yandell and his team
at the Eccles Institute for Human Genetics.
Backdrop's team developed its approach and its software over more than five years, with the benefit of full access to a large set of electronic medical records. This includes almost 4 million patients and over 72 million visits over 10 years. Backdrop's team learned what to do, on a big set of full data.
The core technique, called Poisson binomial-based comorbidity discovery, was summarized in the October 2021 issue of Nature Computational Science. Backdrop Health, Inc. was formed to bring this solution to the commercial market under a license to the intellectual property from The U of U.
A backdrop is not a neural net, trained on a single data set.
Customers use Backdrop's solution on the patient data that they have created or licensed. Customers create their own backdrops against which to understand patients, outcomes, and disease progression. Our business model does not involve the licensing of medical records or any other patient data.
Backdrop's core software -- CoDE and HOPE -- run securely and reliably in the cloud. They're implemented in a way that eliminates the risk of revealing any personal health information while creating or using a backdrop, while allowing full computations on your data.
Committed to improving human health and cutting the cost of health care.