Cross post from the Amazon AWS Startup Blog
In the 10 years since the HITECH Act spurred the healthcare industry to digitize data, and with the ensuing explosion of patient data, the healthcare industry has been looking for more data-driven approaches to understanding the patient experience. The trillions of data points that are being generated daily by care decisions, patient registries, electronic health records, labs tests, and wearables have the potential to help deliver the right treatments and care regimens for the right patient at the right time.
Until recently, the standard way to bring a patient’s information together to understand their journey through the healthcare system was by manually incorporating patient feedback, chart reviews, and analyses of data scattered across different providers, payers, and lab datasets. These costly, time-consuming and manual methods often require reconciling clinical, financial, and transactional data that are also scattered digitally across applications and databases. The data often loses visibility of patients as they “travel” across multiple services, care settings and insurers over the course of a few years, painting an incomplete picture of a patient’s journey.
To meet the ultimate goal of personalized medicine—tailoring each treatment to the specific characteristics of each patient and their condition—it is necessary for patient level datasets to be complete and representative so they can be analyzed holistically. Otherwise we won’t have the full picture of the patient and cannot make the most tailored and specific treatment decisions for them.
The San Francisco-based startup Komodo Health attacks this problem by assembling different elements of de-identified patient-level data to create a complete and representative picture of how patients move through the U.S. healthcare system. Having mapped the journey of over 300 million patients in the U.S. across different doctors and healthcare organizations, insurance plans, with their associated lab and prescription information, Komodo Health has created a multi-layered map that enables seeing patients in a new way.
Komodo Health’s datasets are linked to one another at the anonymized patient level, and by creating these linkages we are able to see the timeline of the patient’s journey and track the associated behaviors, costs and outcomes. With this complete map as our base, machine learning can be deployed to learn from past patient journeys, and make better recommendations and predictions about current and future patients.
For instance, Komodo Health’s platform has been used for compliantly “identifying” providers who are managing patients with undiagnosed rare diseases using complete and provider-identified, clinical encounters. By using Gradient Boosted Trees, we were able to determine the likelihood that a patient would benefit from being tested for a rare genetic disorder, and were able to identify providers in real-time with at least one patient that would benefit from testing—resulting in an approximate 30x increase in rare disease testing.
At Komodo Health, we believe that every patient’s experience with the healthcare system is unique and varied, and creating an accurate and representative healthcare map requires acknowledging that complexity. Our mission from day one has been to improve patient outcomes using big data, combined with human and machine intelligence. The use of machine learning with rich datasets can lead to earlier diagnosis and more personalized care, and works better than the typical trial and error approach of the past. As more patient-level data continues to become available to technology companies, the opportunities to help patients in a meaningful way will only continue to grow.