Healthcare data: what’s next? Part 1: The current wave of healthcare data

A flood of new startups have entered the scene on the premise of using data analytics to improve the quality of the healthcare system and reduce costs. The potential opportunity has gained plenty of buzz: digital health investment is up over 300% since 2011, with healthcare data analytics consistently one of the largest sub-sectors within this category. Vinod Khosla famously has gone as far as saying that over 80% of what doctors do will soon be replaced by computers. We now have hospital systems and insurers alike devoting huge portions of their budgets to analytics and IT solutions, and even IBM’s Watson is jumping into the game to help diagnose some of the toughest cases in the hospital. However, the effort is only just beginning. Healthcare includes an incredibly diverse array of data, from billing records to imaging to DNA sequences, that provide different information about different aspects of a patient’s health and wellness. The various facets of our healthcare system will each require the correct pairing of data and analytical techniques to offer potential improvements and cost savings.

In this article I explore the state of healthcare data analytics today and some of the accomplishments we have seen, as well as some of the shortcomings. In part 2, I will look to the future to see what data types and analytics are going to be leading improvements in our healthcare system tomorrow.

Electronic Medical Records (EMR)

The early days of healthcare data analytics have been focused on analyzing and understanding much of the medical information we have already collected and stored. The advent of government incentive programs and meaningful use has driven a dramatic rise in EMR use. It is now estimated that over 80% of doctors are using some sort of electronic records system. This has created a wealth of data, documenting almost every aspect of the patient’s interaction with the healthcare system.

The integration and analysis of diagnostic information, procedural codes, billing data and common lab tests have enabled companies to provide new insights into the treatment of patients in a variety of different ways. Integrated health care providers such as Kaiser Permanente and Intermountain Healthcare have used patient records to develop new treatment pathways that significantly improve patient outcomes for heart disease and childbirth. Furthermore, payers and other at risk organizations have combined medical data with billing records to improve their population health management efforts and identify classes of patients and diseases that are the largest sources of high cost and low quality care. Nine figure acquisitions, such as Humedica (to United Healthcare) and ActiveHealth Management (to Aetna), show that payers are serious about utilizing data to improve their population health management. Finally, companies like Castlight Health are utilizing medical and billing data to enable self-insured businesses to provide their employees with the tools to compare costs and quality of a wide range of tests and procedures.

Despite these obvious successes, medical records do have several shortcomings. The first is that physicians are often rushed and not well trained therefore they don’t take the time to accurately input information into the EMR. This can lead to low quality data that does not accurately reflect the patient’s health status. Second, and more importantly, there is a limited set of symptoms and medical tests that can be captured in the medical record. Elevated body temperature or pain, for example, can be caused by problems in a wide assortment of underlying processes in the body, each of which may require a different clinical intervention. With only high level symptoms and a limited amount of diagnostic biomarkers, medical records simply do not currently contain the data needed to understand and analyze many of the diseases plaguing our society. The next wave of healthcare data will focus on integrating new data sources that provide clearer evidence of the underlying disease processes, allowing for more insightful analytics and more precise diagnoses and treatments.

Genetics

Since the sequencing of the first human genome in 2001, the cost of DNA sequencing has dropped almost 5 orders of magnitude and DNA sequencing is quickly becoming a common part of clinically accepted lab tests. We can now use genetic tests to determine ancestry, predispositions to various diseases, or even the consistency of earwax!

One of the early areas where genetic testing has gained serious clinical traction is in the area of non-invasive prenatal testing (NIPT). The advent next-generation sequencing platforms with higher sensitivity than older methods have now enabled tests that can analyze small bits of fetal DNA from the mother’s blood, replacing costly and dangerous amniocentesis. Demand for tests that can detect genetic diseases, carrier status, and aneuploidies (such as Down’s syndrome) has exploded as companies such as Natera, Counsyl and Sequenom have burst onto the scene with their own versions of these tests.

The second clinical area in which genomics has gained a large initial acceptance is in cancer. Cancer is a disease predominantly driven by acquired DNA mutations which damage cells’ natural ability to regulate proliferation and growth. An individual’s unique pattern of mutations can affect both their risk of acquiring cancer as well as their likely response to therapies. A well known example is the BRCA mutation which, if present, can increase a woman’s chance of developing breast cancer to over 60%. Furthermore, Foundation Medicine, in which Roche recently paid over $1 billion for a majority stake, provides a product that can recommend different courses of therapy for cancer patients based upon the mutations in their genome.

However, aside from some notable exceptions (in cancer, for example) genomic data is largely static, and the insights gleaned are often not particularly actionable. While one’s genetic makeup may predict the lifetime risk of a particular disease, it rarely informs a particular individual whether or not he or she has actually contracted that disease. Truly personalized medicine will require more immediate and more accurate markers of health and disease that will make diagnoses and treatments fast, accurate, and efficient.

Conclusion

There have been some notable wins in healthcare data analytics up to this point, particularly in organizing and interpreting the medical data that has been readily available. As adoption of these products continues, it will bring massive improvements to the healthcare system by streamlining operations and care pathways, identifying the patients in a population who are most at risk, and giving rise to evidence based medicine. Further improvements will largely be focused on designing great experiences for physicians. Products that empower doctors rather than tell them what to do and allow more time with patients rather than more time in front of a computer will likely be the big winners moving forward.

In part 2 of this article I will focus on new types of data analytics poised to make an impact on healthcare in the near future.