One of the unstated facts of medicine is that we don’t always understand the relationship between treatments and outcomes. For example, the U.S. spends over $2.6 trillion on health care every year, including over $600 billion in unexplained variations in treatments: in other words treatments that cause no differences in outcomes. This is where big data analytics, coupled with the right educational interventions, can make a huge difference to healthcare outcomes.
Figure 1, below, summarizes some of the new data sources in the health sciences.
Figure 1 – New Health Sciences Data Sources (Greenplum/EMC healthcare position paper)
Health care, however, is one of the most regulated fields in public life. Doctors go through some of the longest and most demanding training that exists in the world today. The whole industry is constrained by requirements that require constant training, education and certification, and any “disruptive technologies” cannot raise surprises. When a new technology, therapy, diagnostic or other healthcare product is introduced into such an environment, quality metrics must be applied that determine if the product is truly providing better outcomes.
Typically, when doctors order a treatment, whether it’s surgery or an over-the-counter medication, they are applying a “standard of care” treatment or some variation that is usually based on their intuition, effectively hoping for the best.
Data can make hospital systems more efficient, and avoid preventable complications such as blood clots and hospital re-admissions. It can also help address the challenge of health care “hot-spotting” (1): i.e. identifying people who use an inordinate amount of health care resources. For example, by looking at data from hospital visits, Dr. Jeffrey Brenner of Camden, NJ, was able to determine that just 1% of the hundred thousand people who made use of Camden’s medical facilities accounted for 30% of its costs (2). The proliferation of digital health information, including both clinical and claims information, is creating very large data sets which can be analyzed e.g. synthesizing clinical records and claims data together can help identify patients appropriate for inclusion in a particular clinical trial. These new data sets can also provide insight into improved clinical decision making. (The database of 1.4 million Kaiser Permanente members helped determine that Vioxx, widely used by arthritis patients, was dangerous.(3)) Only by having a huge dataset available from years of electronic health records and tools to properly analyze the data, was this medication pulled from the pharmacy.
Typically, big data refers to a data set that is too large for a typical database software tool to store, manage, and analyze. There is a difference between healthcare and other industry sectors however, since health data is both personal and sensitive, and it also has special security and privacy protections. This makes sharing, aggregating, sorting and analyzing the data challenging. Another difficulty in making the most of big data in healthcare is that those who control different “pools” of data often have different financial incentives. As a consequence, there sometimes tends to be a lack of transparency in performance, cost and quality; and incentives are currently structured so that payers would gain from decreasing revenue to providers, but providers control the data that is needed in order to pay for value. In addition, payers (typically insurance companies and large employers) control another pool, including claims data. Enabling transparency of data would help to identify and analyze sources of variability as well as find waste and inefficiencies. Publishing quality and performance data will also help patients make informed health decisions.
Figure 2, below, shows the cost savings obtainable by various modalities for the prevention (or reduction) of obesity over the average lifetime (75 years) of residents in the United States.
Figure 2 – Role of different modalities in federal obesity-prevention efforts over the average lifespan of a US resident.
Ultimately the primary consumption of healthcare expense in the United States are long-term problems that require lifestyle and environmental changes. The patient is more important than the doctor in these modern ills. But the patients who cost the most and need to make the most far-ranging changes have demonstrated an immunity to advice. This is where continuous education, motivation and support is required in order to “nudge” patients out of unhealthy habits and into healthy ones. The changes a patient would have to make to prevent smoking-related lung disease or obesity-related problems would require a re-evaluation of his course of life, which few can do without adequate support, counseling, education and motivation. The delivery of continuous and aggressive methods for education, outreach and support can play a significant role in improving outcomes.
Technological advances are overcoming many of the traditional obstacles to compiling, storing, and sharing information securely. For instance, EMR systems are now more affordable than in the past, even for large operations, and allow data to be exchanged more easily. In addition to facilitating longitudinal studies and other research, technological advances have made it easier to “clean” data and preserve patient privacy. The new programs can readily remove names and other personal information from records being transported into large databases, complying with all Health Insurance Portability and Accountability Act (HIPAA) patient-confidentiality standards. Some computer systems can even examine information across all data pools — an important feature since there are special combinations that can provide more insights than any individual data set. For example, claims data may show that a patient has tried three treatments for cancer, but only the clinical data show us which was effective in shrinking the tumor. As another example, personal behavior information may show that a patient is taking fewer trips outside the house or looking up information on side effects online, both of which could suggest physical problems or be early indicators of an illness requiring early intervention to prevent a more serious medical episode. But only clinical data will confirm whether the behaviors are truly linked to illness. With new data becoming available, innovators have taken the opportunity to build applications that make it easier to share and analyze information.
By Sunil Maulik, Ph.D., member of ICDK’s advisory board.