Can Hype Spell Hope for Predictive Healthcare Analytics?
Since the mandatory digitalization of health records and IBM’s Watson’s debut, analysts and journalists have written several hundred thousand excited words about what big data and predictive analytics can do for the healthcare industry.
However, Dr. Bonnie Feldman of DrBonnie360 warns in her report named “Big Data Healthcare Hype and Hope” that the expectations of what big data can do for the healthcare industry are currently outpacing the realistic aspirations. While Feldman notes that predictive analytics can slow the unsustainable growth in health care costs and advance important genomics research, she asks to hold the hype a little as significant obstacles inherent to the industry have yet to be overcome.
Put simply, one of the largest limitations on big data healthcare research may be the stringent doctor-patient confidentiality rules. Those rules exist for good reasons: limiting insurance providers’ ability to discriminate and preventing potential employers from making hiring decisions based on medical prospects to name a few. However, those privacy and security concerns make it difficult to build large datasets that would serve as the backbone for a potential predictive analytics system.
With all of that being said, there is plenty of incentive to introduce big data analytics to the healthcare industry, chief among those are managing costs and improving patient care.
Of course, the two are not unrelated. According to the report, in 2009, 2.9 trillion dollars were spent on healthcare in the United States, constituting 17.6% of the GDP. By the year 2025, that percentage is expected to rise to 25%. That kind of increase is unsustainable, and big data can help.
From a financial perspective, big data can provide assistance in the same way it can help finance or retail: by tracking and making predictions based on the transactional, structured data that is produced from patient bills, payments, et cetera.
But certain analytics-based advancements in patient care, such as preventing return trips to the hospital (according to the report, a little over 20% of all patients have to make a return trip within a month) and preventing unnecessary deaths (96 out of 100,000 patients in a hospital die for preventable reasons, such as infections), would also help drive down healthcare costs.
However, researchers are finding difficulty in building the appropriate database. For example, in a survey of 600 healthcare providers and professionals, while 73% pointed to integrating data from multiple sources as their primary goal over the next two years, only 17% felt confident that their analytic and integration needs were going to be met in that time. “The predictive value of analytic tools will not be realized if the data sets being analyzed are low quality or represent irrelevant measures,” the report noted.
As a result, Feldman noticed three trends emerging in medicine research: working with limited datasets, increasing the variety within those datasets, and pooling data with other organizations. The first two are less than ideal but the third represents more of the open source attitude that has been seen in financial sectors.
One potential solution is to approach medicine from the other direction. Instead of making diagnoses based on large databases of patient histories, one could perform genetic analysis on the patient to determine their likelihood of having a particular disease. This area of study, known as genomics, has made significant strides as a result of big data advancements.
However, genomics is still a little way off from solving diagnostic problems. As Dr. Mark Boguski, founder of Genome Health Solutions and doctor at Harvard Medical School noted, “In the past genomics has over-promised and underdelivered with respect to influence on medical practice and improving human health.” However, Boguski believes there is reason to be optimistic for genomics, saying, “we’re now in the ‘third wave’ of genomic medicine which I firmly believe will lead to better health outcomes through precision diagnosis.”
Maybe analysts and journalists jumped the gun a little bit with predictions on what predictive analytics could do for medicine. But with hype, there lies hope, hope that genomics advances naturally and that medical data can be combined without sacrificing privacy and security concerns.