Predictive Analytics in Health Care
One of the most promising uses for big data in health care comes in the form of predictive analytics, in which models are built to detect patterns and predict outcomes. For individualized health, predictive analytics has the potential to tailor health care delivery to each patient’s unique circumstances. Although predictive analytics has already revolutionized several industries (Amazon, for example, uses analytics to recommend products you might like to purchase), its use in health care is less pervasive. In the health policy journal Health Affairs (link available here), two articles address priorities for and issues concerning the implementation of predictive analytics in the health care setting.
In one article, study authors, including Hopkins faculty and inHealth member Suchi Saria, describe several areas that could benefit from predictive analytics. These include having the ability to predict which patients may be high cost, which are likely to be readmitted, and which may experience an adverse event. The article suggests that identifying these patients and recommending appropriate interventions might not only improve care management, but could also save on costs. For example, an algorithm may categorize patients into high or low risk categories and suggest correspondingly different management options, allowing health care institutions to optimize the allocation of their resources. The article concludes with several policy implications of using predictive analytics, including how these tools might be regulated and how patient privacy concerns might be managed.
Patient concerns, along with other legal and ethical issues, were also discussed in another article published in the Health Affairs issue. The authors highlight the importance of incorporating patient perspectives into the governance of predictive analytics. This is especially critical when, as the article notes, patients’ priorities conflict with those of clinicians or those of health care institutions. For example, a conflict of interest may arise if, as a cost-saving strategy, an analytic tool recommends against a treatment desired by a low-risk patient. Not only would this scenario conflict with the patient’s preference, it may also have liability ramifications for the clinician if the patient goes on to have a negative outcome. These types of ethical and legal quandaries require careful consideration.
As these articles suggest, predictive analytics has the potential to revolutionize health care delivery and reduce health care costs, but incorporating a variety of perspectives into the development phase is essential for its success.