Data for the People!

In “The Path to Better Health: Give People Their Data” (Issues, Winter 2021), Jason Cohen makes an important contribution to the discussion of data privacy.

Data privacy in the midst of data integration, data organization, interoperability, and advanced analytics are table stakes for health care organizations—but challenging. The rush to commercialize on personal health data represents a particular risk for underserved populations, who already suffer poor outcomes due to lack of access to health care. There should be an approach to being thoughtful for these populations and filter for critical review of algorithmic bias. Creating a framework for ownership of health data that empowers these populations is essential to ensuring that they receive the benefits of the data science revolution.

Jay Bhatt

Jason Cohen presents some interesting perspectives. Several points in particular jumped out at me.

It is very true that patients making poor decisions is at the center of chronic health problems. With the application of artificial intelligence and data, health tech companies are well positioned to make a difference and deliver personalized patient engagement programs. These engagement platforms can educate patients and drive behavior change to help them adopt healthy habits.

Patients owning their own data and being able to control who gets to use them is a great concept. If such tools are developed and adopted, patients certainly will have a lot more control and power. Some of this is already happening with Apple’s iPhone, Microsoft’s Office 365, and Google’s search queries, where the phone or device is keeping track of communications happening between individuals and the world around them. Big data analysis of the tone of the messages and the time spent on various apps or the content that is consumed can provide leading indicators of a person’s mental state.

Future use of such technology seems positioned to expand.

Madhu Reddiboina

Jason Cohen makes several excellent points, but he does not mention the practical importance of data context. For example, radiologic images require skilled interpretation, and even the image characteristics or “findings,” may then support only a probabilistic measure of the health or prognosis of the patient. Clinicians that use such information to guide patient management are well aware of the reasons the imaging was requested, the context in which such measurements are acquired, ways findings might be interpreted by the local radiologist, and confounding factors specific to the patient, but the future data user doesn’t have that advantage. The data mining algorithms used by a third party years later may not be sufficiently sophisticated or the information in the training data sets may not be available to provide accurate support to the caregiver.

The article by Ben Shneiderman in the same issue, “Human-Centered AI,” discusses these challenges. It is not made clear how care will be better for everyone if each patient owns his or her data, but it seems obvious that countries that have nationalized patient data repositories, such as Norway, offer their citizens a better foundation for clinical practice.

Kirby G. Vosburgh