The ethics of predictive analytics in health care: Balancing patient care with data privacy

The image of the family doctor rushing to a patient’s sickbed on an urgent late night house call is deeply engrained in American culture. For many people, this nostalgic scenario still lingers as the ideal example of what personalized care should be. No doubt, it’s a compelling portrait of individual dedication to the patient. But a future-oriented vision of health care recognizes that the patient can be served even better when they have access to an array of resources and support, with equal urgency paid to preventative care. Modern health care is best delivered not by a single physician arriving at the doorstep at the last minute, but by a whole team working together to address potential trouble spots before they escalate into something more serious.

Predictive analytics is helping Blue Cross NC achieve that vision — not tomorrow, today. We’ve developed an award-winning machine learning platform that looks for patterns across member data that indicate when someone is at increased risk of experiencing a serious health event. Among members with increased risk, it also identifies those who might have missed critical follow-up visits with their primary care provider, those experiencing more than one health condition simultaneously, or who are taking multiple medications. This information helps the care managers on my team take a more proactive, personalized approach to helping members navigate complex health care needs. Our advancements in machine learning capabilities have already led to significant impacts on health outcomes.

It’s important to note, however, that we have made these advancements while respecting that lots of people start to get nervous anytime artificial intelligence is mentioned in the same breath as “health care.” Many are rightly anxious about what could be lost if things aren’t done right: privacy, doctor/patient relationship, individualized treatment, and so on. These concerns are real. We couldn’t move forward on the development of our machine learning without addressing them head on.

From the beginning, we’ve understood that our data analytics and model design must be guided by rigorous oversight, strategic vision, and security protections … and we knew that our work in this arena must ultimately make health care more personalized, more effective and more accessible for all. 

Creating the mechanisms that protect member privacy is absolutely an essential step to developing a predictive analytics tool that serves the best interests of our members — but it’s not the only step.

In addition to building our software platform to have best-in-class security protocols, three aspects of our design and implementation process will ensure that we protect data security, enhance quality of outcomes and maximize impact:

• We’ve taken a collaborative approach to design and implementation: All along, clinicians and our data and analytics team have worked together to build a framework that eliminates statistical bias. As a result, the information that we get from our platform is high quality and actionable.

More than that, the spirit of collaboration is integral to the way we use this data. The platform doesn’t trigger an automated, impersonal advisory that goes directly to the member. Instead, the platform alerts our care managers, who reach out directly. As they learn more about where the member is facing challenges that could undermine their health, care managers can direct the member to resources that can help or take steps to coordinate a multilateral, proactive support strategy.

• We’ve built a platform that collates data from different sources to get the “big picture:” In addition to using our internal data, our predictive technology incorporates census data measuring social vulnerability. This can alert care members of potential situations when a lack of access to drivers of health resources could undermine a member’s health. Pulling all this information together enhances our ability to close gaps in care and minimize health inequities.
• We’ve designed a governance and oversight process to ensure quality and efficacy: Building a data analytics platform is one thing, acting on findings is quite another. As Blue Cross NC develops future model interventions that utilize our predictive analytics tool, these plans will go through a rigorous approval process. A governance council comprising medical, legal and business leaders will ensure that experts evaluate the work across multiple touchpoints and from a variety of perspectives: data security, risk management, bias protection and medical efficacy. This process will protect privacy, ensure accuracy and drive the quality of outcomes.

A future-focused approach to health care recognizes that the most effective care doesn’t arrive at the last minute, it arrives well before a patient’s health has deteriorated to the point of crisis. Unfortunately, that can be difficult in a time when so many communities are experiencing health care workforce shortages. Some estimates on primary care physicians (PCPs) indicate they see, on average, around 20 patients per day, work more than 50 hours per week and spend more than a quarter of that time dealing with non-clinical paperwork. That workload can make it difficult for any individual to keep track of every aspect of their patients’ health care needs.

Machine learning can bring important leverage to this tough environment: information that can make patient support timely and effective. When our care managers have advance notice that an individual would really benefit from more direct engagement, they can find opportunities to address upstream challenges before they escalate into a health crisis. Our predictive analytics platform is empowering our care managers to bring together providers and community organizations to help members find local solutions to unmet drivers of health needs, to initiate dialogue with pharmacists when members have trouble staying on track with their prescriptions, to help members who miss follow-up appointments reconnect with their providers and to connect members to mental health support.

These are remarkable advances — all accomplished with member privacy top of mind.