Machine Learning in the Health Care Industry: Homing In on the KPIs That Matter

Health care is an industry where innovation predominates and demands for greater accountability have intensified. In this enormous, complex, and highly regulated sector, there is much to measure to improve patient outcomes, lower costs, and maintain trust. Importantly, concerns over patient privacy increasingly dominate discussions about using first-party data. Organizations increasingly confront the challenge of keeping their key performance indicators (KPIs) from becoming unwieldy and unmanageable. Dr. John Halamka, CIO of Beth Israel Deaconess Medical Center, hints at the challenge when he observes that in a complex organization like Beth Israel — comprising tens of thousands of doctors, nurses, IT staff, and other employees, not to mention patients — KPIs vary from situation to situation. “For me, there may be four or five what I’ll call high-level key performance indicators,” he says. “But for the IT operation, there are over a hundred.” That kind of misaligned and conflicted KPI overflow is not sustainable.

This KPI overload — along with a tendency to rely on intuitive decision-making — may explain why health care companies have lower levels of engagement with machine learning (ML) than other industries we surveyed. (See Figure 1.)

A Reliance on Intuition

In the winter of 2018, we surveyed 1,600 senior North American marketing executives and managers about their use of KPIs and the role of ML in their marketing activities; 425 were from the health care sector. Sixty-two percent of health care respondents say that their organizations are investing in new skills or training to make marketing more effective in using automation and ML; 63% of respondents in the overall sample responded this way. Seventy-two percent of health care respondents believe that their current functional (marketing-specific) KPIs could be better achieved with greater investment in automation and ML. In the overall sample, that figure was 74%. When asked to describe the proficiency of their organization’s marketing function in data-driven decision-making, 39% of health care respondents said they were more intuitive, while 29% said they were more data-driven (the remaining 32% claimed to be equally intuitive and data-driven).

To date, machine learning remains a point-solution approach rather than a comprehensive health care platform. Within the marketing function, little consensus exists among health care executives around which KPIs matter most for the industry, their organizations, and individual contributors.

This will surely change: In particular, the uncertain failures and successes of health care reform make a fundamental rethink of KPIs both inevitable and necessary. As in other industries, ML will drive that transformation. Health care marketers who recognize the opportunity to narrow their KPIs and to use KPIs as data sources for ML algorithms will be at an advantage, and health care companies already invested in ML will be poised to lead the way.

While health care is, overall, a more intuitive industry with less investment in ML, those companies that are committed to ML are deploying their investments in advantageous ways. For instance, they are significantly more likely to report that they can drill down to see the underlying data aggregated into their KPI components. Of health care companies investing in ML, 82% have this drill-down capacity, compared with only 58% of companies that are not investing.

With greater technological sophistication, however, comes the danger of “KPI creep” brought on by the proliferation of data: As one finds more things to measure, one risks being distracted by information that isn’t critical to organization strategy. When asked how many KPIs they directly manage, 44% of health care executives whose enterprises are investing in ML reported overseeing six or more. In contrast, of executives whose companies are not investing in ML, only 33% oversee as many KPIs. While other factors might be at play here, this data suggests that increased technological sophistication invites more opportunities to measure — an inclination that can work against the need to clearly align KPIs with enterprise goals.

Transition to Data-Driven Decision-Making

Despite greater access to greater volumes and variety of data, health care marketers remain overreliant on intuition to make decisions. The transition to data-savvy decision-making is a basic but essential step. Greater use of ML and more sophisticated data-driven outreach to patients and other stakeholders are essential moves for the industry’s future health. There are no silver bullets here. Developing the will to innovate around data as an asset is crucial. This requires a cultural transformation, not just regulatory reform.

Glenn Thomas, CMO of GE Healthcare, captures the potential: “Ultimately, we want to be massively relevant to our customer and create maximum value for that customer,” he says. “And there can be a tyranny of KPIs that drive you toward the average and to lose sight of the individual customer impact. I think that we’re moving toward a world where the availability of data, and the ability to manipulate and analyze that data, could move us toward the other direction: individual customer KPIs.” Health care marketers who are able to identify a few select KPI outcomes connected to these personalized offerings will have a distinct position, if not a distinct advantage, in their competitive environment.

Privacy concerns among consumers and regulatory limits on what can be shared and with whom will continue to pose key challenges, however. For instance, the desire to protect personal medical record privacy might conflict with the information sharing essential for predictive health care diagnostics.

In many large health care organizations, treating patients in a more holistic way requires internal partnerships, technological improvements, and data sharing between functions. Wider internal use of patient data increases the importance of data governance — how data is collected and used, what it means, and how it is shared. As marketers implement ML approaches to connect with patients and other stakeholders, they will need to be more aware of, and potentially contribute to, their organization’s data governance policies and practices. Data governance itself may well become part of the health care KPI portfolio.

As enterprises take these steps, it will be important to maintain focus on measuring what’s meaningful. Strengthening the connection between KPI parsimony and organizational performance among leaders and marketers is critical to their long-term success.