Caution and Enthusiasm: Using Data and Analytics in Healthcare

As we’ve all read in countless articles and posts over the past few years, the rapidly expanding use of data and analytics in healthcare presents extraordinary potential for improving people’s lives.

While I happen to share that view, I have also seen many technological silver bullets over the years that did not live up to their full potential or resulted in harmful unforeseen side effects.

It is with this cautious enthusiasm that I share the following thoughts to help optimize the effectiveness of data and analytics as a key element in the longstanding central purpose of our healthcare system: Providing high-quality care to all who need it in a cost-effective way.

These are taken from my opening remarks at the 1/31/19 NEJM Catalyst event, “Provider-Driven Data Analytics to Improve Outcomes,” held at Cedars-Sinai.

1.    Transparency in assumptions behind algorithms, AI and data analysis:

One of the key principles of research and science is transparency—the open sharing of methods and assumptions that determines if the research can stand up to scrutiny (and reproducibility) by others. This is essential in preventing bias and premature conclusions from inadvertently harming people. Although artificial intelligence and data analytics may appear to be completely “machine driven” and thus immune from bias or premature conclusions, they are driven at their foundation by underlying, human-based assumptions and strategies. Thus, we need to have the same level of transparency. 

2.    Pragmatism and disciplined thinking are just as important in optimizing data and analytics as in optimizing engagement of staff and use of facilities: 

Endless amounts of data provide endless possible conclusions that can be drawn from the data. Just as healthcare leaders use disciplined thinking to determine how to optimally engage staff and deploy physical resources and equipment, we should use the same high level of pragmatic innovation in determining when and how to use data and analytics.

3.    Adopt clinical and operational changes indicated by data and analytics:

While this one seems like a no-brainer—who wouldn’t adopt changes indicated by data and analytics?—it’s actually one of the most common pitfalls in any organization. The learnings from data and analytics will frequently challenge conventional wisdom. As a result, leaders will then likely hear all sorts of challenges to the data: it’s inconclusive, it’s inaccurate, the methodology was wrong, etc. There may be additional barriers to implementing changes indicated by the data and analytics, especially when those changes affect a lot of people and processes throughout an organization. This is when leadership is crucial: determining the most effective, long-term way to engage people in making change needed by the organization. Having facts and data are necessary, of course, but are not sufficient. Those intangible leadership qualities of listening, engaging, building support and being flexible are crucial to the process. 

4.    Healthcare remains a team sport, with humans at its center: 

Although technology, including data and analytics, is more important than ever in healthcare, it will always be in support of humans: patients and families, physicians, nurses, pharmacists and other healthcare professionals. And these people are much more effective and efficient when they work as a team on behalf of the patient, not as siloed individuals. So as we develop new ways to use data and analytics, we must ensure that it is in support of this.