3 Reasons Why Comparative Analytics, Predictive Analytics, and NLP Won’t Solve Healthcare’s Problems
I had a recent opportunity to engage in an online discussion with a well-known healthcare analytics vendor about the value of comparative analytics, predictive analytics, and natural language processing (NLP) in healthcare. This vendor was describing a beautiful new world of the future, in which comparative data, in particular, would be the cornerstone of our industry’s turnaround. The executive summary of my response: Beware the smoke and mirrors.
My full response is below.
Comparative Data Doesn’t Drive Improvement
We’ve had comparative data for years in the U.S. healthcare system and it hasn’t moved the needle towards better, at all. In fact, the latest OECD data ranks the U.S. even worse than we’ve ever been on healthcare quality and cost. Comparative data, like the OECD, is interesting and certainly worth looking at, but it’s far from enough to drive improvements in an organization down to the individual patient. To drive that sort of change, you have to get your head and hands dirty in your own data ecosystem, not somebody else’s that is at best a rough facsimile of your organization. There are too many variables and variations in healthcare delivery right now that add too much noise to the data to make comparative analytics as valuable as some pundits advocate. We don’t even have an industry standard and clinically precise definition of patients that should be included (and excluded from routine management) in a diabetes registry, much less the other 15 chronic diseases and syndromes we should be managing.
Predictive Analytics Fails to Include Outcomes
We’ve also had predictive analytics supporting risk stratification for years in healthcare, particularly in case management, but without outcomes data, what are we left to predict? Readmissions. That’s a sad state of affairs. Before we start believing that predictive analytics is going to change the healthcare world, we need to understand how it works, technically and programmatically. Without protocol and patient-specific outcomes data, predictive analytics is largely vendor smoke and mirrors in all but a very small number of use cases.
Gaps in Healthcare Industry Data Limits the Effectiveness of NLP
We’ve had NLP for years in healthcare, as well, with essentially no impact on the industry. When I joined healthcare, I brought a deep background in NLP and predictive analytics with me from the military, national intelligence, and credit reporting environments in hopes that we could revolutionize the industry. We’ve made incremental progress, but there are fundamental gaps in our industry’s data ecosystem—missing pieces of the data puzzle—that inherently limit what we can achieve with NLP. Google revolutionized the world of NLP, but Google leveraged a metadata ecosystem that is layered on top of traditional NLP strategies to achieve the revolution. In healthcare, we don’t have the same metadata ecosystem within the current generation of EMRs, much less across EMRs. In today’s EMRs, we have little more than expensive word processors. I keep hoping that the Googles, Facebooks, and Amazons of the world will quietly build a new generation EMR, but with $29B in federal money now squandered on our existing generation of EMRs, there’s very little motivation in the market for innovation.
A Structured Approach: Moving up the Analytic Adoption Model
Does my cynicism and caution imply that we should turn away from comparative data, predictive modeling, or NLP? Absolutely not. I’ve been advocating these analytic tools for 17 years in healthcare. It implies that we should take advantage of the easy value analytic victories, first. Don’t chase the asymptote. Don’t chase the latest fad and vendor hype. Deliberately, but quickly, move your organization up the levels of the Healthcare Analytic Adoption Model. This model draws upon lessons learned from the HIMSS EHR Adoption Model and describes a similar approach for assessing the adoption of analytics in healthcare. It’s going to take at least five years, maybe longer, and a new generation of EMRs, patient reported outcomes systems, and activity based cost accounting systems before we can close the gaps in our data ecosystem to make predictive models and NLP widely valuable in the industry. Comparative data will not be as valuable as it should be until we squeeze the variability out of our healthcare practices and standardize our data definitions of diseases and syndromes. If that happens in five years, I’d be pleasantly amazed. Variability analysis, not benchmarking, might be the most useful application of comparative data in our current healthcare environment.
Beware the vendors that oversell what’s possible, my friends. We are in the uphill climb of the hype cycle right now.