Evolution of Analytics
We’ve certainly had plenty of activity regarding analytics in healthcare over the last decades, and some progress. However, the real era of advance is just starting. Whereas in the past, much of the discussion centered on two areas: data warehouses and business intelligence - each of these initiatives for the most part missed the mark. Data warehouses were too broad and expensive, and suffered from an industry lack of interoperability and data lag. BI initiatives, while featuring increasingly impressive interfaces and visualization, were often only descriptive, too narrow, and departmentally focused. Neither data warehouses nor BI had the clinical/financial and related information and data or powerful-enough algorithms (most relied on statistics) to drive the more comprehensive analysis needed for significant return on investment.
Fast forward to today. We finally have the EHRs in place (driven by MU and especially supported by the consolidation to a few enterprise EHRs) and interoperability is improving (via APIs, FHIR, app store programs). We’re even starting to tap into many other sources of data (e.g., social determinants of health, streaming medical device data, genetic data). We’ve also seen huge strides in predictive algorithms, particularly with ML/AI – made possible by high-performance hardware. And although PHM and value-based care has not really taken-off yet, many leading health systems have at least some initiatives in this space and all signs point us in that needed direction.
So, we’re all set, right? This is just going to continue to unfold. Well, as you know, it’s much harder than that. Inertia, regulation, competing priorities, vendor decisions, costs, and the complexity of healthcare and payment reform all make progress difficult not to mention cyber-security threats, changes in administrations, etc. What does it take to succeed?
Leading the Way
Most of the progress in health IT has come from the top of the market – the academic medical centers, large health systems, and their major vendors. From an HCO perspective, this represents the top 5-15 percent or so of providers in the market. And while these HCOs may cooperate/share with each other to some degree, their vendors have mostly not. Although the rigors of the MU program have whittled the market down to just three major billion-dollar-plus enterprise EHR vendors (Cerner, Epic, and Allscripts), there are still lots of ambulatory and specialty vendors, and competition is severe.
Organizations that have multiple EHRs (and that’s most of them) face tough challenges – particularly when it comes to bringing data together for analytics from EHR and non-EHR sources. Although the R&D expenditures of the leading EHR players in many cases far outpaces the revenue of most of the other vendors, it’s only recently that these enterprise players have turned major attention to analytics – with more experience dealing with in-system data vs out-of-system data sources.
Age and Sweet Spot
Let’s consider the age of companies and their ability to innovate and compete. If they are too young, they don’t have the experience, clients, processes, and partners. If they don’t advance relatively quickly, their investors will pull the plug. If they are too old, and they have too many layers of management, legacy infrastructure, and product history to be supported, or if they are a public company with stockholders/boards that expect a continued return then they may be forcing less than optimal quarterly-driven decisions. At least 10 years old and a few dozen clients is likely a pretty good place to be.
Organic or Acquired?
One area that has tripped up many health IT vendors is growth thru acquisition. Epic and MEDITECH owe a major reason to their success to the control enabled by organic growth. Cerner has advanced thru mainly organic growth – while they have done quite a few acquisitions, including the large acquisition of Siemens a few years ago, they have been better than most in the assimilation process – mostly via using what they acquired to bolster their own products.
Allscripts is the only surviving enterprise EHR company to grow by acquisition and attempt to keep multiple-products alive (it has “merged” with Eclipsys and many others), but it’s been really hard for them technically and culturally with many missteps. McKesson is the poster child for not succeeding with an acquisition strategy - it has sold its health IT capabilities off in pieces. GE, while refocusing on some key areas today, has also failed as an enterprise vendor via an acquisition strategy.
One way around the organic vs. acquired approach is adoption of open source. Health Catalyst, for example, has taken a relatively unique open-source approach when it comes to AI/ML. With advances around the world coming quickly in this space, this gives them the ability to rapidly adopt these technologies without the challenges/limitations of an acquisition.
The Competitive Landscape
Let’s consider at least three categories of companies competing in health IT analytics: healthcare analytics companies, EHRs (with growing analytics capabilities); and large tech companies (e.g., IBM, Google, Microsoft, Amazon). Note: when it comes to user interfaces, visualization, etc., most HCOs and analytics vendors, at least for now, interface their analytics solutions with products like Tableau and QlikView.
Healthcare Analytics Companies: I recently had conversations with two dozen companies known for healthcare analytics based on multiple lists, analyst rankings and media articles. While all had a firm handle on the basics of data aggregation, normalization, descriptive analytics, and the importance of easy-to-use front ends (MedeAnalytics was especially established, as well as Verscend for the payers), when it came to advanced analytics (predictive and prescriptive), the field rapidly shrank, and a number that claimed this used proprietary algorithms (e.g., Ayasdi, Jvion). Most of the companies stated their clients were still struggling with the basics. A few even questioned whether more than a handful of organizations and vendors were even doing advanced analytics. While use of ML may have been mentioned in product marketing, the primary approach was still statistics and manual effort. Health Catalyst and its use of AI/ML is in a leading position here (check out catalyst.ai and healthcare.ai). Medial EarlySign also uses ML - they recently raised significant funding.
EHR Companies: Having over the last few years attended multiple major EHR conferences and sat down with data scientists at these companies, I can tell you that they are further along than most people think. Epic has invested heavily in a portfolio of data warehouse, data sharing (within and across clients) and PHM initiatives. They have platformed/branded their AI efforts and built a significant number of models that are already being shared across many of their clients. Cerner, while not as formalized, also has a data math/science team accessible to clients and Cerner consulting groups and continues to advance their PHM initiatives for their large client base. Health Catalyst got into a bit of a heated situation with Epic a while back when it spoke to greater aspirations that superseded EHR functionality, although this seems to be in the past. Many Health Catalyst clients use Epic, and a number told me they like having options with two vendors both advancing the industry. Health Catalyst is also starting to more fully embrace a partner ecosystem – an approach Allscripts, Athenahealth and Cerner have been especially successful with.
Large Tech Companies: IBM is the most well-known for its Watson Health division and its many healthcare analytics acquisitions. However, its price point and marketing/management overhead present challenges to even the largest of HCOs. Google with Alphabet/Verily certainly has the analytics firepower and cloud resources to delve deep into healthcare but has yet to gain the industry experience to dominate. Microsoft provides advanced tools but leaves it up to its ecosystem to develop solutions. Amazon, while more application focused towards the supply chain and pharma, has been more of a visionary threat than a near-term challenger. Apple, while pursuing Healthcare, has not yet launched an advanced analytics play. Companies like Oracle, SAP, SAS, and other giants have (perhaps sadly) stuck so much to their cross-industry guns that what could have been a strong position for them in healthcare seldom places them on any HCO’s advanced analytics short list.
Future
The good news is that when it comes to advanced analytics, healthcare providers have a growing list of options. While only a handful of healthcare analytics companies have yet to move beyond a descriptive/reporting focus, as their clients advance in their needs, they will likely keep pace. They have an advantage in dealing with disparate data sources. Leading EHR vendors are already making significant investments in advanced analytics, and have the benefit of close integration with user workflow. And although the leading large tech companies may not have deep roots in healthcare, their technical advances are leading the way in how data and analytics can be leveraged at scale.
Ken Kleinberg, Principal of Healthe-Motion, is an independent health IT analyst- he’s spent the last two decades with firms including Chilmark Research, The Advisory Board Company, and Gartner.