Fri, 12/28/2018

There are many great things coming in healthcare for 2019, and before diving into the last part of the “Cutting Through the Hype” series, here is what we have covered so far:

The Introduction:  Cut Through the Hype

  1. Analytics beyond the EHR, here
  2. Real life illustrations that define healthcare’s unique data challenges, here
  3. Data Governance – how to start, organize and build to sustain, here
  4. Leveraging governance to prioritize clinical and business objectives, here

The CEO of HIMSS North America, Hal Wolf, recently shared his thoughts on the challenges of taking increasing mountains of data and turning it into useful information and actionable intelligence across the healthcare delivery continuum. Because the hype of machine learning, AI, and NLP is prevalent in analytics, many have hope that they are the “silver bullet” to decipher this dilemma.  However, we must ask where they really fit in solving the “data rich/information poor” quagmire in which many providers and payers are stuck?  An honest look at the industry recommends a specific life line to pull them from this bog – a strong data strategy that capitalizes on current technology to ascend to the top of value based, consumer driven care. 

Speaking of current technology – let’s review when to build.  In my national work with the Healthcare Data Analytics Association (HDAA), the number of provider organizations I met that were capable of internally developing all technology needed for an enterprise level data strategy could be counted on one hand.  There may be a few payers that have these internal resources in sufficient numbers as well.  However, even though they may have the resources to pull this off with some degree of success (many of them started 10-20 years ago), that doesn’t mean it is the most cost efficient way to build or sustain.  There are also a few niche areas in healthcare that are so specific, it may be more affordable and effective to build something internally, my previous work in solid organ transplant comes to mind.  Generally speaking, because of the uniquely complex challenges associated with healthcare data (many of which were listed in previous blogs), “throwing smart people at the problem” quickly becomes affected by a law of diminishing returns due to turnover and similar issues unless the area of focus is narrow.

When to buy – in their hype cycle publication, Gartner’s Laura Craft describes an enterprise data strategy as a health data curation and enrichment hub, in short, a “Data Orchestration Platform”.  It requires the capability of connecting to 100’s of internal/external data sources then centrally profiling, enriching and cleansing these data with advanced match, merge and mastering across important domains such as patient, provider attribution, workforce and location.  Additionally, in healthcare it is important to effectively and centrally manage the 1.2M codes in all of the clinical terminologies and reference data sets.  Lastly, these data must be democratized by providing access to definitions and data lineage through governance and also publishing up to date, analyst-friendly self-service data marts and metrics cohorts. 

There are real pitfalls to avoid and pathways to development for an effective “Orchestration Hub”.  Ones such strategy is what we, in HDAA, commonly referred to as the “Frankenstack”.  This was an attempt to buy “best of bread” vendor products from various areas such as a data model, an ETL tool, a data quality package, reference data, registries management, geocoding, master data management, and on top of all of that, we would build out our BI and predictive tools.  The drawbacks to this approach were numerous.  Contracting and maintaining software license agreements with all of these vendors was extremely time consuming and expensive, and since they were not designed to work together, the organization is still responsible for developing integration between each tool in the stack (or paying consultant/contractor rates to have this done).  Additionally, it was very counterproductive when issues arose and vendors attempted to blame other product shortcomings vs. their own – there was no “one throat to choke” to get answers or ultimate accountability. 

This is the reason why, when I saw the offering Information Builders had developed in Omni-HealthData, I exclaimed, “you guys don’t know what you’ve got!”  It has everything Gartner described and more, in one integrated platform, developed to work seamlessly together.  Having your “data house in order” will generate ROI organizationally that enables a future investment in AI and machine learning, when the price point of these technologies becomes more palatable.  Additionally, a Data Orchestration Hub in healthcare is exactly what must come first as PwC, in their “Top health industry issues of 2018” report, calls out “an AI tool is only as good as the data it uses for decision-making. Companies should invest in finding, acquiring and creating good data, standardizing it and checking it for errors.”

In my next blog post, I’ll write about my experience speaking at Optum’s recent MinneAnalytics conference and how “Yesterday’s AI is Today’s Common Tech”.