Solving Data Illiquidity in Healthcare

According to the Centers for Disease Control and Prevention, 87 percent of office-based physicians use some form of electronic health record (EHR) system today, and the growing adoption of EHRs within clinical operations has generated a large volume of data.

But healthcare captured less than 30 percent of the potential value the advances in big data made available over the last five years, and “little has been done to unlock and fully utilize the vast stores of data actually contained within (EHRs),” per the McKinsey report “The Age of Analytics.”

>> READ: How to Minimize Data Chaos in Healthcare

While analytics tools work well within select facilities and research communities, these vast data sets and the useful information within them are very complex to combine with data sets from outside organizations. The current state of data illiquidity makes it challenging to share data that operate well within an organization with other participants so they can seamlessly use the information and derive their own valuable insights. Although information transferability is a concern, so is the transferability of analytical models. Heuristics painstakingly developed by teams of operations experts and researchers rarely, if ever, see industrywide adoption.

Consider the complexity in collecting and aggregating a broad medical data set. The three largest EHRs combined still corner less than one-third of the market, and there are hundreds of active EHR vendors across the healthcare landscape, each of them bringing their own unique approach to the information transfer equation. Additionally, because many hospitals use more than one EHR, tracking down records for a single patient at a single hospital often requires connecting to multiple systems. To collect a broader population data set would require ubiquitous connection to all of the hundreds of EHR vendors across the country.

Vendor complexity is just the beginning. Once all the records are aggregated in a single location, the data remain illiquid, as there are dozens of data standards to deal with, such as HL7v2/3, FHIR, HIPAA, LOINC, proprietary standards and more. Add to that the flexibility each vendor has when deciding how they use some of the fields, and whether they deliver summaries versus detailed records. And to complicate matters further, there are no guarantees that the data are digitized, as many EHRs only offer to acquire records as printouts.

Much of this illiquidity stems from the organic evolution of entrenched groups in and around health information. The processes that sprang forth were filled with local administrative needs, which created tremendous waste and inefficiency. To my calculations, there are more than 3 billion combinations and permutations of the health data in question — a comparable level of complexity to our genome. Yet we’re not trying to uncover the mysteries of mankind. We are simply trying to build executable medical records that better serve patients.

As the data-driven world has grown up around healthcare, only now are advanced technologies beginning to break down the information silos. Retrieving, digitizing and delivering medical records is a complex, manually intensive and lengthy endeavor, and technology must be layered within all operations, not only to streamline data acquisition but also to make executable data available at scale and to get population-level data more quickly and affordably.

To unlock the power offered by new advanced technologies, seek a vendor that houses a mix of traditional and emerging technologies, including robotic process automation (RPA), artificial intelligence (AI), computer vision, natural language processing (NLP) and machine learning. All of these technologies serve vital functions:

• RPA can be used to streamline manually intensive and repetitive systematic tasks, increasing the speed and quality at which records are retrieved from the various end-point EHRs and specialty systems.
• AI, NLP and neural networks together can analyze the large volume of images and text received to extract, organize and provide context to coded content, dealing with ambiguous data and packaging the information in an agreed-upon standard.
• Finally, with AI and machine learning, an augmented workforce can be operated to increase quality of the records digitization and the continuous learning of the ecosystem, where every touchpoint is a learning opportunity.

Smarter, faster and more qualitative systems of information exchange will soon be the catalyst leading paradigm-shifting improvements in the U.S. care ecosystem:

• Arming doctors with relevant information about patients
• Increasing claims accuracy and accelerating providers’ payments
• Empowering universities and research organizations with research-grade data sets
• Correlating epidemics with the preparedness of field teams
• Alerting pharmacists with counter-interaction warnings

The first step to arriving at such an era depends almost entirely on adopting these emerging technologies throughout the industry and applying a systematic redesign of many of our operations and systems.

Ultimately, improving information exchange will help us — the healthcare industry professionals — avoid tens of thousands of unnecessary deaths per year, ensure high-quality patient care and extend the lives of Americans at aggregate.

Florian Quarre is the chief digital officer of Ciox Health.

Get the best insights in healthcare analytics directly to your inbox.

Related

The Slow, Frustrating Rise of the Electronic Health Record

Without Analytics, Hospitals Risk Becoming the Blockbuster Video of Healthcare

Targeting Diabetes with Big Data, Machine Learning, Real-Time Informatics