Paving the Way to Healthcare AI with Robust Data

1upHealth’s Chief Technology Officer Shares Perspective on HIMSS24

HIMSS24 is a wrap and as we were all looking with anticipation at how the world’s largest healthcare technology show would play into the newfound enthusiasm for artificial intelligence (AI) technology, we weren’t left disappointed. An accelerated shift towards data and AI was duly noted. Both the exhibition floor and the spotlight stage celebrated the potentially beneficial applications of AI to improve medical diagnosis, accelerate drug discovery, transform patient experience, enrich health data, and perform robotic surgery. 

It was often hard to “condense fact from the vapor[ware] of nuance,” as Neal Stephenson, the sci-fi writer who invented the term “Metaverse,” speculating an AI-augmented future way back in 1992, would say. The amount of excitement for the bright future AI-driven healthcare promises was met with at least as much criticism, substantiated by a fair amount of data-related challenges that need to be overcome before a true AI-driven healthcare revolution can take off. 

2024 is the Year to Derive Business Value from AI

Business value from AI is top of mind for enterprises in 2024. Whereas 2023 was the year of piloting AI technology, this year more organizations are seeking to demonstrate concrete business value from their portfolio of initiatives and bring pilots into production. 

Many vendors and healthcare organizations I spoke to at the conference have AI-enabled products in the final stages of development, aiming to introduce a higher order of intelligence and automation to boost internal operational productivity or deliver personalized customer experience. Others are seeking advice from data partners such as 1upHealth to initiate a business strategy for AI. As they try to keep up with this rapid technological disruption and assess the implications to their workflows and business models, it will be paramount to broaden the focus from open-ended experimentation with AI technologies to more investment in core AI assets and capabilities. 

The Modern Data Stack Paves the Way Toward True Interoperability

The road to advanced healthcare analytics and AI is sprinkled with a myriad of hurdles and potholes, not the least being the lack of robust data assets with proper governance and quality controls. 

During a fireside chat organized by HL7 DaVinci, Erin Landau, product leader at Oscar Health, and I  discussed the role of the modern data stack as a toolbox to fill the potholes and finally pave the way towards true interoperability. We discussed a plethora of technologies, commonly referred to as the modern data stack, that support today’s complex data value chain in many other industries and can help get healthcare there. They meet the requirements to collect, store, process, govern, analyze, and serve highly-diverse, large-scale data in a setting where open formats, shift-left observability, near real-time processing, self-service “dataops”, and fine-grained data access controls governed by federated domains have become the new standard. 

Transparency is a Crucial Component of Success

As more and more healthcare data is reformatted, integrated, and packaged up, we discussed how transparency is becoming an increasingly crucial component of success. Transparency in this context is defined from the perspective of the data consumer or data governance member as: the ability to understand the sources of the data set I am consuming (provenance); how it was transformed and aggregated (lineage); and what is the overall consistency, completeness, validity in terms of reference standards and ontologies (such as FHIR R4, Carin IG, etc.) (data quality); and how it complies to regulations in terms of context and purpose of use, explainability and differential privacy, data sovereignty and locality, etc. (data governance).

Transparency is a challenge in the current system as data comes from different sources and in different formats and travels through different domains – organizations each governed by different policies and standards for provenance, lineage, and quality. Even if AI productization is not a top priority this year, computable interoperability catalyzed by FHIR-native API endpoints and the CMS final rule will remain theoretical unless the industry gets the actual data that would run through those exchanges right.

1upHealth’s Role in the Modern Data Stack  

Several 1upHealth customers on stage and at the booth showcased how they are piloting an AI-powered member experience that is fueled by combined claims and clinical data in our FHIR-native platform. However, to take it to the next level, they rely on stakeholders in the ecosystem to have access to a wider variety of health data, providing the missing pieces in the puzzle to build the data set they need to productize robust AI applications. For example, Primary Record, one of our customers, enables families to connect (their patient records) with their local caregivers community, and elicit crowdsourcing to fill in gaps in the timeline with house visit gaps or other observations that otherwise wouldn’t be captured by EHRs. 

The industry is still early in its digital transformation and enterprise strategies to drive the migration to the cloud, where open standards-based data formats and technologies can be leveraged, are often hampered by a volatile environment mired by intense merger-and-acquisition activity. Healthcare organizations are now looking at the 1upHealth FHIR-native Data Platform to become their system of record where they aggregate robust data from a variety of sources. 

They look to 1upHealth’s thought leadership to be the guiding change agent as they decide what data they retain as they move to the cloud, so it can be served in different ways with the proper quality and governance controls. This includes different serving modes, ranging from FHIR APIs to serve mobile apps or reporting tools, to downstream analytical hyperscalers, such as Snowflake and Databricks, to fuel more advanced volume-intense analytics and AI.

The migration of data from proprietary or custom data systems and pipelines to an open standards-based and cloud-native modern data stack will accelerate the availability of both the quantity and quality of data required for higher orders of intelligence and automation in the industry. With a zettabyte-scale, yet largely untapped reservoir of data growing at an impressive rate of nearly 35% compound annual growth rate, healthcare represents the final frontier for transformative data application.

1upHealth’s Infrastructure Will Improve Lives

1upHealth’s mission is to provide the infrastructure that will enable the creation of the most extensive data cloud designed to leverage the comprehensive, longitudinal, and multifaceted patient records of tens of millions of Americans. This initiative not only aims to unlock new insights into patient care and population health, but also to navigate the complex challenges and seize the vast opportunities of sharing this invaluable data among thousands of stakeholders in the $4 trillion US healthcare economy. Payers, providers, and most importantly, patients have the common intention to explore the incredible potential of data to redefine healthcare, promising a future where we all live healthier, longer lives. Now we have to execute!