Syntropy, Evidium explain their data-centric healthcare tieup
Syntropy is combining its secure data collaboration environment with Evidium’s referenced AI capabilities to make medical knowledge more computational and improve data quality for clinical research and practice. Syntropy’s product development team leader, Michael Vande-Vrede answered my questions about what it all means in practice.
What are the key data quality challenges in clinical research and clinical practice?
Data quality challenges in clinical research and clinical practice fall under many different areas, and we focus on three: data collection and integration, data curation and normalization, and security challenges. Data quality in clinical research often faces challenges such as inaccurate entry due to human error, inconsistencies from varied collection methods, and lack of standardization across sources. Additional issues include data duplication, outdated information, and concerns regarding data security and patient privacy. It's also vital to consider bias in data collection and the need for data provenance in multi-center trials. Addressing these challenges requires a comprehensive governance model to ensure the proper oversight of decisions, robust data pipelines with comprehensive normalization and mapping functionality to bring the data together in a unified manner, and frictionless feedback mechanisms driving a dedication to continuous improvement in data management.
Despite an overabundance of clinical data and evidence, much of it is not fit-for-purpose to provide actionable insights to spur new research knowledge. While it is tempting to think that poor data quality can be overcome by increasing the volume of data or by throwing AI at the problem to leverage data to its full extent, it must be fit for the central question researchers are trying to solve. Our recent collaboration with Evidium combines their Referenced AI capabilities and frontline access to clinical workflow with our secure data curation and collaboration environment to address these challenges by improving data quality and the speed at which valuable information is extracted. Together we have developed an AI Operating System (OS) for healthcare, where we will reshape the complex relationships between data, knowledge, and evidence to convert the insights into AI-ready, scalable intelligence that improves data quality for research, operational insights, and ultimately patient outcomes. AI OS provides essential capabilities to enable trustworthy and steerable AI.
All of this is seamlessly enabled on our platform with security built-in and managed by a rich and powerful governance model.
What does it mean to make knowledge "more computational”?
Making clinical knowledge "more computational" is the process of transforming healthcare data and clinical information and knowledge, no matter what form it is in be it a pdf, text, or image, into a structured, standardized, and machine-computable format. This process involves modeling clinical knowledge, with references back to source evidence, with special attention to the complex relationships between clinical concepts. Multiple AI models, including LLMs, can then directly use and be grounded in this computational knowledge, aiding trustworthiness and impact. Importantly, this helps solve hallucinations in LLMs and provide explainability. We call this “Referenced AI”.
Data Structuring: to make knowledge "more computational" we are structuring the data by converting vast amounts of raw, unstructured clinical data into structured formats, creating a foundation that computational tools can work upon. Structured data makes it easier to run algorithms, data analytics, and AI/ML models, transforming raw information into actionable insights.
Interoperability: clinical data often resides in silos due to various systems used in healthcare. Our recent collaboration emphasizes creating standardized data structures, which promote interoperability. When systems communicate effectively with one another, computational tools can access and analyze data more efficiently.
Data Analysis and Insights Extraction: the aim is not just to structure and standardize data but also to extract meaningful insights from it. By overlaying the structured patient information onto Evidium’s Referenced AI knowledge model, we can detect data quality issues as well as deliver clinical guidelines and knowledge that directly relate to the particular patient and can represent the unique needs of a specific organization’s patients and care model. For example, we can find contradictions in data that make little clinical sense.
By making clinical knowledge "more computational" we are able to harness the power of computational tools to transform raw clinical data into structured, meaningful, and actionable knowledge, ultimately enhancing patient care, research, and the overall healthcare landscape.
What does “AI-ready knowledge at the elemental level” mean? Why is it important?
In the era of AI tools, particularly large language models (LLMs) that have undergone extensive training on wide-ranging data, it's imperative to constrain the solution set and advice to those supported by evidence and established knowledge. This restriction helps to prevent the generation of unfounded assertions and recommendations based on unverified opinions. The Referenced AI's fundamental architecture provides a bedrock of essential knowledge, anchored to actual evidence, including peer-reviewed studies and other forms of published data. It also maps the vital interconnections among clinical concepts. This base of Referenced AI then informs the inferences and recommendations produced when posing questions or determining the next best action using AI tools built upon this data layer. In contrast, LLMs acquire knowledge through next-word prediction techniques and are subsequently refined by thousands of human trainers to attempt to yield "correct" answers to specific questions. While LLMs are potent and valuable, employed in AI operating systems and Referenced AI, they represent just one component of a comprehensive AI architecture.
What is the relationship between data, knowledge, and evidence? How will the AI OS help?
The relationship between data, knowledge, and evidence is intertwined and crucial for advancements in healthcare. Data represents raw, unprocessed information collected from various sources, such as patient records or clinical trials. Knowledge is the insights and understanding derived from analyzing this data. Evidence refers to verified information that supports a particular claim or conclusion validated through peer-reviewed expert analysis and methods, often used to guide clinical decisions, and establish best practices in healthcare. The AI OS developed because our collaboration aims to streamline and enhance this data/knowledge/evidence relationship through a foundation of high-quality source data managed with robust governance. This data can be interpreted against the Referenced AI foundation to infer new knowledge and insights. As this new knowledge is further refined into evidence, that evidence can be put back into the Reference AI architecture to enhance and improve the underlying evidence.
Who are the target customers for this initiative?
We target any organization in the health care space that has a need for end-to-end data management and improvement, with an initial focus on health care delivery. By partnering with Evidium, we extend our services to cover a more extensive range of needs within these organizations. Our collaboration is not limited to just organizing and streamlining the existing data; it also incorporates the deployment of advanced tools at the point of data generation.
Our collaboration is designed to support healthcare organizations throughout the entire data lifecycle—from the initial data capture and electronic record-keeping to the final stages of data analysis and application in evidence-based practice. We understand that healthcare data management is not a static field; it is constantly evolving with advancements in technology and changes in healthcare regulations. As such, our initiative is also aimed at those who are looking to future-proof their operations against such changes. We provide adaptive systems that can adjust to new types of data inputs, regulatory requirements, and analytical techniques, ensuring that healthcare providers can remain focused on delivering high-quality care without being hindered by data management challenges.
What will be the impact and how will you measure it?
We expect the impact to be significant across the customers of both of our organizations. Not only will Syntropy’s data platform enable more efficient research and insights generation for individual researchers and their institutions, but we will see more research collaborations within the industry, decreasing the time to scientific discovery. This improved data will provide a more accurate view of both the patient and the clinical workflows of the organization. This allows Evidium’s Referenced AI to improve wayfinding and next-best-action choices at the point of care delivery. The combination enables a feedback loop driven by high quality data that allows for rapid evaluation of clinical decisions and workflows to rapidly improve care delivery. We have a lot of other exciting initiatives that will shake up how clinical knowledge and evidence is authored, tested, and distributed so stay tuned!
Syntropy itself is a partnership between two companies, which are now entering into a partnership with another company. Why are three players needed and does it make things unwieldy?
Healthcare is an intricate ecosystem fraught with nuanced and complex challenges that necessitate deliberate and well-considered solutions. We are privileged to have three synergistic organizations dedicated to enhancing the scientific methodology and the delivery of patient care. This synergy ensures our endeavors are anything but cumbersome. Each entity contributes unique strengths and shares a commitment to collaborative innovation in areas of shared expertise. Palantir offers a premier data and analytics platform, forming the bedrock of the Syntropy solution, while EMD Digital provides in-depth knowledge in key healthcare sectors, including clinical research and life sciences. Evidium, with its pioneering Referenced AI, ensures reliable and directed AI applications by harnessing structured evidence and profound knowledge, bolstering outcomes through Syntropy's robust data platform and expertise. Collectively, we cover the complete spectrum of the data lifecycle, enhancing data integrity at every phase. This comprehensive approach is instrumental in propelling research, clinical discoveries, and, most crucially, patient care and outcomes, navigating the complexities of healthcare with thoughtfully crafted solutions.
—-
Michael Vande-Vrede is the Vice President, Head of Digital Products at EMD Digital Inc. and runs the product development team at Syntropy. He is driven by his passion to make science faster by enabling collaboration across the healthcare ecosystem. He is not impressed by the AI hype frenzy that has taken over the industry in recent years and firmly believes that without high-quality data and transparency, AI has no chance of moving the needle within the healthcare industry.
Michael was part of the team that directly transformed a 350-year-old chemical company with more than 53,000 employees operating in 60 countries into a leading science and technology company. He was one of the key individuals who spearheaded the work at Sigma Aldrich that culminated in the creation of the largest eCommerce platform in the life science industry, generating more than $1 billion in annual revenue and serving as a catalyst for the acquisition of Sigma-Aldrich by Merck KGaA, Darmstadt, Germany in 2015 for $17 billion.
