Before the landmark drug Herceptin (Trastuzumab) was approved and released to treat metastatic breast cancer, it had plenty of critics who did not believe in it, and it almost died a “premature death” in development. Fortunately, at that time, the analysis of patient data from one of the clinical trials identified a clear benefit in a small group of patients that would otherwise have gone undiscovered: women living with HER2 mutation breast cancer – the hardest type to cure. That discovery turned the odds, and Herceptin went from the lab to clinic and has subsequently revolutionized the field of drug discovery and oncology treatment.
The science and health communities have made huge progress in mining the value from large data sets and seeing its potential to transform healthcare practice and outcomes, especially in the areas of clinical decision support (CDS) and precision medicine.
Yet healthcare – like many other industries undergoing digital transformation – is still facing the challenge of being data-rich but insights-poor.
To obtain actionable insights, “we need to be able to extract, analyze and treat the large scale of data that is currently unstructured, heterogeneous and locked up in silos, and combine them into sufficiently large data sets to be broadly used for AI to improve care,” said Olaf Lodbrok, Senior VP and General Manager of Precision Medicine at Elsevier.
The difficulty of obtaining large data sets is mainly due to legitimate privacy concerns around patient data sharing and the lack of mechanisms to automate the process of standardizing data. Yet, we are making headway in tackling these two fundamental issues, as Dr. Richard Loomis, Chief Informatics Officer of Clinical Solutions at Elsevier, explained:
Advances in interoperability standards, such as HL7 FHIR, combined with regulatory support, present an unprecedented opportunity to unlock the value of historically siloed clinical data, while at the same time presenting new challenges. Elsevier is focused on solving many of these challenges, such as ensuring appropriate de-identification and privacy and security measures, combining data from multiple sources, as well as aligning data longitudinally.
This challenge becomes more complex with the exponential growth of data and medical knowledge, making it impossible for us to humanly keep up, much less apply those insights to clinical practice in a timely and reliable manner.
“With growing digitization that enables us to collect vast amounts of data, we now have to start leveraging the value of data,” said Dr. Ian Chuang, Elsevier’s Chief Medical Officer for Europe, the Middle East, Africa, Latin America and Asia Pacific (EMEALAAP) . “By using machine learning and clinical decision support technology, we can empower knowledge and help augment and make the physician’s decision-making more precise, based on a more complete and holistic understanding of the patient: both the clinical attributes as well as the broader determinants of health and well-being.”
Therefore, if healthcare providers and systems want to realize the full potential of data to change the way healthcare is practiced, the focus needs to be on preparing clinical and patient data for deep data analytics.
Expediting the path from real-world evidence to practice
Much of the data that is being gathered today through routine care comes from multiple sources, such as electronic health records and health tracking devices. This data lacks quality but is broadly available and reflects realistic care settings and comorbidities. To turn this into credible, real-world evidence that supports clinical practice the data first need to be structured, coded and de-identified for AI to even begin.
Dr. Tim Hawkins, Managing Director of Clinical Solutions for EMEALAAP at Elsevier, highlighted that “investment into these areas is critical. Furthermore, effective use of evidence-based clinical decision support tools can help with the proper recording and structuring of clinical data, in addition to helping reduce unwarranted variations in care. Overall the completeness and quality of data produced can be greatly enhanced.”
Hospitals and other players also need to consider how to apply the necessary governance of clinical insights, real-world evidence and AI tools built on such data, as they are applied back into clinical practice. “Putting these processes in place will also be critical for AI to deliver improvements to long-held and complex clinical problems in patient management,” Hawkins added.
Elsevier is well positioned to undertake these processes while also improving the quality of the data, Lodbrok said: “Elsevier has a high-quality and large clinical evidence base and ontologies to better structure and combine large data sets, and to harvest insights from those data sets with our state-of-the-art information analytics capabilities.”
For example, Elsevier’s Medical Graph can predict thousands of medical conditions through data from the disease trajectories of millions of patients.
Turning data into personalized patient care
These process cannot stop at analyzing the data, however. To enable the physician to give personalized care to the patient, the data need to be made actionable with operationalized decision support, pushing the right information to the physician at the right time and place and for the right reason.
“This has the potential to democratize medical knowledge because it isn’t constrained to people, place or time,” Dr. Chuang said. “When I talk to hospitals, they recognize that having the knowledge and insights does not necessarily mean the care experience or outcome will improve, unless there is a systematic way of programming those insights back into the system. We have to pull the information together and be continuously updating what the physicians have access to when it comes to how they order, formulate and prescribe treatment plans. This is where Elsevier’s CDS solutions, such as Order Sets and Care Planning, can support physicians and clinicians to provide more precise and patient-centered care by pushing rapid evidence-based updates.”
This process of merging quality health data sets with information analytics and operationalized decision support must happen more rapidly as medical knowledge increases day by day. If it does not start now, the “bench-to-bedside lag” will continue to grow bigger, with more missed opportunities to advance healthcare outcomes.
Ultimately, what we hope to see more “Herceptin-type” stories.
AI want it now
Health data and AI holds great potential. However, the data is only as good as our ability to store, structure and apply the information in real world.
The next generation of advancing care in the age of AI will demand that healthcare providers and technology partners understand how to deliver and handle data responsibly and securely.
As modern medicine becomes more complex, Elsevier aims to be a supportive partner to physicians, healthcare professionals and hospitals. We are helping them combine health data with scientific medical knowledge to promote lifesaving predictions and enable more precise treatments for the betterment of human health.