COPD Foundation Aims for Model to Spot People Not Yet Diagnosed
The COPD Foundation is partnering with Savana, a machine learning and artificial intelligence company, to develop a way of better identifying people with chronic obstructive pulmonary disease (COPD) who have not yet been diagnosed, or are at high risk of this disease.
De-identified data in electronic medical records will be used to create a model that might predict people likely to have or develop COPD. It will also search for patterns in medication use that may be linked to better clinical outcomes.
“Our new partnership with Savana will complement and extend The COPD Foundation’s ongoing program of research — specifically in the drive to find young individuals with disease and understand how they are diagnosed and treated,” Ruth Tal-Singer, PhD, the nonprofit foundation’s president and chief science officer, said in a press release.
COPD360Net, the foundation’s program to support the development of digital health tools, medical devices, and therapies, will make use of the predictive model that comes from this partnership.
“Through our Digital Health and Therapeutics Accelerator Network, COPD360Net, we are poised to conduct innovative clinical trials where we can utilize the predictive model developed from this relationship to make a difference for our community,” Tal-Singer said.
Artificial intelligence is focused on building machines that are trained to more quickly do tasks that people can do, although much more slowly. To train machines, researchers use deep learning, a type of machine learning that looks for patterns in data.
EHRead, a software technology developed by Savana, uses natural language processing — a branch of computer science that deals with the interaction between computers and written and spoken human language — to understand what is written in electronic medical records.
It also uses deep learning to automatically process, structure, and build a predictive model based on the information available. The free-text, unstructured fields of electronic medical records, in particular, are thought to contain a large amount of useful clinical information, including the signs and symptoms of disease.
As such, they can be a rich source of real-world evidence, helping researchers to better understand how patients behave or respond to a specific treatment in everyday clinical practice.
Importantly, the methodology ensures that all identifiable personal information has been stripped from the records before use, or de-identified, to safeguard the privacy and security of patient data.
“We are absolutely delighted to be supporting the work of The COPD Foundation,” said Ignacio Medrano, founder and chief medical officer of Savana.
“A partnership such as this, with a US-based, not-for-profit health organization, is the first of its kind for Savana. It marks another milestone in our growing international ecosystem of researchers, research institutions and sponsors seeking to use innovative deep [real-world evidence] studies to identify unmet need, gain new insights and answer novel questions in respiratory health and many other specialties,” Medrano added.
A machine learning approach by Savana is also being used in BigCOPData, a European Commission-funded study to identify factors associated with hospitalization in COPD patients across North America and Europe. It aims to create a predictive risk model of hospitalizations in this patient population.