Duke Health partners with nference to reveal EHR insights, improve care

Duke Health has teamed up with nference to utilize the software company’s analytics platform to bolster medical research and patient care. Duke hopes the partnership will work to decrease the amount of time from “bench to bedside” by ushering in new treatments and technologies quicker.

Nference works with health systems to wrangle unexplored or under-explored data from data sources such as electronic medical records, according to the company. New understandings can improve care through nference's artificial intelligence platform by assessing the efficacy of treatment approaches and certain drugs. The longitudinal data of health systems like Duke provide a unique potential for healthcare insights, according to the tech company's executives.

"This strategic partnership with Duke is a key milestone toward developing a unique federated network of leading academic medical centers that will accelerate research, drive new therapeutic and diagnostic discoveries and fuel the creation of new ventures in global healthcare," Murali Aravamudan, co-founder and CEO of nference, said in a statement.

Aravamudan told Fierce Healthcare that nference is working closely with the Preston Robert Tisch Brain Tumor Center at Duke in order to advance the diagnosis and treatment of brain tumor patients.

"We will generate deep molecular data of brain tumor patients paired with the complete longitudinal EMR," Aravamudan said. "This unique dataset will enable discoveries of biomarkers for drug response and assessing the progression of disease, multi-modal non-invasive diagnostics throughout the entire patient journey and new therapeutics including immunotherapies."

A similar partnership exists between nference and Mayo Clinic. The two spun off Anumana, which uses AI and medical algorithms to reveal insights in Mayo Clinic’s patient data.

Anumana has commercialized algorithms for the early detection of diseases that often go undiagnosed until a critical issue is present, such as a weak or thickened heart pump, according to executives. Data explored can even include unstructured formats like providers' notes.   

Mayo Clinic’s dormant data, like raw electrocardiogram (ECG) pulses, were used to craft neural network algorithms in order to enable early detection of heart disease. Anumana’s AI ECG-based pulmonary hypertension early detection algorithm was granted breakthrough-device designation by the U.S. Food and Drug Administration last May.

The pair has demonstrated the effectiveness of the ECG AI to assess left ventricular function through the use of data collected in Apple Watches, as demonstrated in a prospective study. 

“As an academic health system, Duke Health’s mission is to advance the health of our community through innovation and the rapid translation of groundbreaking research,” said A. Eugene Washington, M.D., chancellor for health affairs at Duke and president and CEO of Duke University Health System, in a statement.

The goal of the partnership is to empower Duke Health's faculty, clinicians and staff to push for new medical breakthroughs, share insights and knowledge and better collaborate with peers in healthcare, Washington said.

Nference has been highlighted for its careful de-identification of patient data that still maintains the integrity of the data’s potential for health insights.  

The Cambridge, Massachusetts, company uses a “data behind glass” approach, meaning data do not exit the boundaries of the institution they originate from. Because the data are thoroughly de-identified, the research timeline is accelerated from the current average “bench to bedside” timeline of 17 years, according to the company.

Aravamudan told Fierce Healthcare that "the future of machine learning in healthcare depends upon three factors: multi-modal data, the number of patients and the heterogeneity of the data sources. We believe we are the first ever federated, meaning 'data behind glass,' deployment of multiple institutions' holistic clinical de-identified clinical data." 

During the COVID-19 pandemic, nference’s accelerated timeline was used to help determine emerging variants while aiding providers in caring for patients with the still misunderstood disease.

Along with Mayo Clinic, nference was the first to publish a real-world study of Anosmia/Dysgeusia as symptoms of COVID-19, according to the company. Aravamudan said this insight came from unstructured physicians' notes.

In December 2021, Duke joined the Mayo Clinic in an AI collaboration along with the University of California, Berkeley and DLA Piper with the goal of ensuring the safe deployment of algorithms in healthcare. The collaboration claimed to be the first of its kind and promised an open-source format.

Towards the end of last year, nference launched its new platform, nSights, which collects de-identified patient clinical data from academic medical centers and expanded its relationship with Mayo Clinic. The Mayo Clinic version of the platform has been dubbed Mayo Clinic Platform_Discover.

The new platform currently gathers insights from clinical notes, radiology results, lab tests and ECGs but hopes to add digital pathology and genomics data in the future.

To date, nference has raised $152.7 million in funding, including a $60 million series C round led by Mayo Clinic Ventures.