Novel screening tool enables earlier diagnosis of IPF

A new screening tool, the zero-burden comorbidity risk score for idiopathic pulmonary fibrosis, led to earlier diagnosis of IPF and improved outcomes of disease-modifying therapies and other interventions, according to published data.

“These results demonstrate the crucial role machine learning/AI can play in diagnosis of rare diseases, and that patterns buried in the medical histories can be effectively exploited to substantially improve patient outcomes,” Ishanu Chattopadhyay, PhD, assistant professor of medicine at the University of Chicago, told Healio.

Chattopadhyay and colleagues wrote in Nature Medicine that nonspecific presentation of IPF, a lack of effective early screening tools, unclear pathobiology of early-stage IPF, and the need for invasive and expensive procedures for diagnostic confirmation have hindered early diagnosis.

The researchers tested a new screening tool for IPF in primary care settings that does not require new laboratory tests or recognition of early symptoms.

“Using subtle comorbidity signatures identified from the history of medical encounters of individuals,” the researchers wrote, “we developed an algorithm, called the zero-burden comorbidity risk score for IPF (ZCoR-IPF), to predict the future risk of an IPF diagnosis.”

The risk score utilized a national insurance claims database and was validated on three independent databases.

According to Chattopadhyay, because IPF is a relatively rare disease, affecting fewer than five in 10,000 people, it poses challenges to statistical characterization.

“We developed new EHR processing tools to distill the subtle predictive signatures,” he said.

For the study, researchers included 2,983,215 participants (age range, 45-90 years) with 54,247 IPF diagnoses.

Data indicated that the algorithm attained positive likelihood ratios above 30 at a specificity of 0.99 across different cohorts, for both sexes and for participants with different risk states and a history of confounding diseases.

Furthermore, ZCoR-IPF’s area under the receiver-operating characteristic curve for predicting IPF surpassed 0.88 at 1 year before a conventional diagnosis and was roughly 0.84 at 4 years before a conventional diagnosis.

“The key point of this study is not diagnosis though,” Chattopadhyay said. “The ZCoR algorithm is aimed at primary care clinics, suggesting a universal screen that triggers the pulmonology referral for confirmatory diagnosis. These results are clinically useful. While IPF still has no cure short of a lung transplant, and disease-modifying drugs do not work for everyone, studies have shown that earlier diagnosis by even 1 year reduces the number of hospitalizations significantly.”

For more information:

Ishanu Chattopadhyay, PhD, can be reached at ishanu@uchicago.edu; Twitter: @ishanu_ch.