The providing multidisciplinary ILD diagnoses (PROMISE) study - study design of the national registry of Japan facilitating interactive online multidisciplinary discussion diagnosis.

Background: Multidisciplinary discussion (MDD), in which physicians, radiologists, and pathologists communicate and diagnose together, has been reported to improve diagnostic accuracy compared to diagnoses made solely by physicians. However, even among experts, diagnostic concordance of MDD is not always good, and some patients may not receive a specific diagnosis due to insufficient findings. A provisional diagnosis based on the ontology with a diagnostic confidence level has recently been proposed. Additionally, we developed an artificial intelligence model to differentiate idiopathic pulmonary fibrosis (IPF) from other chronic interstitial lung diseases (ILD)s, which needs validation in a broader population.

Methods: This prospective nationwide ILD registry has recruited patients with newly diagnosed ILD at the referral respiratory hospitals in Japan and provides rapid MDD diagnoses and treatment recommendations through a central online MDD platform with a 3-year follow-up period. A modified diagnostic ontology is used. If no diagnosis reaches more than 50% certainty, the diagnosis is unclassifiable ILD. If multiple diseases are expected, the diagnosis with a high probability takes precedence. If the confidence levels for the top two possible diagnoses are equal, the diagnosis can be unclassifiable. The registry uses tentative diagnostic criteria for nonspecific interstitial pneumonia with organising pneumonia and smoking-related ILD not otherwise specified as possible new entities. Central MDD diagnosticians review the clinical data, test results, radiology images, and pathological specimens on a dedicated website and conduct MDD diagnoses using online meetings with a cloud-based reporting system. This study aims to (1) provide MDD diagnoses with treatment recommendations; (2) determine the overall ILD rates in Japan; (3) clarify the reasons for unclassifiable ILDs; (4) evaluate possible new disease entities; (5) identify progressive phenotypes and create a clinical prediction model; (6) measure the agreement rate between institutional and central diagnoses in ILD referral and non-referral centres; (7) identify key factors for each specific ILD diagnosis; and (8) create a new disease classification system based on treatment strategies, including the use of antifibrotic drugs.

Discussion: This study will provide ILD frequencies, including new entities, using central MDD on dedicated online systems, and develop a machine learning model for ILD diagnosis and prognosis prediction.

Trial registration: UMIN-CTR Clinical Trial Registry (UMIN000040678).