HRCT diagnosis of diffuse parenchymal lung disease: inter-observer variation

Abstract

Background: This study was designed to measure inter-observer variation between thoracic radiologists in the diagnosis of diffuse parenchymal lung disease (DPLD) using high resolution computed tomography (HRCT) and to identify areas of difficulty where expertise, in the form of national panels, would be of particular value.

Methods: HRCT images of 131 patients with DPLD (from a tertiary referral hospital (n = 66) and regional teaching centres (n = 65)) were reviewed by 11 thoracic radiologists. Inter-observer variation for the first choice diagnosis was quantified using the unadjusted kappa coefficient of agreement. Observers stated differential diagnoses and assigned a percentage likelihood to each. A weighted kappa was calculated for the likelihood of each of the six most frequently diagnosed disease entities.

Results: Observer agreement on the first choice diagnosis was moderate for the entire cohort (kappa = 0.48) and was higher for cases from regional centres (kappa = 0.60) than for cases from the tertiary referral centre (kappa = 0.34). 62% of cases from regional teaching centres were diagnosed with high confidence and good observer agreement (kappa = 0.77). Non-specific interstitial pneumonia (NSIP) was in the differential diagnosis in most disagreements (55%). Weighted kappa values quantifying the likelihood of specific diseases were moderate to good (mean 0.57, range 0.49-0.70).

Conclusion: There is good agreement between thoracic radiologists for the HRCT diagnosis of DPLD encountered in regional teaching centres. However, cases diagnosed with low confidence, particularly where NSIP is considered as a differential diagnosis, may benefit from the expertise of a reference panel.