Statistical Models for the Assessment of Knowledge and Skills of Lung Ultrasound Among Respiratory Therapists

Lung ultrasound (LUS) is a bedside imaging tool widely used in intensive care settings for evaluating the pul-monary pathology of various medical and surgical cohorts. It is a radiation-free, portable, easily reproducible and cost-effective method. It provides real-time data which is vital for respiratory care of critically ill patients. The majority of research in both adult and pediatric LUS is focused on the feasibility and reliability of tools in the hands of intensive care and emergency physicians. Non-physicians, in particular respiratory therapists, are also trained in LUS. However, only limited studies are available on the skills and knowledge of LUS among respiratory therapists. The current study explores how age and educational background influence the skills and knowledge of LUS among 30 respiratory therapists. This paper compares five different models for the dependency of the bivariate outcome (skill and knowledge scores) on the bivariate exposure (age and educational background). The five models are independent linear regression, vector linear regression, Plackett copula regression, autoregressive models, and latent variable regression. While the goodness of fit of the models did not differ much, we find that the vector linear regression model and the Plackett copula model provide additional insight into the conditional association of skills and knowledge, and that the autoregressive model slightly outperforms the other four models.