Effects of Pulmonary Rehabilitation on Functional and Psychological Parameters in Post-Acute Sequelae of SARS-CoV-2 Infection (PASC) Patients (preprint)

Background COVID-19 survivors may develop long-term symptoms of fatigue, dyspnea, mental health issues, and functional limitations: a condition termed post-acute sequelae of COVID-19 (PASC). Pulmonary rehabilitation (PR) is a recommended treatment for PASC; however, there is a lack of data regarding PR’s effect on multiple health indices and the factors that influence patient outcomes.Methods We extracted patient demographic, comorbidity, and outcome data from Allegheny Health Network’s electronic medical records. Functionality test results were compared before and after PR, including 6-minute walk test (6MWT), chair rise repetitions (CR reps), timed up and go test (TUG), gait speed (Rehab gait), modified medical research council scale (MMRC), shortness of breath questionnaire (SOBQ), hospital anxiety and depression scale (HADS) and chronic obstructive pulmonary disease assessment test (CAT) scores. Multiple regression analysis was done to evaluate the effect of comorbidities and patient factors on patient responses to PR.Results The 55 patients included in this study had a mean time of 3.8 months between the initial COVID-19 diagnosis and the subsequent PASC diagnosis. Post-PR, patients signficantly improved in 6MWT, CR reps, TUG, Rehab gait, MMRC, SOBQ, HADS, and CAT scores. However, hypertension, diabetes, chronic lung diseases, being an outpatient, and receiving pharmacologic treatments (decadron, decadron + remdesivir, and decadron + remdesivir + tocilizumab) were associated with a poor response to PR.Conclusion Our study supports PR as an integrated model of care for PASC patients to improve several physical and mental health indices. The long-term effects of PR on patients’ functional status should be investigated in the future.

Learning Clinical Concepts for Predicting Risk of Progression to Severe COVID-19 (preprint)

With COVID-19 now pervasive, identification of high-risk individuals is crucial. Using data from a major healthcare provider in Southwestern Pennsylvania, we develop survival models predicting severe COVID-19 progression. In this endeavor, we face a tradeoff between more accurate models relying on many features and less accurate models relying on a few features aligned with clinician intuition. Complicating matters, many EHR features tend to be under-coded, degrading the accuracy of smaller models. In this study, we develop two sets of high-performance risk scores: (i) an unconstrained model built from all available features; and (ii) a pipeline that learns a small set of clinical concepts before training a risk predictor. Learned concepts boost performance over the corresponding features (C-index 0.858 vs. 0.844) and demonstrate improvements over (i) when evaluated out-of-sample (subsequent time periods). Our models outperform previous works (C-index 0.844-0.872 vs. 0.598-0.810).