ABSTRACT
Introduction: The COVID-19 pandemic ushered in a rapid, transformative adoption of telemedicine to maintain patient access to care. As clinicians made the shift from in-person to virtual practice, they faced a paucity of established and reliable clinical examination standards for virtual care settings. In this systematic review, we summarize the accuracy and reliability of virtual assessments compared with traditional in-person examination tools. Methods: We searched PubMed, Embase, Web of Science, and CINAHL from inception through September 2019 and included additional studies from handsearching of reference lists. We included studies that compared synchronous video (except allowing for audio-only modality for cardiopulmonary exams) with in-person clinical assessments of patients in various settings. We excluded behavioral health and dermatological assessments. Two investigators abstracted data using a predefined protocol. Results: A total of 64 studies were included and categorized into 5 clinical domains: neurological (N = 41), HEENT (head, eyes, ears, nose, and throat; N = 5), cardiopulmonary (N = 5), musculoskeletal (N = 8), and assessment of critically ill patients (N = 5). The cognitive assessment within the neurological exam was by far the most studied (N = 19) with the Mini-Mental Status Exam found to be highly reliable in multiple settings. Most studies showed relatively good reliability of the virtual assessment, although sample sizes were often small (<50 participants). Conclusions: Overall, virtual assessments performed similarly to in-person exam components for diagnostic accuracy but had a wide range of interrater reliability. The high heterogeneity in population, setting, and outcomes reported across studies render it difficult to draw broad conclusions on the most effective exam components to adopt into clinical practice. Further work is needed to identify virtual exam components that improve diagnostic accuracy.