AI-guided DVT diagnosis in primary care: protocol for cohort with qualitative assessment.

Abstract

Background: Deep vein thrombosis (DVT), a formation of blood clots within deep veins, mostly of the proximal lower limb, has an annual incidence of 1-2 per 1000. Patients who are affected by multiple chronic health conditions and who experience limited mobility are at high risk of developing DVT. Traditional DVT diagnosis involves probabilistic assessment in primary care, followed by specialised ultrasound scans (USS), mainly conducted in hospitals. The emergence of point-of-care ultrasound (POCUS), coupled with artificial intelligence (AI) applications, has the potential to expand primary care diagnostic capabilities.

Aim: To assess the accuracy and acceptability of AI-guided POCUS for DVT diagnosis when performed by non-specialists in primary care.

Design & setting: Diagnostic cross-sectional study coupled with a qualitative evaluation conducted at primary care DVT clinics.

Method: First, a diagnostic test accuracy (DTA) study will investigate the accuracy of AI-guided POCUS in 500 individuals with suspected DVT, performed by healthcare assistants (HCAs). The reference standard is the standard of care of USS conducted by sonographers. Second, after receiving both scans, participants will be invited to complete a patient satisfaction survey (PSS). Finally, semi-structured interviews with 20 participants and four HCAs, and three sonographers will explore the acceptability of AI-guided POCUS DVT diagnosis.

Conclusion: This study will rigorously evaluate the accuracy and acceptability of AI-guided POCUS DVT diagnosis conducted by non-specialists in primary care.