Real-World Deployment Success of SENSORA™


Summary: Read about the successful real-world implementation of SENSORA™, an AI-driven cardiac disease detection platform, at three high-volume healthcare centers in Miami, FL.

Authors: Valerie Rodriguez, Cody Hitchcock, Caroline Currie, Rosalie McDonough, John Prince, Emileigh Lastowski, Maikel Couto, Jose Martin, Elie Haddad, Keila Hoover, Premium Healthcare, Miami, FL, USA, Eko Health

Background: Valvular heart disease (VHD) poses significant health risks, particularly when left undiagnosed. Traditional screening methods have limitations in early VHD detection. Emerging technologies including AI powered digital stethoscopes show promise, but real world assessments are crucial to gauge their ease of implementation, true accuracy, and thus viability.

Methods: We assessed the implementation success of SENSORA™, an AI enabled cardiac disease detection platform, in identifying VHD at 3 high volume healthcare centers in Miami, FL, from Mar-Oct 2023. Medical Assistants (MAs) were trained to adapt platform-based cardiac screening of all patients during intake. Qualitative data were collected through surveys. Platform usage and success metrics were tracked over time. To assess performance, a cohort of patients aged 50+, screened from Mar-Aug 2023, were analyzed for VHD detection by SENSORA. Echocardiograms confirmed the presence of ≥moderate (mod+) disease and phonocardiogram recordings were annotated for murmur, together comprising the ground truth for clinically significant VHD. Data analysis included descriptive statistics, thematic clustering, and calculation of model performance metrics.

Results: Three MAs and 6 providers were surveyed pre and post deployment. Although 78% did not use AI in current practice, 100% thought the platform was easy to use and added value to their workflow. Challenges were addressed using an iterative co-adaptive design. Since deployment, there has been consistent usage (avg 316 recordings/week) and high (98%) exam success. For the 554 patients that were included in the performance analysis (median age 58 [IQR:54-63], 49.6% female), SENSORA correctly identified 10 patients with mod+ VHD, showing a 100% sensitivity and 84.7% specificity.

Conclusion: This case study highlights the platform’s smooth integration into the MA workflow in a fast paced clinical environment, as well as its excellent performance in disease detection. The successful deployment suggests that AI powered tools could increase the detection of mod+ VHD at the point of care, ultimately reducing the risk of adverse clinical outcomes of undetected VHD.

View Research -->

MKT-0002950