PC3009: Exploring the Real World Performance of an Artificial Intelligence-based Diagnosis Aid for Autism: An Aggregate and De-identified Analysis of Early Canvas Dx Prescription and Output Data Post-market Authorization

Background: On June 2nd 2021 the Food and Drug Administration (FDA) granted market authorization to Canvas Dx, an artificial intelligence-based Software as a Medical Device designed to aid healthcare providers (HCPs) diagnose autism spectrum disorder (ASD) in children at risk for developmental delay, aged 18 through 72 months. The device captures features maximally predictive of ASD from three distinct inputs (see figure 1), and combines them in a mathematical vector for machine learning analysis and classification to produce an output of ‘ASD positive’, ‘ASD negative’ or ‘no result (indeterminate).’ Results from the double-blinded, multi-site, prospective, active comparator cohort pivotal trial were promising. Device output PPV for all study completers was 80.8% [95%CI, 70.3%-88.8%] and NPV was 98.3% [90.6%-100%]. Additional data is needed, however, to learn about device performance when integrated into clinical practice. Early post-market authorization prescription and output data may help shed light on device performance in real-world settings.

Methods: A de-identified aggregate data analysis of the initial Canvas Dx prescriptions fulfilled post-market authorization will be conducted to determine: proportion of determinate device output (ASD Positive or ASD Negative); time associated with device use; key prescriber characteristics; key patient characteristics.

Results: Data collection is ongoing. It is anticipated that data analysis will be complete by October of 2022. The study (PR015) has received an IRB exemption from Advarra IRB.

Conclusion: Results from this analysis may help provide insight into the real-world performance of the device.