Experience with home-based, remote non-stress tests, including automatic decision support for interpretation of reactivity

Abstract

The fetal non-stress test (NST) for antepartum surveillance involves frequent, burdensome in-office visits. INVUTM is a wireless, remote, FDA-cleared monitoring device, designed to perform self-administered NST from home. Here, we present (1) our initial experience with INVUTM home NST monitoring, and (2) a novel machine learning decision support system (DSS) to detect fetal heart rate (FHR) accelerations, with the potential to alert clinicians on NST reactivity. A single-center, prospective, open label study was conducted, in which women with ≥32 week high-risk singleton gestations performed self-administered NST using the INVUTM device at home, replacing in-clinic NSTs. NSTs were reviewed by clinicians via the INVUTM web application who determined clinically interpretability and reactivity. Interpretable sessions were retrospectively analyzed by the DSS for reactivity and compared to clinician assessment (Figure). The time to reactivity as analyzed by the DSS was compared to the session length until reactivity was determined by clinicians. Of the first 30 INVUTM sessions (across 9 subjects), clinicians deemed 27 (90%) interpretable, and 27/27 as reactive (100%). The DSS correctly classified 92.6% (25/27) of the interpretable NSTs compared with clinicians. In the 2 cases where DSS did not match the clinician decision, DSS did not rule the NST reactive, whereas the clinician did. On review, these cases may have been due to clinician averaging the FHR baseline from which accelerations are measured, compared to the conservative nature of the DSS. In the cases deemed reactive by DSS, time to reactivity was reduced by 14.72±13.79 min as compared to clinician determination. Early clinical experience conducting self-administered, remote NSTs at home using the INVUTM device supports the feasibility of bringing this telehealth solution to high risk pregnancy care. In addition, a novel DSS tool accurately identifies reactivity, which could reduce the length of NST sessions and improve clinician workflow.