1 The potential of AI in Health – Referral to a virtual hospital

Abstract

<h3>Aims</h3> To explore the successes and challenges of utilising algorithms with Artificial Intelligence (AI) capability to aid clinicians in identifying patients suitable for a virtual hospital or hospital-in-the-home service. <h3>Methods</h3> The South Australian-based NSQHS-accredited Virtual Hospital has delivered care across five jurisdictions to more than 220,000 patients. An algorithm across the SA Health electronic medical record was developed to create efficiency and support clinicians in identifying patients suitable for the Virtual Hospital. This enabled a ‘pull’, rather than ‘push’ approach based on set criteria of admission date, length of stay, diagnostic-related group (DRG), at-risk behaviours, postcode and ward location to identify patients who occupied a physical bed that may be suitable for virtual care, displayed in a dashboard, with a one-click referral. A qualitative explorative study was performed to understand clinician attitudes to the referral process and the role of algorithms with machine learning capability to support referrals. <h3>Results</h3> Through a grounded theory approach, four themes emerged across nine in-depth interviews with clinicians. These challenges included: (1) Outcomes demonstrated that there was significant value in AI identifying suitable patients for alternative care pathways when clinicians were cognitively overloaded; (2) Broad referral criteria was too ambiguous, and that set pathways were preferred that linked directly to the algorithm; (3) Complex non-clinical care needs were not assessed in the algorithm, including frailty; (4) Interoperability challenges across primary and tertiary ecosystems was a consistent challenge to utilising virtual hospital care, with an inability to track patient outcomes once referred, resulting in clinicians feeling uncertain of the care outcomes. <h3>Conclusions</h3> The biggest hurdle to referring patients to a Virtual Hospital and using AI-supported discharge processes from brick-and-mortar sites appeared to be the system challenges of shifting care virtually once a patient was already onsite. Future clinician engagement was seen as essential in advancing the algorithm to aid decisions in the emergency department or earlier in the patient journey.

Ähnliche Arbeiten