Presentation (CTI)

OBJECTIVES: AI-enabled risk-stratification tools hold promise for reducing elective-care backlogs and improving outcomes, yet real-world adoption is inconsistent and benefits poorly evidenced. This study presents an independent evaluation of an AI-driven risk-stratification tool deployed within an integrated care system, assessing clinical impact, financial value, the conditions for national scaling, and the implications for life-sciences partners reliant on elective pathways.
METHODS: The evaluation separated two use cases: risk stratification combined with prehabilitation, and risk stratification as a standalone prioritisation intervention. For the prehabilitation use case, patient-level tool data were compared with a matched HES control population, assessing length of stay by risk quartile, readmissions, readmission length of stay and complications. For standalone prioritisation, scenario modelling used three years of procedure-level elective and emergency activity to test whether bringing high-risk patients forward could reduce emergency admissions, bed days and complications. Adoption barriers were assessed through engagement with provider teams and review of workflow integration requirements for Federated Data Platform (FDP) deployment.
RESULTS: Real-world adoption was materially constrained by lack of operationally defined use cases: providers were not using the tool as originally intended, and use was largely limited to identifying patients for prehabilitation referral. The evidence review showed that most benefits evidence related to risk stratification combined with prehabilitation rather than risk stratification alone. Scenario modelling indicated that prioritising high-risk patients could generate system benefit through avoided emergency activity and reduced downstream complications, but only if embedded into existing FDP-enabled elective workflows rather than deployed as a standalone product.
CONCLUSIONS: The evaluation demonstrates that the principal barrier to AI impact is not algorithmic performance alone, but the absence of a validated benefits case, workflow-defined use case and deployable operating model. FDP integration provides the infrastructure to standardise adoption, link evidence to operational decision-making and scale AI-enabled elective-care interventions nationally.