Our work has been validated through both intervention studies (Randomised controlled trial in Sweden and UK) and experimental studies. Our patient-facing intervention builds on the methodology of motivational interviewing, shared decision making and structured care coordination. A national randomised control trial (RCT) with the Nuffield Research Trust, started in 2015, has shown consistent results for the past five years. The RCT showed the capability of HN’s intervention to reduce care demand and improve patient reported outcomes (PROMS).
A strength in our approach that is applied across all non-RCT sites is the continuous internal feedback loop between the prediction model and our clinical coaches. Our coaches use a combination of the AI predictions and their clinical judgement and expertise to select and assess patients. Local sites produce quarterly evaluations and evidence from the randomised control trial provide a continuous feed of new findings to our frontline staff to better understand which patients most benefit from our intervention. This becomes an explicit control for any confirmation or automation bias that might happen at the manual screening stage.
HN actively engages with the local health economy to adopt new ideas and ways of working. This includes proactive cooperation with trusts’ frontline staff, CCGs and academic researchers to promote external validation and coproduction. In 2019 HN trialled a new way to evaluate its non-RCT patients, which was an approach suggested by Researchers in Residence at the trust.