Validation and implementation of a Predictive Analytics Model to reduce hospital readmissions in chronic care

Important results from the project

The purpose was to validate and implement a proof-of-concept AI-based decision support for physicians and decision makers to optimize care by predicting hospital admissions. We have 1) identified and inserted the needs of doctors and decision makers in the model (user-centered design), 2) started validation for complex patients with concomitant heart, kidney and diabetes (HND) at Danderyd Hospital HND center and 3) developed a scalability plan that prepares generalization.

Expected long term effects

We have identified the needs and preferences of physicians and decision makers for the AI model. We have begun validation of the predictive decision support model for complex patients with concomitant heart, kidney and diabetes (Heart Nephro Diabetes) at Danderyd Hospital´s HND center from available medical records (VAL database) (delayed due to covid-19) and 3) developed a strategy / scalability plan for the model for predictive decisions that prepares generalization to other sites in Sweden.

Approach and implementation

We have interviewed doctors and decision-makers who both handle patients clinically and are responsible for resource allocation and care planning. We have included their needs and preferences in the AI model. We have begun validation of the model for complex patients with concomitant cardiovascular disease, kidney failure and diabetes (HND) at Danderyd Hospital´s HND center using medical records. We have prepared a strategy / scalability plan for generalization to other environments in Sweden.