A deep learning model to detect cognitive decline from electronic health records


  • Clinical notes in electronic health records can be assessed using a deep learning model to accurately predict which patients will be diagnosed with mild cognitive impairment (MCI).

Why this matters

  • Nearly 6 million people in the US have dementia due to Alzheimer’s disease. However, detecting early stages of Alzheimer’s for early intervention can be challenging.

  • Studies have investigated the use of electronic health record data for early detection of cognitive decline. However, approaches to date have had various limitations and typically been reliant on billing codes or medication use.

  • An investigation of unstructured clinical note data from electronic health records could provide an effective way to identify early stages of cognitive decline.