Developing a prediction model for cognitive impairment in older adults following critical illness

November 29, 2024

IMPACT member, Nicholas M. Pajewski, PhD, is among the researchers who developed and tested a machine-learning model for identifying new cognitive impairment or dementia in older adults after critical illness.

The researchers addressed the need for a prediction model of post-ICU cognitive impairment to guide screening and support resources for those with the greatest need by conducting a cohort analysis using electronic health record (EHR) data from 8,299 patients aged 60 and older admitted to a large academic health system ICU between 2015 and 2021. Using oblique random survival forests (ORSF), they examined the multivariable association of 54 structured data elements available 3 months after discharge with an incident diagnosis of cognitive impairment or dementia over 1 year.

The model showed reasonable discrimination and stability, indicating that machine learning using EHR data can predict new cognitive impairment dementia at 1-year post ICU discharge with acceptable accuracy. The authors call for further studies to understand how this tool may impact screening for CI in the post-discharge period.

Read more here.