All Resources

Using natural language processing to create an AI-based algorithm from EMR data to identify non-traditional caregivers who might benefit from dyadic interventions
Awardee Organization(s): University of Colorado | Kaiser Permanente
Principal Investigator(s): Jennifer Portz, PhD
Official Project Title: Leveraging Patient Portals to Support Caregivers
AITC Partner: PennAITech
Website(s): www.medschool.cuanschutz.edu/general-internal-medicine, www.kpco-ihr.org

Caregivers for people living with dementia (PLWD) make up a diverse group of individuals and can include family, friends, and paid direct care workers. Nearly 30% of caregivers in the United States report caring for a PLWD. While providing care to a loved one can be rewarding, some caregivers may feel caregiver burden and become self-neglectful (e.g., eating poorly, poor exercise habits), which have the potential to lead to poor health outcomes among caregivers. Caregiver-specific interventions are beneficial for improving mental health, confidence in caregiving, and self-care. However, caregivers often experience barriers to such interventions such as time and access. The objective of this pilot is to develop a data-framework for finding caregivers from the electronic medical record (EMR) who can benefit from caregiver and dyadic interventions. PLWD and their caregivers often receive health care services within the same healthcare system and their data hosted within the same EMR. Our previous work found that health outcomes, such as hospitalization, for caregiver-PLWD dyads living in the same household are linked. However, our current model is limited to caregivers living in the same household, often spouses, who share health insurance. Natural language processing can fill this gap by analyzing unstructured EMR data to find patterns among caregivers that will allow us to further identify non-traditional caregivers (e.g., friends, neighbors) and caregivers outside the home (e.g., adult children, extended family members). By automating the process of caregiver identification through the EMR, interventions can be more easily delivered to engage and support the caregiver.