Predicting incident delirium diagnoses using data from primary-care electronic health records

Area of Science:

• Geriatrics
• Public Health
• Health Informatics

Background:

• Delirium risk factors in hospitalized patients are known, but prediction from primary care data is less understood.
• Community-acquired delirium or delirium during emergency hospital admissions requires better predictive models.

Purpose of the Study:

• To identify risk factors in primary care electronic health records (PC-EHR) predictive of delirium.
• To assess the predictive performance of these factors against the cumulative frailty index.

Main Methods:

• A multi-stage study using case-control and retrospective cohort designs.
• Logistic regression and receiver operating characteristic (AUC) analysis were employed.
• Data from the Clinical Practice Research Datalink in England was utilized.

Main Results:

• Fifty-five risk factors were identified, including cognitive impairment, psychoactive drug use, frailty, infection, hyponatraemia, and anticholinergic drugs.
• The developed model accurately predicted 1-year incident delirium (AUC=0.867) and mortality (AUC=0.846), outperforming the frailty index.
• The top 10% of individuals predicted at risk accounted for 55% of incident delirium cases over one year.

Conclusions:

• A PC-EHR-based risk factor model effectively predicts delirium in older adults.
• This model demonstrates potential for identifying individuals at risk and supporting preventive strategies.
• Early identification through primary care data can aid in managing delirium incidence.