November 20, 2024 – A collaborative study by researchers from the Regenstrief Institute, Indiana University, and Purdue University has introduced a low-cost, scalable methodology for the early identification of individuals at risk for developing dementia. While dementia remains an incurable condition, the researchers emphasize that addressing common risk factors can significantly reduce the likelihood of onset or slow cognitive decline.
Early detection of dementia risk is crucial for effective care management and planning. “We aimed to identify individuals likely to develop dementia using a solution that is both scalable and cost-effective for the healthcare system,” said Dr. Malaz Boustani, senior author of the study and a researcher at the Regenstrief Institute and IU School of Medicine.
The researchers’ method leverages existing information—referred to as “passive data”—from patients’ electronic health records (EHR). This “zero-minute assessment” costs less than one dollar and utilizes a decision-focused content selection methodology to create individualized dementia risk predictions or to identify signs of mild cognitive impairment.
By employing machine learning techniques, the study analyzes a subset of phrases and sentences from medical notes written by healthcare providers. These notes, which describe a patient’s health in free text, can include clinician observations, patient comments, vital signs such as blood pressure and cholesterol levels, family member assessments of mental status, and medication histories, including prescriptions and natural supplements.
Identifying dementia risk not only aids patients and their families but also helps healthcare providers access essential resources, such as support groups and programs like the Centers for Medicare and Medicaid GUIDE model, which supports aging individuals in their homes. Additionally, this predictive capability may encourage healthcare providers to reconsider medications that adversely affect cognitive function and to discuss over-the-counter drugs that may have similar effects. Furthermore, understanding a patient’s dementia risk could lead to discussions about newly FDA-approved therapies aimed at altering the course of Alzheimer’s disease.
Dr. Zina Ben Miled, a co-author of the study and an affiliate scientist at the Regenstrief Institute, explained, “Our methodology combines both supervised and unsupervised machine learning to extract relevant sentences from the extensive medical notes available for each patient. This approach not only enhances predictive accuracy but also allows healthcare providers to quickly verify cognitive impairment by reviewing the specific texts that informed our risk assessments.”
The study’s co-author, Dr. Paul Dexter, highlighted the importance of maximizing the clinical value of EHR data, stating, “Regenstrief Institute and Indiana University have been pioneers in utilizing electronic health records since the early 1970s. By applying machine learning methods to identify patients at high risk of dementia, this study showcases the innovative potential of EHRs. Early identification of dementia will be increasingly vital as new treatments are developed.”
The researchers aim to benefit patients and caregivers while providing a cost-effective solution for primary care clinicians, who often face time constraints and may lack the training to conduct specialized cognitive assessments.
Currently, the authors are in the final year of a five-year clinical trial of their risk prediction tool, conducted in Indianapolis and Miami. Insights gained from this trial will inform the enhancement of the dementia risk prediction framework in primary care settings. Future research will also explore the integration of medical notes with additional data from electronic health records and environmental factors.
The findings of this study, titled “Dementia risk prediction using decision-focused content selection from medical notes,” have been published in Computers in Biology and Medicine. The research is supported by a grant from the National Institutes of Health’s National Institute on Aging (R01AG069765), with principal investigators including Dr. Malaz Boustani, Dr. Zina Ben Miled, and Dr. James Galvin.
Related Topics:
-
Canadian Guidelines Expand Treatment Options For Opioid Use Disorders
-
Retirement Can Enhance Social Connections And, In The Long Run, Reduce Feelings Of Loneliness
