Claims, EHR data may help predict mild cognitive impairment in primary care
Click Here to Manage Email Alerts
Key takeaways:
- Researchers are developing a risk algorithm for mild cognitive impairment that can be used by primary care providers.
- Predictive performance of the algorithm was strongest for those aged 50 to 64 years.
Patient data collected through claims and electronic heath records may help identify individuals with mild cognitive impairment in primary care, according to a poster presented at the Alzheimer’s Association International Conference.
"In the US, most cases of [Alzheimer's disease] are diagnosed at the moderate stage or later, and 20% of cases never receive a diagnosis," James E. Galvin, MD, MPH, study co-author and professor of neurology, psychiatry and behavioral sciences at the University of Miami Miller School of Medicine, told Healio in an email. This lessens the opportunity for early intervention. The electronic health record offers a potentially rich resource for information that can help to identify individuals at-risk or who have early disease.
Seeking to develop a risk-predictive model to identify likely cases of mild cognitive impairment (MCI) for referral to specialists, researchers conducted a MarketScan claims data study that included more than 20,000 individuals aged 50 years and older.
They matched a cohort of adults with newly diagnosed MCI (n = 5,185; mean age, 67 years; 42.3% men) with a non-MCI cohort (n = 15,555; mean age, 67 years; 42.3% men), and compared medical conditions linked to MCI between the two groups over a 2-year baseline period.
Researchers used logistic regression and a machine learning method to distinguish MCI from non-MCI individuals and estimated models for all individuals and by age group (50-64, 65-79 and ≥80 years).
According to results, the MarketScan study identified 25 medical conditions with statistically significantly higher frequencies in the MCI cohort. Depression, stroke or transient ischemic attack, obstructive sleep apnea, hypertension and hyperlipidemia were the top five predictors of MCI.
Data additionally showed that the predictive performance of the models was strongest for individuals aged 50 to 64 years, with area under the curve values of 0.75 (50-64 years), 0.69 (65-79 years) and 0.66 (≥80 years).
Researchers are planning two additional studies using electronic health record data, which will compare approximately 200 MCI predictors between cohorts of newly diagnosed MCI and non-MCI individuals. Predictors will expand on findings from the MarketScan claims data study and will be validated by clinical data for cognitively normal individuals.
Data routinely found in [electronic health records] may provide sufficient information to identify patients with potential MCI for cognitive assessment in primary care," Galvin told Healio. "This may facilitate early intervention and treatment particularly with new disease modifying medications."