New electronic health record-based tool detects dementia

A novel electronic health record (EHR)-based tool may help with the early detection of dementia among elderly adults, a recent study has shown.

Researchers performed a retrospective cohort analysis on 4,330 participants (mean age, 80.1±6.6 years; 60 percent female), corresponding to a total of 16,665 visits, who underwent a comprehensive screening process for the detection of dementia. Cognitive screening was performed every 2 years. Patient EHRs were accessed for the determination of important dementia predictors.

A total of 1,015 visits ultimately led to a diagnosis of incident dementia. Almost half of these (49 percent; n=498) were initially unrecognized in the EHRs.

The full predictive model showed good predictive value in both the training cohort (70 percent of the sample; C-statistic, 0.79, 95 percent confidence interval [CI], 0.65–0.81) and in the test cohort (30 percent of the sample; C-statistic, 0.81, 95 percent CI, 0.77–0.84).

Moreover, the final restricted models containing mostly similar factors as the full model also showed good predictive value. At a cutoff of ≥95th percentile in the training cohort, the sensitivity was 22 percent; specificity and positive predictive value were 96 percent and 16 percent, respectively. In absolute terms, when evaluating the top 5 percent of patients, roughly one in six would be predicted to have unrecognized dementia.

In the present study, the researchers developed an EHR-based detection tool called eRADAR (EHR risk of Alzhimer’s and dementia assessment rule). The final full model contained 31 predictors, including age, sex, vital signs and past diagnoses.