Electronic medical records may facilitate early assessment of ADRD risk
Electronic medical records (EMR) may help in the early identification of the likelihood of Alzheimer’s disease and related dementias (ADRD), a recent study has found.
Accessing the Indiana Network for Patient care, researchers conducted a case-control study on ADRD patients and age-, race- and sex-matched controls. The derivation population included 10,504 patients and 39,510 controls, while the validation sample had 4,500 cases and 16,952 controls. Both structured and unstructured data (such as progress, medication and visit notes) were retrieved.
Three logistic regression models were constructed using different time periods before the index date: >1 but <10 years (1–10-year model), >3 but <10 years (3–10-year model), and >5 but <10 years (5–10-year model).
Using only structured EMR information, researchers found that including data closer to the patients’ index date improved model performance. For instance, in the validation cohort, the area under the receiver operating characteristic (AUROC) curve was better in the 1–10-year than in the 3–10-year and 5–10-year models (0.689 vs 0.649 and 0.633).
Moreover, restricting the analysis to only those with complete data improved model accuracy, while the effect of proximity of data to index date remained (AUROC, 0.716 vs 0.688 and 0.686, respectively).
Including the unstructured EMR data into the regression models further boosted their performance. In patients in the validation cohort with complete information, the 1–10-year model had an AUROC of 0.826, superior to that of the 3–10-year (AUROC, 0.792) and 5–10-year (AUROC, 0.764) models.