New Results Help Understand Early Indicators of Neurologic Risk

New research from the McCance Center, currently out for review, covers a new way to measure the risk of developing brain disease.

Importance: Several factors are associated with risk for neurological outcomes. The brain health risk score can be easily computed in a patient encounter and offers patients a measure of their neurological risk.

Objectives:  Our objective was to create a brain health score for a cohort of patients seen in the Sleep Laboratory at the Massachusetts General Hospital (MGH) and quantify how well it predicts neurologic outcomes. We aimed to perform a survival analysis with an 11-year follow-up period and to identify important covariates for each outcome. 

Methods: Clinical data were retrospectively analyzed for a cohort of 17040 adult patients who visited the Sleep Laboratory at MGH in a span of 11 years and 2 months, from Jan. 1, 2010, to Mar. 31, 2021. The study covariates included age, gender, vital signs, laboratory values, employment, marital, tobacco and alcohol status. The study outcomes consisted of intracranial hemorrhage, ischemic stroke, depression, death and composite outcome of dementia, Alzheimer disease and mild cognitive impairment. Stratified random sampling was performed to create a dataset with patients in train (70%) and test (30%). Missing data was imputed by multiple imputation using an ExtraTreesRegressor estimator. One random encounter was selected per patient from a total of encounters for modeling. We trained a generalized linear elastic net model in 5-fold cross validation and with the set of covariates selected, we trained a cox proportional hazards model, with death as competing risk. 1000 bootstrapping iterations were performed to calculate 95% confidence intervals (CI) for the concordance index in test.

Results: The cohort average age was 49 years old, with a majority of males (57%) and White race (78%). The prevalence of the study outcomes was for: intracranial hemorrhage 0.6%, ischemic stroke 12%, depression 15%, composite of dementia 4% and death 5%. The average C-index [95% CI] for 11-years survival prediction in test were for: intracranial hemorrhage 0.78 [0.77-0.79], ischemic stroke 0.68 [0.67-0.69], depression 0.65 [0.64-0.66], composite of dementia 0.78 [0.76-0.79] and for death 0.82 [0.81-0.83]. Higher age was associated with increased risk of the outcome events, except for depression, and being active reduced the risk of the outcome events.

Conclusions: A simple risk score derived from routinely collected data, easily acquired in a patient encounter, is associated with risk of neurological outcomes and death.

Keywords: Electronic health records; chronic disease management; dementia; depression; ischemic stroke; intracranial hemorrhage; mortality.