Assessment of Covariates

Covariates were grouped based on (a) demographics, (b) lifestyle factors, (c) significant cardiovascular disease/risks (CVD)/comorbidities, and (d) neuropsychiatric comorbidities.

In selecting covariates, we first established the upper limit number of variable inclusion by the “1 in 10 rule,” or 1 variable considered for every event (24). We chose candidate variables based on previously demonstrated prognostic performance with the delirium outcome. For example, we grouped demographic variables based on previously established, a-priori knowledge of differences in the delirium outcome such as age (25), sex (26), and educational outcomes (27).

Demographics included age, sex, education, ethnic background, and controlling for number of hospitalizations post-assessment. Age at recent depression assessment was calculated in years based on the participants’ birth dates. Sex and ethnicity were self-reported at baseline. Ethnicity was included as European versus non-European based on the distribution of participants of European descent (94%). Education was based on answering college attendance (yes/no).

Lifestyle factors included the Townsend Deprivation Index (TDI), a material deprivation score classified into higher/lower medians, physical activity (summed metabolic equivalent minutes), alcohol consumption (<4 drinks/≥4 drinks per week), body mass index (BMI, weight [kg] divided by the height squared [m²]), sleep duration was categorized into short (<6 hours/day), normal (6–9 hours), and long (>9 hours) because of the previously demonstrated U-shape associations with delirium or dementia (4,28), frequency of friend and family visits (never vs any), and falls in the last year (none vs any).

Cardiovascular disease (CVD) is based on hypertension, high cholesterol, smoking, diabetes, ischemic heart disease, and peripheral vascular disease. Comorbidities included a previously described morbidity burden (21,29,30) based on the summed presence of any cancers, respiratory, neurological, gastrointestinal, renal, hematological, endocrine, musculoskeletal, connective tissue, infectious diseases/disorders, and classified as none (0)/modest (1–3)/high (≥4) conditions. Cognitive performance was estimated at initial enrollment using a raw processing speed test involving the mean reaction time to identify card matches correctly (31).

The full final model included serum 25-hydroxyvitamin D (25[OH]D, a proxy for vitamin D levels recently linked to delirium within this cohort, categorized into sufficient > 50 nmol/L, low 25–50 nmol/L, and deficient < 25 nmol/L), and pre-existing dementia/Parkinson’s disease, or depression diagnosis/treatment (“any”, from seeing a psychiatrist, or a self-reported/ICD-10 diagnosis).