Predictors of Inpatient Utilization among Veterans with Dementia

Dementia is prevalent and costly, yet the predictors of inpatient hospitalization are not well understood. Logistic and negative binomial regressions were used to identify predictors of inpatient hospital utilization and the frequency of inpatient hospital utilization, respectively, among veterans. Variables significant at the P < 0.15 level were subsequently analyzed in a multivariate regression. This study of veterans with a diagnosis of dementia (n = 296) and their caregivers found marital status to predict hospitalization in the multivariate logistic model (B = 0.493, P = 0.029) and personal-care dependency to predict hospitalization and readmission in the multivariate logistic model and the multivariate negative binomial model (B = 1.048, P = 0.007, B = 0.040, and P = 0.035, resp.). Persons with dementia with personal-care dependency and spousal caregivers have more inpatient admissions; appropriate care environments should receive special care to reduce hospitalization. This study was part of a larger clinical trial; this trial is registered with ClinicalTrials.gov NCT00291161.