Association of Positive Delirium Screening with Incident Dementia in Skilled Nursing Facilities

BACKGROUND AND OBJECTIVE Early detection of delirium in skilled nursing facilities (SNFs) is a priority. The extent to which delirium screening leads to a potentially inappropriate diagnosis of Alzheimer's disease and related dementia (ADRD) is unknown. DESIGN Nationwide retrospective cohort study from 2011 to 2013. SETTING An SNF. PARTICIPANTS A total of 1,175,550 Medicare enrollees who entered the SNF from a hospital and had no prior diagnosis of dementia. EXPOSURE A positive screen for delirium using the validated Confusion Assessment Method (CAM), performed as part of the federally mandated Minimum Data Set (MDS) assessment. MEASUREMENTS Incident all-cause dementia, ascertained throughInternational Classification of Diseases, Ninth Revision(ICD-9), diagnosis in Medicare claims or active diagnoses in MDS. RESULTS Positive screening for delirium was identified in 7.7% of cases (n = 90,449), and most occurred within the first 7 days of SNF admission (62.5%). The overall incidence of ADRD was 6.3% (n = 73,542). Nearly all new diagnoses of ADRD (93.5%) occurred within the first 30 days of SNF admission. Patients who screened CAM positive for delirium had a nearly threefold increased risk of receiving an incident ADRD diagnosis on the same day (hazard ratio (HR) = 2.63; 95% confidence interval (CI) = 1.50-4.63). Among patients who screened CAM positive for delirium, those who were cognitively intact or had mild cognitive impairments were, on average, six times more likely to receive an incident ADRD diagnosis (HR = 6.64; 95% CI = 1.76-25.0) relative to those testing CAM negative. CONCLUSION AND RELEVANCE Among older adults not previously diagnosed with dementia, a positive screen for delirium was significantly associated with higher risk of ADRD diagnosis after admission to a SNF. This risk was highest for patients in the first days of their stay and with the least cognitive impairment, suggesting that the ADRD diagnosis was potentially inappropriate.