Partnering With Family Members to Detect Delirium in Critically Ill Patients*

Objectives: To evaluate the diagnostic accuracy of family-administered tools to detect delirium in critically ill patients. Design: Diagnostic accuracy study. Setting: Large, tertiary care academic hospital in a single-payer health system. Patients: Consecutive, eligible patients with at least one family member present (dyads) and a Richmond Agitation-Sedation Scale greater than or equal to -3, no primary direct brain injury, the ability to provide informed consent (both patient and family member), the ability to communicate with research staff, and anticipated to remain admitted in the ICU for at least a further 24 hours to complete all assessments at least once. Interventions: None. Measurements and Main Results: Family-administered delirium assessments (Family Confusion Assessment Method and Sour Seven) were completed once daily. A board-certified neuropsychiatrist and team of ICU research nurses conducted the reference standard assessments of delirium (based onDiagnostic and Statistical Manual for Mental Disorders, Fifth Edition, criteria) once daily for a maximum of 5 days. The mean age of the 147 included patients was 56.1 years (sd, 16.2 yr), 61% of whom were male. Family members (n= 147) were most commonly spouses (n= 71, 48.3%) of patients. The area under the receiver operating characteristic curve on the Family Confusion Assessment Method was 65.0% (95% CI, 60.0-70.0%), 71.0% (95% CI, 66.0-76.0%) for possible delirium (cutpoint of 4) on the Sour Seven and 67.0% (95% CI, 62.0-72.0%) for delirium (cutpoint of 9) on the Sour Seven. These area under the receiver operating characteristic curves were lower than the Intensive Care Delirium Screening Checklist (standard of care) and Confusion Assessment Method for ICU. Combining the Family Confusion Assessment Method or Sour Seven with the Intensive Care Delirium Screening Checklist or Confusion Assessment Method for ICU resulted in area under the receiver operating characteristic curves that were not significantly better, or worse for some combinations, than the Intensive Care Delirium Screening Checklist or Confusion Assessment Method for ICU alone. Adding the Family Confusion Assessment Method and Sour Seven to the Intensive Care Delirium Screening Checklist and Confusion Assessment Method for ICU improved sensitivity at the expense of specificity. Conclusions: Family-administered delirium detection is feasible and has fair, but lower diagnostic accuracy than clinical assessments using the Intensive Care Delirium Screening Checklist and Confusion Assessment Method for ICU. Family proxy assessments are essential for determining baseline cognitive function. Engaging and empowering families of critically ill patients warrant further study.