Background and objectives:
The use of delirium screening instruments (DSIs) is recommended in critical care practice for a timely detection of delirium. We hypothesize that the patient-related factors "level of sedation" and "mechanical ventilation" impact test validity of DSIs.
Materials and Methods:
This is a prospective, bi-center observational study (clinicaltrials.gov: NCT01720914). Critically ill patients were screened for delirium daily for up to seven days after enrollment using the Nursing Delirium Screening Scale (Nu-DESC), Intensive Care Delirium Screening Checklist (ICDSC), and Confusion Assessment Method for the Intensive Care Unit (CAM-ICU). Reference standard for delirium diagnosis was the neuropsychiatric examination using the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR). Immediately before delirium assessment, ventilation status and sedation levels were documented.
160 patients were enrolled and 151 patients went into final analysis. Delirium incidence was 23.2%. Nu-DESC showed a sensitivity and specificity of 88.5%, a positive predictive value (PPV) of 71.9%, and a negative predictive value (NPV) of 95.8%. ICDSC had a sensitivity of 62.5%, a specificity of 92.4%, a PPV of 71.4%, and a NPV of 89.0%. CAM-ICU showed a sensitivity of 75.0%, a specificity of 94.7%, a PPV of 85.7%, and a NPV of 90.0%. For Nu-DESC and ICDSC, test validity was significantly better for non-sedated patients (Richmond Agitation Sedation Scale (RASS) 0/-1), whereas test validity for CAM-ICU in a severity scale version showed no significant differences for different sedation levels. No DSI showed a significant difference in test validity between noninvasively and invasively ventilated patients.
Test validities of DSIs were comparable to previous studies. The observational scores ICDSC and Nu-DESC showed a significantly better performance in awake and drowsy patients (RASS 0/-1) when compared with other sedation levels. Physicians should refrain from sedation whenever possible to avoid suboptimal performance of DSIs.