Accuracy of Intensivist Prognostications of within-ICU Deterioration and Development of Persistent Critical Illness: A Prospective Cohort Study.

Rationale: Persistent critical illness (PerCI) is costly, rising in incidence, and not reliably predicted with existing risk-prediction tools.

Objective: To assess whether attending intensivists can identify patients at heightened risk of developing PerCI.

Methods: We conducted a prospective longitudinal assessment from August 2020 to January 2023. Intensivists were assessed on each patient within 24 hours of admission to the medical intensive care unit (ICU) and on ICU day three. We measured intensivists' prognostication of within-ICU events (late-onset shock and/or acute hypoxic respiratory failure [AHRF] and PerCI) and self-rated confidence in prognostications. Test characteristics were calculated for both outcomes, at each timepoint, and stratified by self-rated confidence.

Results: 1,295 assessments were completed (response rate: 87.9%), assessing 875 ICU admissions by 18 intensivists. Late-onset shock/AHRF and PerCI occurred in 7.3% and 16.0% of ICU admissions, respectively. C-statistics for intensivist prognostication of late-onset shock/AHRF were 0.5 (95%CI:0.5-0.6) and 0.6 (95%CI:0.5-0.6) on admission and day three, respectively. C-statistics for PerCI were 0.7 (95%CI:0.7-0.7) and 0.7 (95%CI:0.7-0.8), respectively. C-statistics for late-onset shock/AHRF were no different for confident vs unconfident assessments. C-statistics for PerCI were higher for confident vs unconfident assessments (0.8 vs 0.6, p<0.01) on admission, but not different on day three (0.7 vs 0.7, p=0.20).

Conclusions: Intensivist prognostications have poor discriminatory accuracy for the development of late-onset shock/AHRF and moderate accuracy for development of PerCI. Further research is needed to understand what factors influence intensivists' prognostications of within-ICU deterioration and how this information is conveyed to patients and families.