Background Prehospital frailty has been associated with adverse hospital outcomes in critically ill adults. Although frailty assessment in intensive care units depends on patients' surrogates, frailty assessments by surrogates and researchers have not been compared. Objectives To compare agreement and validity between surrogates' and researchers' assessments of frailty in critically ill older adults. Methods An observational cohort study of adults (aged ≥ 50 years) admitted to a medical/surgical intensive care unit was conducted. On admission, patients' surrogates quantified prehospital frailty by using the Clinical Frailty Scale (range, 1-9; scores > 4 defined as frail). Researchers blinded to surrogates' assessments also quantified frailty. Agreement was described with κ scores, McNemar tests, and Bland-Altman plots; validity was compared by using χ2 tests and logistic regression. Results For 298 patients (mean [SD] age, 67.2 [10.5] years), both surrogates' and researchers' frailty assessment scores ranged from 1 to 9, with moderate to substantial agreement between scores (κ ≥ 0.40). Surrogates' frailty assessment scores were significantly lower than researchers' (mean difference, -0.62; 95% CI, -0.77 to -0.48; P < .001). Surrogates were less likely than researchers to identify as frail those patients who experienced adverse hospital outcomes (death, prolonged stay, or disability newly identified at discharge). Conclusions Surrogates identified fewer patients as frail than did researchers. Factors involved in surrogates' assessments of patients' prehospital frailty status should be studied to see if the Clinical Frailty Scale can be modified to facilitate more accurate surrogate assessments.