This study examined whether “observed frailty” can be a valid alternative to accepted frailty scoring systems by examining the correlation between assessment of observed frailty and two currently accepted scores, as well as the ability to predict 3 clinical outcomes—30 day mortality, duration of ICU stay and the number of mechanical ventilation days.

A moderate correlation (rs = 0.4) was observed between the observed frailty score and the Clinical Frailty Score, while the correlation of observed frailty with the Modified Frailty Index was weaker (rs = 0.18) and did not reach statistical significance.

Further, a significant association was observed between the observed frailty score and 30 day mortality, but not with duration of ICU stay or of mechanical ventilation.

These results are consistent with previous research in the field that demonstrated an association between frailty and mortality [22]. A meta-analysis [23] examined the relationship between frailty, measured among other factors by CFS and MFI, and clinical outcomes in intensive care, also revealed that frailty was associated with increased risk of hospital and long-term mortality, but not with duration of ICU stay or mechanical ventilation, in line with the findings of this study.

Moreover, in our study, other functional measures such as pre-hospital activity level and the SOFA score [14] were also significantly associated with 30 day mortality. Conversely, The CFS and MFI scores were not significantly associated with this clinical outcome. These findings may be attributed to differences in the patient population under investigation compared to those on which the MFI and CFS scores were originally validated.

Patients hospitalized in ICU are of different ages and suffer from various medical conditions, usually acute in onset and without prior functional deficits. These factors differentiate the population in this study from the populations examined in previous studies, which included a relatively homogeneous population: The NSQIP study that introduced the MFI score exclusively included patients undergoing surgical procedures and were under 50 years of age, whereas our study encompassed both surgical and medical patients with a wider age distribution; the CSHA-FI study [5], which first introduced the CFS score, also included only patients above the age of 65 while the median age in this study was 62.5 years (IQR 40–70). This may have impacted the perceived frailty scores as frailty may be less expected or less clearly expressed in younger adults.

While subjective assessment of frailty is less reproducible than an assessment based on pre-defined scores, it allows for a rapid evaluation of a patient’s pre-morbid condition, which can be particularly useful when triaging admissions to the ICU. The results of this study suggest that such a subjective assessment may have prognostic value and could reasonably be considered as an additional criterion—alongside other clinical factors—when making ICU admission decisions. Given the differences in patient characteristics—particularly age distribution and baseline functional status—compared to the original CFS and MFI validation cohorts, it should be emphasized that the results of this study should not be interpreted as demonstrating superiority. Nonetheless, observed frailty may serve as a practical alternative or complement to structured frailty scores, especially in situations where time or data availability is limited.

Limitations: this study has several limitations that should be considered. First, this was a single center, retrospective study which may limit the generalizability of its findings. Second, in the population examined in this study, not many participants had poor functional status upon admission and 44% of patients were assessed as having “no frailty”. This may indicate the presence of a selection bias. While this is a basic characteristic of the population hospitalized in the ICU at this institution, it may also limit the generalizability of the findings to other patient populations. Third, while both the MFI and CFS scores have been previously validated, it is plausible their use in the setting of this study was associated with lower predictive values, because they were calculated retrospectively. However, the fact that a moderate correlation (rs = 0.41) was observed between both these scores is an indicator of validity. Moreover, calculating these scores retrospectively enabled a blinded structure, avoiding bias from exposing caregivers (who assessed observed frailty) to an accepted frailty score. Yet, bias may still have been introduced, as a staff member’s estimation of frailty may have influenced their treatment decisions, for example withholding aggressive treatment from a patient perceived to be too frail to receive it. Such an effect may have later translated to an “artificial” observed association between perceived frailty and outcomes.

Lastly, while our sample was adequately powered for the primary outcome, the multiple analyses performed for secondary outcome analysis introduce a risk of type I errors, while the limited sample size may have reduced our ability to detect weaker associations in secondary outcomes, increasing the potential for Type II errors.