Frailty Judgment by Hospital Team Members: Degree of Agreement and Survival Prediction

To the Editor: Studenski et al. propose a most-welcome instrument to quantify physical frailty along with an operational definition,1 but is it possible to restrict frailty to only one dimension?2 Other authors think that frailty is a constellation of many conditions rather than a clinical entity,3 but Fried et al. claim that clinicians can easily distinguish a frail from a nonfrail individual.4 Indeed, in daily geriatric activity, the label “frail” is excessively used to characterize different kinds of older people, depending upon each healthcare professional's definition. The aim of this letter is to analyze the usage of the frailty concept by three healthcare professionals (nurse, resident, and chief resident) in charge of the same patient and their agreement on this concept and to test its validity as a predictor of mortality. To answer these questions, the three above-mentioned hospital team members prospectively and independently categorized 502 consecutively older patients (366 women, mean age±standard deviation=84.8±6.7; 146 men, mean age=82.4±7.2) admitted to a geriatric hospital as frail or not frail in six possible dimensions of frailty (physical, nutritional, sensorial, cognitive, psychoemotional, and social), on the exclusive basis of their own initial and specific clinical evaluation at the time of admission. Intentionally, no a priori definition of frail or robust was given to them. The prevalence of frailty across dimensions according to each of these professionals ranged from 36% to 72% of the admitted patients (Table 1). The physical and cognitive dimensions displayed the highest prevalence, whereas psychoemotional and social were the lowest. With the exception of the psychoemotional dimension, nurses systematically identified a greater proportion of subjects as being frail. Thus, agreement between these professionals ranged from fair (0.21≤kappa statistic≤0.40) to substantial (0.61≤kappa statistic≤0.80). The lowest agreement was on the sensorial dimension and the best on the cognitive dimension. To assess the internal validity of the frailty concept, Cronbach alpha statistic was calculated for each professional category. It showed that the strongest homogeneity in evaluating the six dimensions of frailty came from the chief residents, followed by the nurses and finally the residents, but the results were poor in predicting frailty, with a global reliability varying from 0.62 to 0.69, which is satisfactory for a population analysis but not sufficient for evaluation of individual subjects. In addition, logistic regressions were performed to establish the existing associations between the initial subjective judgment of each professional and objective parameters (clinical and biological) collected during the hospital stay: body mass index (BMI), albumin level, C-reactive protein, Folstein Mini-Mental State Examination (MMSE), and the Charlson Comorbidity Index. Consistently across all professional groups, cognitive frailty evaluation was significantly associated with the MMSE (explained variance of frailty assessed by pseudo coefficient of determination (R2) ranged from 23% to 25%) and nutritional frailty with BMI—but not with albumin (pseudo R2 from 8 to 10%). The four other dimensions all displayed weak association with the clinical parameters, varying across professionals, with pseudo R2 less than 9%. Finally, the association between frailty and overall survival was assessed based on deaths recorded during a 15-month hospital and postdischarge follow-up. Univariate Cox regression models showed that physical, nutritional, and sensorial frailty estimated by nurses and residents was significantly associated with higher mortality rate, with a hazard ratio (HR) ranging from 1.52 to 1.83, and that cognitive frailty estimated by chief residents was also associated with a higher mortality rate (HR=1.46). Adjusting for age, sex, albumin, BMI, C-reactive protein, and number of drugs (the first 3 being significant), only sensorial frailty as assessed by nurses (HR=1.63) and physical frailty as assessed by residents (HR=1.76) were still significantly associated with a higher mortality rate during the 15 months of follow-up. These results seem to indicate that the three healthcare professional categories did not have the same perception of the term frailty. Therefore, the clinical use of this term should be avoided until a better consensus emerges, which could differ depending on whether the evaluation involves an individual or a population and also according to the settings. Studensky's ongoing validation project, as well as other attempts, should provide better detection of physical or global frailty to try to prevent its consequences,1, 4-6 essentially consisting of functional decline and mortality.7, 8 There is an urgent need for such a consensus to be able to better screen the target population and propose appropriate intervention.9, 10 The authors express their thanks to B. Grab and R. Grandjean.