Sensitivity and Specificity of a Short Questionnaire to Screen Frailty in the Community-Dwelling Older Population

Introduction. Frailty represents a major risk factor for death and Use of Hospital Services (UHS) among older adults. A simple tool to detect frailty might permit stratification of the community-dwelling older population according to the risk of negative outcomes. The present study aims at determining the sensitivity and specificity in predicting mortality and UHS of the Short Functional Geriatric Evaluation (SFGE), a short questionnaire to screen for frailty in community-dwelling older citizens. Methods. The study is a secondary analysis of all the data collected through an observational longitudinal cohort study carried out in Lazio region (Italy). The SFGE is compared with the Functional Geriatric Evaluation (FGE) questionnaire to define sensitivity and specificity for mortality and for UHS during the first year following its administration. Results. The SFGE classifies 36.3% of the respondents as frail and shows a sensitivity of 90.4% and a specificity of 78.3% compared to the FGE (area under the ROC: 0.928; CL95%: 0.910 - 0.947; p-value < 0.001). Those respondents identified by the SFGE as frail also include some of those classified by the FGE as pre-frail, who also show a high rate of UHS. The results show that the SFGE score predicts the UHS more accurately than it does the mortality rate. Conclusion. The SFGE identifies as frail a larger portion of the enrolled population than the FGE. Those people so identified show a high rate of UHS. Because of its easy and quick administration, it can be considered a useful primary screening tool but it must be followed up with a more extensive assessment of those identified as frail. The small time needed to fill in the tool and the possibility of administering it by telephone makes the SFGE a useful tool to screen for frailty and to plan the provision of care services at both individual and population level.

Introduction. Frailty represents a major risk factor for death and Use of Hospital Services (UHS) among older adults. A simple tool to detect frailty might permit stratification of the community-dwelling older population according to the risk of negative outcomes. The present study aims at determining the sensitivity and specificity in predicting mortality and UHS of the Short Functional Geriatric Evaluation (SFGE), a short questionnaire to screen for frailty in community-dwelling older citizens. Methods. The study is a secondary analysis of all the data collected through an observational longitudinal cohort study carried out in Lazio region (Italy). The SFGE is compared with the Functional Geriatric Evaluation (FGE) questionnaire to define sensitivity and specificity for mortality and for UHS during the first year following its administration. Results. The SFGE classifies 36.3% of the respondents as frail and shows a sensitivity of 90.4% and a specificity of 78.3% compared to the FGE (area under the ROC: 0.928; CL95%: 0.910 -0.947; p-value < 0.001). Those respondents identified by the SFGE as frail also include some of those classified by the FGE as pre-frail, who also show a high rate of UHS. The results show that the SFGE score predicts the UHS more accurately than it does the mortality rate. Conclusion. The SFGE identifies as frail a larger portion of the enrolled population than the FGE. Those people so identified show a high rate of UHS. Because of its easy and quick administration, it can be considered a useful primary screening tool but it must be followed up with a more extensive assessment of those identified as frail. The small time needed to fill in the tool and the possibility of administering it by telephone makes the SFGE a useful tool to screen for frailty and to plan the provision of care services at both individual and population level.

Introduction
In the context of the rapid ageing of the global population, the concept of older adults' frailty has become a very important issue not merely in clinical settings but also for Public Health. In fact, identifying those elderly who are already frail, or close to become so, represents the first step towards planning appropriate public health interventions and developing suitable prevention programmes [1] [2].
Frailty is a multidimensional and dynamic condition characterized by a reduced ability to deal with acute physical, psychological and socio-economic stressors and/or to perform the normal activities of daily life [3] [4] [5]. It is widely recognized that frailty aggravates the risk of poor outcomes (e.g. death, loss of autonomy, functional impairment and hospitalization) [6] [7] and that it develops within various domains (e.g. physical, psychological, socio-economic, environmental status) [8] [9] [10].
Although numerous frailty tools have already been developed in order to screen older adults living in the community, a gold standard instrument to measure and to screen for frailty has yet to come. A two-step model with two tools, a first and second level questionnaire [11], should be used in order to screen for frailty in the community setting. The first level questionnaire is a screening tool for the community-dwelling older population; the second level questionnaire is a more detailed test, directed at those identified as frail by the first level survey, in order to plan the provision of a personalized care package.
The idea of testing a huge number of elderly citizens, however, raises practical concerns. First of all, such a time-consuming activity can be implemented only if it provides crucial information to improve health, either at an individual or at a population level. Secondly and closely related to the first point, there is the question of the availability of a reliable short questionnaire. Several researchers are committed to find a model for providing effective community care services [12]- [17]; the screening of frailty at population level with a short and reliable tool is a first step to achieve this goal.
The present study aims at defining the short-term sensitivity and specificity in predicting mortality and Use of Hospital Services (UHS) of the Short Functional Geriatric Evaluation (SFGE), a first level questionnaire intended to screen for frailty in community-dwelling older adults.

Methods
The study is a secondary analysis of data gathered for the observational longitu- dinal cohort study carried out in the Lazio region (Italy) [18], by the Biomedicine and Prevention Department of the University of "Tor Vergata", and approved by the Independent Ethical Committee of the University of Rome "Tor Vergata".
The inclusion criteria were to be living in Lazio region, to be more than 64 years old and to be living within the community. Those living in an institution were excluded. The cohort is made up by 1,342 subjects enrolled in 2014, of whom 1,282 completed the one-year follow up.
The Functional Geriatric Evaluation (FGE) questionnaire for the assessment of frailty was administered to all the participants. The FGE questionnaire is a multidimensional tool that deals with biological and non-biological domains (physical and mental health, social and economic resources and functional status). The final score could be divided into two categories: "robust" (FGE > 49) and "frail" (FGE ≤ 49). As shown by previous papers, the frail group has a greater risk of adverse events than the group of the robust; in fact the FGE score has been proved to be an effective predictor of death, of hospitalization and of institutionalization over a span of from 1 to 5 years from the point of observation

Results
The sample was made up of 690 females (53.8%; mean age: 76.6 years with a SD of 7.3) and 592 males (46.2%; mean age: 76.3 years with a SD of 7.1). During the baseline assessment with the FGE questionnaire, 1,008 individuals were classified as "robust" (78.8%) while 272 were classified as "frail" (21.2%).
The results of the SFGE 10 items are summarized in Table 2. A ROC curve was performed in order to choose the appropriate cut-off level for frailty for the SFGE questionnaire. Thus, based on the best combination "sensitivity" (90.4%) against "specificity" (78.3%), the value zero was chosen as the most appropriate cut-off level (area under the curve: 0.928; CL95%: 0.910 -0.947; p-value < 0.001).
As shown in Table 3, the two questionnaires disagree for 26 persons who were "frail" for the FGE and "robust" for the SFGE (category C-2) and for 219 persons who were "robust" for the FGE and "frail" for the SFGE (category C-3). The hospitalization rates of both groups were higher than that of the group classified as "robust" by both questionnaires, which would not have been expected for those citizens classified as "robust" by FGE (Table 4, Kruskal-Wallis Test, p = 0.006; Mann-Whitney U Test, not shown in the table, between categories C-1 and C-3, p = 0.041). In fact, the category C-3 showed a hospitalization rate more similar to the category C-4 (the individuals classified as "frail" by both questionnaires) than to C-1 (the individuals classified as "robust"): the SFGE score that generated the classification was worse in all the SFGE items for the C-3 subjects compared with the C-1 ones.
At the same time the group made up of 26 persons considered "robust" by the SFGE and "frail" by the FGE showed the highest UHS rate (   the "robust" individuals) and the initial functional impairment, as indicated by the higher percentage of individuals not able to take a bath/shower alone. Figure   1 shows the correlation between the FGE and the SFGE scores: 86.9% of the cases which the SFGE classified as "frail" had an FGE score between 50 and 70 (mean age 80.0 years; mean FGE score = 62.03; SD ± 6.88; correlation coefficient value: −0.851; p value < 0.001). This score identifies subjects close to the frail condition, a category which can be defined as pre-frail.
The survival analysis adjusted for gender, age and multi-morbidities shows that while the FGE classification of frailty can be predictive of mortality during the first year of follow up (RR = 6.2; CL95: 3.2 -12.1), the SFGE classification shows no statistically meaningful association (RR = 1.6; CL95: 0.8 -3.1).

Discussion
Public Health should consider frailty a priority, due to its strong relationship with the increase of UHS by older adults and to its association with some chronic degenerative diseases [1] [20] [22]. The screening for frailty allows identification of the population at risk of adverse health outcomes and represents an important support in planning health and social services [23]. Using a simplified frailty index is the most appropriate way to screen the population: the SFGE is a  The FGE is highly predictive of negative outcomes, even in the short-term, so that it represents a very good basis to develop a shorter tool. The FGE classifies as frail the 21.2% of the interviewed, while the SFGE identifies as frail a larger portion of the same population (36.3%).
The SFGE differs from the FGE questionnaire in that it has a smaller number of questions, and in the nature of the final score: the score generated by the socio-economic items of the FGE questionnaire accounts for 26% of the FGE total score but for 47% of the SFGE total score. The higher percentage of frail individuals identified by the SFGE than by the FGE questionnaire then derives from the different balance between physical, mental and functional status relative to socio-economic resources. As shown in the literature, socio-economic resources play a crucial role as determinants of population health [24] [25]; the impact of socio-economic frailty in the generation of higher rates of UHS in an environment with a low level of care services in the community has also been proved [26]. The analysis of non-biological frailty may play an essential role in the as- This study helps to highlight the pre-frail status [35] [36], which is mainly marked by a lack of socio-economic resources, with initial psycho-physical impairment, and is associated with a higher rate of UHS than the robust status. This is in agreement with the observation that a lack of social resources is one of the main determinants of the UHS [36]. These outcomes indicate a potentially major impact on Public Health planning: the provision for social care is in fact able to reduce the hospitalization rate [37], a consequence of the role played by socio-economic resources in the demand for hospital care by older adults. The SFGE questionnaire explores socio-economic areas more extensively than other first level questionnaires. It may thus be better able to predict the demand for hospital care, even if further studies with longer follow up are needed to verify this hypothesis. The SFGE "frail" group may be the target of health promotion and of care intervention aimed at managing frailty at a community level. The SFGE does not show statistical significance in the short-term assessment of the risk of death, probably because of the inclusion in the frail group of the pre-frail subjects with a lower risk of death.

Conclusion
The SFGE is a tool to screen frailty at a community level and has a good short-term sensitivity and specificity for UHS. The relevance of socio-economic resources that account for 47% of the SFGE score makes this tool appropriate for use at a community level, where these resources are strongly correlated with the demand for care services. Further studies are needed to improve the sensitivity of the tool in order to reduce the proportion of false negatives. A public health approach to the management of frailty could also contribute to the medium-and long-term sustainability of health systems by identifying and counteracting the conditions that generate the highest demand for health care. Identifying as many frail individuals as possible becomes of utmost importance when planning interventions aimed at preventing or managing frailty at community level.