Introduction
In Europe, anxiety disorders are one of the most common mental health problems among elderly, with lifetime prevalence estimates ranging from 20.1% in Italy to 32.6% in England [
1,
2]. This prevalence is probably underestimated, considering that subsyndromal manifestations of anxious symptoms probably range from 15 to 52.3% in community-living older adults [
3,
4]. With an increasingly ageing population, the socio-economic costs of mental health problems – such as anxiety disorders – are growing significantly [
3], especially in countries like Italy with low fertility rates and a high life expectancy. As of January 1st 2020, 13.9 million Italians (23.1% of the population) were over 65 years of age, and this percentage continues to increase annually [
5]. By 2050 this percentage is expected to reach nearly 30% in most of the western countries [
6].
Despite its impact and the high prevalence among the elderly population, anxiety remains often undiagnosed [
7,
8]. A possible explanation lies on the specific characteristics of anxious symptoms in later life. Indeed, older adults typically report an increased sensitivity to several types of somatic stimuli and more physical complaints than younger adults, which could be mistaken for psychosomatic symptoms [
9]. Moreover, anxious symptoms are often comorbid with other medical conditions and depressive disorders, worsening their prognosis and making the differential diagnosis difficult [
10,
11]. Anxiety has also a substantial socio-economic burden, due to an increased use of health services among those who report greater symptoms [
12]. Thus, both clinicians and mental health practitioners need valid, reliable measures of anxious symptoms specifically tailored for this population and carefully developed to consider the clinical manifestations of late-life anxiety.
Self-report measures are commonly used to assess anxious symptoms in elderly thanks to their easy administration and to the reduced respondent burden compared to other assessment methods [
13]. In their recent work, Balsamo and colleagues [
13] reviewed 12 self-report instruments that are commonly used to assess anxiety in later life, some of which are specifically tailored for older population, while others have been validated on geriatric population only later. The first group includes the Geriatric Anxiety Scale (GAS) [
14], the Adult Manifest Anxiety Scale (AMAS), [
15] the Geriatric Anxiety Inventory (GAI), [
16] and the Worry Scale (WS), [
17]; while among the second group there are the State-Trait Anxiety Inventory (STAI), [
18] the Beck Anxiety Inventory (BAI), [
19] the Penn State Worry Questionnaire (PSWQ), [
20], and the State-Trait Inventory for Cognitive and Somatic Anxiety (STICSA), [
21]. They varied depending on the design format (Likert scales vs. dichotomous answers), the length (from 16 to 88 items), and the underlying factor structures (unidimensional vs. multidimensional). However, despite their popularity in contexts whereby assessing anxiety in older adults is required, some of these measures lack good psychometric properties, especially those not originally intended for older adults [
13].
Among those instruments specifically tailored for older adults, the Geriatric Anxiety Scale (GAS) is among the most promising ones. Characterized by good to excellent psychometric properties, GAS was developed by Segal and colleagues in 2010. It is composed of 30 items (even though 5 items are intended for clinical purposes and therefore they are not included in the final score) rated on a 4-point Likert scale, ranging from 0 (not at all) to 3 (all of the time). The GAS investigates three dimensions of anxiety, namely Somatic, Cognitive, and Affective ones, and this factorial structure has been confirmed in recent research [
22]. The original version of the instrument has been validated on a large sample of community dwelling older adults [
14] and on a sample of medically ill older adults [
23]. To date, the GAS is one of the most widely-adopted instrument for the measurement of anxiety in geriatric population, as evidenced by its translation in many languages: Arabic [
24] Chinese [
25], German [
26], Italian [
27], Persian [
28], and Turkish [
29]. Furthermore, psychometric properties of the GAS are good: it demonstrated excellent internal consistency (α = 0.88–0.93) in non-clinical population and significant convergent and discriminant validity [
13].
Due to the specificity of the target population, there is a need for development and validation of short-forms of geriatric anxiety questionnaires, that allow researchers and clinicians to collect necessary symptom data with less respondent burden [
30]. Shorter versions may also help in reducing missing data, thereby improving data quality, and reduce the time required by mental health practitioners and clinicians for their administration and scoring [
30]. In addition, the negative consequences of anxiety let emerge the need of short and easy-to-use instruments which could allow its detection at earlier stages, before that deleterious effects of chronic forms of anxiety occur [
8].
A few short versions of GAS are currently available. The most recently developed was a German version of the GAS, named GAS-G-SF [
31]. Validated both in the general population (
N = 242) and medically ill sample (
N = 156), this 9-item version showed good psychometric properties. However, its Somatic subscale exhibited lower reliability than the other two (Cognitive and Affective) subscales. Unlike the GAS-10, the GAS- G-SF was developed to be a multidimensional measure of geriatric anxiety. Specifically, three-factor model without any error covariance was provided an excellent fit by Gottschling et al. [
31].
The 10-item GAS-10 version [
32] was the most widespread short form of the GAS, developed and validated on a large US sample by using the Item Response Theory (IRT) approach. Excellent psychometric properties, including reliability, convergent validity, and unidimensionality, were reported by Mueller et al. [
32]. More recently, Pifer and colleagues [
33,
34] have developed an adapted 10-item version of the GAS (GAS-LTC), specifically tailored for long-term care settings. Starting from the already existing GAS-10, they slightly modified items according to specific cognitive needs of long-term care residents, as well as changed response format from an ordinal Likert-type scale to a dichotomous Yes/No response. Even though the initial validation was conducted in a small sample (
N = 66), it showed good internal consistency and convergent validity.
Overall, both the unidimensional and multidimensional short forms developed so far appear to retain the amount of information captured by the longer version of the GAS, and confirmed their utility in detecting anxiety in older adults.
It is well know that unidimensionality is a desirable requirement for calculating and interpreting a total score of a clinical self-report instrument [
35]. However, health outcomes measures are rarely strictly unidimensional [
36]. This is due to the heterogeneous items that represent the complexity of health constructs [
37,
38]. Although multidimensional instruments potentially reflected the heterogeneity of psychiatric disorders [
39], unidimensional health measures were crucial in measuring the performance of health systems, evaluating outcomes in clinical trials and in everyday clinical practice [
40]. To this purpose, flexible methodological approaches, as well as statistical models have been proposed to assess unidimensionality of health measures (eg. a bi-factor model) [
41].
A brief, specific and unidimensional method of assessment of the severity of anxiety symptoms in older adults seems to be the answer to the main challenges posed by the measurement of anxiety in this population [
13].
Currently, an Italian shortened unidimensional version of the GAS has not been developed yet, despite the current trend of growing percentage of elderly among the Italian general population and the increasing burden of mental disorders, including anxiety.
Therefore, the aim of the present study was to validate the Italian short-form of the GAS-10 in clinical and non-clinical samples. We conducted a series of separate analyses. First, we analysed the properties of the Italian version of the GAS-10 items by assessing within the IRT framework (Structural analyses). Next, in two independent samples of elderly people we investigated the concurrent validity of the Italian GAS-10 through a comparison with other validated instruments (Concurrent validity analysis). Finally, we tested the diagnostic performance of the GAS-10 with respect to Generalized Anxiety Disorder (GAD) (Diagnostic performance analysis).
Discussion
Given the overlapping of symptoms of mental disorders (eg. depression and anxiety) in later life and the comorbidity with physical health problems, it is important to design unidimensional anxiety measures that have been specifically targeted for older adults that account for and measure this unique expression [
13,
23]. This is also supported by the controversial findings derived from empirical studies on the composite ‘general distress score’, which combines depression and anxiety symptoms in a unidimensional domain [
71]. This study aims to filling this gap, adding evidence of the adequacy of the GAS-10 for the assessment of geriatric anxiety through IRT-based evidence.
The IRT analysis provided a clear framework of the good performance of the GAS-10 items in measuring the geriatric anxiety construct. In accordance with Mueller et al. [
32] study, our study showed that the GRM model accurately explained the pattern of responses obtained by the GAS-10 scale. In deep, the item parameters showed that all the GAS-10 items were able to distinguish adequately among older adults with different levels of the trait being measured, and adequately covered the spectrum of the latent trait. Concerning the item difficult values, our findings showed that the GAS-10 items had a medium to high level of difficulty, indicating a decreasing of the probability to endorse the response option “
all the time” to self-statements that quantify the frequency that anxiety symptoms. Nonetheless, the discrimination parameter values in the present study ranged from “low” (item #4) to “moderately high” (item #9), suggesting that the GAS-10 items can discriminate older adults with different anxiety symptoms severity levels. Compared to the cognitive and affective items, the somatic items were found less informative. These findings had already been reported in previous studies employing both the GAS-10 [
32] and other anxiety measures, such as the State-Trait Inventory for the Cognitive and Somatic Anxiety (STICSA) [
21]. It is still open the debate among researchers about the inclusion of physiological symptoms items in anxiety measures, mainly in the elderly, since these could undermine the uniformity of the anxiety construct. Indeed, the content of somatic items in anxiety measures have been criticized due to the possible overlap with the depression symptomatology [
30]. In fact, somatic symptoms (eg. fatigue, aches and pains) have been found to be a prominent part of the clinical presentation in depressed older adults [
72]. In addition, the co-occurrence of physical health problems in older adults makes differentiating between somatic symptoms of anxiety, and depression difficult [
73].
The GAS-10 adequately measured geriatric anxiety ranging from medium to high levels, whereas it was less precise for the lowest levels of the trait. To date, the highest informative point of the GAS-10, or the peak, was observed at 2.5 standard deviation above the mean level of anxiety, supporting its use as a clinical meaningful measure to assess elderly who experienced high levels of mental distress.
Far from previous study the DIF analysis for each item across gender confirmed the invariance of the GAS-10, providing evidence of its ability to scale males and females onto a common metric. Commonly, females tend to score highly in the GAS total score than males, since they are more comfortable in to express mental health symptoms, as well as showed more risk factors [
74,
75]. Potentially, this unbiased version of the GAS-10 could allow an easy and efficient assessment of anxiety among elderly, without the use of differentiated norms by gender. On the contrary, DIF has been found across age, despite its impact was negligible and close to the 2% of the items true score. Adults aged over 70 years scored higher on three items, two of these were drawn from the affective and one from the cognitive domain of the GAS-10. DIF of affective anxiety items in older adults, as well as in the cognitive item measuring daze/confusion, were likely to affect by the higher prevalence of depressed mood symptoms [
76].
Additionally, the GAS-10 has been administered to test the validity and the diagnostic accuracy of the scale in two samples of healthy older adults and outpatients with GAD diagnosis. Similar to the GAS-long form, independent t-test analysis revealed that the GAS-10 was able to capture differences in anxiety symptoms between the two samples. As expected, outpatients with GAD diagnosis displayed statistically higher mean scores of the GAS-10, compared to the healthy sample. Results from correlation analysis, also, provided good support for the validity of the GAS-10. The brief and the long version of the GAS were found highly interrelated, and pairwise comparison analysis corroborated the high degree of overlap between them in correlating with concurrent validity measures.
In line with Gottschling et al. [
31], both the GAS-10 and its long form were found to correlate on average with depression (as measured by the GDS), thus confirming the co-occurrence of anxiety and depression symptomatology among elderly. In addition, this finding was not surprising, since the GDS has been criticized for its lack of unidimensionality, and construct validity [
13].
The diagnostic performance of the GAS-10 was assessed by means of ROC analysis to detect elderly with clinically significant GAD symptoms, as well as with significant anxiety symptoms, as measured by the GAI-SF cut-off scores. Differences in diagnostic accuracy between the long form version of the GAS and the GAS-10 was not found significant, when clinical/non-clinical classification variable has been taken into account. An optimal cut-off value score > 6 for the GAS-10 was selected to maximize the sum of sensitivity and specificity. Therefore all the patients with a total score of 6 in the GAS-10 should be referred for further risk assessment and management. When GAI-SF cut-off has been implemented as classification variable in the ROC analysis, the GAS-10 cut-off remained unchanged, preserving a discrete specificity. Our findings are in line with the previous studies, in which a cut-off score of > 9 optimized the balance of sensitivity and specificity (.60 and .75), for the GAS –long form [
70]. Likewise, diagnostic accuracy of the GAS- G-SF in clinical settings was low, showing a sensitivity of 70% and a specificity of 30% (cut-off > 5) [
31].
Nevertheless, the Italian version of the GAS-10 could be considered a stable measure of geriatric anxiety, with a sufficient discriminant validity in categorizing outpatients diagnosed with GAD, and adults with clinical significant anxiety symptoms (as classified by the cut score of the GAI-SF). Although its use as a screening test might be limited, the identification of a clinical cut-off score of the GAS-10 could help clinicians and researchers to identify older adults with anxiety disorders in resource-constrained settings, where time constrain and fatigue issues affect clinical assessment validity. In this direction, future studies with larger clinical samples are needed to support the clinical utility of the GAS-10.
Some limitations of this study could be addressed in future research. First, participants in the present study were community-dwelling participants and outpatients’ elderly with GAD diagnosis. There is also evidence that medically ill samples of older adults experienced severe levels of anxious distress [
23,
77], and were most likely to overburden healthcare services [
46,
78,
79]. Thus, further studies could address GAS-10 diagnostic accuracy in samples of older adults with GAD diagnosis in comorbidity with medically ill patients. A second limitation was the high prevalence of females in our clinical sample. This datum, however, was similar to real life, since being female gender was found one of the principal factor associated with the pure GAD [
80,
81]. Another limitation of this study was its cross-sectional design which limits causal inferences, and make difficult to understand if the GAS-10 items are a stable anxiety symptoms measure. Further longitudinal studies are required in this direction.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.