What is already known on this topic
- 1Anxiety is relatively common later in life.
- 2Accurately assessing anxiety in later life is of great clinical importance.
- 3Age-appropriate screening tools have maximum efficacy.
Anxiety in later life is relatively common in older populations but remains under-diagnosed and treated. Both primary and comorbid anxiety disorders in later life contribute to overall burden of disease, which is reflected in excess morbidity and mortality. One important means to reduce excess burden and instigate appropriate treatments for older adults with anxiety is to accurately assess this condition more across diagnoses and settings. The introduction of the Geriatric Anxiety Inventory and its variations has made an impact on assessment of anxiety research internationally, contributing to the growing global interest in the topic of late-life anxiety. This article discusses reasons for developing the test, processes of test construction, description of the most recent versions of the test, issues in developing culturally-appropriate normative data, adoption of the tests by researchers and clinicians, and future development plans for the test.
In Australia, the population is ageing, with adults over age 65 currently at 13.6% of the total population, a figure that is expected to rise to approximately 25% by 2056 (Australian Bureau of Statistics (ABS), 2009, 2010). This phenomenon of an ageing population is seen in both the developed and the developing world. It has implications for health care in general, and mental health care in particular, both in terms of the need for well-trained providers of geriatric mental health services as well as the need for specialised assessment tools and age-appropriate, empirically validated psychological interventions (Laidlaw & Pachana, 2009).
Clinically significant anxiety is at least as common in later life as depression (Australian Bureau of Statistics (ABS), 2008; Singleton, Bumpstead, O'Brien, Lee, & Meltzer, 2003), yet anxiety remains less well studied in terms of assessment and treatment strategies in older adults than many other disorders, including depression (Byrne & Pachana, 2010). Methodological and measurement variations contribute to the wide range of prevalence and incident rates of anxiety disorders reported in older populations. Epidemiological studies indicate that the prevalence of anxiety disorders in community-residing older people globally varies from 1.2–15% (Bryant, Jackson, & Ames, 2008). In Australia, the 2007 Australian National Survey of Mental Health and Wellbeing (NSMHWB; Commonwealth of Australia, 2008) found 4.4% of men and 6.3% of women aged 65 years and over met criteria for a current anxiety disorder. Notably, as this survey was limited to community-residing older adults without significant cognitive impairment, it may have underestimated the true prevalence of anxiety disorders in older Australians.
Anxiety disorders are also more prevalent in older adults with chronic general medical conditions and are highly comorbid with depressive disorders (Beekman et al., 2000; Lenze et al., 2001). Anxiety in older adults is associated with poorer treatment outcomes, increased disease burden, deterioration in functionality, and aggravation of symptom expression in a variety of other psychiatric and neurological disease in later life (e.g., Porensky et al., 2009; Seignourel, Kunik, Snow, Wilson, & Stanley, 2008). In fact, for both men and women, anxiety and depression comprise the top causes of non-fatal disability burden in Australia (Begg et al., 2007).
Comorbid depression and anxiety in later life present particular challenges for mental health care providers. Accurate identification of anxiety symptoms is often complicated by their similarity to the symptoms of other mental and physical disorders, as well as to medication side effects common in older populations (Palmer, Jeste, & Sheikh, 1997); when coupled with symptoms of mood disorder, such anxiety symptoms may remain undetected. In addition, the symptoms of agitation or restlessness common in anxiety may be misinterpreted as symptoms of dementia, particularly in the presence of cognitive decline. The potent comorbidity of anxiety and depressive disorders can be conceived as double jeopardy in older adults, impeding effective care and further diminishing quality of life (Beattie, Pachana, & Franklin, 2010). Geriatric healthcare providers need to understand psychiatric comorbidity of this type for accurate diagnosis and to select assessment tools and interventions appropriately.
Measurement of psychiatric conditions in later life is important but poses challenges to the clinician. In general, the psychological assessment of older adults is often challenging due to the frequent comorbidity of mental and physical health problems, presence of multiple medications and medication interactions, as well as age-related sensory and cognitive deficits (Edelstein et al., 2008). Older adults over age 65 are a diverse group, with demographic, psychosocial, health, and interpersonal differences necessitating careful attention to individual differences in assessment and treatment approaches. Age cohort effects also reflect the educational and occupational opportunities that were available to older people. These have the potential to impact on the habitual use of language and interpretation of meanings by older people.
Until recently, the majority of assessment instruments used to study anxiety in later life have been developed for use with younger populations, and thus the psychometric properties of these instruments have remained largely unexplored with regard to older adult samples (Carmin & Ownby, 2010). Against this background, instruments developed specifically for older adults may have many inherent advantages over tests developed for younger populations, including superior psychometric properties and increased acceptability among older cohorts (Edelstein & Segal, 2011). For example, the Beck Anxiety Inventory (BAI; Beck, Epstein, Brown, & Steer, 1988) is dominated by somatic items, making its use in older people with general medical disorders problematic. Assessment of anxiety in later life raises issues specific to older individuals, such as how social, developmental, and medical contexts influence the extent to which specific symptoms or behaviours represent pathology (Carmin & Ownby, 2010). In purpose-built instruments for older adults, not only can reliability and validity with this population be established, but reference to the construct of anxiety in later life and symptom expression in anxiety disorders in later life is made explicit (Edelstein et al., 2008).
There is also the issue of the validity and reliability of psychometric instruments when used on populations other than those specifically targeted in their original development. The development of screening tests, symptom severity measures, and more formal cognitive or affective tests, specific caveats regarding their use with sub-populations of older adults (e.g., those with some cognitive impairment) or their appropriateness for use in specific settings (such as residential aged care) should be specified (Pachana et al., 2010).
Lack of a widely accepted instrument to measure dimensional anxiety symptoms in older adults in community settings as well as in a variety of clinical settings was the impetus for the development of the Geriatric Anxiety Inventory or GAI (Pachana et al., 2007). Reasons for developing the test, processes of test construction, description of the most recent versions of the test, issues in developing appropriate norms, adoption of the tests by researchers and/or clinicians, and future development plans for the test are discussed in the following sections.
The guiding purpose in developing the GAI was to design a simple, clinician-friendly instrument that could be easily administered to older adults in a variety of settings, including the bedside. The GAI was designed to measure symptom severity, not as a diagnostic tool. As such, it measures the experience of dimensional anxiety across the range of anxiety disorders rather than reflecting diagnostic criteria for any specific anxiety disorder. The GAI was designed as a brief screening tool, and also to measure change in symptom severity before and after an intervention.
In the development of the original 20-item GAI (Pachana et al., 2007), over 100 potential items were generated with reference to existing anxiety scales such as the State–Trait Anxiety Inventory (Spielberger, Gorsuch, & Lushene, 1970) and the Short Anxiety Screening Test (Sinoff, Ore, Zlotogorsky, & Tamir, 1999). Special attention was paid to coverage of symptoms across all anxiety disorders, rather than with a focus on any single disorder. Items were chosen to reflect the primary domains covered in existing inventories: fearfulness, worry, meta-worry (i.e., worry about worry), thoughts about anxiety, somatic symptoms of anxiety, and anxious mood. This item pool was then reduced to 60 items through consultation with a reference group consisting of geropsychologists, geriatric psychiatrists, and community-dwelling healthy older people.
The psychometric properties of these 60 items were tested in 452 normal older adults and 46 older patients attending a psychogeriatric service (full details of these studies can be found in the original article describing development of the GAI, Pachana et al., 2007). Cronbach's alpha coefficient of the original 60 items was calculated at 0.90. Each item was then correlated with the total scores to identify those 30 items that were most highly correlated with the total score. For the sake of brevity in clinical settings, the scale was reduced to 20 items by eliminating long items or those that were potentially problematic across a variety of settings and populations (e.g., those with mild cognitive impairment; items with lower item total correlations were discarded if a choice between two similar items was made).
Items on the GAI are all scored in a single direction to avoid a possible “confusion factor” obscuring performance on the scale. This decision was based on a methodological literature (e.g., Green, Goldstein, Sirockman, & Green, 1993) suggesting that, in factor analyses with questionnaires given to older populations, reverse-scored items often form their own factor, labelled by one author as the “confusion factor.” An assessment period of one week and a dichotomous response scale, modelled on the Geriatric Depression Scale (GDS; Yesavage and colleagues, 1983) was chosen so as to maximise utility in normal clinical practice. The GAI was designed to have language that would be easily understood, with short items amenable to being read out loud to patients. The instrument's administration and scoring were kept deliberately simple to facilitate its use by a wide range of mental health professionals across an assortment of clinical settings.
The psychometric properties of this new scale were found to be sound (Pachana et al., 2007). Cronbach's alpha for the 20-item GAI was 0.91 among healthy community-dwelling older adults and 0.93 in the psychogeriatric sample. GAI score was not related to age (r = −0.12, p = .42), gender (F(1, 44) = 0.59, p = .45), or cognitive function (r = 0.08, p = .61). Concurrent validity with a variety of other measures including the State–Trait Anxiety Inventory (state version; Spielberger et al., 1970) was demonstrated in both the normal and clinical samples using Pearson product–moment correlations: GAI × STAI-S (r = 0.80, p < .00). Inter-rater and test–retest reliability were found to be excellent. Test–retest reliability was assessed by asking participants to complete the scale at a 1 week interval (r = 0.91, p < .00). Inter-rater reliability was assessed by having a second rater score the GAI on the basis of an audiotape of participant responses to an oral administration of the GAI (r = 0.99, p < .00).
Receiver operating characteristic (ROC) analysis indicated a cut-point of 10/11 for the detection of DSM-IV Generalized Anxiety Disorder (GAD) in the psychogeriatric sample, with 83% of patients correctly classified with a specificity of 84% and a sensitivity of 75%. A similar ROC analysis to identify the optimum GAI-20 cut-point to identify patients with any anxiety disorder (not shown) found an optimum cut-point of 8/9, which correctly classified 78% of patients with a sensitivity of 73% and a specificity of 80%. It was considered useful to provide cut-off scores both for all anxiety disorders as well as specific cut-off scores for GAD. Considered broadly, GAD is one of the least reliably diagnosed anxiety disorders, often going undiagnosed (Portman, Starcevic, & Beck, 2011).
The GAI subsequently has been used in a number of empirical studies of anxiety in older adult populations (e.g., Andrew & Dulin, 2007). Our own research group has investigated the psychometric properties of the GAI in older adult populations in a variety of clinical settings (e.g. Pachana, Woodman, Byrne, 2007). We are attempting to both confirm the properties and functionality of the GAI while also contributing to empirical data about anxiety in later life.
For example, generalised anxiety in memory clinic attendees was explored using the GAI (Byrne, Pachana, Arnold, Chalk, & Appadurai, 2008). A consecutive series of 100 community-residing memory clinic attendees and their caregivers at an academic medical centre were recruited. Generalised anxiety was measured on the GAI and a standardised battery of measures. Again, the utility of the GAI in detecting clinically significant levels of anxiety symptoms was confirmed, but now in a population in which over two thirds had dementia (47% Alzheimer's Disease; 3% Vascular Dementia; 16% mixed), with an additional 15% showing signs of mild cognitive impairment. This sequentially recruited memory clinic sample consisted of 58 female and 42 male patients with a mean age of 76.9 years (SD 7.8; range 61–95). Mean standardised Mini Mental State Examination score was 23.2 (SD 5.1; range 5–30) in this group. Mean GAI score was 4.97 (SD 5.32; range 0–20) with a Cronbach's alpha of 0.92. As expected, GAI score was correlated with Neuropsychiatric Inventory (NPI) anxiety severity (r = 0.31; p = .01) and NPI anxiety distress (r = 0.32; p = .00). The GAI was also correlated with Barthel score (r = −0.28; p = .01) and a leisure activities inventory score (r = −0.32; p = .01). From this study, we concluded that the mean levels of memory clinic attendees' self-reported anxiety were much higher than those reported by normal community-dwelling older volunteers but somewhat lower than those reported by psychogeriatric patients. Higher levels of anxiety were reported by patients with greater levels of physical dependency, whereas lower levels of anxiety were reported by patients with greater involvement in leisure activities. This has implications for treatment as well as the structuring of diversional activities in residential aged care settings.
In another study, the psychometric properties and health correlates of the GAI were explored by Byrne et al. (2010) in a cohort of community-residing older women in Brisbane (n = 286). The GAI demonstrated sound internal consistency and demonstrated good concurrent validity against the state half of the Spielberger State–Trait Anxiety Inventory and the neuroticism domain of the NEO five-factor inventory. GAI score was significantly associated with self-reported sleep difficulties and perceived memory impairment but not with age or objective cognitive function. Women with current DSM-IV diagnosed GAD had significantly higher GAI scores than women without such a history. In this cohort, the optimal cut-point to detect current GAD was 8/9. Although the GAI was designed to have few somatic items, women with a greater number of general medical problems or who rated their general health as worse had higher GAI scores.
Anxiety is often under-recognised in residential aged care settings, so our group (Boddice, Pachana, & Byrne, 2008) undertook to ascertain whether the GAI would have utility among more cognitively intact residents of such facilities. In this study, the predictive validity of the GAI in residential care settings was examined. Results indicated that classification of presence or absence of anxiety symptoms by the GAI was not significantly associated with an individual's cognitive status, either in a residential care sample or a comparative healthy older community-dwelling group. In addition, data supported the predictive validity of the GAI in residential care settings with respect to diagnosis of anxiety disorders. An Italian version of the GAI also supported its use in a mildly cognitively impaired population (Rozzini et al., 2009). Thus, the GAI may be a useful measure to assess anxiety symptoms in residential aged care settings with relatively cognitively intact residents.
Other research groups have studied the psychometric properties of the GAI, including Cheung (2007), Smith, Ingram, Brighton, and Adams (2009), and Yochim, Mueller, June, and Segal (2011). In particular, Diefenbach, Tolin, Meunier, and Gilliam (2009a) noted that the GAI demonstrated the strongest and the Beck Anxiety Inventory the weakest psychometric properties, with the GAI also showing excellent inter-rater reliability to support verbal administration and good ease of use in the field (specifically, among older adults receiving home care services from clinicians).
The GAI has been translated into over two dozen languages spread across five continents. Care has been taken with such translations both to maximise utility, reliability, and validity across languages and cultures, as well as to ensure response formats are consistent and understandable. The standard translation approach of a translation and independent back-translation checked for accuracy in the original language has been employed with all translations. In some cases, two independent back-translations have been made, as in Martiny, Cardoso de Oliveira, Nardi, and Pachana (2011). In addition, consultation with geriatric mental health specialists in the countries in question, as well as consultations as appropriate with the authors of the original inventory, has occurred.
Specific translations of the GAI for individual countries have been encouraged. To that end, there are US Hispanic as well as Spanish (Spain) versions of the instrument, for example. This trend has been duplicated in France and French-speaking Canada and Portugal and Portuguese-speaking Brazil. Also, care with respect to culture-specific translations of the instructions for participants filling in the instrument has been taken (e.g., participants in commonwealth countries have been told to “tick” boxes while North Americans are instructed to “check” boxes). While such attention to detail may appear trivial, a number of tests are in common use across cultures, and older adults in particular may question the relevance of tests if it appears the items or content do not pertain to their particular circumstances (Edelstein et al., 2008).
It is important to note that items in the original GAI are not all translated verbatim to other languages. The original GAI has a number of items that reflect more colloquial expressions of anxiety by older adults. For the most part these reflect gastrointestinal discomfort (e.g., “I have butterflies in my stomach”). These have been translated into feelings of tightness in the chest in the Portuguese (Portugal) (Ribeiro, Paúl, Simoes, & Firmino, 2011) GAI, and feeling like ants are in one's stomach for the Spanish (Spain) (Márquez-González, Losada, Fernández-Fernández, & Pachana, 2011) versions of the GAI. This ensures that all items reflect how older adults themselves would describe anxiety symptoms while still allowing for important cross-cultural comparisons of anxiety symptoms in healthy and clinical populations.
The psychometric properties of the GAI remain largely consistent across translations. For example, Cronbach's alpha for the Spanish (Spain) translation (0.91) and Portuguese (Portugal) translation (0.96) are similar to the Australian GAI obtained alpha of 0.91. The Italian translation has a relatively lower alpha (0.76) but similar test–retest reliability of 0.86 (compared to the original scales test–retest reliability of 0.91). Cut-off scores were similar as well; for example, ROC analyses indicated the same cut point of 8/9 for all anxiety disorders, with a sensitivity of 88% and a specificity of 80%, in the Portuguese GAI (Ribeiro et al., 2011).
It should be noted that large cross-cultural studies of anxiety are absent in the literature, whereas several such studies (e.g., Dewey, de la Cámara, Copeland, Lobo, & Saz, 1993) have been undertaken for depression. Particularly in countries with less well-developed geriatric mental healthcare systems, accurate knowledge of incidence rates of late-life anxiety would assist in service planning and provision.
The GAI has been used in a variety of international studies examining specific relationships between anxiety and other variables or outcomes in older adults, as well as sequelae of anxiety symptoms, disorders, and interventions in clinical populations. For example, Andrew and Dulin (2007) sought to examine the influence of experiential avoidance (EA) as a moderating variable between reported physical health problems and anxiety and depression among older adults in New Zealand. In previous studies in younger adults, EA has been associated with a number of psychological disorders. In this study the GAI, the GDS, and the Acceptance and Action Questionnaire (AAQ) measure of EA were administered to 208 individuals between the ages of 70 and 92 years old. Data suggested that EA explained 8% of the unique variance in depression, 20% in anxiety, and that EA moderated the relationships between self-reported health and both depression and anxiety. This study also found that relationships involving EA were more pronounced with anxiety as compared with depression, thus pointing to EA as an important consideration when examining self-reported health and anxiety in particular in older populations.
Several recent studies have used the GAI to examine anxiety in older cardiac patients. Current data suggest that symptomatic depression and anxiety are under-recognised in heart failure patients and that mental health screening may be important for receipt of care (Cully, Jimenez, Ledoux, & Deswal, 2009). Paukert, Lemaire, and Cully (2009) examined 104 older veterans over the age of 60 with heart failure to determine the relative importance of demographic, physical, and psychiatric parameters and various aspects of coping in predicting depressive symptoms in this group. The GDS and GAI were used to screen for depression and anxiety, respectively, such that the final sample consisted of equal numbers with and without significant levels of anxiety and depression. Correlational analyses indicated that depressive symptoms were more significantly associated with various poor outcomes from heart failure (e.g., physical limitations and maladaptive coping) than were symptoms of anxiety.
In a prospective study of the prevalence, recognition, and treatment of depression and anxiety in ambulatory patients with heart failure, Cully et al. (2009) found the prevalence of depression (as defined by GDS score ≥6) was 41.8%, and the prevalence of anxiety (as defined by GAI score ≥9) was 25.3%. Of their 158 patients with a positive GDS or GAI result, 57.5% had a diagnosis or medical-record notation for depression and/or anxiety in their chart. However, 60.3% of those with a recognised psychiatric syndrome received mental health treatment during the 18-month period of the study. As Cully and colleagues note, once depression and/or anxiety were documented in the medical record, patients were highly likely to receive mental health treatment. This again underscores the importance of having specific instruments to detect anxiety in older clinical populations of interest.
Finally, many international groups have used the GAI in investigations of the impact of anxiety symptoms, and specific interventions for anxiety symptoms and disorders, in older adults. For example, the prevalence and impact of anxious depression (i.e., major depressive disorder, MDD, occurring concomitant with generalised anxiety symptoms) among older adults, with a particular focus on racial an ethnic variations in presentations, was undertaken by Diefenbach, Disch, Robison, Baez, and Coman (2009b). Interviews were conducted with 218 Puerto Rican and 206 African Americans age 60 and above living in the United States. The impact of depression and anxiety was most pronounced for the Puerto Rican participants who also reported poorer subjective health and more substantial disability (independent of depression or anxiety status). These results highlight the importance of conducting culturally sensitive assessments of depression and anxiety among older adults using age appropriate instruments such as the GAI.
Diefenbach, Tolin, Gilliam, and Meunier (2009c) also have examined the efficacy of cognitive-behavioural therapy (CBT) for late-life anxiety. The authors describe their home-based CBT programme developed specifically for late-life anxiety, outlining their experience partnering with a community care management organisation. The GAI was demonstrated to be useful in showing change in participants who completed the treatment. The authors highlight the need for future research efforts with respect to treatments tailored for older adults experiencing anxiety. Here again, instruments which can successfully document change after interventions are important.
Our own research group has similarly documented the utility of the GAI in demonstrating positive change in an older adult population treated for simple phobia in a group therapy format (Pachana et al., 2007). Likewise, the GAI performed well in a study by Welch et al. (2010) examining the effectiveness of a group CBT intervention with older adults living in smaller rural and remote communities.
While several instruments, including the GAI, now exist to measure anxiety in older adults, there remains a need for a very brief self-report scale to measure anxiety symptoms in epidemiological surveys in primary care and in acute geriatric medical settings. With this brief in mind, our group undertook the development of the Geriatric Anxiety Inventory—Short Form (GAI-SF) based on the GAI (Byrne & Pachana, 2011). The five best performing items of the GAI on ROC analyses were selected to construct a measure that would have similarly high discriminative properties to the original 20-item GAI scale. Employing a cross-sectional methodology, a randomly selected group of community-residing older women (n = 284; mean age 72.2 years) were administered DSM-IV diagnostic interviews using the Mini International Diagnostic Interview, fifth edition (MINI-V). A score of three or greater was optimal for the detection of DSM-IV GAD as determined by ROC analyses. At this cut-point, sensitivity was 75%, specificity was 87%, and 86% of participants were correctly classified. GAI-SF score was not related to age, MMSE score, or level of education. Internal consistency was high, with a Cronbach's alpha of 0.81 and adequate concurrent validity against the State–Trait Anxiety Inventory (rs = 0.48, p < .001). The GAI-SF appears to show promise as a brief measure of dimensional anxiety suitable for use in epidemiological studies, as well as in primary care and acute geriatric medical settings.
Finally, an informant version of the GAI for use in persons with dementia is in development. The Informant Questionnaire for Anxiety in Dementia (IQAD) has been developed based on the ten most robust items from the GAI. These have been field tested in various smaller cohorts; a large field trial is currently under way.
With increasing numbers of older people seeking treatment, mental health professionals will need to develop greater awareness, understanding, and appreciation of psychiatric disorders in later life to deliver optimally effective services to this population (Laidlaw & Pachana, 2009). Knowledge of incidence and prevalence rates of psychiatric illnesses most affecting older adults, as well as specific and empirically validated assessment tools and treatment approaches which are effective with these disorders, are required for effective clinical practice with this population (Calleo & Stanley, 2008).
Anxiety in later life is relatively common in older populations but remains under-detected and treated. One important means to reduce excess burden and instigate appropriate treatments for older adults with anxiety is to more accurately assess this condition across diagnoses and settings. The GAI has proven useful in research into the prevalence of anxiety in specific populations, in examining relationships between anxiety and other variables of interest in older populations, and in measuring symptom reduction in intervention trials. Translations of the instrument have shown the need for care in adapting screening instruments for use in international contexts.