A Systematic Review of the Psychometric

Properties of the Geriatric Anxiety Inventory*

Alexandra Champagne,1 Philippe Landreville,1,2 and Patrick Gosselin3,4

RÉSUMÉ
Le Geriatric Anxiety Inventory (GAI) et sa forme courte (GAI-SF) sont des échelles utilisées internationalement pour évaluer
les symptômes anxieux chez les aînés. L’objectif de cette étude était de conduire la première revue critique des propriétés
psychométriques de ces outils. Les études pertinentes (n = 31) des deux versions du GAI ont été extraites de bases de données
électroniques ainsi que d’une recherche à la main. La qualité des études a été évaluée par la grille COSMIN. Le GAI et le GAI-SF
présentaient une consistance interne ainsi qu’une fidélité test-retest adéquates. La validité convergente présentait des
corrélations élevées avec des mesures d’anxiété généralisée alors que de faibles corrélations étaient retrouvées avec celles
incluant des symptômes somatiques. Un chevauchement important a été trouvé avec des mesures des symptômes dépressifs.
Alors qu’il n’y a pas de consensus quant à la structure factorielle du GAI, le GAI-SF est unidimensionnel. Malgré de bonnes
sensibilité et spécificité pour détecter l’anxiété, les scores-frontières recommandés variaient considérablement. Le GAI et le
GAI-SF sont des instruments présentant des propriétés psychométriques satisfaisantes. Afin d’élargir leur utilisation,
certaines d’entre elles nécessitent toutefois un examen plus approfondi. Cette revue souligne l’importance de porter attention
quant à certaines lacunes méthodologiques qui ont été retrouvées dans les études.

ABSTRACT
The Geriatric Anxiety Inventory (GAI) and its short form (GAI-SF) are self-reported scales used internationally to assess
anxiety symptoms in older adults. In this study, we conducted the first critical comprehensive review of these scales’
psychometric properties. We rated the quality of 31 relevant studies with the COSMIN checklist. Both the GAI and GAI-SF
showed adequate internal consistency and test-retest reliability. Convergent validity indices were highest with generalized
anxiety measures; lowest with instruments relating to somatic symptoms. We detected substantial overlap with depression
measures. While there was no consensus on the GAI’s factorial structure, we found the short version to be unidimensional.
Although we found good sensitivity and specificity for detecting anxiety, cut-off scores varied. The GAI and GAI-SF are
relevant instruments showing satisfactory psychometric properties; to broaden their use, however, some psychometric
properties warrant closer examination. This review calls attention to weaknesses in the methodological quality of the studies.

1
School of Psychology, Université Laval, Québec
2
Centre d’excellence sur le vieillissement de Québec, Québec
3
Department of Psychology, Université de Sherbrooke, Sherbrooke
4
Institut universitaire de première ligne en santé et services sociaux – Centre intégré universitaire en santé et services sociaux de
l’Estrie - CHUS (CIUSSS de l’Estrie- CHUS), Sherbrooke
*
The authors wish to thank Alexandra Michel in the searching and assessment of methodological quality of studies.

Funding Sources None.

Conflict of Interest Authors have no conflicts of interest to declare.
Manuscript received: / manuscrit reçu : 01/05/2019
Manuscript accepted: / manuscrit accepté : 22/04/2020
Mots-clés: vieillissement, Geriatric Anxiety Inventory, propriétés psychométriques, validité, fidélité, COSMIN
Keywords: aging, Geriatric Anxiety Inventory, psychometric properties, validity, reliability, COSMIN
La correspondance et les demandes de tirés-à-part doivent être adressées à : / Correspondence and requests for offprints should be
sent to:
Alexandra Champagne, Psy.D.
School of Psychology
Université Laval
2325, rue des Bibliothèques
Québec, QC G1V 0A6
([email protected])

Canadian Journal on Aging / La Revue canadienne du vieillissement: Page 1 of 20 (2020) 1

doi:10.1017/S0714980820000185
2 Canadian Journal on Aging Alexandra Champagne et al.

Introduction conducted on the psychometric properties of the GAI

The Geriatric Anxiety Inventory (GAI; Pachana et al., and GAI-SF in various populations and in different
2007) is one of the few appropriate screening measures languages. Reviews of this literature (Balsamo, Cataldi,
for assessing anxiety in elderly individuals (Creighton, Carlucci, & Fairfield, 2018; Creighton et al., 2018;
Davison, & Kissane, 2018; Therrien & Hunsley, 2012). It Dissanayaka, Torbey, & Pachana, 2015; Edelstein
is a self-report scale composed of 20 items designed to et al., 2008; Lin et al., 2016; Pachana & Byrne, 2012;
assess anxiety symptoms over the past week using a Therrien & Hunsley, 2012) generally conclude that these
dichotomous (yes/no) response format. Prior to the measures present sound psychometric properties
despite some disparities in the results (e.g., factorial
final 20-item version of the GAI, scale developers gen-
structure, convergent validity) and possible problem-
erated a pool of items, either formulated de novo or
atic issues (e.g., problematic items, divergent validity,
adapted from existing anxiety scales (Pachana, Byrne,
cross-cultural issues). However, none of the reviews
et al., 2007). Items were chosen by the developers if they
evaluated the methodological quality of the studies they
reflected general anxiety and covered primary domains
examined, which makes it difficult to determine the
of existing scales (e.g., fearfulness, worry, cognitions
appropriateness of the findings and to offer guidelines
about anxiety). Developers were careful to limit the
to improve research in this area. Moreover, each review
inclusion of somatic symptoms that may overlap with examined only a small portion of studies in light of the
symptoms of general medical conditions. The develop- available empirical evidence on the GAI and GAI-SF.
ers reduced the pool of items by consulting with a This is not surprising because most reviews are not
reference group and psychometric testing with pilot dedicated exclusively to the GAI; some are specific to
samples. The GAI was normed with samples of certain populations or settings (e.g., Parkinson’s dis-
community-dwelling seniors and older adults receiving ease; residential aged care facilities), and new studies
psychiatric services. Total score on the GAI ranges have only been published recently.
between 0 and 20 and higher scores indicate greater
anxiety symptoms. Cut-off scores of 11 and 9 and above The goals of this review were to summarize existing
were recommended to detect generalized anxiety dis- evidence on the psychometric properties of the GAI and
order (GAD) and other anxiety disorders respectively. GAI-SF, to assess the methodological quality of the
In the original study, the GAI presented sound psycho- studies and to provide guidance for future psychomet-
metric properties with internal consistency coefficients ric validation studies. To our knowledge, this was the
of .91 and .93 for the non-clinical and psychogeriatric first systematic review that examined the methodo-
logical quality of studies that were conducted on the
samples respectively (Pachana, Byrne, et al., 2007). Con-
psychometric properties of these instruments.
vergent validity varied between .44 to .70 for measures
that assessed anxiety or related constructs. Retest reli-
ability was .91 for the psychogeriatric sample. Methods
This review targeted published studies that reported
A 5-item short form (GAI-SF; items 1, 6, 8, 10 and 11 of
data on psychometric properties of the GAI and
the GAI long form; Byrne & Pachana, 2011) was further
GAI-SF in older adults. The methodology was guided
developed to make the use of the instrument more
by the Preferred Reporting Items for Systematic
practical in primary care and acute geriatric medical
Reviews and Meta-Analyses (PRISMA) statement
settings. The short-form items were chosen among
guidelines (Moher et al., 2015; Moher, Liberati, Tet-
those of the standard GAI based on parameters such
zlaff, & Altman, 2009).
as scale cohesion (item-total correlation), endorsement
rate, and ability to distinguish participants with GAD Hunsley and Mash (2008) have provided criteria for
from those without the diagnosis. A cut-off score of 2/3 rating some psychometric results. Internal consistency
out of a maximum total score of 5 on the GAI-SF proved is considered “adequate” with Cronbach α values of
optimal for the identification of GAD. The GAI-SF .70–.79, “good” with α values of .80–-.89, and “excel-
presented satisfactory results with respect to internal lent” with α values equal or greater than .90. Test-retest
consistency (α = .81), retest reliability (r = .80), and reliability is considered “adequate” when correlations
convergent validity indices (r = .48–.88) (Byrne & are of at least .70 over a period of several days to several
Pachana, 2011). The popularity of the GAI is well weeks, “good” when they are of at least .70 over a
reflected by the fact that it has been translated into more period of several months, and “excellent” when at least
than 24 languages (Pachana & Byrne, 2012). This they are of .70 over a period of a year or longer.
includes English and French for Canada, which makes
it a convenient tool for use at a national level. Search Strategy
Since the publication of Pachana, Byrne, et al.’s (2007) We conducted a literature search using Pubmed, Psy-
first study on the GAI, much research has been cINFO, CINAHL, EMBASE, and Google Scholar as
Properties of the Geriatric Anxiety Inventory La Revue canadienne du vieillissement 3

these databases are representative of the literature pub- For each measurement property, an overall score is
lished on this topic. We made additional efforts to locate determined by taking the lowest rating of any of the
relevant studies through a handsearching process. The box items (worst score counts method; Terwee et al.,
keywords “Geriatric Anxiety Inventory” in the title and 2012). Quality assessment of studies was independently
abstract section of the databases was what we used to performed by the first author and a research assistant.
filter relevant studies. We decided on this approach Discrepancies were resolved through a discussion.
after conducting different tests (e.g., with broader key- When necessary, a third reviewer made the decision.
words like “anxiety” or “assessment” or by including
them in the “any field” section), which considerably
broadened the number of non-relevant articles Data Collection Process
retrieved. The search was restricted to articles published Data extraction was conducted by the first author.
between January 1, 2007 (the GAI was developed in Extracted data included basic information about study
2007 [Pachana et al., 2007]) and December 31, 2018. We demographics (e.g., publication year, country in which
retained articles according to the following criteria: the study was conducted, language in which the instru-
(a) written in English or French, (b) presented original ment was administered) as well as sample characteris-
empirical research, and (c) expressed the primary tics (e.g., type of sample, sample size, mean age). When
objective of exploring the psychometric properties of available, we collected data on the different measure-
the GAI and/or the GAI-SF. We excluded the following ment properties (e.g., results, statistical methods used,
types of articles because either the information pro- time interval, comparator instruments) defined in the
vided was limited or the articles were frequently non- COSMIN checklist. More specifically, these properties
peer reviewed: unpublished manuscripts, editorials, were: internal consistency, test-retest reliability, meas-
dissertations, theses, randomized controlled trials, case urement errors, content validity, structural validity
reports, and published abstracts. The first author and a (factor analysis), hypothesis testing, cross-cultural val-
research assistant independently screened the titles and idity, criterion validity, and responsiveness.
abstracts of the retrieved studies to determine their
eligibility. When a disagreement emerged between the
two reviewers, a discussion ensued in order to reach a Results
consensus. When necessary, a third reviewer made the Search Results
decision.
As shown in Figure 1, the database search retrieved a
total of 485 articles. Duplicates (n = 232) were removed
Quality Assessment and of the 253 remaining records, we excluded 222.
We assessed the methodological quality of the included The main reasons for exclusion were that the GAI or
studies with the “COnsensus-based Standards for the GAI-SF was not the topic of interest (n = 179) or that the
selection of health status Measurement Instruments” article was not presented as an original published
(COSMIN) checklist (Mokkink, Terwee, Knol, et al., manuscript (i.e., conference proceeding; n = 21). Four
2010; Mokkink, Terwee, Patrick, et al., 2010). The COS- articles were excluded based on language (i.e., were
MIN checklist consists of eight boxes that each refer to a not written in English or in French). Thus, we retained
specific measurement property (i.e., internal consistency, a final list of 31 articles for the purpose of the current
reliability, measurement error, content validity, struc- review.
tural validity, hypothesis testing, cross-cultural validity,
and responsiveness). Each box contains 5 to 18 items that
Methodological Quality of the Included Studies
assess methodological standards, and items are scored
on a 4-point rating scale (i.e., poor, fair, good, or excel- The results of COSMIN ratings for the 31 studies
lent) using specific criteria. For example, the fifth item on retained are displayed in Table 1. The studies assessed
the internal consistency box assesses whether the uni- an average of 2.7 psychometric properties out of the
dimensionality of the scale was verified. Criteria pro- nine COSMIN criteria. Most of the COSMIN boxes
posed by the COSMIN checklist for rating this item were rated as having “poor” (43.5%) or “fair” (40%)
follow: a factor analysis was performed in the study quality. The most frequent reasons for these ratings
population (excellent); the authors refer to another study were low sample size or a lack of information concern-
in which factor analysis was performed in a similar study ing the number of missing items and how they were
population (good); authors refer to another study in handled. This information corresponds to key criteria
which factor analysis was performed but not in a similar because it is assessed in almost all COSMIN boxes.
study (fair); factor analysis was not performed and Only 11.8 per cent of the rated boxes were rated as
contains no reference to another study (poor) (for more having “good” quality, and 4.7 per cent as having
information on rating, see https://www.cosmin.nl/). “excellent” quality.
4 Canadian Journal on Aging Alexandra Champagne et al.

Idenficaon
Records identified through database Additional records identified
searching through other sources
(n = 484) (n = 1)

Number of duplicates removed

(n = 232)
Screening

Records screened and

assessed for eligibility Records excluded
(n = 222)
Eligibility

(n = 253)
Included

Studies included in
synthesis
(n = 31)

Figure 1: Flow diagram of study selection

Study and Participant Characteristics Sample recruitment source was categorized as either
non-clinical (e.g., community-dwelling seniors), psychi-
Basic characteristics of the studies retained for the cur-
atric (e.g., in-patient, outpatient, or institutionalized
rent review and their samples are presented in Table 2.
patients, or individuals with a psychiatric diagnosis),
Psychometric properties of the GAI were examined by
medical (i.e., having a medical diagnosis or receiving
22 studies, while only one study investigated the prop-
medical care), or mixed (i.e., different sources of recruit-
erties of the GAI’s short form, and eight studies exam-
ment in the sample). Of the 31 selected studies, nine
ined both forms. The latter studies generally extracted
used mixed samples (non-clinical and/or psychiatric
GAI-SF scores from the GAI. We examined psychomet-
and/or medical). Other studies’ recruitment sources
ric properties of 15 versions of the GAI: Brazilian Por-
were for the most part exclusively non-clinical (n = 9),
tuguese, Chinese, Czech (long and short forms), English
medical (n = 9), or, in a smaller proportion, psychiatric
(long and short forms), French Canadian (long and
(n = 4 studies). Mean scores on the GAI varied between
short forms), Italian (long and short forms), Norwegian
.58 to 16.3, with a mean of 5.5. Those for the GAI-SF
(long and short forms), Portuguese (long and short
ranged between .17 to 3.64, with a mean of 1.8.
forms) and Spanish.
The 31 retained studies provided data for 8,174 patients
who completed the GAI and/or the GAI-SF. Sample
Reliability
sizes ranged from 32 to 1,318 patients. Most studies had
samples composed mainly of women (on average, Internal Consistency. The alpha coefficient of the GAI
64.9% of the samples were composed of women). Par- ranged between .71 and .97 with a mean of .91, and
ticipants were aged between 52 and 94 years old, between .61 to .84 with a mean .80 for the GAI short
excluding participants in the study by Matheson et al. form (see Table 2). According to the COSMIN checklist
(2012) that included young adults aged 37 years old and results, internal consistency was mostly (63%) rated as
older. Mean age of the participants was 72.5 years. poorly assessed. The items rated as “poor” referred
 Properties of the Geriatric Anxiety Inventory
Table 1: Methodological quality of each study per measurement property

Internal Measurement Content Structural Hypothesis Cross—Cul- Criterion

Study Consistency Reliability Error Validity Validity Testing tural Validity Validity Responsiveness

Brazilian Portuguese Version (GAI-BR)

Massena et al. (2015) Poor Poor — — — Fair — Fair —
Chinese version (GAI-CV)
Yan et al. (2014) Good — — — Good — Poor — —
Guan (2016) Fair — — — Fair — Poor — —
Dow et al. (2018) Poor — — — — — Poor — —
Czech version
Heissler et al. (2018) Fair — — — — — Poor — —
English version
Cheung (2007) Poor — — — — — — — —
Pachana et al. (2007) Poor Fair — Excellent — — — Fair —
Boddice and Byrne (2008) — — — — — — — — —
Diefenbach, Tolin, Meunier, and Gilliam (2009) Poor Poor — — — — — Fair —
Byrne et al. (2010) Poor — — — — — — Fair —
Cheung et al. (2012) Poor — — — — — — Fair —
Matheson et al. (2012) Poor Poor — — — — — Fair —
Bradford et al. (2013) Poor — — — — Poor — Fair —
Gerolimatos, Gregg, and Edelstein (2013) Poor — — — — Good — Good —
Diefenbach et al. (2014) Fair — — — Fair Fair — —
Gould et al. (2014) Poor — — — — — — — —
Ball et al. (2015) Poor — — — — — — — Poor
Johnco, Knight, Tadic, and Wuthrich (2014) Good — — — Good — — Good —
Gould et al. (2016) Poor — — — — — — — —
Kneebone et al. (2016) Poor Poor Fair — — Fair — Fair —
Creighton et al. (2018) Poor — — — — — — Fair —
English version – short form (GAI-SF)
Byrne and Pachana (2011) Poor Poor — — — — — Excellent —
French Canadian version (GAI-FC)

La Revue canadienne du vieillissement

Champagne et al. (2016) Fair Fair — — Fair Fair — — —
Italian version (GAI-It)
Rozzini et al. (2009) Poor Fair — — — — Poor — —
Ferrari et al. (2017) Poor — — — — Fair — Fair —
Norwegian version
Bakkane Bendixen et al. (2016) Excellent Poor — — Excellent — — — —
Molde et al. (2017) Fair — — Poor Fair — Fair Fair —
Portuguese version (GAI-PT)
Ribeiro et al. (2011) Good Poor — — Good — Poor Good —
Silva et al. (2016) Poor Poor — — — — — — —
Spanish version
Marquez-Gonzalez et al. (2012) Fair — — — Fair — Poor — —
Mababu and RuizSánchez (2016) Fair — — — Fair — — — —

5
 6
Table 2: Characteristics of the retained studies on the GAI and GAI-SF and their reliability coefficients

Canadian Journal on Aging

Internal
Gender Mean Age & Consistency Test-retest
Version Study Country Type of Sample Subjects (n) Female (%) Range (years) Mean Score (SD) (α) Reliability

Brazilian Portuguese Massena et al. Brazil Mixed Mixed sample from com- 82 72.2 8.77 .91 .85
version (GAI-BR) (2015) munity and outpatient 1 week
Long form psychogeriatric clinic
(n = 72)
Chinese version Yan et al. (2014) Beijing Nonclinical Community-dwelling 59.4 70.8 2.17 (4.19) .94 —
(GAI-CV) seniors (n = 1,047)
Long form Guan (2016) Beijing Nonclinical Community-dwelling 59.4 71.4 — .94 —
seniors (n = 1,318)
Dow et al. (2018) Australia Nonclinical Community-dwelling 66 76.9 — .95 —
Chinese immigrants 60–92
(n = 87)
Czech version Heissler et al. Czech Repub- Nonclinical Community-dwelling 52 75.5 Men: 2.27 (2.85) .85 —
Long form (2018) lic seniors (n = 485) Women: 3.44
(3.74)
Czech version Heissler et al. Czech Repub- Nonclinical Community-dwelling 52 75.5 Men: .64 (1.11) .75 —
Short form (2018) lic seniors (n = 485) Women: 1.08
(1.37)
English version (GAI) Cheung (2007) New Zealand Psychiatric Geriatric psychiatry 63 75.5 7.59 (6.5) — —
Long form patients (n = 32) 66–85
Pachana et al. Australia Mixed Community-dwelling 64.4 71.7 2.3 (3.8) .91 —
(2007) – Nonclinical (n = 452) 60–90
– Psychiatric Patients attending a 74 78.8 5.22 (5.83) .83 .91
psychogeriatric 66–94 1 week
service (n = 46)
Boddice and Australia Mixed Community-dwelling 52 75.8 — — —
Byrne (2008) – Nonclinical seniors (n = 31)
– Medical Older adults living in 62.9 82.8 2.3 (4.2) — —
nursing homes (n = 27)
Diefenbach United States Medical Older home care recipi- 83.3 76.6 4.63 (5.57) .93 .95
et al. (2009) ents (n = 66); data on 6592 1 to 2 weeks
the GAI available only
for a subset of the

Alexandra Champagne et al.

sample (n = 35)
Byrne et al. (2010) Australia Nonclinical Community-residing older 100 71.7 2.33 (4.05) .92 —
women (n = 286) 60–86
Cheung et al. New Zealand Medical Older adults with chronic 44 72.7 3.3 (4.6) .92 —
(2012) obstructive pulmonary
disease (n = 55)
Continued
Table 2: Continued

Properties of the Geriatric Anxiety Inventory

Internal
Gender Mean Age & Consistency Test-retest
Version Study Country Type of Sample Subjects (n) Female (%) Range (years) Mean Score (SD) (α) Reliability
Matheson et al. Australia Medical Parkinson’s disease 41 66.2 5.03 (6.06) .95 .99
(2012) patient (n = 58) 37–88 2 weeks
Bradford et al. United States Medical Patients with mild to 23 79 5.8 (5.7) .92 —
(2013) moderate dementia
(n = 41)
Gerolimatos United States Medical Nursing home residents 52 69.6 7.92 (6.17) .92 —
et al. (2013) (n = 75) 52–94
Diefenbach United States Mixed Older adults with mild 55.6 76.7 5.71 (5.37) .91 —
et al. (2014) dementia (n = 45)
– Medical Elderly individuals with 70.9 70.1 5.73 (5.64) .92 —
cognitive impairment
but no dementia
(n = 55)
– Nonclinical Nonclinical group (n = 50) 56 69.5 1.42 (2.44) .83 —
Gould et al. United States Mixed Total sample (n = 110); 57.3 75.2 1.51 (3.07) .89 —
(2014) data on the GAI avail-
able only for a subset
of the sample (n = 74)
– Nonclinical Nonclinical anxiety — 1.19 (2.42)
(n = 67)
– Psychiatric Current anxiety disorder — 4.57 (6.16)
(n = 7)
Ball et al. (2015) United States Psychiatric Elderly patients with gen- 78 71.6 14.3 (4.1) .81 —
eralized anxiety dis-
order (n = 291)
Johnco et al. Australia Mixed Total sample (n = 256) 62.8 67.5 — .93 —
(2014) 60–88
– Psychiatric Clinical geriatric partici- 60.4 67.5 11.08 (4.86) .85 —
pants with co-morbid 61–88
anxiety and unipolar

La Revue canadienne du vieillissement

mood disorder
(n = 197)
– Nonclinical Nonclinical control group 71.2 67.6 0.58 (1.32) .71 —
(n = 59) 60–86
Gould et al. United States Medical Older veterans who 0 78.5 1.94 (2.77) .82 —
(2016) attended a geriatric
primary care out-
patient clinic (n = 50)
Kneebone et al. England Medical In-patients with stroke 48.1 Mdn: 79 (IQR = 4.41 (5.80) .95 .53
(2016) (n = 81) 14.5)
Creighton et al. Australia Medical Nursing home residents 66.7 85.4 4.46 (5.79) .95 —
(2018) (n = 180)

Continued

7
Table 2: Continued

Internal

8
Gender Mean Age & Consistency Test-retest

Canadian Journal on Aging

Version Study Country Type of Sample Subjects (n) Female (%) Range (years) Mean Score (SD) (α) Reliability
English version Byrne and Australia Nonclinical Community-residing older 100 72.2 2.5 (4.2) .81 .80
(GAI-SF) Pachana women (n = 284) 60–87 1 week
Short form (2011)
Gerolimatos United States Medical Nursing home residents 52 69.6 2.23 (1.72) .73 —
et al. (2013) (n = 75) 52–94
Diefenbach United States Mixed Older adults with mild 55.6 76.7 2.09 (1.81) .77 —
et al. (2014) dementia (n = 45)
– Medical Elderly individuals with 70.9 70.1 2.05 (1.85) .80 —
cognitive impairment
but no dementia
(n = 55)
– Nonclinical Nonclinical group (n = 50) 56 69.5 0.56 (0.99) .61 —
Johnco et al. Australia Mixed Total sample (n = 256) 62.8 67.5 — .84 —
(2015) 60–88
– Psychiatric Clinical geriatric partici- 60.4 67.5 3.46 (1.48) .67 —
pants with co-morbid 61–88
anxiety and unipolar
mood disorder
(n = 197)
– Nonclinical Nonclinical control group 71.2 67.6 0.17 (0.62) .72 —
(n = 59) 60–86
French Canadian Champagne Canada Nonclinical Community-dwelling 73.5 74.6 4.10 (5.51) .94 .89
version (GAI-FC) et al. (2016) seniors (n = 331) 1 week
Long form
French Canadian Champagne Canada Nonclinical Community-dwelling 73.5 74.6 1.31 (1.66) .83 .85
version et al. (2016) seniors (n = 331) 1 week
(GAI-FC-SF)
Short form
Italian version Rozzini et al. Italy Mixed Total sample: outpatients 56 71.2 3.2 (3.8) .76 .86
(GAI-It) (2009) – Medical with mild cognitive 1 week
Long form impairment (n = 57)
– Psychiatric Patients with anxiety 62 69 11.5 (1.2) — —
(n = 44)
– Nonclinical Patients without anxiety 55 71.8 2.4 (2.5) — —
(n = 13)
Ferrari et al. Italy Mixed Mixed sample of out- 60.5 72.7 11.3 (6.5) .93 —

Alexandra Champagne et al.

(2017) – Psychiatric patients from psychi- 65–91
– Medical atric services and from
a clinic for cognitive
disorders (n = 76)
Italian version Ferrari et al. Italy Mixed Mixed sample of out- 60.5 72.7 3.1 (2.1) .77 —
(GAI-It SF) (2017) – Psychiatric patients from psychi- 65–91
Short form – Medical atric services and from
a clinic for cognitive
disorders (n = 76)
Continued
Table 2: Continued

Properties of the Geriatric Anxiety Inventory

Internal
Gender Mean Age & Consistency Test-retest
Version Study Country Type of Sample Subjects (n) Female (%) Range (years) Mean Score (SD) (α) Reliability

Norwegian version Bakkane Bend- Norway Mixed Total sample: patients 67 75.7 8.5 (6.6) .92 for the 1st —
(GAI) ixen et al. who were admitted to factor
Long form (2016) a department of geri- .85 for the
atric psychiatry 2nd factor
(n = 428)
– Psychiatric Patients with a diagnosis 67.6 75.6 11.1 (6.0) — —
of depression (n = 220)
Patients with a diagnosis 64.3 76.9 5.9 (6.1) — —
of nonorganic psych-
osis (n = 68)
– Medical Patients with a diagnosis 70.6 73.3 5.5 (5.9) — —
of dementia (n = 140)
Molde et al. Norway Psychiatric Psychogeriatric mixed 67.9 75.7 8.2 (6.5) .94 —
(2017) in-and-out patient 62–78
sample (n = 543)
Norwegian version Molde et al. Norway Psychiatric Psychogeriatric mixed 67.9 75.7 — .84 —
Short form (2017) in-and-out patient 62–78
sample (n = 543)
Portuguese version Ribeiro et al. Portugal Mixed Total sample (n = 217) — — — .96 —
(GAI-PT) (2011)
Long form – Nonclinical Community-dwelling 56.6 73.9 With PD: 16.3 (4.9) .97 .99 (ICC)
seniors (n = 152) 59–92 Without PD: 4.1 2 weeks
(5.4)
– Psychiatric Patients with depression 71.9 70.5 15.2 (5.58) —
(n = 32) 55–85
Patients with anxiety 47.8 72.3 With AD: 14.8 (4) — —
disorders (n = 23) 56–89 With GAD: 16.1
(4.7)
– Medical Patients with an early 80 74.6 11.9 (5.7) — —
Alzheimer’s disease 63–88

La Revue canadienne du vieillissement

(n = 10)
Silva et al. (2016) Brazil Medical Patients registered in a 78.2 72.8 9.2 (4.89) .89 .58
primary care setting 60–91 30 weeks
(n = 55)
Portuguese version Silva et al. (2016) Brazil Medical Patients registered in a 78.2 72.8 3.04 (1.44) .62 .97
(GAI-SF) primary care setting 60–91 30 weeks
Short form (n = 55)
Spanish version Marquez- Spain Nonclinical Community-dwelling 75.5 71.7 Female: 8.30 .91 —
Long form Gonzalez seniors (n = 302) (5.62)
et al. (2012) Male: 6.43 (6.12)
Mababu and Spain Nonclinical Community-dwelling 61 67.6 — .83 —
RuizSánchez seniors (n = 652) 60–89
(2016)

Note. AD = anxiety disorder; GAD = generalized anxiety disorder; IQR = interquartile range; MDN = median; PD = psychological distress.

9
10 Canadian Journal on Aging Alexandra Champagne et al.

mostly to the absence of information on missing items scale. Molde et al. (2017) performed a content analysis
and unidimensionality of the scale. on data retrieved from a panel of older adults and a
group of clinical psychologists and psychiatrists who
Test-Retest Reliability. Test-retest coefficients (mostly Pear-
were invited to comment on the items. Content validity
son’s r and the intraclass correlation [ICC]) ranged
was rated as “excellent” for the study by Pachana,
between .53 to .99 with a mean of .79 for the GAI, and
Byrne, et al. (2007) and “poor” for the one by Molde
between .80 to .97 with a mean of .90 for the short form
et al. (2017), according to COSMIN criteria. The latter
(see Table 2). Aside from the two lowest coefficients of the
study obtained such a rating because it is not clear
long form (r = .53 and .58; Kneebone, Fife-Schaw, Lin-
whether all items were assessed to determine whether
coln, & Harder, 2016; Silva et al., 2016), the lowest coef-
they comprehensively covered the construct of interest
ficient was .85. These large differences in the coefficients
in regard to its theoretical foundation, and whether
obtained are difficult to explain, and authors did not
items were relevant to the purpose of the instrument.
comment on their results. In general, the interval of time
between the two administrations of the scale was one to Convergent Validity. Convergent validity has been estab-
two weeks, except in the study of Silva et al. (2016) in lished between the GAI and GAI-SF and a variety of
which the interval was 30 weeks. Surprisingly, this longer other instruments that also assess anxiety and related
interval generated a coefficient of r = .58 for the GAI and constructs (e.g., symptoms of GAD, worry, general
the highest coefficient for the short form (r = .97). anxiety, or both anxiety and symptoms of depression
at the same time). As shown in Table 3, correlations vary
Test-retest reliability was chiefly rated as poorly
assessed (72.7%) according to the COSMIN checklist between .25 to .86 for the GAI and between .55 to .79 for
because of a lack of information concerning missing the short form. Convergent validity with GAD scales
items, stability of participants, and similarity of test appear to be the highest (r = .65 to .86). Data are scarce
conditions between the two administrations. The COS- on the association between the GAI and measures that
MIN checklist asks whether there were any important assess other anxiety disorders. Available evidence
flaws in the study design or method, and in light of reveals only a moderate relationship (r = .56) with a
certain retest research recommendations, there are sev- measure of post-traumatic symptoms (Gould et al.,
eral other weaknesses present. In the majority of the 2014). The weakest associations were found for scales
studies, little information was provided on sampling that contain somatic items such as the Hamilton Anx-
and rationale for major decisions that were made (e.g., iety Scale (HAMA) (r = .25; Ball, Lipsius, & Escobar,
length of the retest interval). Polit (2014) has suggested 2015) and the Beck Anxiety Inventory (BAI) (r = .28;
that seeking input from patients or experts regarding Gould et al., 2014). In contrast, the GAI focuses pre-
the stability of the construct being assessed can help dominantly on psychological symptoms. Another low
support decisions regarding retest interval. Park, Kang, correlation was found with the State-Trait Anxiety
Jang, Lee, and Chang (2018) have recommended that Inventory [STAI]-subscale state) (r = .28; Massena, de
the sample size be about five times the number of items, Araújo, Pachana, Laks, & de Pádua, 2015). The authors
which was not the case for any of the studies since they explained this result as due to a possible bias in the
generally assessed the retest reliability on a subgroup of formulation of the questions of the STAI-state, where
the sample. Moreover, the attrition rate for the retest symptoms were assessed according to participants’
assessment was rarely reported although there is evi- feelings at the time of the interview rather than those
dence that high rates of attrition can depress reliability experienced over the past week.
estimates (Polit, 2014). Although the COSMIN checklist
Convergent validity was not evaluated in depth with
prioritizes the use of the ICC to analyze retest reliability,
the COSMIN checklist because only two items referred
Vaz, Falkmer, Passmore, Parsons, and Andreou (2013)
to it in the hypothesis test box. For the purpose of this
made a case to consider measurement error indices such
review, we rarely used these items to assess convergent
as the coefficient of repeatability (CoR) or the smallest
validity since most of the retained studies did not
real difference (SRD) over coefficients like the Pearson’s
provide hypotheses to test. Despite this, the general
r and the ICC.
trend was that studies provided a poor description of
the constructs measured by the comparator instrument.
Validity In addition, it was not always clear whether the com-
parator instrument was an established and validated
Content Validity. Only two studies addressed content
instrument for use with elderly individuals.
validity. This very low number could be explained by
the fact that validation studies may have assumed that Divergent Validity. We assessed divergent validity in
items of the GAI and GAI-SF are relevant and compre- some studies by examining the association with a meas-
hensive. As the developers of the GAI, Pachana, Byrne, ure of depression symptoms. Correlations ranged
et al. (2007) thoroughly evaluated the content of this between .28 to .86 for the GAI and between .37 to .63
Properties of the Geriatric Anxiety Inventory La Revue canadienne du vieillissement 11

for the short form (see Table 3). The lowest correlations Only items 1 – “I worry a lot of the time” – and 2 – “I find
(r = .28) between the GAI and the Hospital Anxiety and it difficult to make a decision” – were always related to
Depression Scale – Depression Scale (HADS-D) are cognitive symptoms and items 12 – “I get an upset
explained by the fact that patients with major depres- stomach due to my worrying” – and 18 – “I sometimes
sive disorder were excluded (Ball et al., 2015) and by the feel a great knot in my stomach” – were always associ-
low prevalence of depression symptoms (Kneebone ated with physical symptoms. This variability may be
et al., 2016). These results suggest that there may exist due to the type of sample (i.e., three studies used non-
different patterns of divergent validity where highly clinical samples; one, a mixed sample of psychiatric and
uniform samples with low rates of depression symp- medical patients; and one, composed of elderly people
toms could facilitate distinction from anxiety symptoms with cognitive impairment) and cultural differences
assessed with the GAI and GAI-SF. because four versions were used (Norwegian, English,
Spanish, and Chinese).
Diefenbach, Bragdon, and Blank (2014) and Bakkane
Bendixen, Hartberg, Selbæk, and Engedal (2016) shed Four studies investigated the factor structure of the GAI-
new light on the association between the GAI and GAI- SF and all confirmed its unidimensionality (Champagne
SF and measures of depression. Diefenbach et al. (2014) et al., 2016; Diefenbach et al., 2014; Johnco et al., 2014;
found that depressive symptoms were more strongly Molde et al., 2017). Most items with high factor loadings
correlated with the “central nervous system hyperar- referred to cognitive symptoms of anxiety.
ousal” factor and to a lesser extent with “gastrointes-
Criterion Validity. At first, Pachana, Byrne, et al. (2007)
tinal symptoms”. These results suggest that there may
recommended a GAI cut-off score of 9 for the identifi-
be a certain response pattern in patients with greater
cation of any anxiety disorder and of 11 for the detection
co-morbid depressive symptoms. Bakkane Bendixen
of GAD. Further studies suggested cut-off scores that
et al. (2016) found that in comparison to those with
varied between 3 and 13 out of 20 for the identification
dementia or psychosis, a group of patients with depres-
of an anxiety disorder (see Table 5). Multiple factors can
sion present a different pattern of results on the GAI;
explain this variability such as the type of sample (non-
that is, with a higher total score and a higher endorse-
clinical vs. clinical), the proportion of patients who
ment of 18 of the 20 items (except items 3 and 18).
actually met the criteria for an anxiety disorder, cultural
Factorial Validity. The GAI was first described as being differences in the expression of anxiety, and the external
unidimensional although no factor analysis was pre- criterion used for the diagnosis. The much lower cut-off
sented to support this assumption (Byrne & Pachana, score of 3 found by Cheung, Patrick, Sullivan, Cooray,
2011; Pachana, Byrne, et al., 2007). Ten studies investi- and Chang (2012) may be attributable to differences in
gated the factorial validity of the GAI and half of them the nature of the sample as their participants had
confirmed the one-factor structure (Champagne, Land- chronic obstructive pulmonary disease; the mean score
reville, Gosselin, & Carmichael, 2016; Johnco, Knight, on the GAI was low (M = 3.3; SD = 4.6) as was the
proportion of participants with an anxiety disorder
Tadic, & Wuthrich, 2014; Molde et al., 2017; Ribeiro, Paul,
(25.5%). Test sensitivity values for the GAI ranged
Simoes, & Firmino, 2011; Yan, Xin, Wang, & Tang, 2014).
between 30 and 100 per cent; while specificity values
The other five studies that investigated the factorial ranged between 43 and 100 per cent. The area under the
validity of the GAI found a two-factor structure ROC curve (AUC) ranged between 79 and 98.1.
(Bakkane Bendixen et al., 2016), a three-factor structure
For the GAI short form, a score of 3 or more was
(Guan, 2016; Mababu & RuizSánchez, 2016; Marquez-
originally found to be optimal for the detection of
Gonzalez, Losada, Fernandez-Fernandez, & Pachana,
GAD in a non-clinical sample (Byrne & Pachana,
2012), and a four-factor structure (Diefenbach et al.,
2011). Results of subsequent studies were similar with
2014) (see Table 4). The identified factors can be
optimal thresholds at 2 to 3 out of 5 for the identification
grouped into three categories: (a) cognitive symptoms
of an anxiety disorder. Sensitivity varied between
(includes the following factors: worries, excessive
72 and 100 per cent and specificity ranged between
worry symptoms, decision-making symptoms, and
35 and 98.3 per cent. The AUC ranged between 78 and
mental anxiety), (b) physical symptoms of anxiety
95.4.
(includes the following factors: central nervous system
hyperarousal, arousal and somatic symptoms), and With regard to the different diagnostic parameters, the
(c) negative anxiety. Cognitive and physical symptoms performance of the standard and short forms of the GAI
of anxiety were found across all five studies. In contrast, seemed quite comparable. According to the COSMIN
negative anxiety, which refers to the motives and checklist, we largely rated criterion validity as “fair” for
behaviours related to anxiety disorders, was found only different reasons (e.g., no information on how missing
by Guan (2016). Most of the GAI items were not con- items were handled; unclear if the criterion was a “gold
sistently associated with the same symptom category. standard”).
 12
Table 3: Convergent and divergent validity of the GAI and GAI-SF

Canadian Journal on Aging

r

Convergent Validity GAI GAI-SF Study

Anxiety Measure
ASI = Anxiety Inventory Status .85 Rozzini et al. (2009)
BAI = Beck Anxiety Inventory .28–.75 .58 Diefenbach et al. (2009); Gould et al. (2014); Massena et al. (2015); Pachana et al.
(2007); Silva et al. (2016); Yan et al. (2014)
GADS = Goldberg Anxiety and Depression Scale – Anxiety scale .57 Pachana et al. (2007)
GAI = Geriatric Anxiety Inventory .77–.94 Byrne and Pachana (2011); Champagne et al. (2016); Gerolimatos et al. (2013);
Heissler et al. (2018); Johnco et al. (2015); Silva et al. (2016)
GAI-SF = Geriatric Anxiety Inventory – Short Form .77–.94 Champagne et al. (2016); Gerolimatos et al. (2013); Heissler et al. (2018); Johnco et
al. (2015); Silva et al. (2016)
GAS = Geriatric Anxiety Scale .60–.82 Cheung (2007); Pachana et al. (2007) Gould et al. (2014)
HADS-A = Hospital Anxiety and Depression Scale – Anxiety Scale .51–.71 .61 Ball et al. (2015); Creighton et al. (2018); Dow et al. (2018); Ferrari et al. (2017);
Kneebone et al. (2016)
HAMA = Hamilton Anxiety Scale .25–.47 Ball et al. (2015); Gould et al. (2014)
RAID = Rating Anxiety in Dementia Scale .61 Creighton et al. (2018)
SAS = Self-Rating Anxiety Scale .52 Yan et al. (2014)
SRQ-20 = Self-Reporting Questionnaire .74 .55 Silva et al. (2016)
STAI = State-Trait Anxiety Inventory .61–.69 Cheung (2007); Massena et al. (2015); Matheson et al. (2012); Ribeiro et al. (2011)
STAI-S = State-Trait Anxiety Inventory – subscale state .28–.80 .48–.50 Byrne and Pachana (2011); Byrne et al. (2010); Ferrari et al. (2017); Massena et al.
(2015); Pachana et al. (2007)
STAI-T = State-Trait Anxiety Inventory – subscale trait 55 .53 Ferrari et al. (2017); Massena et al. (2015)
Anxiety/Depression Measure
GHQ = General Health Questionnaire .76 Ribeiro et al. (2011)
Generalized Anxiety Disorder Measure
GAD-7 = Generalized Anxiety Inventory-7 .86 .79 Champagne et al. (2016)
GADQ-IV: Generalized Anxiety Disorder Questionnaire for .65 Diefenbach et al. (2009)
DSM-IV
GADSS = Generalized Anxiety Disorder Severity Scale .84 Diefenbach et al. (2009)
Intolerance to Uncertainty
IUI = Intolerance of Uncertainty Inventory .62 .58 Champagne et al. (2016)
Neuroticism Measure
NEO-N = NEO Five-Factor Inventory- neuroticism .63 Byrne et al. (2010)
Posttraumatic Stress Disorder Measure
PCL-C = Posttraumatic stress disorder checklist-civilian version .56 Gould et al. (2014)

Alexandra Champagne et al.

Worry Measure
BMWS = Brief Measure of Worry Severity .78 Diefenbach et al. (2009)
GWS = Geriatric Worry Scale .86 Diefenbach et al. (2009)
PSWQ = Penn State Worry Questionnaire .70–.79 Pachana et al. (2007) Diefenbach et al. (2009); Gould et al. (2014)
PSWQ-A = Penn State Worry Questionnaire – Abbreviated .60–.79 .56–.79 Champagne et al. (2016); Diefenbach et al. (2009); Johnco et al. (2015)
WSOA-R = Worry Scale for Older Adults Revised .53 .53 Champagne et al. (2016)

Continued
 Properties of the Geriatric Anxiety Inventory La Revue canadienne du vieillissement 13

Sensitivity to Change and Responsiveness

Byrne and Pachana (2011); Champagne et al. (2016); Diefenbach et al. (2009); Gerolimatos et al. (2013);
To our knowledge, only Ball et al. (2015) explicitly
assessed sensitivity to treatment of the GAI in a clinical
controlled trial. They concluded that the GAI is a useful
tool for monitoring the outcome of treatment. Accord-
ing to the COSMIN checklist, responsiveness was rated
as “poor” because no analyses were conducted between
the score on the GAI and the gold standard to demon-
strate the good performance of the former. Although it
wasn’t their primary aim, there are studies that support
the sensitivity to change of the GAI in the treatment
Diefenbach et al. (2014); Johnco et al. (2015); Ribeiro et al. (2011)

monitoring of anxiety or specific phobia (Pachana,

Woodward, & Byrne, 2007; Welch et al., 2010).

Problematic Issues
Cross-Cultural Adaptation. Simple translation of a ques-
tionnaire is insufficient if it is to be used with a popu-
Ball et al. (2015); Kneebone et al. (2016)

lation from another country, culture, or language. In

such cases, cross-cultural adaptation of the instrument
Marquez-Gonzalez et al. (2012)

is recommended (Gjersing, Caplehorn, & Clausen,

2010). Studies conducted with other language versions
Massena et al. (2015)

of the GAI generally report information on the transla-

Pachana et al. (2007)

Pachana et al. (2007)

tion process but most do not report having culturally

Gould et al. (2014)

adapted the instrument.

Potentially Problematic Items. Some authors have pointed
out that item 12 – “I get an upset stomach due to my
Study

worrying” – is problematic and not precise enough

(Champagne et al., 2016; Molde et al., 2017; Yan et al.,
2014). This item may present a limitation due to its
GAI-SF

.37–.63
.37-.47

cultural validity. Alternatively, it may be difficult for

older adults to consider an upset stomach as a conse-
quence of worrying. It is also possible that this symptom
r

is endorsed only by some individuals because it refers to

.31–.86
.38–.79
GAI

–.34
.49
.56

.28

.58

a more severe level of anxiety.

Other authors have questioned whether certain items
truly assess the construct of interest. For example, item
CES-D = Center for Epidemiologic Studies – Depression scale

PANAS = Positive and Negative Affect Schedule – Negative

2 – “I find it difficult to make a decision” – was reported

PANAS = Positive and Negative Affect Schedule – Positive

as being insufficiently precise and having low corrected

item-total correlations (Gould et al., 2014; Yan et al.,
HADS-D = Hospital Anxiety and Depression Scale –
GDS-A = Geriatric Depression Scale – Abbreviated

2014). Researchers have proposed that this item may

instead assess decision-making abilities that aren’t neces-
sarily related to worry. For the same reasons, authors
have questioned whether item 18 “I sometimes feel a
BDI-II = Beck Depression Inventory-II

great knot in my stomach” assesses anxiety or a symp-

GDS = Geriatric Depression Scale

tom of a general medical condition (Gould et al., 2014;

Heissler, Kopecek, & Stepankova Georgi, 2018; Yan et al.,
Negative Affect Measure

2014). Gould et al. (2014) have also indicated that item 7 –

Positive Affect Measure
Table 3: Continued

Depression Measure

“I often feel like I have butterflies in my stomach” – and

Divergent Validity

Depression Scale

14 – “I always anticipate the worst will happen” – may be

endorsed for reasons unrelated to anxiety.
Results from Rasch models in Molde et al. (2017) iden-
tified substantial item overlap between item-pairs 10 (“I
often feel nervous”) and 15 (“I often feel shaky inside”),
14 Canadian Journal on Aging Alexandra Champagne et al.

Table 4: Factors associated with each item of the GAI

Bakkane Bendixen Diefenbach Marquez-Gonzalez Mababu and

Item et al. (2016) et al. (2014) et al. (2012) RuizSánchez (2016) Guan (2016)

1 Worries Excessive worry symptoms Cognitive symptoms Cognitive symptoms Mental anxiety
2 Worries Decision-making symptoms Cognitive symptoms Cognitive symptoms Mental anxiety
3 Physical symptoms NA Cognitive symptoms Cognitive symptoms Mental anxiety
4 Worries CNS hyperarousal symptoms Arousal symptoms Arousal symptoms Mental anxiety
5 Worries CNS hyperarousal symptoms Cognitive symptoms Cognitive symptoms Mental anxiety
6 Worries Excessive worry symptoms Arousal symptoms Arousal symptoms Mental anxiety
7 Physical symptoms Gastrointestinal symptoms Somatic symptoms Somatic symptoms Mental anxiety
8 Worries Excessive worry symptoms Cognitive symptoms Cognitive symptoms Physical anxiety
9 Worries Gastrointestinal symptoms Cognitive symptoms Cognitive symptoms Mental anxiety
10 Physical symptoms CNS hyperarousal symptoms Arousal symptoms Arousal symptoms Mental anxiety
11 Worries CNS hyperarousal symptoms Cognitive symptoms Cognitive symptoms Mental anxiety
12 Physical symptoms Gastrointestinal symptoms Somatic symptoms Somatic symptoms Physical anxiety
13 Worries Gastrointestinal symptoms Arousal symptoms Arousal symptoms Mental anxiety
14 Worries Excessive worry symptoms Cognitive symptoms Cognitive symptoms Negative anxiety
15 Physical symptoms Gastrointestinal symptoms Somatic symptoms Somatic symptoms Negative anxiety
16 Worries Excessive worry symptoms Cognitive symptoms Cognitive symptoms Negative anxiety
17 Worries Excessive worry symptoms Cognitive symptoms Cognitive symptoms Negative anxiety
18 Physical symptoms Gastrointestinal symptoms Somatic symptoms Somatic symptoms Physical anxiety
19 Worries Excessive worry symptoms Cognitive symptoms Cognitive symptoms Negative anxiety
20 Physical symptoms CNS hyperarousal symptoms Arousal symptoms Arousal symptoms Mental anxiety

Note. CNS = central nervous system; NA = not available.

7 (“I often feel like I have butterflies in my stomach”) Dissanayaka et al., 2015; Edelstein et al., 2008; Lin
and 18 (“I sometimes feel a great knot in my stomach”), et al., 2016; Pachana & Byrne, 2012; Therrien & Hunsley,
10 (“I often feel nervous”) and 13 (“I think of myself as a 2012). However, these reviews were mostly dedicated
nervous person”), 6 (“Little things bother me a lot”) and to the GAI long form and examined in specific popula-
9 (“I can’t help worrying about even trivial things”), tions or settings (e.g., Parkinson’s disease; residential
1 (“I worry a lot of the time”) and 8 (“I think of myself as aged care facilities). Moreover, these reviews examined
a worrier”), and 16 (“I think that my worries interfere only a small sample of studies. The study that reviewed
with my life”) and 17 (“My worry often overwhelms the largest number of articles on the GAI included only
me”). To a certain extent, the same phenomenon of 18 (Balsamo et al., 2018), a number that is almost half of
redundancy was observed for items 8, 10, and 11 of what our own search retrieved.
the GAI short version. Thus, the detected item overlap
The goals of our study were to summarize existing
suggests that there may be redundant items in the GAI
evidence on the psychometric properties of the GAI
and the GAI-SF that do not provide any additional
and GAI-SF, to assess the methodological quality of the
information because of their similar content (Molde
studies and to provide guidance for future psychometric
et al., 2017).
validation studies. For the current review, we identified
Floor Effects. Some authors have made assumptions 31 studies that purposely studied the psychometric
about the possible presence of floor effects in the GAI properties of these scales. As in reviews by other
and GAI-SF. Yan et al. (2014) and Johnco et al. (2014) researchers (Balsamo et al., 2018; Creighton et al., 2018;
hypothesized the presence of floor effects when they Dissanayaka et al., 2015; Edelstein et al., 2008; Lin et al.,
observed that the GAI may be less suitable for elderly 2016; Pachana & Byrne, 2012; Therrien & Hunsley, 2012),
people with low-level anxiety as this would mean that we generally found appropriate psychometric properties
they would not endorse several items suggesting high- for the GAI and GAI-SF among various clinical and non-
level anxiety and serious outcomes. clinical populations of older adults. However, we also
mostly found low levels of methodological quality in the
studies retained for this review.
Discussion To our knowledge, this is the first systematic review to
Since their development, the GAI and, to a lesser extent, examine the methodological quality of research con-
the GAI-SF, have undergone extensive psychometric ducted on the psychometric properties of the GAI and
testing in a wide range of populations and countries. GAI-SF. The majority of the COSMIN boxes were rated
These tools have been the subject of different reviews as “poor” or “fair” (83.5% of all boxes) for the different
(Balsamo et al., 2018; Creighton et al., 2018; psychometric properties evaluated. The results of
 Properties of the Geriatric Anxiety Inventory
Table 5: Criterion validity and cut-off point of the GAI and GAI-SF

Study Cut-off Sn Sp PPV NPV AUC % correct Criterion

Brazilian Portuguese Version (GAI-BR)

Massena et al. (2015) 13 to detect GAD 83.3 84.6 — — 90 — MINI
English version
Diefenbach et al. (2009) 9 to detect GAD or an anxiety 87.5 95.5 87.5 95.5 89.5 — ADIS-IV
disorder NOS
Byrne et al. (2010) 9 to detect GAD — — — — 93 93 MINI-V
Cheung et al. (2012) 3 to detect an anxiety disorder 85.7 78 57.1 94.1 83 — MINI
Matheson et al. (2012) 7 to detect an anxiety disorder 87.5 85.7 — — 91 — MINI-Plus
Bradford et al. (2013) 8 to detect an anxiety disorder 58 93 94 56 69.1 — MINI
10 to detect an anxiety disorder 62 93 94 57 80.9 —
(when rated by collaterals)
Gerolimatos et al. (2013) 9 to detect an anxiety disorder 100 60 — — 79 65.3 Psychiatric diagnosis based on an
assessment that is part of routine care.
Johnco, Knight, Tadic, and 9 to detect an anxiety disorder 69.5 100 — — 98.1 — ADIS-IV
Wuthrich (2014)
Kneebone et al. (2016) 7 to detect anxiety 88 84 — — 84 — SCID-I-RV
Creighton et al. (2018) 9 to detect GAD 90 86.3 45 98.6 93 86.7 MINI
English version – short form (GAI-SF)
Byrne and Pachana (2011) 3 to detect GAD 75 86.8 — — 80 86 MINI-V
Gerolimatos et al. (2013) 2 to detect an anxiety disorder 100 46.2 — — 78 53.3 Psychiatric diagnosis based on an
assessment that is part of routine care.
Johnco et al. (2014) 3 to detect an anxiety disorder 78.1 98.3 — — 95.4 — ADIS-IV

Italian version (GAI-It)

Ferrari et al. (2017) Unspecified 30 57 31 55 — — DSM-IV diagnosis after clinical
psychiatric evaluation
74 43 36 79 — — ICD-10 diagnosis after clinical
psychiatric evaluation
76 88 95 54 — — STAI-Trait
— —

La Revue canadienne du vieillissement

74 81 93 46 STAI-State
55 71 60 67 — — HADS-Anxiety
Italian version- short form (GAI-It-SF)
Ferrari et al. (2017) Unspecified 70 35 45 69 — — DSM-IV diagnosis after clinical
psychiatric evaluation
75 44 36 81 — — ICD-10 diagnosis after clinical
psychiatric evaluation
74 76 89 52 — — STAI-Trait
72 69 88 44 — — STAI-State
82 71 69 83 — — HADS-Anxiety
Norwegian version
Molde et al. (2017) Unspecified 77 79 — — 86 78
Norwegian version- short form
Molde et al. (2017) Unspecified 79 68 — — 82 74

Continued

15
 16 Canadian Journal on Aging Alexandra Champagne et al.

Note. ADIS-IV = Anxiety Disorder Interview Schedule for DSM-IV; AUC = area under the curve; GAD = generalized anxiety disorder; ICD-10 = International Classification of
Diseases, 10th Edition; MINI = Mini International Neuropsychiatric Interview; NOS = not otherwise specified; NPV = negative predictive value; PPV = positive predictive value;
studies with low methodological quality were not
ignored in this review. A “poor” or “fair” methodo-

psychiatrist and/or by a clinical

Clinical diagnosis established by a
logical quality score indicates a certain risk of bias in
the results but does not necessarily mean that findings
are in fact biased or invalid. For example, the most
frequent reasons for these ratings were low sample size
and a lack of information on the number of missing items
psychologist

and how they were handled. Even though information

on missing items was lacking, it still could have been
Criterion

correctly managed in an effort to avoid introducing bias

into the study results. It is also important to note that
results of studies that presented a “poor” versus “good”
or “excellent” level of methodological quality according
% correct

to the COSMIN checklist generally presented compar-

—

able results in this review. Nonetheless, our main

finding – that the GAI and GAI-SF generally show
adequate psychometric properties based on methodo-
logically weak studies – is important.
AUC

91.6

92.7
—

We recommend the use of these tools and emphasize

SCID-I-RV = Structured Clinical Interview for DSM-IV Axis 1 Disorders: Research; Sn = sensitivity; Sp = specificity.

the need for better designed research on the validity and

reliability of the GAI and GAI-SF. Researchers should
NPV

always assess the psychometric properties of the instru-

ments they use. Psychometric properties of an instru-
ment are specific to the population and purpose for
which it was intended and thus, they are not the prop-
PPV

erties of the instrument per se (Hunsley & Mash, 2008).

—

Not all psychometric properties received the same level

of attention, which in turn influences the confidence
that can be placed in the results (e.g., internal consist-
80.4

80.4

85.9
Sp

ency was evaluated in 30 studies vs. two for content

validity). Internal consistency was assessed in the
majority of the studies and coefficients mostly fell above
the acceptable threshold recommended by Hunsley and
84.4 to
95.7
88.8

Mash (2008) for evidence-based assessment. Since the

Sn

90

alpha values are influenced by the number of items of a

scale, it is not surprising that lower coefficients were
11 to identify patients with psy-

found for the GAI-SF (Streiner, 2003). Retest reliability

group without psychological
9 to detect clinically significant

psychiatric disorder (clinical

9 to identify patients with any

samples) versus community

community group without

was assessed in only 11 studies. Apart from the two

chological distress versus

psychological distress

lowest coefficients found for the long form (r = .53 and

.58; Kneebone et al., 2016; Silva et al., 2016), the lowest
coefficient was .85. The two studies that presented the
lowest coefficients also exhibited poor methodological
anxiety

distress

quality for the reliability assessment according to the

Cut-off

COSMIN checklist. However, other studies also pre-

sented poor quality but higher retest coefficients. Most
studies presented a retest coefficient that was con-
sidered as acceptable according to the recommenda-
Portuguese version (GAI-PT)

tions of Hunsley and Mash (2008).

Table 5: Continued

Ribeiro et al. (2011)

Results concerning test-retest reliability as well as con-

vergent and criterion validity suggest that the GAI and
GAI-SF both assess rather stable components of anxiety.
The good convergent and criterion validity found for
measures of GAD symptoms suggest a capacity to
Study

evaluate trait anxiety (stable tendency to experience

Properties of the Geriatric Anxiety Inventory La Revue canadienne du vieillissement 17

anxiety), and a low correlation found with a measure of specific cut-off score for the detection of an anxiety
state anxiety supports this hypothesis. The higher retest disorder can be established because of significant
reliability of the short form versus the long form found variability in the results.
in Silva et al. (2016) also leads us to question whether the
Some psychometric properties of the GAI were some-
items of the short version refer to more stable anxiety
times found to be slightly better than those of the GAI-
symptoms.
SF, but most authors concluded that the results were
Inconsistency in divergent, factorial, and criterion val- nevertheless comparable. Most studies that assessed the
idity was found across studies. Whereas some authors psychometric properties of the short form had extracted
have concluded that moderate to high correlations with data from the GAI long form. Although we can only
a measure of depressive symptoms are evidence of poor speculate about the consequences of this procedure, it is
divergent validity, others have argued that it has not possible that the psychometric properties of the GAI-SF
been well established that anxiety and depression are differ when administered independently because of
completely independent disorders in the elderly popu- context, primacy and recency, and warm-up effects.
lation considering the overlap of symptoms (Cassidy, When validating a brief scale, it should be considered
Lauderdale, & Sheikh, 2005). The associations found are as a completely new measure and thus submitted to
not specific to the GAI and GAI-SF, but are rather independent validation procedures (Smith et al., 2000).
characteristic of other measures commonly used to Until further data from an independent assessment of
assess anxiety in the elderly population (Therrien & the short version become available, the findings of this
Hunsley, 2012). review suggest that psychometric properties are not a
major issue when choosing between the GAI and the
Further, conflicting results were found for the factorial
GAI-SF. The choice of one instrument or the other
structure of the GAI concerning the unidimensionality
depend on the user’s needs. For example, the short
and multidimensionality of the scale. However, the fact
form may be the best option in specific situations such
that three of the four studies that presented the highest
as when time is limited, when there is a demanding
methodological quality (either “good” or “excellent”)
clinical context (e.g., acute geriatric settings), with eld-
for structural validity concluded to the unidimension- erly people who are easily fatigued or distracted, when
ality of the scale leads us to support this result as well. multiple questionnaires are administered to patients, or
Mababu and RuizSánchez (2016) suggested that the when patients are frequently monitored.
different factorial structures found in previous studies
for the GAI could be due in part to the dichotomous
response format. Among the possible impacts of a
dichotomous scale are a decrease in the percentage of Suggestions for Future Research
explained variance and lower loadings (Lozano, Future efforts to validate the GAI and GAI-SF should
García-Cueto, & Muñiz, 2008; Velicer, DiClemente, & include paying particular attention to the previously
Corriveau, 1984). Molde et al. (2017) also proposed identified problems and aiming to achieve a higher
different explanations for the lack of factorial consist- degree of methodological quality. Since content validity
ency: different cultural response styles, differences in is considered to be the most important psychometric
semantics due to translation processes, different sam- property according to COSMIN and that it was hardly
ple characteristics, and true cultural differences in the tested in previous research, more studies should
structure of anxiety across countries. address this situation. Researchers should not assume
that the culture and scales’ content are equivalent. Also,
The unidimensionality of the GAI raises the question as
the ability of the GAI and GAI-SF to distinguish
to whether it reflects all manifestations of anxiety in a
between anxiety and depression symptoms is limited.
context where the GAI was designed to assess a range
Therefore, it would be interesting to further examine
of anxiety presentations (Pachana, Byrne, et al., 2007).
this issue by going beyond standard correlational ana-
There is currently a consensus on the unidimensional-
lyses (e.g., by using the heterotrait-monotrait ratio of the
ity of the GAI-SF, which is not surprising for a 5-item
correlation method, by comparing the answer profiles
scale. An obvious issue when designing the short form
of depressed and non-depressed elderly individuals, or
of an instrument is to ensure that the target content
by identifying specific items that spark confusion as to
domain is still adequately represented despite the
the true nature of symptoms [i.e., related to depression
reduced number of items (Smith, McCarthy, & Ander-
or anxiety]).
son, 2000). This does not seem to be the case with the
GAI-SF as it is composed largely of items that relate to The appropriateness of the GAI and GAI-SF for moni-
cognitive symptoms. Evidence on criterion validity toring treatment change also requires further attention.
shows that the GAI and the GAI-SF can screen for Considering that the GAI and GAI-SF were developed
probable cases of anxiety disorders. However, no to assess a range of anxiety disorders rather than a
18 Canadian Journal on Aging Alexandra Champagne et al.

specific disorder, it would also be interesting to define International Psychogeriatrics, 27(9), 1533–1539. https://
cut-off scores to identify different severity levels of anx- doi.org/10.1017/S1041610215000381
iety symptoms (i.e., no symptoms; mild, moderate, and Balsamo, M., Cataldi, F., Carlucci, L., & Fairfield, B. (2018).
severe anxiety). The usefulness of the GAI as a screening Assessment of anxiety in older adults: a review of self-
tool for various anxiety disorders should also be docu- report measures. Clinical Interventions in Aging, 13,
mented. Some previously described results (e.g., high 573–593. https://doi.org/10.2147/CIA.S114100
alpha values [Panayides, 2013], possible floor effects, the
Boddice, G., & Byrne, G. J. (2008). The clinical utility of the
overrepresentation of cognitive symptoms in factorial
geriatric anxiety inventory in older adults with cognitive
structures, problematic items due to item overlap) lead
impairment. Nursing Older People, 20(8), 36–39. doi:10.
us to think that the GAI and GAI-SF may not comparably 7748/nop2008.10.20.8.36.c6809
cover the various areas of anxiety. Thus, a review of the
performance of each item is required to identify those Bradford, A., Brenes, G. A., Robinson, R. A., Wilson, N., Snow,
that lack precision and those that adequately measure A. L., Kunik, M. E., … Amspoker, A. B. (2013).
anxiety. A reformulation of items may also be necessary Concordance of self- and proxy-rated worry and anxiety
symptoms in older adults with dementia. Journal of Anxiety
in order to improve the scale’s validity.
Disorders, 27(1), 125–130. doi:10.1016/j.janxdis.2012.11.001

Strengths and Limitations Byrne, G. J., Pachana, N. A., Goncalves, D. C., Arnold, E., King,
R., & Khoo, S. K. (2010). Psychometric properties and health
Our findings are subject to several limitations. Despite correlates of the Geriatric Anxiety Inventory in Australian
great effort having been made to identify all relevant community-residing older women. Aging Mental Health, 14
studies for this review, it is still possible that some were (3), 247–254. doi:10.1080/13607861003587628
not included. Exclusion of non-English and non-French
Byrne, G. J., & Pachana, N. A. (2011). Development and
papers may have introduced a selection bias. Although a
validation of a short form of the Geriatric Anxiety
notable strength is that the literature search and the Inventory - the GAI-SF. International Psychogeriatrics, 23
quality assessments were conducted by individuals (1), 125–131. https://doi.org/10.1017/S1041610210001237
working independently by using the COSMIN checklist,
data extraction was completed by the first author only. Cassidy, E. L., Lauderdale, S., & Sheikh, J. I. (2005). Mixed
Some notable strengths of this review are that the meth- anxiety and depression in older adults: Clinical
characteristics and management. Journal of Geriatric
odological quality of the research was assessed and that
Psychiatry and Neurology, 18(2), 83–88. https://doi.org/
the coverage of the review is the broadest to date.
10.1177/0891988705276060
Champagne, A., Landreville, P., Gosselin, P., & Carmichael,
Conclusion P.H. (2016). Psychometric properties of the French
To our knowledge, this was the first systematic review to Canadian version of the Geriatric Anxiety Inventory.
be conducted on the psychometric properties of the GAI Aging & Mental Health, 22(1), 40–45. doi: https://
and GAI-SF. The data provided by this review support doi.org/10.1080/13607863.2016.1226767
the recommendation of these tools as screening meas- Cheung, G., Patrick, C., Sullivan, G., Cooray, M., & Chang, C. L.
ures for anxiety in older adults, especially in countries (2012). Sensitivity and specificity of the Geriatric Anxiety
and in versions wherein psychometric properties were Inventory and the Hospital Anxiety and Depression Scale
shown to be adequate. However, this review also high- in the detection of anxiety disorders in older people with
lights weaknesses in the methodological quality of chronic obstructive pulmonary disease. International
research in this area. Researchers are encouraged to Psychogeriatrics, 24(1), 128–136. https://doi.org/10.1017/
continue to assess the psychometric properties of the S1041610211001426
GAI and GAI-SF and apply standards such as the COS- Cheung, G. (2007). Concurrent validity of the Geriatric
MIN checklist for study design and result reporting. Anxiety Inventory in late-life depression. International
Psychogeriatrics, 19(2), 333–335. doi:10.1017/
S1041610206004340
References
Bakkane Bendixen, A., Hartberg, C. B., Selbæk, G., & Engedal, Creighton, A. S., Davison, T. E., & Kissane, D. W. (2018). The
K. (2016). Symptoms of anxiety in older adults with assessment of anxiety in aged care residents: A systematic
depression, dementia, or psychosis: A principal review of the psychometric properties of commonly used
component analysis of the geriatric anxiety inventory. measures. International Psychogeriatrics, 30(7), 967–979.
Dementia and Geriatric Cognitive Disorders, 42(5–6),
Diefenbach, G. J., Bragdon, L. B., & Blank, K. (2014). Geriatric
310–322.
anxiety inventory: Factor structure and associations with
Ball, S. G., Lipsius, S., & Escobar, R. (2015). Validation of the cognitive status. The American Journal of Geriatric
geriatric anxiety inventory in a duloxetine clinical trial for Psychiatry, 22(12), 1418–1426. https://doi.org/10.1016/
elderly adults with generalized anxiety disorder. j.jagp.2013.04.014
Properties of the Geriatric Anxiety Inventory La Revue canadienne du vieillissement 19

Diefenbach, G. J., Tolin, D. F., Meunier, S. A., & Gilliam, C. M. Hunsley, J., & Mash, E. J. (2008). A guide to assessments that
(2009). Assessment of anxiety in older home care recipients. work. NY, United States: Oxford University Press.
Gerontologist, 49(2), 141–153. doi:10.1093/geront/gnp019
Johnco, C., Knight, A., Tadic, D., & Wuthrich, V. (2014).
Dissanayaka, N. N. W., Torbey, E., & Pachana, N. A. (2015). Psychometric properties of the Geriatric Anxiety
Anxiety rating scales in Parkinson’s disease: A critical Inventory (GAI) and its short-form (GAI-SF) in a clinical
review updating recent literature. International and non-clinical sample of older adults. International
Psychogeriatrics, 27(11), 1777–1784. https://doi.org/ Psychogeriatrics, 11, 1–9. https://doi.org/10.1017/
10.1017/S1041610215000885 S1041610214001586

Dow, B., Lin, X., Pachana, N. A., Bryant, C., LoGiudice, D., Kneebone, I. I., Fife-Schaw, C., Lincoln, N. B., & Harder, H.
Goh, A. M., & Haralambous, B. (2018). Reliability, (2016). A study of the validity and the reliability of the
concurrent validity, and cultural adaptation of the Geriatric Anxiety Inventory in screening for anxiety after
Geriatric Depression Scale and the Geriatric Anxiety stroke in older inpatients. Clinical Rehabilitation, 30(12),
Inventory for detecting depression and anxiety 1220–1228.
symptoms among older Chinese immigrants: an
Australian study. International Psychogeriatrics, 30(5), Lin, X., Haralambous, B., Pachana, N. A., Bryant, C.,
735–748. doi:10.1017/S1041610217002332 LoGiudice, D., Goh, A., & Dow, B. (2016). Screening for
depression and anxiety among older Chinese immigrants
Edelstein, B. A., Woodhead, E. L., Segal, D. L., Heisel, M. J., living in Western countries: The use of the Geriatric
Bower, E. H., Lowery, A. J., & Stoner, S. A. (2008). Older Depression Scale (GDS) and the Geriatric Anxiety
adult psychological assessment: Current instrument Inventory (GAI). Asia‐Pacific Psychiatry, 8(1), 32–43.
status and related considerations. Clinical Gerontologist:
The Journal of Aging and Mental Health, 31(3), 1–35. Lozano, L. M., García-Cueto, E., & Muñiz, J. (2008). Effect of
https://doi.org/10.1080/07317110802072108 the number of response categories on the reliability and
validity of rating scales. Methodology, 4(2), 73–79.
Ferrari, S., Signorelli, M., Cerrato, F., Pingani, L., Massimino,
M., Valente, S., … De Ronchi, D. (2017). Never too late to Mababu, R., & RuizSánchez, G. (2016). Factorial validity of
be anxious: validation of the Geriatric Anxiety Inventory, the Spanish version of the Geriatric Anxiety Inventory
Italian version. La Clinica terapeutica, 168(2), e120. doi: 10. (GAI): Empirical analysis of its structure and
7417/CT.2017.1992 dimensions. Revista de Psicopatología y Psicología Clínica,
21(3), 201–208.
Gerolimatos, L. A., Gregg, J. J., & Edelstein, B. A. (2013).
Assessment of anxiety in long-term care: Examination Marquez-Gonzalez, M., Losada, A., Fernandez-Fernandez,
of the Geriatric Anxiety Inventory (GAI) and its short V., & Pachana, N. A. (2012). Psychometric properties of
form. International Psychogeriatrics, 25(09), 1533–1542. doi: the Spanish version of the Geriatric Anxiety Inventory.
10.1017/S1041610213000847 International Psychogeriatrics, 24(1), 137–144. https://
doi.org/10.1017/S1041610211001505
Gjersing, L., Caplehorn, J. R., & Clausen, T. (2010). Cross-
cultural adaptation of research instruments: Language, Massena, P. N., de Araújo, N. B., Pachana, N., Laks, J., & de
setting, time and statistical considerations. BMC Medical Pádua, A. C. (2015). Validation of the Brazilian
Research Methodology, 10(1), 13. Portuguese version of Geriatric Anxiety Inventory—
GAI-BR. International Psychogeriatrics, 27(7), 1113–1119.
Gould, C. E., Beaudreau, S. A., Gullickson, G., Tenover, J. L., https://doi.org/10.1017/S1041610214001021
Bauer, E. A., & Huh, J. W. (2016). Implementation of a
brief anxiety assessment and evaluation in a Department Matheson, S. F., Byrne, G. J., Dissanayaka, N. N., Pachana, N. A.,
of Veterans Affairs geriatric primary care clinic. Journal of Mellick, G. D., O’Sullivan, J. D., … Marsh, R. (2012).
Rehabilitation Research and Development, 53(3), 335–344. Validity and reliability of the Geriatric Anxiety Inventory
doi:10.1682/jrrd.2014.10.0258 in Parkinson’s disease. Australasian Journal on Ageing, 31(1),
13–16. https://doi.org/10.1111/j.1741-6612.2010.00487.x
Gould, C. E., Segal, D. L., Yochim, B. P., Pachana, N. A., Byrne,
G. J., & Beaudreau, S. A. (2014). Measuring anxiety in late Moher, D., Liberati, A., Tetzlaff, J., & Altman, D. G. (2009).
life: A psychometric examination of the Geriatric Anxiety Preferred reporting items for systematic reviews and
Inventory and Geriatric Anxiety Scale. Journal of Anxiety meta-analyses: The PRISMA statement. Annals of
Disorders, 28(8), 804–811. http://dx.doi.org/10.1016/ Internal Medicine, 151(4), 264–269.
j.janxdis.2014.08.001
Moher, D., Shamseer, L., Clarke, M., Ghersi, D., Liberati, A.,
Guan, M. (2016). Factor structure of the Chinese version of the Petticrew, M., … PRISMA-P Group. (2015). Preferred
geriatric anxiety inventory. Annals of General Psychiatry, reporting items for systematic review and meta-analysis
15, 4. https://doi.org/10.1186/s12991-016-0092-4 protocols (PRISMA-P) 2015 statement. Systematic
Reviews, 4(1), 1.
Heissler, R., Kopecek, M., & Stepankova Georgi, H. (2018).
Geriatric Anxiety Inventory (GAI) and its short form Mokkink, L. B., Terwee, C. B., Knol, D. L., Stratford, P. W.,
GAI-SF: Czech normative study. Ceskoslovenska Alonso, J., Patrick, D. L., … de Vet, H. C. W. (2010). The
Psychologie, 62(5), 462–476. COSMIN checklist for evaluating the methodological
20 Canadian Journal on Aging Alexandra Champagne et al.

quality of studies on measurement properties: A in mild cognitive impairment. International Journal of

clarification of its content. BMC Medical Research Geriatric Psychiatry, 24(3), 300–305. doi:10.1002/gps.2106
Methodology, 10(1), 22.
Silva, L. d. S. V. e., Oliveira, G. M. d., Yokomizo, J. E., Saran, L.
Mokkink, L. B., Terwee, C. B., Patrick, D. L., Alonso, J., F., Bottino, C. M. d. C., & Yassuda, M. S. (2016). The
Stratford, P. W., Knol, D. L., … de Vet, H. C. W. (2010). Geriatric Anxiety Inventory in primary care:
The COSMIN checklist for assessing the methodological Applicability and psychometric characteristics of the
quality of studies on measurement properties of health original and short form. Archives of Clinical Psychiatry
status measurement instruments: An international (São Paulo), 43, 103–106.
Delphi study. Quality of Life Research, 19(4), 539–549.
Smith, G. T., McCarthy, D. M., & Anderson, K. G. (2000). On
Molde, H., Hynninen, K., Torsheim, T., Bendixen, A., Engedal, the sins of short-form development. Psychological
K., Pachana, N., & Nordhus, I. (2017). A bifactor and item Assessment, 12(1), 102–111. https://doi.org/10.1037/
response analysis of the geriatric anxiety inventory. 1040-3590.12.1.102
International Psychogeriatrics, 29(10), 1647–1656.
Streiner, D. L. (2003). Starting at the beginning: An introduction
Pachana, N. A., & Byrne, G. J. (2012). The Geriatric Anxiety to coefficient alpha and internal consistency. Journal of
Inventory: International use and future directions. Personality Assessment, 80(1), 99–103.
Australian Psychologist, 47(1), 33–38. https://doi.org/
10.1111/j.1742-9544.2011.00052.x Terwee, C. B., Mokkink, L. B., Knol, D. L., Ostelo, R. W.,
Bouter, L. M., & de Vet, H. C. (2012). Rating the
Pachana, N. A., Byrne, G. J., Siddle, H., Koloski, N., Harley, methodological quality in systematic reviews of studies
E., & Arnold, E. (2007). Development and validation of the on measurement properties: A scoring system for the
Geriatric Anxiety Inventory. International Psychogeriatrics, COSMIN checklist. Quality of Life Research, 21(4), 651–657.
19(1), 103–114. doi:10.1017/S1041610206003504
Therrien, Z., & Hunsley, J. (2012). Assessment of anxiety in
Pachana, N. A., Woodward, R. M., & Byrne, G. J. A. (2007). older adults: A systematic review of commonly used
Treatment of specific phobia in older adults. Clinical measures. Aging & Mental Health, 16(1), 1–16. https://
Interventions in Aging, 2(3), 469–476. doi.org/10.1080/13607863.2011.602960

Panayides, P. (2013). Coefficient alpha: Interpret with caution. Vaz, S., Falkmer, T., Passmore, A. E., Parsons, R., & Andreou,
Europe’s Journal of Psychology, 9(4), 687–696. P. (2013). The case for using the repeatability coefficient
when calculating test–retest reliability. PloS One, 8(9),
Park, M. S., Kang, K. J., Jang, S. J., Lee, J. Y., & Chang, S. J. (2018). e73990. https://doi.org/10.1371/journal.pone.0073990
Evaluating test-retest reliability in patient-reported
outcome measures for older people: A systematic Velicer, W. F., DiClemente, C. C., & Corriveau, D. P. (1984).
review. International Journal of Nursing Studies, 79, 58–69. Item format and the structure of the Personal Orientation
Inventory. Applied Psychological Measurement, 8(4),
Polit, D. F. (2014). Getting serious about test–retest reliability: 409–419.
A critique of retest research and some recommendations.
Quality of Life Research, 23(6), 1713–1720. Welch, T., Welch, M., Baer, J., Dias, J., Gurney, C., Van Dale, B.,
… Psiurski, S. (2010). Removed but not out of reach:
Ribeiro, O., Paul, C., Simoes, M. R., & Firmino, H. (2011). Seniors with depression in smaller center, rural, and
Portuguese version of the Geriatric Anxiety Inventory: remote communities. Psychiatric Annals, 40(12), 616–623.
Transcultural adaptation and psychometric validation.
Aging & Mental Health, 15(6), 742–748. https://doi.org/ Yan, Y., Xin, T., Wang, D., & Tang, D. (2014). Application of
10.1080/13607863.2011.562177 the Geriatric Anxiety Inventory-Chinese version (GAI-
CV) to older people in Beijing communities. International
Rozzini, L., Chilovi, B. V., Peli, M., Conti, M., Rozzini, R., Psychogeriatrics, 26(3), 517–523. https://doi.org/
Trabucchi, M., & Padovani, A. (2009). Anxiety symptoms 10.1017/S1041610213002007