Internal Validity and Reliability of the GAD-7 Test in Latin America

Abstract

Generalized anxiety has significantly increased in the general population during and after the COVID-19 pandemic, highlighting the need for rapid screening tools. In this context, the present study analyzed the psychometric properties and internal consistency of the Generalized Anxiety Disorder Scale (GAD-7) in healthcare workers and the general population in Latin America. A cross-sectional e-health study was conducted, surveying 11,279 Latin Americans online using snowball sampling. The sample included healthcare professionals, hospital populations, community members from various occupations, and university students from six countries (Argentina, Bolivia, Ecuador, Chile, Colombia, and Peru). Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) were performed separately for each country. Additionally, the instrument's internal consistency was evaluated by calculating McDonald's W index and item-total correlations based on the final items. The EFA revealed a unidimensional structure comprising the seven items of the instrument, which explained between 62.8% and 66.1% of the variance (KMO = between .900 and .910; p < .000). The CFA confirmed adequate fit indices for each country. The omega index ranged from 0.85 (Peru: CI = 0.800-0.884) to 0.95 (Argentina-Bolivia: CI = 0.901-0.985), and item-total correlations were high, ranging from .642 to .869, demonstrating the instrument's reliability. In conclusion, the findings of this study indicate that the GAD-7 is a valid and reliable instrument for assessing generalized anxiety symptoms in the Latin American population.

Introduction

Generalized Anxiety Disorder (GAD) manifests itself with excessive anticipatory and persistent worry about various events; it is usually accompanied by restlessness, irritability, muscle tension, nervousness, sleep, and cognitive disturbances.^1,2 Its etiology is unknown, but it seems to be associated with specific genetic and environmental predispositions. ³

The COVID-19 pandemic severely impacted mental health globally, leading to a significant increase in the prevalence of disorders such as Generalized Anxiety Disorder (GAD).^4-6 Various studies have reported high rates of anxiety, reaching 30.5% among healthcare professionals ⁷ and between 31.4% and 36.9% in the general population. ⁸ Although the figures for younger individuals were lower (20.5% to 25.2%), they remained alarming. ⁸ This increase underscored the urgent need for reliable and rapid tools to identify GAD and its association with other psychiatric disorders, given its clinical and social significance.^5,9

A crucial factor in the development and persistence of GAD is chronic stress, defined as an aversive emotional state caused by prolonged exposure to threatening conditions. ¹⁰ This type of stress not only increases the allostatic load, affecting multiple biological systems such as the hypothalamic-pituitary-adrenal axis, but is also linked to psychiatric comorbidities, functional impairment, and a higher risk of suicide. ¹¹ During the pandemic, chronic stress acted as a global catalyst, ¹² exacerbating anxiety levels and highlighting the need to understand its effects better and implement effective strategies for its detection and management.

In this context, the available evidence has highlighted the performance of various instruments designed to quickly and reliably identify Generalized Anxiety Disorder (GAD). Among them, the GAD-7, consisting of seven questions, has been established as an effective tool for briefly and accurately assessing generalized anxiety symptoms over the past two weeks.^13,14

Although previous studies have analyzed the psychometric properties of the GAD-7 in different populations, most of this research has focused on Europe and North America.^15-23 That has left a significant gap in regions like Latin America, where unique cultural, social, and economic factors may influence the manifestation and assessment of GAD. In this regard, we consider this to be the first study to examine the psychometric properties of the GAD-7 in a large and diverse sample from six Latin American countries, representing a significant contribution to the literature on anxiety in this region.

In response to the alarming increase in mental disorders during and after the pandemic, this study focuses on analyzing the psychometric properties of the GAD-7 in a large and representative sample from Latin America. This analysis aims to provide essential evidence on its applicability in diverse cultural contexts, expanding the understanding of GAD in historically underrepresented populations.

Materials and Method

An e-Health Study 52 was conducted, using a snowball sampling methodology, collecting 11 279 online questionnaires from a large Latin American sample between 2022 and 2023. Through a snowball methodology and using social networks and personal and institutional e-mails, the general population over 18 years of age in 6 countries in the region (Argentina, Bolivia, Ecuador, Chile, Colombia, and Peru) were invited to complete a Google form. The average age of the participants was 31.57 years (SD = 19.81); 52.4.6% female and 47.8% male. Of the participants, 22.6% were healthcare personnel, followed by other professions (19.3%), university students (17.6%), the general population (26.1%) and outpatients (14.4%).

In the context of the COVID-19 pandemic, an international research consortium focused on analyzing the mental health of the general population in Latin America and the Caribbean. Inter-institutional agreements were made, a working group was consolidated, and a research team was trained in each member country of the study. Then, institutional mailing databases were collected; subsequently, the online questionnaire was sent by e-mail, social networks, and WhatsApp, and it was requested to be disseminated among the community. Measurements began on May 12, 2022, and ended on November 27, 2023.

Instrument

An automated Google form was used to inform the study's objective, informed consent, 10 demographic questions, and the GAD-7 test 53. This is a brief and quick administration test; it has seven items that favor the early detection of generalized anxiety from the DSM-V criteria 54. Its score ranges from 0 to 3 for each item, with the response options being “Never” “Several days” “More than half of the days” and “Almost every day”. The GAD-7 total score can range from 0 to 21, with a score ≥10 indicating a generalized anxiety disorder; it also allows grading the severity of the disorder.

Stadistical Analyses

Construct validity was assessed through exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). EFA was conducted following the sampling adequacy criteria of KMO and Bartlett's test of sphericity, using the non-orthogonal Oblimin method. CFA was performed using the weighted least squares method adjusted for mean and variance (WLSMV).

For goodness-of-fit evaluation, the following indicators were employed: the root mean square error of approximation (RMSEA ≥ .900), the general fit index (GFI ≥ .900), the standardized root mean square residual (SRMR > .050), the normalized fit index (NFI ≥ .950), and the relative fit index (RFI ≥ .950) (17). Additionally, the normalized parsimonious fit index (PNFI ≥ .500) and the parsimonious goodness-of-fit index (PGFI ≥ .500) were included.

Lastly, the instrument's internal consistency was analyzed using McDonald's W index and item-total correlations, with a minimum required value of .300. All analyses were performed using the R software, version 1.3.1056.

Formal Aspects

The ethical standards established in the Declaration of Helsinki of 1975, revised in 2008, were followed. All participants provided their informed consent digitally. Upon accessing the link to the virtual questionnaire, each participant reviewed the study's objectives and, after providing their consent, completed the form.

Participants did not receive financial compensation for their participation. However, they were provided with a results report accompanied by a guide containing clinical guidelines for addressing anxiety. The online evaluation of the instrument took approximately 2 min.

This international and multicenter research project was developed under the coordination of an interinstitutional consortium led by Universidad de La Costa (Barranquilla, Colombia). The project received ethical approval under code No. 089-2021 and is registered with the research code INV.140-02-004-15.

Results

Table 1 shows the descriptive mean, standard deviation, kurtosis, asymmetry, and confidence intervals based on the mean of the items comprising the GAD-7 scale administered in the 6 Latin American countries. The means ranged from 0.95 (Peru, Item 5) to 1.59 (Bolivia, Item 4). Likewise, the skewness and kurtosis values ranged from −2 to +2, suggesting that the items follow approximately normal distributions.

Table 1.

Descriptives of the GAD-7 Test.

Items	Countries	M a	SD b	Kurtosis	Asymmetry	Confidence Interval
						Lower.L c	Upper L. d
1	Argentina	1,41	0,91	−0,78	0,12	1,34	1,47
	Bolivia	1,40	,088	−0,76	0,09	1,33	1,41
	Chile	1,44	0,92	−0,79	0,17	1,38	1,5
	Colombia	1,25	0,84	−0,54	0,34	1,08	1,17
	Ecuador	1,33	0,87	−0,58	0,25	1,27	1,39
	Perú	1,29	0,92	−0,79	0,23	1,22	1,35
2	Argentina	1,2	0,9	−0,74	0,28	1,14	1,27
	Bolivia	1,15	0,87	−0,15	0,59	1.08	1,20
	Chile	1,18	0,9	−0,6	0,39	1,12	1,24
	Colombia	1,01	0,85	−0426	0,43	0,98	1,13
	Ecuador	1,12	0,86	−0,51	0,39	1,07	1,18
	Perú	0,99	0,88	−0,23	0,67	0,92	1,05
3	Argentina	1,43	0,95	−0,85	0,16	1,36	1,49
	Bolivia	1,14	0,82	−0,13	0,58	1,03	1,29
	Chile	1,47	0,9	−0,78	0	1,41	1,53
	Colombia	1,38	0,87	−0,64	0,18	1,28	1,42
	Ecuador	1,42	0,9	−0,79	0,04	1,36	1,48
	Perú	1,28	0,96	−1,04	0,10	1,21	1,36
4	Argentina	1,58	0,95	−0,92	0,22	1,51	1,65
	Bolivia	1,59	0,93	−0,83	0,18	1,47	1,63
	Chile	1,57	0,96	−0,97	0,02	1,51	1,64
	Colombia	1,47	0,96	−0,16	0,21	1,39	1,56
	Ecuador	1,53	0,88	−0,74	0,11	1,47	1,59
	Perú	1,46	0,98	−1,02	0,07	1,38	1,53
5	Argentina	1,13	0,87	−0,64	0,32	1,07	1,19
	Bolivia	1,14	0,91	−0,73	0,42	1,08	1,22
	Chile	1,16	0,91	−0,71	0,36	1,1	1,22
	Colombia	1,06	0,90	−0,48	0,60	0,98	1,16
	Ecuador	1,09	0,84	−0,27	0,5	1,04	1,15
	Perú	0,95	0,86	−0,34	0,60	0,89	1,01
6	Argentina	1,41	0,93	−0,82	0,17	1,35	1,48
	Bolivia	1,25	0,90	−0,79	0,31	1,17	1,40
	Chile	1,44	0,95	−0,87	0,24	1,38	1,5
	Colombia	1,51	0,92	−0,79	0,29	1,39	1,50
	Ecuador	1,34	0,91	−0,7	0,28	1,28	1,4
	Perú	1,25	0,94	−0,72	0,37	1,18	1,33
7	Argentina	1,32	1,01	−1,03	0,22	1,25	1,39
	Bolivia	1,44	1,09	−1,02	0,19	1,38	1,53
	Chile	1,33	1,03	−1,09	0,23	1,26	1,4
	Colombia	1,35	0,91	−0,82	0,31	1,29	1,48
	Ecuador	1,3	0,98	−0,94	0,25	1,24	1,37
	Perú	1,18	1,05	−1,00	0,45	1,10	1,26

Abbreviation:

^a M = Mean

^b SD = Standard Deviation

^c Lower.P = Lower Limit

^d Upper. L = Upper Limit.

Exploratory Factor Analysis

The statistical analysis determined a unidimensional structure, made up of the 7 items, in each of the countries where the GAD-7 was administered. Furthermore, in Argentina, the test explained 63.34% of the variance of the data (KMO = .906; p < .000); in Bolivia, it was 65.71% (KMO = .914; p < .000); in Chile, 64.44% (KMO = .917; p = .000); in Colombia, it was 62.84% (KMO = .910; p = .000); in Ecuador, the variance explained was 64.81% (KMO = .925; p < .000) and in Peru, 66.1% (KMO = .900; p < .000), respectively. The factor loadings of most of the items exceeded .720, complying with Stevens’ recommendations 55.

Confirmatory Factor Analysis

We observed that the root mean square error of approximation of population values (RMSEA = .061), the general fit index (GFI = .998) and the adjusted general fit index (aGFI = .995) have good goodness-of-fit values, as does the standardized root mean square residual (SRMR = .058). In the comparative incremental means of fit, the normalized fit index and relative fit index have higher values (NFI = .998; RFI = .998); providing a percentage increase in goodness-of-fit over the base or null model. In the parsimony fit means, adequate results are obtained, the normalized parsimonic fit index (PNFI = 0.665) and the parsimonic goodness-of-fit index (PGFI = 0.553). Table 2 shows the factor loadings and communalities for the 7 items of the GAD-7. The factor shows factor loadings ranging from .66 to .92.

Table 2.

Internal Consistency, Factor Loadings and Communalities of GAD 7.

Items	Countries	ME a	SD b	Item-total correlation	Factorial loading	Community
1. ¿In the last 15 days, have you had feelings of nervousness, anxiety, nervousness?	Argentina	1,41	0,91	,840**	.86	0,73
	Bolivia	1,40	0,88	,836**	.84	.73
	Chile	1,44	0,92	,841**	.87	0,75
	Colombia	1,25	0,84	,729**	.90	0,77
	Ecuador	1,33	0,87	,837**	.85	0,71
	Perú	1,29	0,92	,700**	.78	0,73
2. ¿In the last 15 days, have you been unable to avoid or control your worries?	Argentina	1,20	0,90	,842**	.92	0,82
	Bolivia	1,15	0,87	,847**	.89	0,80
	Chile	1,47	0,90	,842**	.89	0,79
	Colombia	1,01	0,85	,642**	.89	0,69
	Ecuador	1,12	0,86	,839**	.84	0,7
	Perú	0,99	0,88	,805**	.77	0,78
3. ¿In the last 15 days, have you been excessively worried about different things or situations?	Argentina	1,43	0,95	,807**	.84	0,71
	Bolivia	1,14	0,82	,868**	.88	.79
	Chile	1.47	0,90	,812**	.83	0,69
	Colombia	1,38	0,87	,703**	.91	0,77
	Ecuador	1,42	0,90	,811**	.85	0,72
	Perú	1,28	0,96	,698**	.77	0,79
4. ¿In the last 15 days have you had difficulty relaxing?	Argentina	1,58	0,95	,871**	.88	0,76
	Bolivia	1,59	0,93	,854**	.89	.88
	Chile	1,57	0,96	,869**	.90	0,81
	Colombia	1,47	0,96	,782**	.93	0,87
	Ecuador	1,53	0,88	,862**	.90	0,82
	Perú	1,46	0,98	,832**	.91	0,79
5. ¿In the last 15 days have you been so restless that you cannot sit still?	Argentina	1,13	0,87	,822**	.82	0,66
	Bolivia	1,14	0,91	,841**	.85	.80
	Chile	1,16	0,91	,827**	.85	0,72
	Colombia	1,06	0,90	,837**	.84	0,74
	Ecuador	1,09	0,84	,816**	.82	0,67
	Perú	0,95	0,86	,801**	.70	0,71
6. ¿In the last 15 days, have you been easily angered or irritable?	Argentina	1,41	0,93	,700**	.72	0,49
	Bolivia	1,25	0,90	,696**	.68	.51
	Chile	1,44	0,95	,703**	.68	0,46
	Colombia	1,51	0,92	,699**	.78	0,72
	Ecuador	1,34	0,91	,692**	.68	0,46
	Perú	1,25	0,94	,691**	.74	0,46
7. ¿In the last 15 days have you felt afraid, as if something bad might happen?	Argentina	1,32	1,01	,702**	.68	0,45
	Bolivia	1,44	1,09	,812**	.81	.69
	Chile	1,33	1,03	,703**	.66	0,44
	Colombia	1,35	0,91	,686**	.90	0,72
	Ecuador	1,30	0,98	,701**	.79	0,61
	Perú	1,18	1,05	,830**	.89	0,58

Abbreviation:

^** = p < .000

^a ME = Mean

^b SD = Standard Deviation

Finally, taking into account the AFE and AFC models, internal consistency analysis was performed by country; showing values of 0.85 in Peru (CI = 0.800-0.884); 0.88 in Chile (CI = 0.861-0.913); 0.89 in Colombia (CI = 0.867-0.933) and Ecuador (CI = 0.892-0.943); and, finally, 0.95 in Argentina (CI = 0.911-0.985) and Bolivia (CI = 0.901-0.975). Thus, it was found that the Omega index ranged between 0.85 (C. I = 0.800-0.884) and 0.95 (C. I = 0.901-0.985); which means that the scale is reliable. The item-total correlations show a range from .642 to .869 (Table 2).

Discussion

This study evaluated the psychometric properties of the GAD-7 to identify symptoms of generalized anxiety in a large Latin American sample. First, we found that the instrument demonstrated good psychometric performance in each country where it was used. Each item in the instrument showed adequate factor loadings and good saturation levels, with no item displaying low communalities (range 0.46-0.88). These results confirm that the GAD-7 is a unidimensional instrument, with all items aimed at the objective assessment of generalized anxiety. This has been demonstrated in various global^24-35 and regional studies.^36-40

The available evidence supports that the GAD-7 is a clinically valuable instrument, applicable in a variety of contexts. In South American countries, it has been used to detect GAD in healthcare professionals in Colombia and Bolivia during the COVID-19 pandemic,^36,40 in otolaryngology specialists in Hispanic America, ⁴¹ and in patients from public hospitals in Ecuado. ⁴² Additionally, it has been evaluated in Chilean adolescents, ⁴³ Peruvian university students, and the general population,^44,45 as well as in a large sample of adults and older adults in Argentina. ⁴⁶ These studies demonstrate the validity and reliability of the GAD-7 for detecting anxiety in various population contexts, corroborating similar findings reported in other countries and continents.^19-23

In the context of rural and intercultural populations, the GAD-7 was recently adapted linguistically, along with an analysis of its psychometric indicators in an urban Quechua-speaking population in Peru. ⁴⁷ This study confirmed the unidimensionality of the instrument and reported solid internal consistency values (α = 0.896; ω = 0.894). Building on this, our research group ⁴⁸ was the first to report the diagnostic accuracy of the GAD-7 in identifying generalized anxiety symptoms in a clinical sample consisting of a rural Quechua-speaking indigenous population in the same country. In this case, with a cutoff score of 11 points, the GAD-7 demonstrated excellent sensitivity (92.56%) and specificity (89.35%) values. These findings underscore that the items of the GAD-7 are clear, comprehensible, and invariant, ensuring its applicability in both urban contexts and rural or intercultural settings.^49,50

The GAD-7 has also been used to detect other anxiety disorders due to its ability to evaluate long-term changes in symptoms, including post-traumatic stress disorder, social anxiety, and panic disorder. ⁵¹ In a large sample of English patients with various psychiatric disorders, the GAD-7 proved to be an effective tool for identifying the severity of anxious symptoms, unaffected by racial, demographic, or cultural characteristics. ⁵² Similarly, Doi et al ⁵³ analyzed the performance of the GAD-7 in a Japanese adult population with anxiety and depressive disorders, demonstrating its cross-cultural validity. The results indicated that the instrument's factor loadings were equivalent to those of the original version and that the items’ behavior remained stable, with no significant differences between groups with and without anxiety or between those with anxiety and major depression.

On the other hand, the GAD-7 demonstrated strong internal consistency across the analyzed countries, with Omega indices ranging from .88 to .95. These findings confirm that the GAD-7 is a reliable instrument for detecting generalized anxiety without the need to reduce or eliminate items. This is consistent with scientific evidence from various context,^54-56 supporting its reliability and validity for assessing GAD symptoms. Several studies in Latin America reinforce this perspective. For instance, in workers from diverse organizations in Puerto Rico ³⁸ and Colombian physicians during the COVID-19 pandemic, the GAD-7 showed high reliability values (.92). ³⁶ In a large sample of Argentinians, construct validity, internal consistency, and external validity were assessed, achieving high fit indices (IFI, NNFI, and CFI above .99) and strong internal consistency (α = .88; Ꞷ = .92). ⁴⁶ Similarly, Chilean adolescents from private subsidized schools reported a Cronbach's alpha of .86 and a Spearman-Brown coefficient of .82. ⁴³ Together, evidence from other languages and cultural contexts further supports the existence of optimal psychometric indicators for the GAD-7.^14,49,57-60

The GAD-7 is an extremely efficient instrument that, with only 7 items and less than one minute of application, allows for the identification of anxiety symptoms and comorbid conditions, making it a highly valuable clinical tool. However, there are other tests with adequate psychometric indicators, such as the COVID Anxiety Scale, ⁶¹ the Anxiety Symptoms Questionnaire (ASQ), ⁶² or the Hospital Anxiety and Depression Scale (HADS). ⁶³ While these tools are useful, many require significantly more administration time or lack the clinical specificity necessary for accurate diagnosis. In contrast, the GAD-7, in addition to being quick and objective, has been validated across various contexts, languages, and populations.^45,64,65

Despite the positive results, this study has some limitations. First, although we evaluated a large sample of participants from the region, the sampling was not probabilistic, which affects the generalization of the results. Secondly, the premorbid health status of the participants was not known, nor was it possible to include clinical samples to establish case-control comparisons or establish causal relationships. However, in our opinion, the most important limitation is that no invariance analysis was performed according to sex and country, to provide more evidence in favor of the test. Although these data will be revealed in another study, it is important to highlight the importance of showing adequate test behavior in each country comprising the general population.

Therefore, we can conclude that the GAD-7 is a reliable and valid instrument for assessing generalized anxiety symptoms across diverse cultural contexts in Latin America. Its brevity and efficiency make it a valuable tool for both clinical and research settings, particularly in regions with limited mental health resources. Future research should focus on exploring its applicability in underrepresented populations, such as indigenous communities, rural areas, and economically vulnerable groups, to enhance its cultural adaptability. However, its implementation faces challenges, including the need for additional cultural adaptations and the limited availability of trained mental health professionals. Addressing these barriers will be crucial to maximizing its impact in the region.