Depression is a mental disorder characterized by a persistently sad or irritable mood, loss of interest or pleasure, decreased energy, sleep or appetite disturbances, concentration difficulties, and feelings of worthlessness or guilt, lasting at least two weeks, with a significant impact on daily functioning (1–3).The COVID-19 pandemic exacerbated the burden of emotional disorders across all regions, with an increase in cases of major depressive disorder in the general population (4, 5). In Latin America, these effects were intensified due to inequity in access to healthcare, unemployment, food insecurity, and pressure on health systems (6).

From a clinical and public health perspective, depression is associated with multiple comorbid conditions such as insomnia (7), anxiety (8), eating problems, and suicide (9), which increase the severity of the condition and reduce therapeutic effectiveness if not detected in a timely manner (10). In addition, depression may coexist with conditions such as burnout and chronic stress, especially in contexts of work overload and lack of institutional support (11, 12). It also negatively impacts quality of life and life satisfaction, work performance, and overall individual functioning, becoming a key determinant of global disease burden (13, 14).

In this context, early detection of depression is a priority in clinical, community, and epidemiological settings. Rapid screening instruments, such as the Patient Health Questionnaire-9 (PHQ-9), have proven to be effective, valid, and low-cost tools for identifying depressive symptoms in accordance with DSM-5. criterio (15, 16). This questionnaire has become one of the most widely used tools globally due to its brevity, ease of administration, and usefulness in resource-limited settings (17, 18).

In Latin America, various studies have supported the validity and reliability of the PHQ-9. In Colombia, among a university population, adequate internal consistencies were reported (α = 0.80; ω = 0.81) and concurrent validity with the HADS-D (ρ = 0.64) and PHQ-2 (ρ = 0.70) (9). In Peru, a large-scale population study confirmed a unidimensional structure with good fit (CFI = 0.936; RMSEA = 0.089; SRMR = 0.039), high reliability (α = ω = 0.87), and invariance by sex and age groups (19). In Argentina, high internal consistencies were observed (α = 0.87) and adequate diagnostic accuracy (AUC = 0.87) in primary care (20, 21). In Mexico and Chile, stable performance of the PHQ-9 has been documented in the general population and clinical samples, with good internal consistency and validity in local contexts (6, 22). In Puerto Rico, although two-dimensional and bifactor models were explored, the authors concluded that the most parsimonious interpretation of the scores is unidimensional, also confirming its invariance by sex (23). Additionally, in Spain, high internal consistency has been reported (ω = 0.89), good sensitivity (88%) and specificity (80%), as well as factorial stability in primary care; although both unifactorial and two-dimensional models showed acceptable fit (CFI = 0.95; RMSEA = 0.08), the high correlation between factors (r = 0.86) supported the parsimony of the unidimensional model, which also demonstrated invariance by sex, age, marital status, educational level, employment status, and over time (24).

Although some studies have explored alternative models to the unifactorial structure, including two-factor structures (25), bifactor models in Spanish-speaking university populations (21), and even four-factor solutions in community samples (26), the accumulated evidence indicates that such configurations tend to show very high correlations between factors and low reliability in the subfactors. The most recent meta-analysis (10) integrated the available evidence and confirmed that, although two-factor or bifactor models may show acceptable fit, the high correlation between factors and low subfactor reliability limit their usefulness; in contrast, the unidimensional model proves to be more parsimonious and stable, consolidating itself as the most robust option for clinical practice and cross-cultural research.

In this regard, the present study is based on the hypothesis that the PHQ-9 assesses a predominantly unidimensional construct, allowing the total score to be interpreted as a global measure of depressive symptoms and enabling the evaluation of its metric equivalence across countries and sexes. This is particularly relevant in Latin America and the Caribbean, where studies on the factorial invariance of the PHQ-9 remain scarce (16, 23). Therefore, the aim of this study was to assess the structural consistency and comparability of the PHQ-9 in a multinational sample of adults from the region, in order to provide evidence supporting its clinical and epidemiological use in diverse contexts.

Table 1 shows the descriptive statistics of the PHQ-9 items. Item 9 has the lowest mean value, while item 4 has the highest. Regarding skewness and kurtosis values, the vast majority are close to 0, with the exception of item 9, which presents the highest skewness and kurtosis values. Considering the content of this item, which reflects thoughts of self-harm, the high positive skewness and leptokurtic kurtosis indicate that the vast majority of respondents fall into the “not at all” category.

Confirmatory factor analysis was then conducted to examine the internal structure of the test. The models analyzed and their fit indices are presented in Table 2. For the PHQ-9, the unidimensional structure without correlated errors (Model 0) showed adequate fit indices according to the CFI, TLI, and SRMR; although the RMSEA presented a value close to the threshold of poor fit. Given the content overlap among somatic items (sleep–energy–appetite) and the modification indices, a minimal set of residual covariances was allowed (3, 4, 5). This decision is methodologically acceptable when local dependence is documented and there is theoretical justification (47, 48). Comparability by sex and country was assessed through multigroup CFA.

Figure 1 presents the factor loadings of the model that showed the best fit for the PHQ-9. The factor loadings were greater than.68, and the values of the correlated errors ranged from.23 to.37.

In addition, the equivalence of the PHQ-9 factor structure was evaluated by sex and country. The results are presented in Table 3. In the measurement invariance analyses of the PHQ-9 factor structure, the fit indices ranged from good to adequate across all levels examined (configural, threshold, metric, scalar, and strict). Furthermore, the differences between the fit indices were below the threshold established in the data analysis section, indicating that the factor structure of the instrument demonstrates measurement invariance and is equivalent for both groups (men and women). The same applies when using country as the comparison variable. The structure (configural), thresholds, factor loadings (metric), intercepts (scalar), and residuals (strict) are equivalent across groups.

Concurrent validity analysis was conducted by correlating the PHQ-9 with two measures related to depressive symptomatology: occupational burnout, assessed through the Mini-Z, and generalized anxiety symptoms, measured with the GAD-7. To interpret the correlations, the following thresholds were considered: r = .20 as weak, r = .50 as moderate, and.50–.80 as high (49). The results show a high positive correlation between the PHQ-9 and the GAD-7 (r = 0.792, p < 0.01), which supports that both instruments measure related but distinguishable constructs, as anxiety and depression often present comorbidly. Likewise, a high correlation was observed between the PHQ-9 and the Mini-Z (r = 0.635, p < 0.01), suggesting that high levels of occupational burnout are related to higher levels of depressive symptoms. (see Supplementary Data 3: Correlation Matrix).

Finally, McDonald’s Omega coefficient was used to assess the internal consistency of the instrument, yielding a value of 0.89, which indicates an adequate level of reliability.

The findings of this study strengthen the psychometric evidence supporting the PHQ-9 as a clinically valid, robust, and culturally adaptable tool for assessing depressive symptoms across a broad sample of countries in Latin America and the Caribbean. Consistent with previous global research (50–52), our results confirm a unidimensional structure of the instrument, aligned with DSM-5 criteria. This configuration enables a concise and effective evaluation of depressive symptomatology without compromising diagnostic sensitivity.

The factorial stability of the unidimensional PHQ-9 model across the 15 included countries is particularly noteworthy. Despite cultural, linguistic, and socioeconomic differences among these contexts, the instrument demonstrated excellent fit, supporting its structural robustness and comparative utility across countries. This finding is consistent with multicultural studies that have consistently reported a predominant unidimensional structure (6, 21). However, some research has proposed alternative models with bifactorial or bidimensional configurations (23, 53). Although these approaches aim to capture structural variations in specific contexts, their results present relevant limitations.

In the study by Costantini et al. (53), the factor loadings associated with the specific factors were weak and inconsistent, which reduced the interpretability of the bifactor model. Similarly, Rosario-Hernández et al. (23) identified an acceptable fit for the two-dimensional model in Puerto Rican workers; however, the low loadings on the subfactors and the risk of overfitting limited its applicability, partly due to the reduced sociodemographic diversity of the sample. Other studies have also explored alternative models with heterogeneous results. Yu et al. (54) reported marginal fit for the unidimensional model (RMSEA = 0.122), which partially improved when allowing for error covariances; whereas Doi et al. (55), found that the bifactor model achieved superior fit indices (CFI = 0.980; RMSEA = 0.083) compared to the unidimensional model; although the subfactors showed weak loadings and the sample was recruited online without clinical diagnostic verification.

In contrast, our results consistently supported the unidimensional structure, reinforcing its practical utility for screening, without ruling out that alternative models may provide complementary information in specific contexts. Unlike the aforementioned studies, our work was based on a large and heterogeneous sample from Latin America and the Caribbean, which allowed us to confirm a stable unidimensional structure, with excellent fit and adequate cross-cultural validity. These findings are consistent with the most recent meta-analysis (10), which integrated evidence from 40 countries and demonstrated that, although bifactor or two-dimensional models may show statistically acceptable fit, the high correlations between factors and the low reliability of subfactors limit their practical usefulness.

Moreover, previous evidence indicates that the instrument with a unidimensional structure has demonstrated invariance across sociodemographic variables such as age, sex, educational level, and socioeconomic status (19). Therefore, the accumulated evidence supports the parsimony, stability, and invariance of the unidimensional model, confirming that the PHQ-9 should be interpreted primarily as a global measure of depressive symptoms. Thus, the alignment between our findings and the strongest international evidence helps close the conceptual debate, consolidating the unifactorial model as the most robust solution for research and screening in Latin America and the Caribbean.

It is worth noting that, like other studies using non-probabilistic sampling, our work is subject to the limitation of selection bias. However, unlike those investigations, the factorial structure and multigroup invariance were empirically evaluated using the dataset itself, which provides additional evidence of structural stability and comparability across countries and sexes. This approach partially mitigates concerns about the influence of such bias on the main results.

On the other hand, the high reliability coefficients observed both at the global level and within country and sex subgroups support the internal consistency of the PHQ-9 in LAC. These values exceed international standards established for brief screening scales (25, 56) and are consistent with studies conducted in various countries, including those in Asia, Europe, and the Americas (10, 14, 57, 58), The consistency of these results supports the psychometric stability of the instrument across diverse sociocultural contexts.

On the other hand, the multigroup invariance analysis represents a significant contribution. The results show that the PHQ-9 meets the criteria for configural, metric, and scalar invariance by both gender and country. These levels of invariance, which are not always achieved in comparative studies (59), allow for valid comparisons between groups and support its use in multicenter research. The ability to assess depressive symptoms equivalently across diverse populations is essential for advancing global mental health research and for the development of evidence-based public policies.

Compared to these findings, international evidence on the cultural invariance of the PHQ-9 remains limited in many regions. In Africa, for example (60),, found configural and metric invariance across samples from South Africa, Ghana, and Kenya, supporting its cross-cultural equivalence. Similarly, in rural areas of Asia (61),, reported a stable factor structure and good internal consistency in impoverished populations in India. In Europe (24),, documented invariance by age, sex, and educational level in a Spanish population.

These findings reinforce the need for culturally validated instruments in regions such as Latin America and the Caribbean, where mental health inequities and limited resource availability hinder access to timely diagnostic evaluations (22, 62). In this context, the PHQ-9 stands out as a useful tool due to its brevity, reliability, and ease of use-even in high-demand services with limited technical capacity. Its implementation is feasible in primary care, community programs, schools, correctional facilities, and public health campaigns, and it is particularly relevant for vulnerable populations such as migrants, individuals with chronic illnesses, workers exposed to stress, and rural communities, where early diagnosis and preventive interventions are a priority (63–65).

Beyond its clinical applicability, the confirmation of PHQ-9 invariance supports its usefulness for multicenter, longitudinal, and clinical studies in the region. This finding aligns with international evidence on its cross-cultural validity (51, 59) and is complemented by studies in Indigenous populations of Peru and Bolivia, where it has demonstrated strong properties in intercultural contexts (65, 66). The present work, along with previous research conducted by our team in clinical, general, and Indigenous samples (32, 67, 68), contributes to strengthening a regional diagnostic validation agenda aimed at mental health equity.

Finally, the results show that the PHQ-9 demonstrates concurrent validity with other scales measuring constructs related to depressive symptomatology, such as anxiety and occupational burnout. Its strong correlation with the GAD-7 replicates previous findings in multicultural contexts, which have documented consistent associations between depressive and anxious symptoms (69, 70). Complementarily, its moderate correlation with the Mini-Z supports its utility in identifying psychosocial distress in the workplace. In a study involving intensive care unit staff, Jackson et al. (71) reported that higher PHQ-9 scores were significantly associated with elevated levels of burnout (β = –0.32), confirming its sensitivity to occupational stress. Taken together, these findings reinforce the value of the PHQ-9 as a reliable tool for comprehensive mental health screening in clinical and workplace settings.

This study presents some limitations that should be considered when interpreting the findings. First, although participants from fifteen LAC countries were included, the sample is not representative of the entire region due to the non-probabilistic snowball sampling method. This limits the generalizability of the results to national or underrepresented rural populations. Nevertheless, the geographic breadth and sociodemographic heterogeneity of the sample provide a robust framework for regional interpretations. Likewise, the use of digital self-report sampling may have biased the sample toward individuals with greater technological access, higher educational levels, urban residence, and recruitment channels that may have varied between countries. This type of selection bias has also been noted in previous studies, reinforcing the need to interpret it as a shared limitation in the literature. While these characteristics limit population representativeness, the relative homogeneity of the sample, particularly regarding educational level and language proficiency, may have contributed to the internal consistency and invariance observed. Therefore, the diversity achieved and the robustness of the findings should be interpreted as evidence of internal validity within this sample, and their generalization to broader or less educated populations should be made with caution.

Second, the cross-sectional design prevents establishing causal relationships and assessing the temporal stability of the PHQ-9 or its sensitivity to change. These limitations could be addressed through longitudinal studies that examine its structural consistency and responsiveness to interventions. Additionally, the use of self-report measures may introduce biases, such as social desirability or comprehension difficulties, especially in contexts with low health literacy. Nevertheless, the brief format and accessible language of the PHQ-9 help mitigate these risks. Likewise, no formal diagnostic verification through clinical interviews was included, as the main objective was to assess the metric equivalence of the PHQ-9 across different countries and genders. The use of standardized self-reports such as the PHQ-9, which has been widely validated in population-based and cross-cultural studies, is an accepted strategy in this type of research. However, future studies should incorporate structured diagnostic interviews to complement self-reports and strengthen the clinical validity of the findings.

Third, although invariance by sex and country was confirmed, other dimensions of cultural diversity such as Indigenous populations, people of African descent, migrants, or individuals with low educational attainment were not explored. These groups require specific studies to assess the psychometric and cultural equivalence of the instrument.

Future research should incorporate qualitative or mixed-method approaches to explore the subjective meaning of the items, especially in contexts where somatic expressions of emotional distress are predominant. It is also essential to examine the instrument’s performance in various clinical settings, such as primary care, prisons, rural communities, and populations with chronic illnesses.

The present study provides robust evidence on the structural validity, internal consistency, concurrent validity, and factorial invariance of the PHQ-9 in Latin America and the Caribbean. The results confirm a unidimensional structure with excellent fit, high reliability, and equivalence by sex and country, supporting its use as a comparative instrument in multicenter and transnational contexts. These findings consolidate the PHQ-9 as a psychometrically sound and culturally adaptable tool for assessing depressive symptoms in diverse populations across the region.