Quechua-Language Screening Tool Proves Reliable for Detecting Abuse Against Women

Deep in the Peruvian Andes, in the Puno region where the Collao variant of Quechua has been spoken for centuries, a quiet statistical revolution is underway. A team of Peruvian researchers has rigorously tested whether a short eight-question screening tool, originally developed in English to detect violence against women, can work just as powerfully when translated into an Indigenous language. Their answer, published in BMC Psychology, is a resounding yes, and the way they arrived at it offers a masterclass in modern psychometric science.

Violence against women remains a pervasive problem across rural Peru, where geographic isolation, limited healthcare infrastructure, and language barriers often leave victims without any formal route to help. In communities where Spanish is a second language at best, screening instruments written in Spanish or English simply cannot capture the experiences of Quechua-speaking women. The Woman Abuse Screening Tool, known as the WAST, is an eight-item questionnaire grouped into a single factor, designed to be administered quickly in clinical and community settings. But a translated questionnaire is only as good as its psychometric properties, the statistical evidence that it measures what it claims to measure, reliably and accurately, in the new linguistic context.

The researchers, led by Oscar Mamani-Benito of Señor de Sipán University, administered the Collao Quechua version of the WAST to 300 Quechua-speaking women from the Puno region. Rather than relying on a single statistical technique, they deployed an integrative multi-method approach that combined three complementary analytical frameworks: Exploratory Graph Analysis, Confirmatory Factor Analysis, and Item Response Theory modeled through a graded response model. This triangulation matters because each method interrogates a different aspect of the instrument. Network analysis examines structure, factor analysis examines fit, and IRT examines how each individual question performs at different levels of the underlying trait being measured.

Exploratory Graph Analysis, the newest of the three techniques, treats the questionnaire as a network in which each item is a node and the statistical relationships between items are the edges connecting them. Algorithms then detect how many distinct communities, or dimensions, exist within the network. For the Collao Quechua WAST, the analysis identified a highly stable one-dimensional structure, with a structural consistency value of exactly 1. In plain terms, all eight questions form a single cohesive cluster that measures one unified construct, and this structure is not a statistical fluke but a robust feature that persists under repeated estimation.

Confirmatory Factor Analysis then put the one-factor hypothesis to a formal test. The model showed an excellent fit to the data, with a Comparative Fit Index of 0.985 and a Tucker-Lewis Index of 0.979, both far above the conventional 0.95 threshold that psychometricians consider evidence of good fit. The Standardized Root Mean Square Residual came in at 0.062, comfortably within acceptable limits. Factor loadings, which indicate how strongly each question relates to the underlying construct, ranged from 0.741 to 0.944, values that sit at the high end of what screening instruments typically achieve. Reliability, estimated with the Omega coefficient, reached 0.935, meaning that roughly 93.5 percent of the variance in total scores reflects the true underlying level of the measured construct rather than measurement noise.

Item Response Theory added a third layer of scrutiny. Under a graded response model, each question is characterized by its discrimination parameter, which reflects how sharply the response probability changes as the latent trait increases, and by its thresholds, which mark the trait levels at which respondents shift between response categories. The IRT model presented good global fit, with fit indices of 0.971 and 0.935, adequate item discrimination, ordered thresholds, and an absence of relevant local dependence, the statistical redundancy that occurs when two questions effectively ask the same thing twice. The test’s information function, which quantifies how precisely the instrument measures at each point along the trait continuum, was highest at medium to moderately high levels of the latent trait, around zero to two standard deviations above the mean, precisely the range where clinicians most need accurate detection.

Empirical reliability at those critical trait levels reached 0.899, a value that confirms the instrument’s precision where it counts most. This is a crucial detail for a screening tool. Instruments can be globally reliable yet imprecise in the specific score range where clinical decisions are made. The WAST in Collao Quechua avoids that trap, delivering its sharpest measurement in the score range associated with meaningful levels of the latent construct, which in this context corresponds to the experiences the tool is designed to flag.

The implications reach far beyond statistical elegance. For Quechua-speaking women in the Peruvian highlands, a validated screening instrument in their own language could transform how violence is detected in health posts, community centers, and outreach programs. Healthcare workers who currently lack any culturally and linguistically appropriate assessment tool now have evidence that this eight-question instrument functions with high reliability and adequate item performance in the Collao variant. The findings support the WAST’s validity as a screening instrument for detecting violence in Quechua-speaking women, potentially opening a door for victims who have historically been invisible to formal support systems.

The study also carries broader lessons for global health measurement. Most psychological instruments are developed in English and translated into a handful of major languages, leaving thousands of linguistic communities without validated tools. The multi-method framework used here, combining network psychometrics, confirmatory factor analysis, and IRT, offers a template that can be replicated for other Indigenous languages, from Aymara in the Andes to languages across the Amazon and beyond. As Indigenous-language validation efforts gain momentum worldwide, this study from Puno demonstrates that rigorous modern statistics can travel across languages just as effectively as the instruments they evaluate, provided researchers are willing to apply the full analytical toolkit rather than settling for a single test of fit.

Ethical safeguards framed the research throughout. The study was approved by the ethics committee of Señor de Sipán University under code 1578-CIEI, and all participants provided informed consent before completing the questionnaire. Participation was voluntary and anonymous, with confidentiality ensured in accordance with the Declaration of Helsinki. The research was fully funded by Señor de Sipán University, and the authors declare no competing interests. Behind the fit indices and information functions stand the Quechua-speaking women of Puno who participated in the study, and it is their experiences that the validated instrument is now equipped to capture, in their own words, with statistical precision that could finally make their invisible struggles measurable and actionable.

Subject of Research: Psychometric evaluation of the WAST in the Collao Quechua: an integrative multi-method approach based on CFA, IRT, and network analysis

Article Title: Psychometric evaluation of the WAST in the Collao Quechua: an integrative multi-method approach based on CFA, IRT, and network analysis

Article References: Mamani-Benito, O., Sánchez-Villena, A., Bejarano-Benites, Z., Pacheco-Vizcarra, M. Y., Farfán-Solís, R., & Cruz-Campos, C. (2026). Psychometric evaluation of the WAST in the Collao Quechua: an integrative multi-method approach based on CFA, IRT, and network analysis. BMC Psychology. https://doi.org/10.1186/s40359-026-05574-3

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05574-3

Keywords: Psychometric, evaluation, WAST, Collao, Quechua, integrative, multi-method, approach, based, network, analysis, scientific research