Resumen La valoración del nivel de riesgo de violencia permite brindar medidas de seguridad a las víctimas y disminuir la revictimización. El objetivo fue validar la escala de predicción de riesgo de violencia grave contra la pareja (EPVR) en población mexicana. Método. Se aplicó la EPV-R modificada a 591 víctimas de violencia, clasificadas en dos grupos: violencia grave y no grave. Se analizó la consistencia interna, la concordancia inter-observadores, la capacidad discriminativa de los ítems y la eficacia diagnóstica. Resultados. Se encontró una confiabilidad interna adecuada (alfa de Cronbach de 0.79); 20 ítems presentaron capacidad discriminativa entre violencia grave y no grave acorde al índice de discriminación y la prueba de chi-cuadrada (p < 0.05). Se observó una concordancia adecuada inter-observador con base en el índice Kappa (0.79, p = 0.001) y el coeficiente de correlación interclase (0.96, p = 0.001). A partir de una puntuación de 21, se encontró un AUC de 0.86 (p = 0.001), una sensibilidad de 70%, una especificidad de 87% y una eficacia diagnóstica de 79%. Conclusión. La escala EPV-R demostró ser un instrumento fiable y válido para estimar el riesgo de violencia grave contra la pareja, en la ciudad de Chihuahua, México.
Abstract Mexico present high rate of violence against women, which comes from the partner. The risk level assessment of intimate partner violence allows providing security measures to the victims and reducing the re-victimization. The aim was to validate the Severe Intimate Partner Violence Risk Prediction Scale (EPV-R) in the urban context of Chihuahua, Mexico. Methods. The scale was applied in 591 victims of violence, classified into two groups: severe and non-severe violence. An experts committee included three new items: psychological violence, economic violence and aggressor links with organized crime. The Internal consistency, the inter-observer concordance, the discriminative capacity of each of the items and the diagnostic efficacy were analyzed. All test were performed at the 95% of confidence level. Results. Adequate internal consistency (Cronbach’s alpha of 0.79) and good agreement between inter-observers (Kappa coefficient 0.79, p < 0.001; Intraclass Correlation Coefficient = 0.96, p = 0.001) were found. Of all 23 items, 20 showed adequate association capacity between severe and non-severe violence groups (p < 0.05), five items presented high capacity, 6 items moderated capacity and 11 items showed low capacity, assigning them a value of 3, 2 and 1 points respectively. The total score mean of the scale was higher in severe group than non-severe group, 22.7 ± 5.9 vs 12.1 ± 7.2 respectively (t = 14.6, p = 0.001). After several cutoff scores tested, a value of 21 combine the best sensitivity, specificity and diagnostic efficacy of the instrument with 70%, 87% and 79% respectively; also, an AUC of 0.86 was found (p = 0.001). From this viewpoint, three levels of severe violence risk were determined: low (0-8), moderate (9-20) and high (20-36). Conclusion. The EPV-R is a reliable and valid instrument to estimate the risk of serious violence against the couple in the urban context of Chihuahua, Mexico.