A Bayesian dynamic model for the analysis of 1522 helpline data

Violence against women (VAW) remains a widespread human rights violation, and its quantification is a complex task due to women reluctance in declaring violence episodes. In order to increase awareness of this issue, the Italian government proposed several initiatives: among the others, the 1522 anti-violence helpline serves as a major tool in preventing and combating violence against women, acting as the first step in disclosing abuse. We examine data on valid calls to the 1522 helpline over time for the period 2013- 2024, aiming to identify the trend and key shifts in service usage. We propose a generalized Negative Binomial dynamic regression model accounting for covariate effects: the model is estimated in a fully Bayesian framework using an Integrated Nested Laplace Approximation (INLA) approach to overcome the related computational issues. We find patterns that may reflect an increased awareness of the problem over time and an increased awareness of available resources.