Lizzy

Frontline

Lizzy

Lizzy is a web app used by professionals to identify high risk victims of intimate partner violence

Berlin, Germany

DEU

For-profit, including B-Corp or similar models

Globally, 1 in 3 women experiences physical/sexual violence by a partner in their lifetime, with severe consequences: chronic physical injuries, trauma, economic instability, and in extreme cases, homicide (often preceded by escalating abuse). Survivors face systemic barriers to safety, including fear of retaliation, stigma, financial dependence, and mistrust of institutions. A core problem is the invisibility of risk dynamics. Abuse often occurs in private, with patterns of coercive control (e.g., isolation, threats, psychological manipulation) that outsiders—or even survivors—may not immediately recognise as precursors to high-risk violence. Healthcare systems frequently fail to identify high-risk cases due to fragmented responses, implicit biases, or lack of training, leading to misjudgments (e.g., dismissing threats as “non-urgent”). As such, emergency workers without a risk assessment to support them are very inaccurate at identifying victim risk levels. Without accurate risk identification, interventions are reactive rather than preventive. This perpetuates cycles of harm: survivors are revictimised, perpetrators face no accountability, and communities bear the social and economic costs of trauma. However, in the counter-violence sphere, the most used tool to identify high-risk victims of intimate partner violence has been proven to be incorrect in 48% of cases.

Lizzy is a web app built alongside practitioners at Charité Research Hospital that operates as two question-and-answer surveys. It is usually administered by a victim support worker alongside the victim in an emergency, social, or legal setting. The first survey aims to identify ongoing abuse and risk levels. The second is completed when a potentially high-risk victim is identified by the first survey, which tells the victim and support worker the likelihood of revictimisation. At AUC 0.98 (98%), Lizzy is currently the most accurate diagnostic tool for domestic abuse globally, and at AUC 0.84 (84%), she is the most accurate predictive tool. It is also the only assessment that outputs results detailing ongoing coercive control. The models at the back end of the Lizzy web app achieve their accuracy through several nationally bespoke studies built to track nationally representative victims and perpetrators of domestic violence. Victims responding to Lizzy feed in answers to a model that recognises patterns of behaviour from prior perpetrators and victims and, in turn, outputs a risk level to emergency services and the victim based on the historic patterns that align with the victim's experience.

Our target population is women who need institutional support. However, as a side project, we have also completed an update that works with male victims. The average risk assessment has a predictive value of 58% (AUC 0.58). In the latest studies, victims are better at predicting their own risk levels than the average tool. This means victims are being asked to relive their experiences of abuse and answer questions that may churn out inaccurate answers. The outcome is an incorrect risk level, and in turn, a wrong risk classification. Women in Germany, and hopefully one day beyond, who complete our tool will be more likely to get the support they deserve because they will have been given a more accurate risk classification.