The context for this paper is that domestic violence, a form of gender-based violence, is declared a major public health issue and that as a major health issue it is not inconceivable to imagine that there is an information infrastructure to monitor health consequences of domestic violence in health populations. The International Classification of Disease (ICD) is the global standard in operation for translating health experiences into categorical codes for intra and international statistical analyses of health. In this paper I present a qualitative analysis of ICD classifications of violence in operation and the degree of presence of the form of gender-based violence known as ‘domestic violence’ in hospital-based administrative health data in England. Drawing on a sociology of diagnosis I consider the epidemiological account of domestic violence in health data as ‘diagnosis-as-consequence’ and introduce qualitative interview data from two critical points along the ‘diagnosis-as-process’ to ‘diagnosis-as-consequence’ trajectory that troubles ‘domestic-violence-as-category’ and thus the health data. These two critical points are the interface between patient, clinician and health record and the interface between health record, clinical coder and code generation. Through my analysis of ‘domestic-violence-as-category’ and category thresholds I begin to unravel the diagnostic work of complex systems – the bodies, agencies and people that intersect and by degrees work to absent the presence of domestic violence in health experiences from health data.