Introduction: Self-harm entails high costs to individuals and society in terms of suicide risk and healthcare expenditure.The risk of future self-harm is higher among those who have already engaged in self-harm as compared with the general population.However, previous self-harm is not a necessary nor sufficient criterion for the prediction of future self-harm.Effective risk assessment of self-harm patients requires identification of factors that reliably predict repetition of self-harm.This systematic review aims to identify risk factors for repetition of self-harm after an index self-harm presentation.Methods: The review included prospective studies published in any language, which examined risk factors for repetition of self-harm among self-harm patients, irrespective of suicidal intent.A systematic search located 116 relevant studies.These were quality-assessed and synthesised in narrative form.Results: The majority of the studies were conducted in Europe.Most studies were of medium to high quality but very few studies used multiple sources to detect repetition.Strong predictors of repetition were: being aged 20-54, high levels of hopelessness; previous psychiatric treatment; personality disorder; alcohol misuse; previous self-harm; and self-injury involvement in the index episode.Under-researched but emerging risk factors were: lower problem-solving ability; lower sense of coherence; external hostility; sociopathy; suicidal ideation; regretting surviving; being a victim of violence; not reporting work or relationship problems; homelessness; epilepsy; and endorsing one of the Manchester Self-Harm Rule criteria.There was mixed evidence as to the predictive power of the following factors: gender; history of violence/criminal record; suicidal intent; leading a suicide note; lethality of the index episode; social class; depressive disorder; psychotic disorder; substance misuse; unemployment; living alone; marital status; physical illness; sexual abuse; and self-esteem.Despite extensive study, familial suicide, education level, Beck Depression Inventory scores, any psychiatric morbidity, and wish to die failed to reliably predict repetition.Composite scales to predict repetition tended to derive from univariate analysis and had high sensitivity but poor specificity.Discussions: Selection of data items in the reviewed studies tended to be driven by practical constraints rather than by theory.Nonetheless, it seems that many of the emerging risk factors were related to diathesis factors rather than stress factors with some notable exceptions.Given the emphasis on risk assessment in guidelines for the management of self-harm, the current review will enhance prediction and reduce repetition of self-harm.