We leveraged the biological evidence that a first pregnancy ending in miscarriage is considered a quasi-exogenous shock to fertility and linked electronic health records to estimate adjusted associations between miscarriage and self-harm and psychiatric outcomes. In a random cohort of 1.2 million women aged 16 to 50, all first recorded pregnancies between 01/01/2004 and 31/12/2017 were identified using data linked from health registries in England, UK. Each first pregnancy was subsequently categorized into one of two mutually-exclusive groups: miscarriage vs continued pregnancy using valid medical definitions. For each outcome-specific model, we excluded women with a prior recorded diagnosis of the same outcome before first pregnancy. Our empirical strategy relied upon methods under the selection on observables assumption (logistic regression, the augmented-inverse probability weighting estimator, and entropy balancing) to estimate the effects of miscarriage. Miscarriage was associated with higher adjusted odds of self-harm at 6 months (OR 2.30), depression at 6 months (OR 1.50), and anxiety at 6 months (OR 1.25), with the self-harm association persisting up to three years (OR 1.60). Associations with self-harm differed by area-level deprivation: no statistically significant association was observed in the least deprived quintile, whereas elevated odds were observed in more deprived quintiles. Targeted interventions such as counselling aimed at ensuring that women who miscarry have access to healthcare services are required to mitigate possible harms caused by early pregnancy losses.