Objectives Potential clusters of self-harm within schools and peer groups remain a public health concern, yet there is limited evidence quantifying the burden within schools. We aimed to estimate the proportions of pupils and schools affected by self-harm between 1^stSeptember 2018 and 31^stAugust 2019 and to examine the characteristics of affected schools and pupils.

Methods We used data from ECHILD, a linked de-identified database of records from schools (National Pupils Database) and hospitals (Hospital Episodes Statistics) in England. For all pupils enrolled in years 7–11 (aged 11/12–15/16 years) in 2018/19 we estimated the proportion of pupils within each school and year group who had one or more emergency hospital admissions with an injury related to self-harm (including self-injury, poisoning, drug/alcohol misuse or violence) in 2018/19. We also estimated the proportion of pupils with a personal history of self-harm prior to the 2018/19 academic year. We used sex-specific multilevel logistic regression models to examine characteristics associated with self-harm including; school year, ethnic group, quintile of deprivation, a history of self-harm hospital admissions, current or historic special educational needs (SEND) provision, school size and the proportion of pupils within the school that were female.

Results A total of 0.5% (13,878 of 2,907,075) of pupils were admitted with self-harm in 2018/19. Self-harm was more common in girls (78% of all presentations were in girls) and at older ages; <0.1% of pupils in year 7 vs 0.7% in year 11. 84% (3,271 of 3,893) of schools had one or more pupils with a self-harm admission during 2018/19. Of these schools, the mean number of pupils admitted with self-harm was 4.2 (SD; 3.1). Similar results were evident for the 0.5% (14,438 of 2,907,075) of pupils with a history of self-harm admissions. In adjusted regression models self-harm was most strongly associated with a history of self-harm (girls; odds ratio and 95% confidence interval; 22.7 (21.7–24.1), boys; OR 15.8 (13.6–18.4)). In addition, pupils who were; in older school year groups, of white ethnicity, had current or historic SEND provision, had higher proportions of female peers or were in smaller schools were more likely to have a self-harm admission. Higher levels of deprivation were associated with self-harm in girls, but not consistently in boys. For both girls and boys, within school variation accounted for 6% of total variation in self-harm (intraclass correlation coefficient 0.06 (95% CI; 0.05–0.08)).

Conclusions This is the first study with national coverage of England to examine the burden of hospital admissions with self-harm within schools. Although our study identifies only the tip of the clinical iceberg of self-harm, our results indicate that the vast majority of schools are affected. Our results indicate some evidence of clustering associated with school, however, the numbers of affected pupils within each school are relatively low, particularly among boys. Individually targeted interventions (e.g. to girls with a history of self-harm) may therefore be more effective than universal strategies for reducing self-harm admissions.