Understanding variations in self-harm rates between deprived areas of London

Student thesis: Doctoral ThesisDoctor of Philosophy


Studies comparing rates of self-harm between areas have consistently found socio-economic deprivation to be the area-level exposure most strongly associated with higher rates, with social fragmentation and urbanness also generally associated with greater self-harm. In this context, the consistently low rates of admissions following self-harm seen in public health data for London appear paradoxical. This study uses mixed methods to investigate rates of self-harm within London, how they vary between different small areas and how area context influences these variations.
I aimed first to review the literature on associations between area socio-economic conditions and self-harm rates, at small-area level. I then aimed to explore spatial variation in small-area rates of general hospital admission following self-harm, initially in South East London then in London as a whole. I examined the extent to which these rates are associated with socio-economic deprivation, adjusting for other population and area characteristics associated with self-harm, and whether spatial patterning of self-harm admissions is explained by differences in admission practices between hospitals. Next, I aimed to use qualitative methods to explore the reasons why a case-study area with high levels of deprivation might have low rates of hospital attendance for self-harm. My final aim was to integrate the findings of the qualitative and quantitative studies and use them to identify additional area and population level factors that may be associated with hospital admission following self-harm across Greater London and explore these in the quantitative data.
The initial quantitative studies in this thesis used disease mapping techniques which aggregate outcome data at small-area level to calculate and map area-specific rates of self-harm. Bayesian hierarchical models were used to smooth estimates of standardised rates for small areas and to conduct ecological regression analyses. These methods were applied to routine administrative data on hospital admissions following self-harm from Hospital Episode Statistics and clinical electronic patient record data on Emergency Department (ED) attendances following self-harm. The studies included 8327 individuals admitted, and 12,577 individuals attending EDs, following self-harm in South East London and 59,510 individuals admitted following self-harm in Greater London, between 2007 and 2018. These studies were used to identify a case study area in South East London with persistently high levels of deprivation but low rates of self-harm ED attendance and admission. I conducted focus groups and interviews with people who had in-depth local knowledge of this area, focusing on experiences of stress, distress and self-harm within the community. I analysed this qualitative data using reflexive thematic analysis. The final study used Greater London data to calculate and compare rates of admission following self-harm for different racial and ethnic groups and applied disease mapping techniques to explore the associations between the ethnicity of small-areas’ populations and their self-harm rates.
The disease mapping studies found that self-harm admission rates were highly spatially patterned across both South East London and London as a whole at small-area level, with lower rates close to the city centre. Socio-economic deprivation was the area exposure most strongly associated with self-harm rates but did not explain the spatial patterning seen. Analysis of ED attendance data demonstrated substantial differences in rates of admission for those presenting with self-harm between hospitals, which appeared to be related to difference in hospital practices. These differences explained some, but not all, of the spatial patterning seen, however they did not change the direction or size of associations with area-level exposures. Within the qualitative case study area, participants described the ways that the stressors they experienced were related to the multiple marginalised identities inhabited by much of the population, in particular being racialised as Black, and to community wide experiences, especially exposure to violence. These stressors on the community act to create norms that prioritise toughness and self-reliance both collectively and within individuals. Alongside negative community experiences of contact with mental health services, this made individuals in the area less likely to identify themselves as mentally ill by harming themselves in ways recognised clinically as “self-harm” or seeking help when they had done so. The final study found substantial differences in rates of admission for self-harm between different racial groups, with the lower rates in Black and Asian people than White people in both sexes at all ages. These differences increased with adjustment for socio-economic deprivation.
The work in this thesis suggests multiple explanations for the spatial patterning of self-harm rates seen in London. The service use data currently used within public health to capture self-harm as an outcome may result in self-harm being less likely to be counted in some areas and among some population groups. This has the potential to underestimate need in some marginalised communities. The measures of area exposures that lead to an expectation of higher rates of self-harm in London may not capture the social environment within the city well. They may fail to capture protective aspects of communities in deprived, racially and ethnically diverse areas. Finally, self-harm as defined clinically and in research is probably a less useful outcome as an indication of the stress and distress in the population of London than in other places. A wider range of harmful outcomes may need to be studied simultaneously to capture the impact of individuals’ social environment on their mental health.
Date of Award1 Jul 2021
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorStephani Hatch (Supervisor), Matthew Hotopf (Supervisor) & Ioannis Bakolis (Supervisor)

Cite this