Suicide and self-harm in Britain: researching risk and resilience using UK surveys

Background
Since the early 1990s national UK surveys have sensitively collected information from large, probability samples of the general population about experiences of non-suicidal self-harm, suicidal ideation, and
suicide attempts. These surveys are available to researchers for further analyses, and the opportunity to link the responses of survey participants to their subsequent mortality outcomes and other routinely
collected data has been developing.

Study aims
In this study we support further suicide prevention research by identifying resources and developing methods. We have revisited a range of datasets, linked survey participants to mortality outcomes, and
analysed data to identify common themes from across the surveys. We have worked alongside people with lived experience to find out whether results resonated with them, and what they saw as the implications for policy and practice. The aim has been to elicit methodological and substantive insight, within a framework of consultation.
Our research questions were:
 What survey datasets are available that provide insight into who self-harms, has suicidal thoughts or makes a suicide attempt in the UK?
 Across these datasets, what common factors predict non-suicidal self-harm, suicidal thoughts, suicide attempts, or suicide?
 How do statistical findings produced in analyses like these resonate with the lived experience of people affected by suicidal distress? What implications do they see arising from the statistics?
The primary objective was to draw out, from multiple sources, consistent patterns in trends, prevalence, subgroup variation and risk. Given that poor mental health has long been established one of the strongest
risk factors for self-harm and suicide we situated this study in a mental health research context.

Methods
The study consisted of three linked strands of work.
Strand 1 involved identifying and reviewing the content and scope of cross-sectional, national surveys from the past two decades that have asked people about suicidal thoughts, attempts and non-suicidal self-harm. The datasets were analysed using descriptive analyses, multiple logistic regression modelling, and latent class analysis to a) produce trends in the nature and extent of non-suicidal self-harm between 2000 and 2014/5, and b) to identify common factors that predict suicidal thoughts, attempts and non-suicidal selfharm.
Strand 2 consisted of pooling waves of Health Survey for England (HSE) data from 1994 to 2010. Only waves including the General Health Questionnaire (GHQ-12) were retained, to ensure all cases had data
on mental health. The combined sample was linked to participants’ mortality outcome (that is: whether they were alive in 2013 and whether they had taken their own life) where participants had consented to data linkage at the time of interview. Multiple regression analyses were undertaken to identify longitudinal predictors of subsequent suicide.
Strand 3 involved depth interviews with twenty people carried out in a location of their choice, facilitated by experienced qualitative researchers. Participants in the consultation were midlife men who had experienced self-harm or suicidal crisis, family members bereaved by male suicide, or professionals working to support men in crisis. Topic guides were developed and interviews began with discussion of the Strand 1 statistical findings. Interviews were audio-recorded (with participants’ permission), transcribed verbatim and managed using the Framework approach for thematic analysis.

Public involvement in the research
Central to this study was this involvement of people with lived experience. They have informed how the statistical findings have been prioritised, presented and interpreted and they have led on formulating policy
implications. The study also involved working with third sector and survivor-led organisations, to guide the findings towards addressing the information needs of those sectors. A complete list of participants
recommendations is provided in the separate Technical Appendices.

Equality and diversity
We started with a health inequalities framework and a focus on the social determinants of health, including mental health. An aim of the research was to identify subgroups in the population at elevated risk of
suicidal thoughts, attempts and non-suicidal self-harm. The statistical analysis involved examining variation by sex, ethnicity, age, sexual identity, relationship status, health status, disability, and subgroups
characterised by economic disadvantage and poverty. In addition to protected characteristics, groups exposed to particular types of adversity were also addressed. The Strand 3 consultation work focused on
some of the groups highlighted as experiencing particular disadvantage. To ensure sufficient numbers to address issues in a meaningful way, we focused the consultation on those who were male and in midlife,
and ensured inclusion of men with a non-heterosexual identity. This was done in part to address policy need to understand these under-researched, high-risk groups.

Results: Trends in self-harm, 2000-2014
Analyses of three waves of the Adult Psychiatric Morbidity Survey (APMS) show that people have become more likely to report non-suicidal self-harm. The proportion of the population aged 16 to 74 reporting this
increased from 2.4% in 2000 and 3.8% in 2007, to 6.4% in 2014. The increase was evident in both men and women and across age groups. An upward trend since 2000 is also evident in suicidal thoughts. These
results are striking, but consistent with research based on people in contact with services indicating an increase in self-harming behaviours in England since 2000.

Results: Risk and context
Analyses of the surveys and mortality linked data, spanning more than 230,000 people, identified features common in many of the lives of people affected by suicidal thoughts and self-harm behaviours. Because
most of the surveys analysed were cross-sectional, conclusions are drawn about associations and context, but not causality.