Archives

New possibilities created by crime survey wave integration

The Crime Survey for England and Wales (CSEW) and its predecessor, the British Crime Survey (BCS), are widely used by both academics and government to assess the level of crime and its impact on society. While the survey has run since 1982, combining the multiple years of the survey can be complex and mistakes are easily made. As a researcher in criminology who frequently uses the CSEW and its predecessor, I have produced detailed Stata code to combine data from multiple survey years to support other researchers who also analyse the CSEW in Stata (or would like to start). I worked with the UK Data Service (UKDS) and the Office for National Statistics (ONS) to share the code and develop guidance for its use.

With this code, you can specify what you need, namely, which years of the Crime Survey you want to merge and if you want the adolescent and young adult panels, the bolt-on datasets that provide uncapped codes, and/or if you want to use the ethnic minority booster samples. As a result, the code can be easily tailored for each researcher’s needs.

By combining multiple survey sweeps, analysts can examine temporal trends. A combined file also enables analysts to look at low prevalence offences, population groups, or consequences, that do not have a high enough frequency in a single year.

Two examples are given below on how this integrated dataset provides new and exciting opportunities.

The code can be downloaded via this link: https://reshare.ukdataservice.ac.uk/856494/

Example 1: Revealing gender and age differences in trends in experiencing violence

We used our integrated crime survey dataset to examine temporal trends in different types of violence, and whether these varied by gender and age.

After a rise in violent crime in the 1980s, there was a decade of steady decline followed by a decade of stability (blue line, Figure 1a). However, for other crimes, which can also be considered violent, the patterns observed are different. After a short period of decline in the 1990s, sexual violence against women remained relatively stable until around 2010 when it began to increase, reaching the 20 years high by 2020. Additionally, there has been a sharp rise in threats reported by women in the last 5 years of data, making threats almost as prevalent as at its peak in the late 1990s.

The trends in violent crime for men follow a broadly similar pattern as for women, but at a higher rate. Unlike women, however, men did not experience an increase in threats in the more recent period.

Figure 1. Prevalence of violence by type of violence and gender, 1982 to 2020

a) Proportion of women experiencing violence by type of violence

b) Proportion of men experiencing violence by type of violence

Source: Authors’ analysis using CSEW/BCS data from 1982 to 2019/2020.

Notes: Weighted proportions. Violent crime includes the following offences: Serious wounding, other wounding, common assault, attempted assault, serious wounding with sexual motive, other wounding with sexual motive. Sexual violence includes the following offences: rape, attempted rape and indecent assault. Due to low frequencies, sexual violence is not reported here for men.

Figure 2 reveals that there has been major change in the age profile of victims over the past 40 years. 16- to 19-year-olds were almost 3 times as likely to become a victim of violence as people aged 30 to 39 in the mid-1990s. But violence against this group has declined rapidly since then: while they continue to be the group that is most likely to be victim of violence with 7.2% annual victimization in 2020, this used to be over 28% in the mid-1990s. While risk of violence has declined for all the ages under 40, the shift has been the largest for the younger groups.

Relatively few people over 50 become victims of violence compared to younger age groups in each time period. However, closer inspection reveals there is a significant increase in the risks of violence among the older age groups (60-69 and 70 and older) since the late 1990s, and particularly since 2015.

Overall, the age profile of victims has shifted massively over the decades, there is now much less variation in rates between age groups.

Figure 2 Prevalence of violence (including violent crime, threats, robberies, and sexual violence) by age group, 1982 to 2020.

Source: Authors’ analysis using CSEW/BCS data from 1982 to 2019/2020.

Example 2: Investigating smaller groups: Differences in wellbeing impact between intimate partner perpetrators

Our integrated crime survey dataset allows for the study of minority groups that are relatively small or forms of violence that are not often reported.

For example, only by combining twenty years of the crime survey (2001 to 2020) do we have sufficient sample size to study the impact physical intimate partner violence has on wellbeing and health, and how it differs between various types of intimate partner perpetrator.

Firstly, it is important to note that physical violence by any type of intimate partner has a higher risk of high emotional impact (Figure 3a) and a higher risk of injury (Figure 3b) than violence by other types of perpetrators.

Figure 3a below shows that the emotional impact reported by female victims is higher when the violence was committed by a current or former spouse/partner compared to if it was done by a current or former boy/girlfriend. Women were more likely to say they were ‘very much’ affected by the violence when it was committed by a current or former spouse/partner. It could be that the proximity of spousal relationships, which are often cohabitating, and their average longer duration account for some of the greater report impact. However, in contrast to emotional impact, figure 3b (below) shows that women are more likely to get an injury(ies) by violence by current spouses than by former spouses.

Overall, this study highlighted that physical violence by an intimate partner has a more severe wellbeing and health impact than violence by others, but also the need to differentiate intimate partner violence and abuse by not only the type of violence/abuse but also the type of intimate partner.

Figure 3 Estimated emotional wellbeing and risk of injuries for women following physical intimate partner violence, differences between intimate partner perpetrators.

a) Respondent’s reported emotional impact (showing the highest category).

b) Respondent’s reported physical health impact (showing the risk of injury)

Source: Authors’ analysis using CSEW/BCS data from 2001 to 2019/2020.

Notes: Respondent’s self-assessed emotional impact measured in four categories: not impacted, little impact, quite a lot, very much impacted. Respondent’s self-assessed risk of injuries is measured in three categories: no force was used, force was used but no injury was sustained, force was used that led to an injury. Figures are based on average marginal effects following ordered logit models controlling for key (socio)demographics. Significance was tested in additional models.

What the merger code does and doesn’t do

The Stata code enables users to merge the raw CSEW/BCS datasets. Consequently, at the moment, this code does not harmonize variables that change (slightly) over different years. Considering the measurement of many variables changes over the years, the users of this combined file need to make their own decisions on what operationalisations work best for their research and for the years they use.

Most of the time new variable names are used when a new measurement is used. However, for a few variables, different measurements seem to be used in different years, but they have the same variable name (for instance for household income variables such as tothhin2). In the current code, these variables are treated as being the same. Therefore, users need to carefully check the variables that they use for the relevant years.

Next, this code does not work in the secure researcher environment as provided by UKDS or ONS because the datasets in these environments have different names and the structure of the folders is different.

Overall, the merger code will save researchers precious time in combining the surveys that they want to use. As we have shown here, combining survey sweeps can benefit the study of trends in victimisation. The code can also be used for studying groups or crimes that are too rare to study using only a single sweep, therefore, this code may provide an incentive for studying marginalised groups and specific crimes, contributing to new insights into victimisation.

Citation for merged code

Blom, Niels (2023). Code for Merging Waves of the Crime Survey of England and Wales and the British Crime Survey, 1982-2020. [Data Collection]. Colchester, Essex: UK Data Service. 10.5255/UKDA-SN-856494

Examples in this blog are from

Blom, N., Obolenskaya, P., Phoenix, J., and Pullerits M. (2023, September 11-13). Differentiating intimate partner violence by perpetrator relationship type. Types of crimes committed and consequences for victims’ health and wellbeing by different types of intimate partner perpetrators [Conference Presentation]. European Conference on Domestic Violence, Reykjavik, Iceland.

Obolenskaya, P. & Blom, N. (2023, September 6-9). The rise, fall and stall of violence in England and Wales: how have risks of violence changed for different groups? [Conference Presentation]. EuroCrim 23^rd Annual Conference of the European Society of Criminology, Florence, Italy.

Data reference

Office for National Statistics. Crime Survey for England and Wales, 2001-2002 to 2019-2020 and British Crime Survey 1982 to 2001 [data collections]. UK Data Service SN: 8812, 8608, 8464, 8321, 8140, 7889, 7619, 7422, 7252, 6937, 6627, 6367, 6066, 5755, 5543, 5347, 5324, 5059, 4787.

For further information, please contact Niels at niels.blom@city.ac.uk

Photo by Andre Lichtenberg

New Data Assessment Tool: Mitigating Risk of Bias – Ethnicity and Migration

At UK Prevention Research Partnership (UKPRP) funded consortium VISION we have dedicated a workstream to studying the data gaps, analytical biases, and systemic blind spots that arise around questions of race, ethnicity and migration. We acknowledge that all research data is socially constructed. Asking questions about that process of construction can help researchers be aware of the biases and distortions that will arise when data are used uncritically and unreflectively.

Over the first two years of the project, we have considered the gaps in our own data, drawn on expertise and insight from our research team and associates, and drawn on a diverse array of methodological and disciplinary backgrounds. This research has allowed us to develop a tool to support researchers on our project and beyond to mitigate the risk of introducing or reproducing bias regarding ethnicity and migration in data, analysis and reporting findings. The tool initially responds specifically to administrative and survey data but can be adapted for use with any dataset.

The tool was conceived following consultation with VISION colleagues regarding the construction of ethnicity and migration data across the consortium. The objective was to produce the best quality data and analysis possible that could account for diversity and diverse experiences in the population. The VISION consortium is concerned with measuring violence: ethnicity and migration status are key areas where violence is underreported, which creates gaps in data. We found that existing datasets are unable to properly reflect diversity and inequality in the population and therefore cannot fully explain different experiences. Thus, we sought ways to prevent reproducing biases and data gaps within our analyses.

A collaborative workshop held by the Ethnicity and Migration research group initiated an iterative process of tool development. A meeting with the UKPRP Community of Practice for Race and Ethnicity specified feedback to the tool design. Extensive research of existing literature took place between September 2022 and March 2023 and was summarised into a companion document, both providing citations and further information.

A draft tool and companion document was shared with specialist service IMKAAN (a by-and-for service for minoritized women and girls at risk of violence) in July 2023 for consultative feedback.

I am pleased to announce the tool is now ready for circulation and use.

The tool comprises three parts:

The first part offers guiding questions to assess the quality of a data set at the adoption stage, with regard to how well the dataset mitigates the potential biases that may be produced during data collection regarding ethnicity and migration. Researchers can apply the questions within this section to any dataset they are adopting, although it was designed particularly with survey data and administrative data in mind because these are the main datasets used in the VISION project.
The second part of the tool guides the researcher in a reflective process that is intended to allow the researcher to assess the potential for internally held biases or structural and systemic biases to which they have been exposed affecting the data analysis.
The third part asks researchers to consider the impact of reporting findings. The wording of publications might affect how the finds are interpreted or reported in the media, cited by other researchers, or circulated. Findings might also be misused. This final section of the tool offers some techniques for mitigating the misinterpretation or misuse of findings (although we acknowledge that this is often outside of researcher control).

This document will evolve over time, therefore, we are keen for your feedback. If you download the tool and guide and use them, please let us know how you get on! Is the tool easy to use? Is the guide clear? How effective were they in helping you mitigate bias when working with your data? Please let us know by contacting Andri at Alexandria.innes@city.ac.uk. We look forward to hearing from you!

Download the assessment tool

RiskOfBias_web Download

Download the companion guide

VISION-TADRoB_web Download

Dr Annie Bunce receives award at Lancet Public Health Science conference

Dr Annie Bunce, VISION Research Fellow, was awarded Best Oral Presentation at the Lancet Public Health Science conference in London this November. She presented on the Prevalence, nature and associations of workplace bullying and harassment with mental health conditions in England: a cross-sectional probability sample survey.

Annie’s research, conducted with VISION colleagues Ladan Hashemi, Sally McManus, and others, presents the first nationally representative findings on the prevalence of workplace bullying and harassment in England for over a decade. Annie analysed data from the 2 014 Adult Psychiatric Morbidity Survey (APMS) to demonstrate: the prevalence of workplace bullying and harassment (WBH) in the working population in England; the nature of WBH experienced, who it was perpetrated by and the types of behaviour it involved; and associations between the experience of WBH and indicators of adverse mental health.

The study is unique in that the APMS makes robust assessments of mental health – operationalising diagnostic criteria – which provides an accurate assessment of clinical need. Implications for employers, policymakers, health services and researchers are outlined.

For the article, please see: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(23)02066-4/fulltext

Please contact Annie at annie.bunce@city.ac.uk for further information.

Photo by Icons8 Team on Unsplash

Causal discovery for studying sexual abuse and psychotic phenomena

Sexual abuse and bullying are associated with poor mental health in adulthood. Elucidating putative causal relationships between affective and psychotic symptoms may inform the development of therapies. Causal diagrams can help gain insights, but how?

Given a causal diagram, usually represented as a directed acyclic graph (DAG), and observational data from the variables on the graphs, many analytical methods (especially adjustment techniques) allow us to estimate the effect that intervening on a variable is expected to have on another.

In real-world problems, we rarely have a complete picture of an underlying structural mechanism regulating the relationship among different variables. Causal discovery is a technique leveraging statistics and machine learning tools to uncover plausible causal relationships from data, with little to no prior knowledge of them. While learning causal structures from purely observational data relies on unrealistic assumptions (especially causal sufficiency and faithfulness), a causal discovery exercise may help us identify the most promising scenarios to prioritise when designing interventional studies.

In a recent article, now available open access in Psychological Medicine, Dr Giusi Moffa, Statistician affiliated with the University of Basel, Switzerland and colleagues used state-of-the-art sampling methods for inference of directed acyclic graphs (DAGs) on data from the English Adult Psychiatric Morbidity Surveys, to investigate sexual abuse and psychotic phenomena.

The analysis sought to model the interplay among 20 variables, including being a victim of bullying or sexual abuse and a range of psychotic (e.g. paranoia, hallucinations and depression) and affective symptoms (e.g. worry and mood instability) while accounting for the sex of the participant. To respect temporality, we imposed some prior constraints on the DAG structure: childhood sexual abuse and bullying referred to events that were temporally antecedent to the assessment of the psychological variables, and hence they only admit incoming edges from sex and each other.

Contrary to expectations, the procedure favoured models placing paranoia early in the cascade of relationships, close to the abuse variables and generally upstream of affective symptoms. A possible implication is that paranoia follows from early abuse involving bullying or sexual exploitation as a direct consequence. Overall, the results were consistent with sexual abuse and bullying driving a range of affective symptoms via worry. As such, worry may be a salient target for intervention in psychosis.

Check out the paper for a more thorough discussion of the findings (joint work with Jack Kuipers, Elizabeth Kuipers, Paul Bebbington and VISION member Sally McManus).

This is a repost of a blog available on LinkedIn: https://www.linkedin.com/pulse/causal-discovery-studying-sexual-abuse-psychotic-phenomena-moffa

Paper available open access: https://www.cambridge.org/core/journals/psychological-medicine/article/sexual-abuse-and-psychotic-phenomena-a-directed-acyclic-graph-analysis-of-affective-symptoms-using-english-national-psychiatric-survey-data-erratum/CF603075EBBD5D75E60F327CE01C4050

For further information about the approach: giusi.moffa@unibas.ch

VISION member awarded UKDS Impact Fellow focused on the socioeconomics of violence

We’re delighted that one of VISION’s core researchers, Dr Niels Blom, has been awarded a prestigious UK Data Service (UKDS) Fellowship.

The award will be used to improve the reach and impact of Niels’ research on violence and abuse and its relationship with job loss, health, and wellbeing. He is using several UKDS datasets, including the UK Longitudinal Household Survey and the Crime Survey for England Wales, to understand the link between violence, particularly intimate partner violence, and its socioeconomic, wellbeing, and health impact.

For more information about Niels, his work, and what he hopes to get out of the Fellowship scheme, see his blog on the UKDS website.

The UKDS is funded by the UKRI and houses the largest collection of economic, social and population data in the UK. Its Data Impact Fellowship scheme is for early career researchers in the academic or the voluntary, community, and social enterprise (VSCE) sector. The focus in 2023 is on research in poverty, deprivation, the cost of living crisis, housing and homelessness, using data in the UK Data Service collection. The purpose of the programme is to support impact activities stemming from data-enhanced work.

For further information on the UK Data Service please see: UK Data Service

To read Niels’ blog please see: UK Data Service Data Impact Fellows 2023: Niels Blom – Data Impact blog

Or contact Niels at niels.blom@city.ac.uk

Photo by Alina Grubnyak on Unsplash

Measuring violence using administrative data collected by specialist domestic and sexual violence and abuse support services

Interpersonal violence, which can include various forms of domestic and sexual violence and abuse (DSVA) is a leading cause of death, particularly among young adults. In the UK, specialist DSVA services provide much-needed support to victim-survivors of these types of violence, and some provide support for perpetrators to change their behaviour. To monitor and support their work, specialist services collect data on violence. This data has the potential to improve understanding of violence but presents unique challenges.

In this review, VISION researchers Dr Annie Bunce, Dr Sophie Carlisle and Dr Estela Capelas Barbosa describe and discuss some of the key challenges facing the data collected by specialist services.

Inconsistencies in data collection arise due to the differing remits and priorities of specialist services, which mean violence and abuse are defined and measured in slightly different ways by these organisations. Particularly, the review highlights the significant variation in outcomes and outcome measurement tools used to evidence the effectiveness of services and interventions.

Specialist support services collect valuable data on many and multiple types of violence, the wide impacts of violence on victim-survivors’ lives, and information about perpetrators. As the data are not collected for research purposes, a considerable amount of work is often required to make the data suitable for statistical analysis. Critically, the piecemeal and insecure funding of specialist services limits their capacity to collect and analyse data.

Together these issues make it challenging to collate data from specialist services and use it to inform measurements of violence.

The researchers recommend the development of a core outcomes framework, exploration of methods for linking specialist services data with other sources of administrative data on violence, and sustainable funding for third sector specialist support services.

For further information please see: Social Sciences | Free Full-Text | The Concept and Measurement of Interpersonal Violence in Specialist Services Data: Inconsistencies, Outcomes and the Challenges of Synthesising Evidence (mdpi.com)

Or contact Dr Annie Bunce at annie.bunce@city.ac.uk

Photograph by Claudio Schwarz on Unsplash

Varying definitions and measurements of violence limit reduction strategies

Violence reduction is a United Nations (UN) sustainable development goal (SDG) and is important to both the public health and criminology fields. The collaboration between the two has the potential to create and improve prevention strategies but has been hampered by the usage of different definitions and measurements.

In this paper, VISION researchers Dr Niels Blom, Dr Anastasia Fadeeva and Dr Estela Capelas Barbosa explore the definitions and measurements of violence by the World Health Organization, UN, and Council of Europe to arrive at a harmonized framework aligned with the SDGs.

Violence and abuse are defined by these organizations as intentional actions that (are likely to) lead to harm, irrespective of physicality or legality. When recording violence and abuse, health- and justice-based administrative systems use different codes which cannot directly be translated without resorting to broad overarching categories.

The researchers propose a framework to record violence that includes individual and event identifiers, forms of violence and abuse (including physical, sexual, and psychological), harm, and individual and event characteristics.

For further information please see: Social Sciences | Free Full-Text | The Concept and Measurement of Violence and Abuse in Health and Justice Fields: Toward a Framework Aligned with the UN Sustainable Development Goals (mdpi.com)

Or contact Niels at Niels.Blom@city.ac.uk

Photo by Parsa on Unsplash

Unlocking violence information from clinical text

Blog by Dr Lifang Li, Research Associate with UKPRP VISION, Kings College London

Clinical Record Interactive Search (CRIS)

In 2008, the Clinical Record Interactive Search (CRIS) system was launched. CRIS removes personal identifiers from the health records of the South London and Maudsley NHS Trust, making them available for use in mental health research. The platform operates under a governance framework that prioritises patient anonymity and places patients at the centre of its operations. The use of exceptionally large volumes of records with unprecedented levels of detail has the potential to revolutionise mental health research.

The CRIS Violence application

The CRIS violence application is computer software that finds clinical text that refers to interpersonal violence, including the presence of violence, patient status (i.e. as perpetrator, witness or victim of violence) and violence type (domestic, physical and/or sexual) using Natural Language Processing (NLP). NLP uses pattern matching and statistical techniques to automatically process natural human language. It was developed by Riley Botelle, Professor Robert Stewart and their colleagues to the identification and classification of experiences of violence in narrative records, described in their 2022 paper “Can natural language processing models extract and classify instances of interpersonal violence in mental healthcare electronic records: an applied evaluative study”. Recently, after a thorough validation process, the CRIS team has started to run the violence application routinely, alongside many other NLP applications (e.g., to find suicidality, agitation, medications, anxiety) that are available for CRIS. Structured output from these, now including various violence-related variables, is saved back into the CRIS database, from where authorised health researchers can access it.

How does it serve the researchers and clinicians?

By accurately identifying the presence of violence, different types of violence, and patient status, the application is enabling researchers to examine how experiences of violence are correlated with various mental health problems, outcomes and treatment trajectories, and how these relate to patients’ characteristics (such and age, gender, and ethnic group), and account for health inequalities.

Future work

Given the possibility that psychological abuse and economic abuse may also occur in patients and are recorded in the health record by clinicians, our work as part of the VISION consortium involves updating the current violence application to identify mentions of these, allowing us to extend the violence research possible using CRIS.

Illustration: Nina Rys / Shutterstock.com

Mental health and wellbeing data – webinar for researchers

This webinar focused on quantitative analysis of secondary data, to provide insight into population mental health and its social determinants. It took place on Teams Monday, 6 March 2023, at 14:00-15:30.

Speakers included VISION researcher Sally McManus, who discussed England’s main mental health survey, the Adult Psychiatric Morbidity Survey (APMS). The survey series covers anxiety and depression, alongside items on violence and abuse.

This webinar formed part of a series organised by Understanding Society, UK Data Service, Centre for Longitudinal Studies (CLS) and the National Centre for Research Methods (NCRM). The Data Resource Training Network is a collaboration between a number of ESRC-funded resource centres working together to promote the value and use of social science data.

Photo credit: Photo by Erol Ahmed on Unsplash