Flashcards
PSR Lecture 8 - Social experiments
Cards in this deck (9)
- 1The social sciences is a broad area which consists of many sub-fields and sociology is a subfield of social science. • Social science is about understanding social interactions, the deep rules enabling and constraining these, and how interactions influence behaviour and institutions • Sociology especially in health, is almost exclusively qualitative, aiming to understand meaning and agency using interviews and observations. Qualitative research is really useful for understanding how people construct meaning of the world, what their behaviours mean to them, what their experiences mean to them and understanding people's agency how people make decisions.Sociology and causation:
• But sociological theory has traditionally been interested in cause and effect too
• Marx theorised how capitalist relations of production bring about class conflict within and beyond the workplace
• Durkheim theorised and researched how industrial societies engender anomic communities with increased rates of suicide. Sense of rootlessness and Normlessness, and sort of sense of social isolation, and how that was associated with increased risks of suicide
• Bernstein theorised how schools raise cultural boundaries between middle class staff and working class students engendering the latter’s educational disengagement - 2The limits of qualitative research: • Qualitative research could explore some aspects of these phenomena but can’t easily examine causality. You could touch on causality, but it can't develop very definitive conclusions about causality. • Interviews could explore how people describe their experiences of and views on class and disengagement. • Observations could explore how social interactions occur in factories, communities and schools and their immediate consequences • But interviewees’ and observers’ viewpoints are narrow and they can’t report on the broader consequences of these phenomena.Quantitative research and causation:
• Quantitative research is needed to consider (but not directly observe) cause and effect. Useful role to play in shedding light on causality.
• Tries should test questions about causality by testing hypotheses derived from theory, and then using data to test the hypothesis.
• It can explore whether changes in one thing are followed by changes in another - to offer a window on possible causation (albeit an indirect and imperfect one).
• It can do this because it can collate enough comparable data across time and space to see if patterns of association are occurring as they would be predicted by theories. - 3Misunderstanding causation: • Comparisons offer insights into causation but these can be misleading • Comparisons might be confounded by another important factor - it might be urbanism rather than capitalism per se that causes class antagonism • The data collected might be biased - industrial societies associated with higher rates of suicides, but it might be that they might collect better data on suicide rather than having more suicides • Causation might be in the opposite direction than that assumed The least-worst design for examining cause and effect in human phenomena is the randomised controlled trial (RCT)RCTs:
• Adjust for any baseline imbalances (before randomisation)
• Randomise individuals or clusters
• Reduce confounding (randomisation), bias (e.g. blinding fieldworkers), reverse causality (longitudinal data) (distinguish between causes and effects in time, so it gives you more information to allow you to come to a conclusion about whether the intervention has caused the outcomes or not).
• Allows obtaining balance on measured and unmeasured factors - 4A cluster randomised controlled trial is a type of randomised controlled trial in which groups of subjects (as opposed to individual subjects) are randomised. • Can reduce contamination e.g. condom provision in schools. You could randomise individuals to get the condoms or not, but the kids will likely share the condoms with their friends or use them when they have sex with other people who may be their friends in schools, and so contamination would occur. So it's better to do a cluster RCT then.RCTs and ethics:
Ethical to withhold intervention from control group? Yes if:
-genuine uncertainty about effectiveness for that population (if there isn’t uncertainty why do a trial?)
-if control group still gets standard care rather than nothing. You wouldn't want people to be disadvantage by taking part in a trial and being allocated to the control group, and that could happen where there's a certain minimum standard of care operating in a context and you don't offer that to the control group, so that would be unethical.
-if participants give informed consent; need to know they're taking part in a trial and they need to understand that it's 5050 to intervention assignment or something like that. - 5RCTs and politics: • RCTs aim to minimise bias in effect estimates • But can be biased in terms of: -what interventions evaluated (abstinence or sex-positive sex education?) -what populations involved (clinical interventions and trials dominated by white men?) -what do control group get (local or universal standard of care in drug trials in LMI countries?). In such cases, trials have failed to deliver an adequate minimum standard of care to the control group. There has been huge augments in the literature about whether in that situation the control group is entitled to the local standard care or to a more international standard of care, and increasingly, people have argued successfully if you're taking part in a trial which is being sponsored by a high income country, then you should be entitled to the minimum standard of care that's available in that high income country. -what outcomes are measured (sex education as preventing pregnancy or promoting pleasure?) • But these biases are not necessary features of RCT or any other designRCT and ‘positivism’:
• Often accused of being positivist
• Positivism involves: thinking that knowledge is built up from sensory data; using same methods in social as natural sciences; treating theory as string of variables
BUT
• Trials build knowledge by testing hypotheses informed by theory not the other way round i.e. hypothetico-deductive not inductive approach
• Social science trials use qualitative alongside quantitative data unlike e.g. agricultural trials
• Trials can be informed by rich theory - 6Social sciences and natural sciences: Although randomized controlled trials can be used in the social Sciences and the natural Sciences RCT's that are used in the social Sciences are very different from those used in the natural Sciences. For example, in agriculture or pharmacology. For example, most trials of social interventions have large process evaluations which use qualitative research, asking participants about their experiences of an intervention and that would obviously never happen in the in a trial of an agricultural intervention where you wouldn't ask plants about their experiences of an interventionAn example: INCLUSIVE RCT of whole-school bullying prevention:
• Aimed to prevent bullying in English secondary schools
• Our intervention involve teacher training on restorative practice, staff-student action groups to review local data and plan local actions, plus a social/emotional skills curriculum
• Theory of change informed by Bernstein
– improving staff-student and student-student relationships will increase student belonging and reduce bullying
- this will be strongest students of low socio-economic status
• Trial aimed to test hypotheses informed by this theory - 7The theory being that bullying is caused partly when people don't feel apart the school community don't want to obey the norms of the schools and might engage in bullying as well as other respected factors as a sort of sign of their status as somebody outside the conventional school community. The theory is that an intervention trying to improve these relationships will have greatest impact for kids as low social economic status, because for these kids, a sense of belonging at school isn't the default necessarily.-Recruited 40 secondary schools
-Then randomly allocated 20 schools to receive an intervention and 20 schools to carry on as normal
-Longitudinal: 3 year intervention with follow-ups at 24 and 36 month
-Fieldworkers blind to allocation - 8Results: • We found higher engagement/belonging and lower bullying (and lots of other health benefits e.g. smoking, drinking, drug use, student wellbeing and psychological quality of life, as well as increasing student self reported sense of engagement and belonging in the school environment) at 36 but not 24 months; slow-burn intervention. • Because it was an RCT, this couldn’t be explained by confounding by baseline factors (there were minimal baseline differences due to it being an RCT, and where there were baseline differences, we adjusted for the covariates statistically), by biased data (our fieldworkers were blind to allocation) or by reverse causality (our data were longitudinal; the intervention came first and the outcomes came second). • We found that effects were not moderated by students’ socioeconomic status (but were affected by gender, with boys getting more benefits, and baseline need).There might have been some bias in that we obviously couldn't blind students to allocation 'cause the students knew their intervention schools. So it is possible that they were biased and gave us information they thought they wanted us to hear. On the other hand, a lot of students hadn't actually been that aware that an intervention was taking place in their schools because the intervention was often in the background, changing the way the school functioned without being branded as a big intervention that everyone knew the name of, so it's unlikely that bias was important.
- 9Conclusions: • RCTs offer the least biased way to assess intervention effects via randomisation, bliding and longitudinal design. • They are ethical and aren’t positivist but they are political in terms of what gets trialled and how. • RCTs need to continue to focus on pragmatic potentially scale-able interventions rather than turn into academic sociological experiments Primary purpose of RCT = used to identify whether potential public health interventions are effective and can be scaled up. But where possible they can as a secondary aim contribute to assessing and refining social science theory
Information
- Description
- public health Flashcards on PSR Lecture 8 - Social experiments , created by -- -- on 24/12/2021.
- Created by
- -- --
- Views
- 1
- Rating