Saturday, October 24, 2020

Gender and other bias

I teach a session to MRes students entitled "Synthesising research evidence", looking at the process of doing a systematic review.    As part of that we consider bias, both in finding evidence and in the individual pieces of evidence.

So, when searching, there are issues around whether all studies are written up, and whether the "standard" databases are good at indexing journals published in Africa, South and Central America, and Asia (research I did for a poster at ICLC2019, the International Clinical Librarians' Conference, indicates they are not, so we also look at alternative sources like Global Index Medicus and African Journals Online).

Then, within the studies there can be bias introduced by the randomisation process, or by the cohort not being representative of the general population, to give two examples.   

For years I have sort of thought of the second one, for example, thinking about if a trial of an intervention is conducted in Sweden, do the results apply to patients in the UK?    But I think I was just thinking of whether there's anything about the population in general that would make the study not apply to the UK.   I don't think I have ever thought specifically about gender or even ethnic bias in the study - so, are all the participants men, or all white?    

As part of my involvement in decolonisation and diversification, this year I have thought about it, so I added gender bias to the list of biases we thought about in the session, and a student mentioned ethnicity bias.

There was research reported recently that found that a minority of COVID-related trials did not specify gender as a recruitment criterion. I know that my midwifery colleagues are concerned that a lot of research does not include pregnant women.   

So I wondered, do the "standard" risk of bias tools, and critical appraisal checklists, look at gender bias and ethnicity bias?

A quick look (at RoB 2, the SIGN checklists for RCTs and for cohort studies, and the CASP checklists for the same) suggests they do not, as such.    They do ask questions about whether the group of participants in a study are representative, but don't give examples of what this might mean.   I have always assumed it was along the lines of my Sweden example above.  They ask, for example, in an RCT, whether both arms of the trial are comparable.    If both groups are 100% men, or 100% white (or 100% young), then they are comparable.

So, should these tools and checklists ask specifically?

And if they did, would we be able to answer the question from the information in the published report?

And when, in evidence based practice, we think about whether the evidence we have found is applicable to our situation, do we think about gender, ethnicity and age bias?    I never have, specifically, but in next year's Finding the evidence, the companion MRes session where we do think about EBP, we will!

No comments: