Epi in the News

Thanks Sujit,

There is no mention of potential confounding by research field. It is possible that research fields with a higher proportion of female researchers have a lower rate of publication than other research fields which both have a lower proportion of female researchers and a higher rate of publication. The analysis should be stratified by research field to examine this possibility. If this is found to be the case, then the association between female mentee/mentor relationships and fewer citations would be confounded by research field and the adjusted relationship between female mentee/mentor relationship and fewer citations might no longer be significant.

Thank you for brining this article, an interesting discussion though.

Well let's start from the beginning. In here the exposure will be : being mentored by female. Outcome: impact of the scientist on the field. I couldn't spot the effect modifier though! not sure if we need to refer to the full article or I just simply didn't get it.

Important limitation will be how the outcome was measured. Is the number of citation per se enough indicator to assess the impact. Was there a comparison group e.g. group of females mentored by males or males mentored by females. Were any other confounding factors taken into account e.g. length of career life for the female mentees, any change in jobs later on , any other factors contributed to lower impact other than having a female mentor.

I believe this design can be improved if we studied as four groups:
male mentor and female mentors each with either male or female mentee and examine the differences as well as stratify for some of the confounding factors that were mentioned.

Wow.. another mind-blowing read.. The authors did what they did, I am just amazed they found a journal who was willing to publish this study, obviously aware of the immense criticism they would be subjected to... Maybe they thought they needed any sort of press- even if it is negative. It's a little difficult to keep aside emotions when reviewing this article. Surely the authors must have realized the way the world would read their paper, regardless of however good their intentions might have been while conducting the study.

I agree with Judith here... they have not specified if they controlled for the field. They have looked at authors of multiple fields that could by-default have gender differences. They have then mentioned that they approached a random sample of 2000 (out of some million)- of whom only 167 responded to the survey (which is about 8%). Its not clear how representative is this 8% and what were the characteristics of the 92% who did not respond.

I think the results here just purely reflect the advantage that males have experienced over females since forever. They have looked at studies over 100 years. Males have dominated the field of publication for so many years and still do across all fields- so if the outcome measures are based on citations and publications, how can the results of this study ever be un-biased?? Sadly it just serves as a reminder that no matter what the female gender does, there will be some who likes to dig up the past and say " well u see- it's not working... here's the twisted proof for the same".
So if I were to design a study to see the effect of gender on anything- I would start in a time period where I am confident that all genders have got an equal opportunity or get to start from the same level... and not include the period where one gender already has the high ground.

I guess there is less of epidemiology in my post today. This topic hits close to home for me, being in a surgical field (where gender inequality is the norm and the story of my career)

Fathima

I was concerned about how the exposure and the outcome were defined;
1. Exposure was “early career informal mentorship in academic collaborations”. This was defined as co-authorship on a paper within first 7 years of academic work.
2. Outcome was junior author/’protegee” performance post mentorship. The authors defined this as scientific impact of the protegee during senior years without their mentors (senior years were defined as the time period beyond 7 years after their first academic publication).

Firstly, co-authorship may not mean that you gain any form of mentorship at all so defining even an informal mentorship by co-authorship strikes me as problematic. They attempted to analyse their mentor-protegee pairs further by sampling 2000 scientists who they had identified as junior author with a survey to ask them about their relationship with their mentors but only got a response from 167 (8.35%). Not sure that you can generalize much based on the limited response rate. There may be some selection bias.

As regards the outcome, I found the time frames that they used for defining junior and senior researchers quite arbitrary. What exactly is the difference between year 7 and 8 of an academic career?

Interesting read with fairly provocative conclusions. Am surprised that Nature Communications published this paper!

Hi all,

This is a thought provoking one, and i totally agree with the thoughts shared so far.

I would say that confounding is one of the potential limitations of this study. The authors do not speak much on controlling for the potential confounding effect of several important factors as 1) Career duration 2) Any careers breaks for instance maternity leaves 3) Research field. Without having this adjusted analysis, i wonder how the they came to these controversial conclusions.

On design, i would implement a case control study with a more objective outcome measure or rather a list of criteria that would be considered as career success.

@Fathima, the struggle is real and you are not alone.

UPDATE:
https://retractionwatch.com/2020/12/21/nature-communications-retracts-much-criticized-paper-on-mentorship/