Core Problems of Gender Data Gap Challenge

We organized gender data gap challenges in health into a set of core problems. Each core problem reflects a subset of related issues central to the gender data gap challenge. From unconscious bias to data accessibility, these problems are barriers that prevent gendered data from being produced, used, and applied. Each of these core problems is an umbrella for several sub-issues, and many of these core problems intersect with each other. One core problem can also lead to the other problems. Below, we have provided the six core problems:

Social norms, patriarchal structures, and bias render gendered data invisible or ignored
Gender data is not collected consistently across domains
Lack of granularity of data across domains / data is not disaggregated
Accessibility of data is restricted due to privacy and sensitivity concerns
Data are under or mis-analyzed
Data is not applied to policy and practice, and accountability systems are not in place to ensure that data is appropriately used

Have we missed any core problems? Are there any other significant problems which hinder gendered data collection? What innovations are addressing these core problems?

Hi @Pavel, @Andrea, @clausdh, @WD_Research, @EVSwanson,
Do you’ll have any insight on Karan’s questions? You can share with the community. Thanks.

Would you look into the gender bias in designing products? Many consumer products are designed for men’s proportions for example, such as cars. Smart products that gather data may be hindered by this fact.

There is a mindset of mistrust in many, which feeds into data not being shared in the first place because of concerns about misuse.

Yes. Equally, there are concerns related to the ways in which data are not only analysed and used, but also collected (i.e. surveys that pose poorly/insensitively phrased questions, which may sound intrusive or inappropriate/trigger survivors of trauma, violence and discrimination etc).

Hundreds of millions of dollars have been wasted funding a small group of guys out of MIT to implement health data solutions. This group doesn’t understand consent - period! Without consent, there’s no foundation for future machine learning or AI.
No consent means that now, when we need to mobilize the population for cures and stratify the population for risk, we have worthless de-identified genomes to work with. To deploy consented genomes at scale is super easy, dirt cheap, and can be done by reusing mapping software hospitals, community clinics and governments have already have paid for.

Short answer. VC’s and NIH keep funding to men from MIT and the Broad who don’t want to be bothered with consent.

If you want a foundation for AI and ML to work with health data, try funding women that build solutions for consenting genomes at a global scale :slight_smile: i.e. DNA Compass

For Core issue #1, using simulated awareness training (one possible use case, researchers/clinicians undergo VR/AR simulations where they are in the body of a woman in both clinical and day to day environments) could sensitize the people who gather the data to social bias and potential gender-based tendencies (e g. women may be less likely to report symptoms, less likely to ask for a second opinion, less likely to question the doctor - and medical professionals may be more likely to dismiss women’ health issues especially for older and post- menopausal women - and therefore these data are not gathered). My examples are for women but this can similarly be applied to trans and other genders.

From my experience in gender based medical data collection projects is -
Women are not believed when it comes to degree of pain or symptoms, with these being attributed to “hysteria” or feminine weakness, when the opposite is more often true.
Test results for women are often not believed and this results in re-testing for the same attributes. I have seen multiple x-ray, CT Scans and ultrasound investigations for an ultimate diagnosis of Endometriosis, which can only be diagnosed via laparoscopy.
Diagnosis is taught in a male oriented frame where anatomical differences are glossed over.

Most of this is medical professional attitude, being that in the teacher or the student.

This generates a lack of trust and therefore data can be withheld due to that lack of trust, resulting in data pools that are gender biased, reinforcing the sterotypes.

I’d also consider this:
7. The behavioral economics of the exchange (information for services) are not often in the women’s self-interest.

Many data collection initiatives never stop to consider why a woman would want to share her data? (And not some feel good social justice reason, but true self interested reason)

For men, when there are so many drugs and solutions under research and brought to market, the answer is pretty easy: sharing information = solving their problem. The trust issue also plays a role here.

you have conceptualized the problem incorrectly. It is not lack of gender data that is the problem. It is lack of data on isolated populations, especially stigmatized populations. Why is this refocusing necessary? First, consider that isolated populations usually are low status and low economic power. Among ultra orthodox Jews women require their husband’s permission to divorce according to Jewish law. The women may be suffering higher rates of spousal abuse but the data are anecdotal since these women are isolated. In Alaska and North Dakota, although physician salaries are among the highest in the country, we have a shortage of physicians. Data on the health status of these people, men and women and children, is probably not as reliable or accurate as in highly urbanized states such as New Jersey… Data on populations that are mostly in rural areas such as Indian tribes whose members live on the reservation are probably full of gaps for similar reasons… We have a shortage of pediatricians in this country. Women are the primary caregivers of our children. Obviously if they cannot find a pediatrician the health problems of their children are under and/or inaccurately reported . we may not make any real progress on gender data gaps until we make progress on data gaps about isolated populations whether they be different ethnic backgrounds, different religions etc.,

This is a great thread. There are many aspects of lack of gendered data. I agree also with the above comment that product design has male as model and that is a concern. Something as simple as the bar to hold on a bus or metro is generally too high for most women. Medical studies are often conducted on men and the same standards applied to women.
Also on issues of violence against women, reported crime data is very inadequate in many countries as we know there is tremendous under reporting. The DHS kind of survey in the UK is a model that provides victimisation data. In India we have a similar one called the National Family Health Survey which focuses on health indicators but also domestic and intimate partner violence. We need to look at other forms as well.

One point I like to contribute: sex (and gender) is almost always a proxy for other variables which are due to biology and/or socialization more or less associated with it, e.g. physical, social or psychological characteristics.

It is important to directly measure those characteristics. As an example, think of exoskeletons. The application of exoskeletons depends on strengths and body measurements of its users but not on their sex. In practice, due to bias in design processes, the exoskeletons are regularly optimized for the average male user. So: sex and also other diversity characteristics should be treated as proxies and it should be tried to measure their effects more directly. However, sex data needs to be collected to learn more about its correlations with the more direct variables.

@AlexandraW, @Esther_Colwell, @DNAtimes, @CrazyMike2500, @CHardaker, @Vrabec, @Kalpana, @clestrie, @boblf029, @DrLiliaGiugni - Thank you so much for this insightful perspective. Your feedback is very helpful in deciding the research direction ahead.

As Kalpana notes, clinical studies in some cases are limited to men. Men don’t have the same degree of hormone fluctuation as women, which likely contributes to consistent study results. However, excluding women means that dosages are tailored for men (and probably too high for many women who tend to be physically smaller on average), and ignores the possibility that women’s responses to drugs and side effects experienced may differ at different points in their menstrual cycles. Then when women have side effects that did not previously surface in the clinical study (of men), those experiences are minimized or dismissed by their health care provider.

Good point @stephaniel. Thanks for sharing. @DrLiliaGiugni, @clestrie, @Esther_Colwell, @Vrabec . @erickson, @Tsion - We would love to hear your thoughts around any existing or potential innovations, which are trying to solve any of the core problems which we have listed or mentioned in the thread above.

@Vrabec thank you for your suggestion of a seventh core problem — indeed, the behavioral economics of the exchange are important to measure, and data rights and ownership are important areas to consider as part of this overall evaluation.