Leaders in Dermatology on Diversity in Research and Technology

Leaders in Dermatology on Diversity in Research and Technology

Show Video

So I like to kick off this session with a question about innovation and the pace of innovation with new tools like artificial intelligence becoming more rapidly available, and what steps must be in place to ensure that algorithms are trained with the data that represents the whole population, especially in the field of dermatology. So merriment like to see if you can answer that question first. Sarah, thanks so having us. It's a great pleasure to be here today and discuss the challenges and the opportunities to bring technology to help patients and also to be inclusive. The challenge of technology development is that it's happening so fast. I think there are days left behind when it comes to dermatology.

If you compare it with the other fields of medicine, like PET scan, MRI, MRI, C.T. scan, there are many tools that can help our doctors to be better at what they do every day. But when you look at their mythology, unfortunately, it's super archaic and it's very subjective and we need to have better tools to help it. Improving outcomes for our patients and also helping our doctors in the front lines, serving these patients globally. The I think the lack of digital data

when it comes to different skin types is really the problem that is leading to bias in building technologies and having tools to help all of our patients. And I keep mixed happy to discuss this in more detail, but I think it's basically having digital tools in these centers that can help first digitize the ecosystem Vonnie Quinn gather data and now expects from our technology tools to be fair when it comes to helping our doctors to all patients equally and things. And what are some of the valuable steps that other innovators in this space can learn from monopolizing you all? Prioritize efforts to improve diverse representation in clinical research. I think having your science is really helpful to understand the basics of the tools you are building and also understand where it's going to stay. There are the biases. I think you need to be the best advocate

for your users and patients and doctors when it comes to building tools, engineering tools to help them. So understanding where is the bias, where it is going to fail? There is the lack of enough data and they say be fair to the technology. You can expect from the technology what you trained, specifically A.I., but you trained this tool.

You can not basically just have one hundred thousand one hundred thousand images from one center or two centers and expect this tool built in this limited data set toward in real life settings. I would say look at the data. Do your research evaluated the real clinical settings. It is not about published data, just and tables of scientific papers. We know these are studies.

These have very limited scope. And then you compare it to real life settings. As a scientist, as a computer scientist, though, my piece, she was a machine gun in the competing science also. I had training in hematology, so I know it was going to fail if I don't have enough representation from those cohorts of patients in my data. So this is very, very important to understand from the beginning when you're building your tools and also to be upfront and say, I know my data is not comprehensive in these segments. So I will do risk analysis. I will do risk mitigation. And I will have a proper communication

with my users. And now that you have identified the problems, it's all about building partnerships. The doctors are really champions in these centers. Want to bring technology to life and help their patients. They're more than happy to partner with industry partners, then help.

So both my science and my industry have basically, you know, work hand in hand. And then it's helpful to understand the challenge, the situation, all the failure cases and also Bloomberg Technology with partnerships and research specifically centers who can help you identify and resolve those challenges? Thanks, Miriam Inductor Burgess. Can you share some thoughts here about the potential harm of inadequate diversity and about developing some of these new technologies? Yes, developing some of these technologies have been a little skewed in a way, and I'd like to say when we talk about it's Patrick skin type, we're not trying to trying to imply that this stands for racial, you know, ethnic groups, which more often than not it is used in that incorrect manner. What it is used for is the minimal era theme windows.

And so if we look at someone who has brown skin, say my skin color or darker, a lot of times they will identify themselves as skin color, just like Miriam's someone your skin color can still classify themselves, his skin color. So you get the false sense of that there is enough skin color in your sampling and that you don't understand or realize it's not more stratified into what color skin? Are you referring to when you say skin type for 5 and 6? There's so many different hues in that spectrum and there is a mark pigment score which is sometimes used because patients can then identify against this this classification of ten different colors. In which one are you closest to? So you're not using incorrect data to determine if a person is skin off color. And so that's one of the issues. I think most of the time when they say they have plenty of skin color in it, it's probably skin type 4 and it's less than 3 percent. So it's not even that high when you want to stratify that data. So there's a long way to go in, including people of color of all skin colors from the dark as to the lightest person.

And that's what we're finding is a big issue. An all because some of the sampling and questions you're asking is, are you skin of color? That's it doesn't go any deeper than that. So, Miriam, you as you work to bring technologies to the public. Working with regulators, can you share any insight into how the FDA prioritizes the importance of diverse representation? What should innovators now when they go to the FDA to try and get a new technology cleared or approved? It's been a great experience. And not only FDA must solve their regulatory bodies and we are working, but where they actually have this ISE their mandate. It's all about fairness and being

inclusive. And also, you cannot expect your technology work for only if you just like about every specific, for example, three skin types. It means it's going to fade if you're a user. Your doctor is not good at identifying as that. The skin takes three or four.

Right. Because you say it's not going to work for skin type for another user doesn't know what is the skin type. It's not like, you know, a real set guideline.

Fully understood. We don't have best practices. So there are many, many challenges in building tools that can actually work and in life said things again in clinical settings. And what we have seen from the FDA and other regulatory by these is having this mandate to make sure we understand and fully understand and identify those challenges. You have risk analysis, you have risk mitigation strategies. And again, basically based on the ISE, you identify. OK.

My system doesn't have enough representation from a skin type 5, five and six and a skin cancer melanoma. This it, of course, is challenging because we don't have many patients with dark skin. I mean, skin tight, five and six melanoma. But unfortunately, there are so patients who die from melanoma with a skin tight five and six. So when it comes to the representation of data, it's actually important to say there. I have fair contribution of the state in my data set.

There is my system validated what is a clinical performance and if there's not enough data, that's the reality because we don't have enough really digital information out there about skin CAC 5 6 patients. So we need to collect more than your mandate is to collect more and your mandate is to have a milestone that you say we are committing to get to this level by this time. For example, this is one of the strategies that I believe should be in place. The other one is, yes, melanoma is relatively rare, but how about other skin problems, General? Dermatology is not like a skin cancer. You have E. coli, you know, patients are even more

patients. Sometimes there are some conditions up there. So you need to make sure you have the right digitization, workflow, implementation, data gathering, building tools and data validation in place. So we have seen this mandate from all organizations requesting that technology companies either do already have that in place or commit to build it. I think this is probably one of the best moves we can see from regulators. But the good thing is that it's not not

only not ignore, but also it is in a high life. Now, I also actually I also agree with Dr. Vegas that it's not just to say it's skin color. My skin is a skin type for and I'm skin up color. Right. But it's different when you look at skin

DAX. I've been six. So we need to be more precise. We need to be basically more understanding about the impact of different pigmentation, skin types, different conditions. And this is not easy.

I'm telling you, from technology perspective, different skin conditions will have different representations during the disease life, but different skin types. But it has to start from somewhere. And one more point to add, I believe technology will be most helpful in those cases. It is not that difficult to identify RTS melanoma and you're not advanced over the same melanoma and skin type 1 patient, but it's very difficult when it comes to skin type 5 and 6. And that's where we need technology to help it.

So those are some of the conversations we've had with different glassy bodies, including the FDA. And also, again, risk analysis and risk mitigation and making sure labeling the tool is to defect to the truth that our doctors will see in the clinic. So it's not underestimating is I'm overestimating the proponents. So there is a lot of good doctors. I was just saying that there are a lot of issues that come way before we start thinking about artificial intelligence. We have to think about on, you know, the cases. I remember when I was a resident there. They said you hardly see psoriasis in black patients.

And I'm like, really? No, not that's not true. So one, it's it's getting more physicians and of color in the mix. And I think that's where you lack maybe the technology or interests. I'm not sure which one it is. But there are plenty of like Howard University is a center of skin of color. There have been tons of Kodachrome and pictures and things like that, because maybe your system won't recognize a pigmented basal cell, which is a common basal cell that we see in skin of color. It can look like a separate care tosses.

And more than 50 percent of our residents today who were training in dermatology do not know how to recognize skin cancers and certain basic dermatological issues in skin color. And so that's where we need to to make people more aware. Is our physicians because they're misdiagnosing. So if they're going to miss diagnose the case, how are they going to use a eye to substantiate it? If there's nothing there to compare it to. So. So it really has to start way back in

the chain here in that we have to get our residents up to date with. Recognizing and correctly diagnosing skin color conditions, you have a pool of all that data that goes towards a A.I. and then you can collectively put all the data together. I think, you know, when we look at the census report in 2018, I think it was fourteen point six percent blacks or African-Americans or those who identified as that. We're not talking about immigrants or

any one like that, but just black Americans, African-American, which is about 47 million people. And you're telling me we can't get photographs of different conditions in these, you know, individuals. We see, you know, we only have 3 percent skin of color dermatologists. So we have to start there and getting more dermatologists out there, getting more people into research where we can identify. We have atlases and other photography and images of these conditions.

But right now, we don't. We have probably 1 percent that we utilize in Atlas and things and and other digital images where doctors or residents can go to and say, I want to look up. Like, for instance, if you look up beauty, you will see all white images come up. So is that my beauty or what?

You know, you can Google anything and you will get a skewed look at what the definition means. So that's where we have to start in order for you to gather all the data, to put it into some type of a A.I. technology. And so do I.

Part of this conversation is also a willingness to participate in actual research. And Dr. Burgess, I was hoping you could comment on what are some of the factors at play for the underrepresentation you mentioned. Certain populations don't even meet their level in terms of the US demographics. What's at play with this

underrepresentation and what do you think can be done based on some of the conversations you have about what works to encourage people to spend time to be part of this conversation? Well, I think that when we look at the population of blacks or African-Americans in the U.S., it used to be more urban areas. And really the latest census says that a lot of people are moving towards smaller communities and so on. You know, maybe not to Kansas or Oklahoma in the Midwest, but say they used to congregate more in Atlanta. Now it's the suburbs of Atlanta. There's plenty of physicians in these areas who can be recruited to run clinical trials. I think part of the issue are the

sponsors or the pharmaceutical industry. They don't cater to skin of color. I can name maybe five researchers in the whole U.S. who are African-American and maybe five Latinos and and maybe three Asians who are at the top of their game in clinical research. They're not many of us. But one suggestion would be the sponsors to hold forums or hold many, you know, research one on one courses to train more people to become clinical researchers or principal investigators in their research. Right now, I'm conducting a study by a major pharmaceutical company, and it's only involving skin tight five and six.

Well, the reason is this, because the first time they did it and it's on TV, it's advertised on TV. And what I tell patients, would you like to enter the trial? Everyone is very happy because it's a very expensive biologic that's used for psoriasis. And they just did not include those patients in the studies. And these medications are very expensive and where most average individuals can't afford it. And definitely if you don't have health care benefits, you definitely can't afford it.

These are patients that are willing and able and will do anything to get it in a trial to clear up their psoriasis or what have you. But they may be medical on medical assistance like Medicaid or sometimes Medicare. And these are the patients that they're not going after or seeking. And that's one of the issues. It's it's who can afford care in this country and who has access. And so maybe the round out this conversation, Mary Mel, I'll start with you at the National Academy of Sciences, offered a few recommendations to encourage diversity. Tax credits for restricting development,

fast track criteria, exemption, exemption from some FDA drug application fees. What do you think about some of the recommendations that have been offer and what would you offer as well in addition to those recommendations? And Dr. Burgess, I'll close out that conversation with some thoughts from you as well. First, I couldn't agree more with Dani

Burger about the lack of training. Then it causes. It basically goes to the root of the problem. The like, if they are working on regulations and approvals, we are comparing a ISE systems that gays are human experts. If our human exports are not good at identifying the good case and treating that, it means that there will be actually a bias inherent bias there. Just as our baseline is not good. So anything in addition to that will be

a and that won't be good either. So we need more investment in the earliest stages of the ecosystem in the life of basically training and educating and chemical usage and and all of that. When it comes to CNS and other incentives to provide technology for clinical research and also after regularly blocking approvals for market authorization, I believe from my eagle view the challenge is lack of digitization.

If you don't have data, you don't have the basics of what you will have for improving the technology and building technology. So I think having incentives for implementing digital solutions in the centers that can contribute that information will be important. Having incentives, maybe more supports for centers who can recruit patients, build inclusive, basically perspectives, diverse backgrounds is important. And it's not just about skin off color. I can tell you there are many, many biases in the system when it comes to economy level. When it comes to age, when it comes to a gender, when it comes to skin color, we need to work hand in hand.

And across all those biases in the ecosystem, when we have digitization, we have enough clinical research and then hand-in-hand working with industry partners to bring technology to life. So we just need more of those programs to support. And again, it goes back to our human experience. Research scientists in those centers

working really, really hard to help these patients. They need more tools. They need more resources. If our doctors don't have time to put this simple data in their EMR, how you can expect them to have the data to build future A.I. tools. How doctors are under-resourced.

I know this is a very, very important challenge when it comes to quality of life for our doctors. They just need more support. And then that is just wrong. Yes. If it's a mandate, we need to invest in those areas. ISE competing priority. I understand that, but it has to happen.

Dr. Burgess, your final thoughts here? Yes. Well, you know, when we talk about seeing mass that we talk about the government, when we look at the Tuskegee experiment, that was government related. So I don't know how much

some individuals really trust the government when it comes to that. I think clinical research should be on a volunteer basis wanting to further the advancement of science and not something that is paid for in the instance of certain studies. They got free medication or what have you. I think education is a very important part of letting people know what it involves. I just heard something today about HDL. You know, what was it? We looked at it always is. It was the good cholesterol, but it may

not be, you know, when it comes to African-Americans. So until we know all of the information that may effect different genders, I would support the community by possibly education, helping people to get to research clinic or research sites and publicizing. I know being in Washington, D.C., we get publications on clinical dot gov, which clinical research dot gov, which anyone who wants to participate in a research can kind of just look it up.

But we're fortunate in Washington, D.C. to have a facility like that. And and but there are other universities and colleges who do a ton of research. And it's just not it's not communicated there to the community. But I really think that. Some people would love to participate if

they're getting a medication that they would never be able to afford. Some of the phase for studies are there already FDA approved. They're looking for further safety and efficacy patients to love to be involved in those. So it just depends on what aspect of

clinical research you're talking about. But I don't think you should ignore the minority communities just because of something like the Tuskegee experiment, because a lot of those people are still some elders are still around who actually remember that. But it's educating individuals and also cold heart. Visit CAC concordance. Visits with people of color, with physicians of color is very important because then the explanation that trust is is usually there versus someone who's a disc can coordinate relationship. Patient physician relationship may not

understand or may not trust that physician. So that's why we're trying to increase the numbers of minority dermatologists out there, because that's those are the patients who trust are going to be concordance relationships. Well, you both have given us a lot to think about, and I feel like we could have a part two to this conversation.

I hope we get the chance to do so. I want to thank Dr. Burgess and Dr. Grams today for the conversation, too. Thank you very much. Thank you for having me.

2022-12-08 02:46

Show Video

Other news