Dealing with Inequality - the Promise and Pitfalls for Social Media and AI for Prostate Cancer in Disadvantaged Populations "Presentation" - Stacy Loeb
November 15, 2024
At the 2024 Advanced Prostate Cancer Consensus Conference (APCCC), Stacy Loeb discusses the potential and challenges of social media and artificial intelligence in addressing the global surge in prostate cancer cases. The presentation addresses significant concerns including medical misinformation, poor readability of AI-generated content, and limited representation of diverse populations in online resources, emphasizing the need to ensure accuracy, comprehensibility, and equity in digital health information.
Biographies:
Stacy Loeb, MD, MSc, PhD (Hon), Urologic Oncologist, NYU, Langone Health, Manhattan Veterans Affairs, New York, NY
Biographies:
Stacy Loeb, MD, MSc, PhD (Hon), Urologic Oncologist, NYU, Langone Health, Manhattan Veterans Affairs, New York, NY
Read the Full Video Transcript
Stacy Loeb: I don't have any relevant disclosures. You've heard a wonderful introduction from Nick James earlier on the Lancet Commission. And, of course, two of the recommendations for dealing with the surge in cases include artificial intelligence and social media. So very nice section on that in this document highlighting the great potential, but also the risks. So I'll briefly highlight that here.
And one of the reasons this is so important is just because of the massive rise in use of the internet globally. These are some of the 2022 data from the World Bank showing in Latin America and sub-Saharan Africa, the rising use of the internet. But even more widespread than that are mobile cellular subscriptions. You can see these extremely high numbers. It is basically ubiquitous worldwide. And so even if we just reach people through voice or text messaging, these platforms can really be leveraged.
So this was really the conclusion of a recent review of mobile health in low- and middle-income countries, where, in fact, in some countries, mobile connections are actually more readily accessible than electricity and clean water. So these are really useful ways to do things like reminders for medications, behaviors, and also things like AI algorithms and apps are very easy to scale once they've been developed with minimal costs compared to things like increasing the number of health care providers.
Now, in high-income countries, these online networks can also be a great way to reach minoritized groups. For example, these are the latest US statistics on social media platforms. And you can see there's the highest use among Hispanic and Black US adults. So really, through the spectrum of prostate cancer care, there's many potential roles for social media and artificial intelligence, beginning with raising awareness—because low awareness is one of the main reasons for high mortality from prostate cancer.
Also, a lot of utility in screening and diagnosis, and we've heard multiple times in this meeting about AI-assisted radiology and pathology. For treatment selection, one of the big emphases in the Lancet Commission is on AI use for tailored advice based on cloud-based medical records. But these networks also allow for other opportunities, such as virtual tumor boards, which have already been used in India and in the US to help improve treatment management selection.
Another great opportunity is the use of telemedicine, and this has even been done through networks like WhatsApp. And finally, for survivorship, social networks can be very important to get advice and support from the community. These are a few examples of our work in the US using social media. For example, we did a prospective study showing that a podcast series was a useful way to provide public education about prostate cancer genetics. Also, in collaboration with the Prostate Cancer Foundation, we did a prospective study of a Facebook ad campaign to raise prostate cancer awareness.
These methods are also being used globally. For example, this is a campaign that's discussed in the Lancet Commission. It's called Project Pink Blue in Nigeria, and this is a campaign to raise awareness for cancer. As I mentioned, these networks can also be used for telemedicine. So just as an example, this is a very interesting study in Mexico using WhatsApp to provide oncology care for older adults aged 65 and above. And they were able to do geriatric assessments and chemotherapy prescriptions using WhatsApp.
Here's another interesting study out of China, where one of the very popular social networks is WeChat. And this study showed that doing pelvic floor training through WeChat resulted in improved continence after prostatectomy. So very interesting to see use in survivorship. But there is a dark side, and this was exemplified during the Ebola outbreak in 2014 in Africa, where they found that tweets with misinformation about Ebola spread way more—three times as far—as medically accurate tweets.
Unfortunately, this is also the case with prostate cancer. In fact, our group has looked at English-language content on most of the popular social platforms and found that, unfortunately, misinformation is really widespread about prostate cancer. Unfortunately, it appears that the same is true in content in other languages. As part of an R01 project, we've been looking recently at prostate cancer content in Spanish from Latin America.
And in preliminary data from 100 Spanish-language posts, we're seeing 22% with misinformation, such as this example here: three plants that can cure your cancer that the pharmacies don't want you to know about. Now, AI actually may do a little bit of a better job. So this is a study that we published earlier this year looking at the top five search queries about the top five cancers using four different AI chatbots.
And actually, here the information was pretty good quality, and we didn't see any misinformation. But there were other problems. It wasn't easily actionable for lay health consumers or easily understandable. And there may be some additional concerns. This is some unpublished preliminary data where we did the same queries using the free version of ChatGPT versus the paywalled version, and actually, the free version had significantly worse readability than the results from the paywalled version, raising concerns about health equity.
Another concern is representation of diversity. In this study, we examined the top 150 websites in a Google search and the first 150 videos about prostate cancer on YouTube. And unfortunately, of all the over 1,500 people featured in this content by a consensus panel of stakeholders, only 9% of the people were perceived as Black and 1% as Hispanic or Latinx.
And this can actually have significant ramifications. So we did focus groups with Black patients with prostate cancer in the US who said that because they don't see any Black men on the websites, they thought they were less likely to get prostate cancer, which is actually the opposite.
Same thing with videos online about prostate cancer clinical trials—Black adults are also underrepresented in this content, which is very concerning given the underrepresentation of Black patients in clinical trials. So in conclusion, there is certainly great promise from social media and artificial intelligence to help with the surge in cases by raising awareness, clinical care, and support. But there are really some significant concerns, including poor readability, misinformation, and limited representation of diversity.
So in order to really harness the benefits, we need to ensure accuracy, comprehensibility, and equity in the provision of digital information and services. Thank you so much to the groups who have supported our work and our collaborators, and for allowing me to participate in this session.
Stacy Loeb: I don't have any relevant disclosures. You've heard a wonderful introduction from Nick James earlier on the Lancet Commission. And, of course, two of the recommendations for dealing with the surge in cases include artificial intelligence and social media. So very nice section on that in this document highlighting the great potential, but also the risks. So I'll briefly highlight that here.
And one of the reasons this is so important is just because of the massive rise in use of the internet globally. These are some of the 2022 data from the World Bank showing in Latin America and sub-Saharan Africa, the rising use of the internet. But even more widespread than that are mobile cellular subscriptions. You can see these extremely high numbers. It is basically ubiquitous worldwide. And so even if we just reach people through voice or text messaging, these platforms can really be leveraged.
So this was really the conclusion of a recent review of mobile health in low- and middle-income countries, where, in fact, in some countries, mobile connections are actually more readily accessible than electricity and clean water. So these are really useful ways to do things like reminders for medications, behaviors, and also things like AI algorithms and apps are very easy to scale once they've been developed with minimal costs compared to things like increasing the number of health care providers.
Now, in high-income countries, these online networks can also be a great way to reach minoritized groups. For example, these are the latest US statistics on social media platforms. And you can see there's the highest use among Hispanic and Black US adults. So really, through the spectrum of prostate cancer care, there's many potential roles for social media and artificial intelligence, beginning with raising awareness—because low awareness is one of the main reasons for high mortality from prostate cancer.
Also, a lot of utility in screening and diagnosis, and we've heard multiple times in this meeting about AI-assisted radiology and pathology. For treatment selection, one of the big emphases in the Lancet Commission is on AI use for tailored advice based on cloud-based medical records. But these networks also allow for other opportunities, such as virtual tumor boards, which have already been used in India and in the US to help improve treatment management selection.
Another great opportunity is the use of telemedicine, and this has even been done through networks like WhatsApp. And finally, for survivorship, social networks can be very important to get advice and support from the community. These are a few examples of our work in the US using social media. For example, we did a prospective study showing that a podcast series was a useful way to provide public education about prostate cancer genetics. Also, in collaboration with the Prostate Cancer Foundation, we did a prospective study of a Facebook ad campaign to raise prostate cancer awareness.
These methods are also being used globally. For example, this is a campaign that's discussed in the Lancet Commission. It's called Project Pink Blue in Nigeria, and this is a campaign to raise awareness for cancer. As I mentioned, these networks can also be used for telemedicine. So just as an example, this is a very interesting study in Mexico using WhatsApp to provide oncology care for older adults aged 65 and above. And they were able to do geriatric assessments and chemotherapy prescriptions using WhatsApp.
Here's another interesting study out of China, where one of the very popular social networks is WeChat. And this study showed that doing pelvic floor training through WeChat resulted in improved continence after prostatectomy. So very interesting to see use in survivorship. But there is a dark side, and this was exemplified during the Ebola outbreak in 2014 in Africa, where they found that tweets with misinformation about Ebola spread way more—three times as far—as medically accurate tweets.
Unfortunately, this is also the case with prostate cancer. In fact, our group has looked at English-language content on most of the popular social platforms and found that, unfortunately, misinformation is really widespread about prostate cancer. Unfortunately, it appears that the same is true in content in other languages. As part of an R01 project, we've been looking recently at prostate cancer content in Spanish from Latin America.
And in preliminary data from 100 Spanish-language posts, we're seeing 22% with misinformation, such as this example here: three plants that can cure your cancer that the pharmacies don't want you to know about. Now, AI actually may do a little bit of a better job. So this is a study that we published earlier this year looking at the top five search queries about the top five cancers using four different AI chatbots.
And actually, here the information was pretty good quality, and we didn't see any misinformation. But there were other problems. It wasn't easily actionable for lay health consumers or easily understandable. And there may be some additional concerns. This is some unpublished preliminary data where we did the same queries using the free version of ChatGPT versus the paywalled version, and actually, the free version had significantly worse readability than the results from the paywalled version, raising concerns about health equity.
Another concern is representation of diversity. In this study, we examined the top 150 websites in a Google search and the first 150 videos about prostate cancer on YouTube. And unfortunately, of all the over 1,500 people featured in this content by a consensus panel of stakeholders, only 9% of the people were perceived as Black and 1% as Hispanic or Latinx.
And this can actually have significant ramifications. So we did focus groups with Black patients with prostate cancer in the US who said that because they don't see any Black men on the websites, they thought they were less likely to get prostate cancer, which is actually the opposite.
Same thing with videos online about prostate cancer clinical trials—Black adults are also underrepresented in this content, which is very concerning given the underrepresentation of Black patients in clinical trials. So in conclusion, there is certainly great promise from social media and artificial intelligence to help with the surge in cases by raising awareness, clinical care, and support. But there are really some significant concerns, including poor readability, misinformation, and limited representation of diversity.
So in order to really harness the benefits, we need to ensure accuracy, comprehensibility, and equity in the provision of digital information and services. Thank you so much to the groups who have supported our work and our collaborators, and for allowing me to participate in this session.