National Library of Medicine | Predicting Diabetic Retinopathy (Audio Described Version) @NLMNIH | Uploaded July 2024 | Updated October 2024, 11 hours ago.
Vision loss from diabetic retinopathy brings substantial personal, health, and financial costs. The condition poses unique challenges in medically underserved communities, where patients may have trouble meeting the American Diabetes Association’s guidelines on annual eye examinations. Additionally, patients may be unaware that they have latent retinopathy because they are not yet experiencing any symptoms. Therefore, it is important to develop methods to assist clinicians in caring for diabetic patients.
Dr. Lola Ogunyemi’s team has performed research that demonstrates the potential of certain machine learning methods for identifying otherwise-underserved patients at high risk of diabetic retinopathy. These methods use clinical data collected during the patients’ care. Refinement of the methods to improve their sensitivity and specificity will be an important next step. This work was funded by the National Library of Medicine under grant 1 R01 LM012309.
Non-AD version - youtu.be/WL4_SR3XGks
Transcript:
[Lola Ogunyemi] What if you had somebody who had diabetes and was slowly going blind and they didn't know it?
[Lauren P. Daskivich] Diabetic retinopathy is damage caused to the blood vessels of the retina by diabetes. Diabetic retinopathy is also the number one cause of blindness in working-age adults in the United States. But luckily, it is treatable and even reversible if you catch it early enough.
[Ogunyemi] Obviously, we don't want any American anywhere going blind because they have diabetes. But in the area that surrounds Charles Drew, it's called Service Planning Area 6 in Los Angeles County, we have 1.1 million people, and there are not enough health care providers
to serve the needs of the community.
[Daskivich] And so we'd basically be watching patients going blind in front of us despite our attempts to treat it. And so I knew I really saw a need that we could get to this disease earlier.
[Ogunyemi] And so one of the things that we're trying to do is to use informatics, to use AI and telehealth, to reduce those disparities, to increase access for those patients. So we applied to the NLM and proposed that we could use machine learning to figure out who was at risk for diabetic retinopathy and didn't know it and thought they were fine.
[Daskivich] Much of artificial intelligence has been focused on grading the images for diabetic
retinopathy and providing a disease result. However, the exciting piece of the work that we're doing is actually looking at non-imaging data. It's looking at established risk factors and demographic data, lab data that we may already have in the chart, and using that to develop predictive models for diabetic retinopathy to tell us who may be at highest risk for having this disease right now. And then we're able to target our limited resources for outreach on those patients that can really benefit from it.
[Ogunyemi] What we did for the project was do this in batch mode, not individually. And so we were able to look at a population of about 31,000 patients who had not been screened, applied the machine learning to the data on all 30,000. We came up with about 7,500 who were considered to be at very high risk of retinopathy by our models. And then we had bilingual health educators and research assistants reach out to them to explain why it was important to come in.
[Daskivich] And so software like this really enables us to continue to maximize our capacity by finding the patients that really need to see us the most.
[Ogunyemi] Having the NLM say, hey, let's try this. Let's look at this issue that's really impacting an underserved population right now, and let's fund this. That was amazing.
[Daskivich] This project would not have existed without the National Library of Medicine. I think it's an amazing validation of the type of innovative work that we can do here in the safety net, which is so impactful for the most vulnerable patient populations that we have in the United States.
#humanhealth #retinopathy #diabetic @charlesdrewuniversity8395 #audiodescription #audiodescriptions
Vision loss from diabetic retinopathy brings substantial personal, health, and financial costs. The condition poses unique challenges in medically underserved communities, where patients may have trouble meeting the American Diabetes Association’s guidelines on annual eye examinations. Additionally, patients may be unaware that they have latent retinopathy because they are not yet experiencing any symptoms. Therefore, it is important to develop methods to assist clinicians in caring for diabetic patients.
Dr. Lola Ogunyemi’s team has performed research that demonstrates the potential of certain machine learning methods for identifying otherwise-underserved patients at high risk of diabetic retinopathy. These methods use clinical data collected during the patients’ care. Refinement of the methods to improve their sensitivity and specificity will be an important next step. This work was funded by the National Library of Medicine under grant 1 R01 LM012309.
Non-AD version - youtu.be/WL4_SR3XGks
Transcript:
[Lola Ogunyemi] What if you had somebody who had diabetes and was slowly going blind and they didn't know it?
[Lauren P. Daskivich] Diabetic retinopathy is damage caused to the blood vessels of the retina by diabetes. Diabetic retinopathy is also the number one cause of blindness in working-age adults in the United States. But luckily, it is treatable and even reversible if you catch it early enough.
[Ogunyemi] Obviously, we don't want any American anywhere going blind because they have diabetes. But in the area that surrounds Charles Drew, it's called Service Planning Area 6 in Los Angeles County, we have 1.1 million people, and there are not enough health care providers
to serve the needs of the community.
[Daskivich] And so we'd basically be watching patients going blind in front of us despite our attempts to treat it. And so I knew I really saw a need that we could get to this disease earlier.
[Ogunyemi] And so one of the things that we're trying to do is to use informatics, to use AI and telehealth, to reduce those disparities, to increase access for those patients. So we applied to the NLM and proposed that we could use machine learning to figure out who was at risk for diabetic retinopathy and didn't know it and thought they were fine.
[Daskivich] Much of artificial intelligence has been focused on grading the images for diabetic
retinopathy and providing a disease result. However, the exciting piece of the work that we're doing is actually looking at non-imaging data. It's looking at established risk factors and demographic data, lab data that we may already have in the chart, and using that to develop predictive models for diabetic retinopathy to tell us who may be at highest risk for having this disease right now. And then we're able to target our limited resources for outreach on those patients that can really benefit from it.
[Ogunyemi] What we did for the project was do this in batch mode, not individually. And so we were able to look at a population of about 31,000 patients who had not been screened, applied the machine learning to the data on all 30,000. We came up with about 7,500 who were considered to be at very high risk of retinopathy by our models. And then we had bilingual health educators and research assistants reach out to them to explain why it was important to come in.
[Daskivich] And so software like this really enables us to continue to maximize our capacity by finding the patients that really need to see us the most.
[Ogunyemi] Having the NLM say, hey, let's try this. Let's look at this issue that's really impacting an underserved population right now, and let's fund this. That was amazing.
[Daskivich] This project would not have existed without the National Library of Medicine. I think it's an amazing validation of the type of innovative work that we can do here in the safety net, which is so impactful for the most vulnerable patient populations that we have in the United States.
#humanhealth #retinopathy #diabetic @charlesdrewuniversity8395 #audiodescription #audiodescriptions
