false
Catalog
Removing Disparities in Diabetes Care | Recording
Removing Disparities in Diabetes Care
Removing Disparities in Diabetes Care
Back to course
[Please upgrade your browser to play this video content]
Video Transcription
Hello. Welcome, everyone. Welcome to today's webinar, and thank you for being here to hear more about removing disparities in diabetes care. I'm Dr. Julie Pak, and I'm Chair-Elect of the American Diabetes Association Public Health and Epi Interest Group Leadership Team. We are so fortunate today to hear from four experts to provide insights on this topic. Thank you to our four presenters for joining us today. After a few brief announcements, we'll welcome Dr. Sarah Cromer first to discuss disparities in access to novel diabetes therapies. Then, Dr. Colette De Jong will discuss out-of-pocket costs for novel diabetes therapies under Medicare Part D. That presentation will be followed by Dr. Serena Goh's presentation, iSmile, AI-Based Individualized Social Risk Management in Type 2 Diabetes Healthcare. And our final presenter today will be Dr. Ananta Adala, who will discuss reducing disparities in pediatric diabetes, a roadmap to equitable care. Today's webinar will be recorded, and you will receive an email with the links to the recording. We encourage you to share your experiences and send questions to the presenters. Speakers will take questions as a panel at the end of the event. Please don't wait until the end, though, to send in your questions since we'll be monitoring the Q&A chat and preparing the questions for the Q&A session throughout. Instead, please type your questions there, and please be sure to use that Q&A box. And also, please feel free to use the chat box to share your experiences with other audience members. We will be using the chat box to send you links during this announcement segment. I wanted to take a moment to thank all of the members of the leadership team for their work throughout the year to provide opportunities to the interest group members. This group recently hosted a webinar on diabetes risk scores and improving access to diabetes care in underserved populations. Please see the link to the recordings in that chat or at the link on the screen. Stay updated also with the events for this interest group on the Diabetes Pro Member Forum. This is an ADA member-exclusive forum where all members of the interest group can connect. Next, I wanted to highlight other ADA webinars that are scheduled for the next month. To register for these, please go to the link in the chat and on the screen. Now I'd like to go ahead and introduce today's first presenter, Dr. Sarah Cromer. So Dr. Sarah Cromer is an assistant in medicine in endocrinology, diabetes, and metabolism at Massachusetts General Hospital. She received her medical degree from Baylor College of Medicine, completed internship and residency at New York Presbyterian Hospital, Columbia University Medical Center, and completed fellowship at Massachusetts General Hospital. Dr. Cromer's research focuses on understanding and addressing disparities in the care of patients with type 2 diabetes, including applications of epidemiologic methods to understand the ways in which race, ethnicity, and socioeconomic status impact the incidence treatment and response to treatment of type 2 diabetes. To study this, she has explored numerous observational data sets, including national health surveys, clinical trial databases, hospital and community-based biobanks, electronic health records, and insurance data. I'm going to turn it over now to Dr. Cromer. Thank you so much. Thank you very much, Dr. Paik, and thank you all of you for attending. Today we'll be talking about disparities in access to novel diabetes therapies, and in particular GLP-1 receptor agonists and SGLT2 inhibitors. I would like to highlight one of my disclosures, which is that my spouse works for a Johnson & Johnson company. Notably, their pharmaceutical group makes canin cliflizaner and buccano, which is one of the SGLT2 inhibitors, so relevant to this talk. I also receive research support from the ADA and indirectly through the NIDDK. So, as many of you likely know, the treatment paradigm for type 2 diabetes has changed drastically in the past few years after the publication of data showing a cardiovascular and renal benefit for GLP-1 agonists and SGLT2 inhibitors in patients who have type 2 diabetes and either atherosclerotic cardiovascular disease, chronic kidney disease, or heart failure. The most recent consensus statement from the American Diabetes Association and the European Association for the Study of Diabetes now recommends that these medications be used regardless of baseline vaccine control for all patients who have these indications, as shown on the left side of this diagram. However, as with many newly approved medications or new indications for medications, the uptake and implementation of recommendations is not immediate and often is rolled out in a way that is not equitable. And this has been documented for a number of diseases, but notably for cardiovascular disease. For example, disparities have been reported in the use of novel agents for the treatment or anticoagulation in patients with AFib. For example, a study in 2018 found that Black and Hispanic individuals were less likely to receive the direct acting oral anticoagulants like rivaroxaban and were more likely to receive the more cumbersome warfarin or to receive no oral anticoagulation at all. A similar study that looked at a large heart failure cohort found that Hispanic individuals were less likely to receive multiple elements of guideline-directed care, including sasubitril valsartan, mineralocorticoid receptor agonists, and gold doses of beta blockers. And these strategies address the topic of pharmacoequity, which is a field that's been advanced by the gentleman in the bottom right of this slide, Dr. Atibe Essien, and pharmacoequity has the goal of ensuring optimal medication use and access for all individuals. And so with that topic in mind, we sought to understand the current patterns of use of the novel diabetes medications, GLP-1s and SGLT-2s in people who had type 2 diabetes and also had a cardiovascular indication, either atherosclerotic cardiovascular disease or heart failure. And particularly, we wanted to understand whether there were any racial and ethnic or socioeconomic disparities in the use of these medications. Also today, we'll hope to appreciate the role of pharmacoequity as it may be helpful to narrow preexisting health disparities. And so to study this, my team and I used data from the Medicare Fee-for-Service Insurance Claims Database. For those of you who are unfamiliar, Insurance Claims Databases includes all information that is gathered by an insurance company in order to bill. So for example, a patient goes to a doctor's office or a hospital or an emergency room, a number of billing diagnoses are placed using ICD codes. And then anytime a patient goes to a pharmacy and fills a prescription, a bill is placed for that prescription. And so there is information on diagnoses and medications, but usually in these insurance claims databases, there is not information from doctors, nodes, and usually not any lab values. Using this data set, we identified 4 million people who were older adults over the age of 75 who had type 2 diabetes and who developed either incident atherosclerotic cardiovascular disease or incident heart failure during their time enrolled in this insurance data set. We excluded individuals who had end-stage kidney disease because they would not be a candidate for either GLP-1s or SGLT-2s. And we excluded those who had either previously used or were currently taking GLP-1s or SGLT-2s. And we studied this group starting in July 2016, which was a time after which both GLP-1s and SGLT-2s had strong clinical data suggesting that they had cardiovascular benefits for this group. And so what I really want to highlight here is that 100% of the patients that we studied in this study had a strong indication for GLP-1s and SGLT-2s, and none of them had firm contraindications to both medications. Our key exposure that we wanted to study was race, ethnicity, and then zip code level social deprivation index. The social deprivation index is an area or a neighborhood level measure of socioeconomic deprivation as measured by a conglomeration of a number of factors related to race, ethnicity, income, education, and other factors. We also examined a number of covariates, which could potentially be confounders in the relationship between, for example, race, ethnicity, and receipt of GLP-1 or an SGLT-2. These included demographic characteristics like age, sex, region, and year. They also included comorbidities, including a number of metabolic diseases and cardiovascular diseases, as well as relative contraindications to either GLP-1s or SGLT-2s, for example, cholethiasis for GLP-1s. And in the comorbidities group, we also looked at frailty and comorbidity indices that captured kind of a general picture of a patient's health. We also adjusted for medications that patients were using, including all other available diabetes medications and a number of cardiovascular and lipid medications. And lastly, we adjusted for healthcare utilization measures, including number of PCP endocrine or cardiology visits, number of emergency room visits or hospitalizations, and number of total medications used. And we were studying the time to risk to filling of at least one prescription for either a GLP-1 or an SGLT-2. So because this is a time-to-event analysis, our unadjusted analyses are incidence rates, and our adjusted analyses use cost-proportional hazard models. And in examining this cohort, we considered someone's start of follow-up time as the day that they were diagnosed with either atherosclerotic cardiovascular disease or incident heart failure. We looked back about 180 days prior to examine their overall health, get an idea of their comorbidities and their demographics, and to make sure they had not recently been taking SGLT-2 or a GLP-1. And then we followed them for up to 180 days into the future to monitor for whether they received one of these medications, censoring for death, disenrollment from their insurance company, or admission to a nursing home, at which point the billing data becomes much less reliable. And of the approximately 4 million people that we studied, the mean age was 75.6 years, a little over 50% of the individuals were women, and 80% were non-Hispanic white. We included people who had either atherosclerotic cardiovascular disease or CHF in the cohort, and the vast majority of them had atherosclerotic cardiovascular disease, 84%, whereas only 24% had heart failure. In raw analyses, we found that people who initiated GLP-1 receptor agonist or SGLT-2 inhibitors were more likely to be younger and more likely to be male. We also saw that the rates of receipt of these medicines increased over time with higher representation in the initiators group in 2019 than in 2016. And we also saw lower rates of prescribing among non-Hispanic black individuals. However, one of the most striking findings from our kind of raw analyses and baseline characteristics was the overall very low rate of receipt of these medications. So I'll remind you that 100% of the individuals studied had a strong indication and no firm contraindications to these medications, yet only 1.7% of them received one of these medications within six months after their diagnosis of either atherosclerotic cardiovascular disease or CHF. And to give you an idea of what this looks like in terms of incidence rates, the overall rate of initiation of one of these medications was 0.34 events per 100 person months observed. That number did go up over time, actually more than doubled between 2016 or 2018. But unfortunately, that is still nowhere near what it should be considering that 100% of these patients had a clear indication for these medicines. So overall, we're still looking at very, very low rates of prescribing. And then when we looked at analyses which were adjusted for demographics, comorbidities, medication use, and healthcare utilization, we found that there were lower rates of filling of these medications in people who were older and in women. We also saw lower rates of prescribing in non-Hispanic black and other race and ethnicity individuals. And lastly, we did see a small but significant association between living in an area of higher socioeconomic deprivation and lower rates of medication use. There were a number of other trends that were somewhat interesting in relation to who did and did not start these medications during the follow-up period. One trend that we saw was that people who had diagnoses of kind of early stages of disease like non-alcoholic cutting liver disease or hypertension had higher rates of filling, whereas those who had more advanced disease like cirrhosis, atrial fibrillation, higher frailty indices, or more frequent emergency and hospital admissions were less likely to receive the medication. We also found that those who had more complex diabetes as marked by either microvascular complications or use of any of the other diabetes medications were more likely to receive these medicines. And then we saw that people who were receiving other guideline recommended medications, for example, people who were receiving ACE inhibitors, beta blockers, and spironolactone were more likely to receive these medications, whereas people who were receiving kind of second line or, you know, less indicated medications, for example, thiazides or calcium channel blockers were less likely to receive these medications. And then lastly, both endocrine and cardiology subspecialty visits were associated with higher rates of receipt of GLP-1s or SGLT-2s. We additionally wanted to understand whether these disparities and associations with use were similar in those with atherosclerotic cardiovascular disease or CHF as compared to just the combined cohort. The most striking thing in comparing these two cohorts is that the rate of prescribing was actually much lower in the CHF cohort than in the cardiovascular disease cohort with an incidence rate of 0.13 initiations per 100 person months, which was about two-thirds lower than in the cardiovascular cohort. That said, trends we had seen in the overall cohort were similar in this group. We again saw that older individuals and women were less likely to receive these medications. Black individuals were less likely to receive these medications. And then people of other race, ethnicity, or of higher socioeconomic deprivation were less likely to receive a GLP-1 or an SGLT-2 in the cardiovascular disease cohort. And while there were trends for lower receipt of these medications in those groups in the CHF cohort, we weren't able to appreciate them, which may have to do with the fact that the cohort was being smaller and having a lower event rate. So we had lower power to detect differences in that group. Lastly, I wanted to share a little piece of good news, which is that we wanted to understand what was happening to these disparities over time, whether they were widening or narrowing. And fortunately, we did see that there was narrowing of racial and ethnic disparities in use of these medications between 2016 and 2019, both of which had a significant race-by-year interaction p-value. We also saw somewhat of a trend for narrowing of the disparities by social deprivation index, but this did not reach statistical significance. And to put our study in context, this and similar questions have been studied in a number of large databases, including commercial and VA insurance databases in the US, a primary care database in the UK, and then a large cohort from Denmark. Notably, it's been studied in populations like ours, in which although patients have a clear indication for these medicine, but it's also been studied in populations of kind of a more general group of people with type 2 diabetes, some of whom have cardiovascular indications and some of whom do not. But kind of across the board, we see overall fairly low rates of use, generally less than 10% of any eligible population is using these medications. And then racial, ethnic, and socioeconomic differences in the use of these medications pop up again and again in these cohorts, including in the international cohorts, as you can see in this graph on the right. And then the last thing, which is somewhat interesting about the way these medicines are used, is that many of these studies that looked at a general type 2 diabetes population with and without clear indications for the medicines actually found lower rates of use of these medications in those with cardiovascular or renal indications for them, suggesting that the individuals who stand to benefit the most from these medicines are actually less likely to receive them. And so what does all of this mean for health disparities? Well, because we know of a benefit from these medications, we believe that failure to treat with these medicines can be expected to result in failure to prevent major adverse cardiovascular events, or MACE, in at-risk patients. And one study used data from clinical trials to create models estimating the number of MACE events which could be prevented if all patients were treated with guideline-directed care. This study particularly studied three medication classes. They looked at statins, ACEs and ARBs, and then a combination of GLP1s and SGLT2s to understand, first of all, what percent of patients with indications for these medicines were on them, either one of them, two of them, or all three of them. And if everyone who needed them was prescribed these medications, what degree of kind of MACE improvement could we see? And so they studied a population of about 150,000 individuals with type 2 diabetes and atherosclerotic cardiovascular disease. They found that overall only about 10% of the patients were taking either a GLP1 or an SGLT2. And of the three therapies they studied, also including statins and ACEs and ARBs, they found that GLP1s and SGLT2s were the most underused therapy and therefore had the greatest potential benefit if we could bring, you know, everyone who needs these therapies into a place where they were taking these therapies. And then lastly, this group used modeling studies to estimate the number of MACE events that could be prevented if everyone who was not on these medicines was given these medicines. And they estimated that of these 150,000 people, 2,200 MACE events could be prevented over three years if everyone who needed an SGLT2 or a GLP1 was placed on one. So in conclusion, we see that significant racial, ethnic, and socioeconomic disparities already exist in the use of GLP1 agonists and SGLT2 inhibitors, including in populations with strong indications for treatment. Based on medication performance in clinical trials, we can also see that modeling suggests that these disparities in medication use will translate to disparities in clinical outcomes. By contrast, however, we believe that improving pharmacoequity may help to narrow disparities in diabetes and cardiovascular outcomes, and our study at least showed that there may be some hope for improving equity over time. With that, thank you all for coming, and I'm happy to take any questions at the end of the webinar. Great. Thank you so much, Dr. Cromer. As a reminder, as Dr. Cromer just mentioned, please submit your questions in the Q&A panel in the Zoom platform for the speakers. Next, I'd like to introduce Dr. Colette De Jong. Dr. Colette De Jong is a Clinical Fellow in Cardiovascular Medicine at the University of California, San Francisco, a graduate of Brown University and the University of California, San Francisco School of Medicine. Dr. De Jong completed an Internal Medicine Residency and Chief Residency at UCSF, as well as a two-year Editorial Fellowship at JAMA Internal Medicine and a Research Fellowship at the UCSF Center for Healthcare Value. Dr. De Jong's research focuses on expanding access to effective therapies among low-income groups with a specific interest in Medicare Part D. Her work on out-of-pocket costs for SGLT2 inhibitors, GLP-1 receptor agonists, and other novel medications has appeared in JAMA Internal Medicine and JAMA Cardiology. Thanks so much, Dr. De Jong. I'll turn it over to you now. Wonderful. Thank you so much, everybody. Thanks, Dr. Cromer, for that fantastic talk, and it's a real honor to be here with you today to talk about this important topic of improving disparities in diabetes care. Dr. Cromer presented such compelling data about the disparities in access to these highly effective novel diabetes therapies, and there's so many factors and drivers of that from prescribers writing the prescription to adherence issues, access to the medications. Now, of course, with the Zempik in the news, does the pharmacy actually have it in stock? And I want to zero in on just one aspect that could be driving some of those disparities, which is out-of-pocket costs for our Medicare patients under Medicare Part D. My spouse is employed by a medical device company in the cardiology space, and I have no other disclosures. So an outline of what I want to talk about in the next 10 to 15 minutes. I think it's useful to have a kind of bird's-eye view of Medicare Part D. It's very complicated, and hopefully getting less complicated due to some exciting regulatory changes that are coming in the next year or two. But I think it's quite relevant to providers for patients with diabetes, given how much it impacts their access to these therapies throughout the year. So I want to go through out-of-pocket costs, kind of where we are now, where we have been for the past few years. And then, like Dr. Cromer, there's actually some really good news to share, which is that the Inflation Reduction Act has made enormous changes to out-of-pocket costs for SGLT2 inhibitors, GLP-1, insulin, and other therapies. So we'll go through some of those changes and highlight some key takeaways for clinicians and pharmacists. So Medicare Part D is the voluntary prescription drug benefit for Medicare beneficiaries. There are 65 million Americans with Medicare. One in five of them have diabetes. And over three-fourths of Medicare beneficiaries elect to enroll in a prescription drug plan. And they pay a separate premium for this. They vary widely. The Part D plans are individual costs, Medicare Advantage, or standalone Part D plans that contract with Medicare, essentially, to provide this prescription drug benefit. And they're working within specific regulations and kind of bounds set by Medicare, but they operate, you know, as private individual entities. So Part D coverage, I think one thing that makes it complex is that patients move through phases of coverage throughout the year. And now, and in the past, since Part D was introduced, these have had huge implications for how much patients have to pay at the pharmacy, and it can vary a lot month to month. So basically, if I'm a patient with Medicare Part D insurance, starting in January, I'll pay a deductible, which is standardized. So in 2022, it's $480. I would then enter a standard coverage period where I pay a specific copay, like, you know, $10 for generics, $50 for brand name drugs, something like that. Once total drug costs reach a certain threshold, patients enter the notorious and infamous donut hole or coverage gap. And as soon as patients enter this phase, their costs really skyrocket to 25% of the sticker price of all of their medications. Once out-of-pocket costs reach another threshold, patients enter this catastrophic coverage phase where their costs drop to 5% of the sticker price of all their medications for the rest of the year. And I think an important thing to think about is that, you know, when Part D was first introduced, maybe an expensive medicine was $500 or $1,000. But as we'll talk about in a moment, you know, SGLT2 inhibitors, GLP-1, have become standard agents in primary care, first-line agents for diabetes, and they cost upwards of $5,000 a year. And so when you think about paying even a quarter of the sticker price of those medications for a condition as common as diabetes, chronic kidney disease, or heart failure, this program kind of topples down. It's just not built for that. And so what we found was that patients were very, very quickly moving through these phases, pay their entire deductible in January, quickly spend only a couple months in the standard coverage period and enter the donut hole, before June, and then spend a lot of time in catastrophic coverage. And so in the way that Part D was designed, there's actually no limit on how much patients can pay. For example, a patient on difamidus, a cardiovascular medication that costs $250,000 a year, you know, even if they enter catastrophic coverage earlier in the year, 5% of that sticker price can amount to enormous out-of-pocket expenditures for patients, and there was kind of no cap on it. In 2019, one and a half million beneficiaries entered this catastrophic coverage phase and spent nearly $4 billion out-of-pocket. And Dr. Cromer presented, you know, these guidelines as well. So I'll just highlight again, this is the ADA guidelines from 2016, when, you know, these drugs were kind of newer on the field and there was more aqua poise in selecting a second line agent after metformin between, you know, glipizide, pioglitazone, and some of these novel agents. This is the 2023 guidelines. And so as you can see, GLP-1 and SLT-2 have effectively become first line with metformin for, you know, cardiovascular disease, heart failure, glycemic control, weight loss. And that's thanks to the really tremendous benefits, including mortality benefits of these medications. So in this paper in Jim Internal Medicine, we projected costs for a standard diabetes regimen of metformin and ACE inhibitor and nostatin, plus one of these agents. So plus glipizide, the total price, total sticker price of all those medicines was only $250 per year. And so patients wouldn't even meet their deductible and would pay the full 250 per year. Pioglitazone, other TZDs, $355. And so if these three novel oral classes of medications, you can see that they're, you know, sometimes in excess of 30 times more expensive and patients would be paying up to eightfold increase in out-of-pocket costs in our projections. So what that means is, you know, if I was a patient who was on metformin, ACE inhibitor, and nostatin and glipizide, my total out-of-pocket costs would go from $250 to almost $2,000 when my provider changed my glipizide to Ozempic, for example. And one thing I really want to highlight is that it's not just the annual spending that I think matters to patients and therefore, you know, matters to us as providers. It's the fact that the copays can be unpredictable and variable. And that makes it really hard for patients to kind of plan for the future. And every time that I think a patient goes to the pharmacy and is charged a high, unexpected or kind of significantly increased copay, it's kind of an independent risk factor when that patient could fall off the therapy, be unable to pick up their SGLT2. And that messaging may or may not get back to the provider. And it's kind of a risk factor for self-discontinuation of these therapies that not only have huge benefit, including mortality benefit, but, you know, if you're not going to be able to afford SGLT2 and your provider doesn't know, maybe you're not on glipizide, maybe you're not even on metformin. And so it kind of is taking up space for patients when maybe they're kind of not on anything because they can't afford those costly medications. So just to underscore that again, you know, the high watermark of monthly copays is an important thing to keep in mind for our patients. And it can really drive cost-related disparities in access to these therapies. I think this is kind of well-known, I'm sure to everyone on the call and, you know, has really been on the news a lot recently, but three in 10 adults couldn't take a medicine as prescribed in the U.S. due to cost in 2019. And Medicare, like other insurance companies, uses these tiered copay structures. So, you know, $0 or up to $10 for generic, whereas for brand name drugs, it's often more, you know, 40 or even over $100 for brand name medication each month. And that serves a few purposes, you know, in part it's to actually share the cost between the insurers and the patients, and in part it's to kind of put a little bit of pressure on the patients to ask your doctor, hey, could I get a generic instead? Is there a cheaper alternative? But there's a few problems with that. You know, all of us on the call who prescribe medications or pharmacists know that it's really hard to know what the out-of-pocket requirement is gonna be when we prescribe a medication. It's almost impossible to know, I feel, without really spending a lot of extra time trying to find out. Most patients don't feel comfortable telling their doctor or their prescriber that they couldn't afford something and they left it at the pharmacy. And furthermore, a lot of these new agents, by definition, they're breakthrough, important, innovative drug classes, and they won't have generics for 10 years or 15 years from entry. And it's well-known that these high out-of-pocket costs can reduce adherence. At one diabetes center, one in four patients had difficulty taking insulin due to costs that may even be an underestimate, I think, in our anecdotal experience. And some Medicare patients qualify for this low-income subsidy, which really helps with costs, but it's only for those who make under $20,000 a year. And so you can imagine people making 25,000, you know, $40,000 a year, where these kinds of sticker prices are truly prohibitive. And of course, we could spend another hour just talking about insulin, but, you know, despite this drug being over 100 years old, thanks to, you know, these important kind of incremental improvements in the formulation of insulin, the delivery, the injectables, it kind of continues to be patented, and it also has this biologic status that has made generic entry difficult. Although I think that's going to change as California and other states are looking into manufacturing their own insulin. So finally, for the really, really good news, and I just want to underscore this so much, the Inflation Reduction Act is going to tremendously improve access to medications for our patients with Medicare. Starting in, kind of phasing in in 2024 and 2025, the IRA is going to eliminate the donut hole in catastrophic coverage, and it's going to introduce a $2,000 cap on out-of-pocket spending each year. There's also going to be a $35 cap on monthly co-pays for insulin, which has been now extended by some insulin manufacturers to commercially insured patients as well. And then starting in 2026, Medicare will be able to start negotiating the price of 10 high-cost drugs, of which it's well-expected that some SGLT2 inhibitors, some insulin, will make that list of the 10 costliest drugs in America. Things for us to keep in mind as clinicians and pharmacists is that it's not just the annual outlay, it's the monthly high watermark of costs. So even if my cap per year is $2,000, if I can't afford to pay $500 in January for a deductible, I might still have to discontinue my Jardians, for example. There's been discussion of some kind of program to help smooth these costs across the year, but that's still in development. And I think in the meantime, it's going to be important to still kind of invite conversation about monthly co-pays from our patients. You might wonder, well, gosh, who's going to pick up the tab if patients are going to be saving a tremendous amount of money? And really it's the individual Part D plans that are kind of getting a little bit squeezed by the Inflation Reduction Act. And so we need to watch out for increased premiums. We need to watch out for more restrictive formularies, although by law, they're generally required to cover at least one or two drugs in each class. It'll become even more important to find out which SGLT-2 is on formulary or which SGLT-1 is on formulary for the Part D plan that our patient happens to have out of the hundreds of thousands in the country. And so just to kind of underscore things that we can do, I think inviting conversation about costs is super important knowing that the data suggests that many patients, especially older adults, don't necessarily feel comfortable bringing that up. There's various methods to try to find out which drug is covered. It's built into some EMRs. Sometimes I'll send a test prescription and just tell patients, hey, if it costs more than X, just don't pick it up. Send me a message and we'll try a different one. Good Rx and manufacturer websites can have some great resources. And then I think as Part D plans potentially get more restrictive in their formularies, it's going to be important to kind of find ways to support patients in competitive shopping for these plans. So just to summarize, there's been enormous advances in diabetes medications. Up until now, they have really not been well covered at all for our Medicare patients. But fortunately, the Inflation Reduction Act has some extremely exciting improvements, including a $2,000 cap on spending and a $35 cap on out-of-pocket costs for insulin. And it's important for us to continue inviting patients to talk to us about costs so that we can be in the loop if a patient's monthly costs go way up unexpectedly and so we can adapt. So thank you so much to Dr. Pack, Dr. Cromer, Nikki, Logan, and everyone for having me. And we'll look forward to your questions at the end. Thank you. Great. Thank you so much, Dr. DeJong for those insights. Next, I'd like to welcome Dr. Serena Goh, who is an assistant professor in the Department of Pharmaceutical Outcomes and Policy at the University of Florida. She conducts pharmaco-epidemiologic and pharmaceutical health services research focusing on diabetes, with the goal of promoting equity in diabetes treatment, care, and outcomes. Her research draws on real-world data, example, electronic health records and insurance claims data, and advanced analytics, such as AI, machine learning, and causal principle modeling to assess comparative effectiveness and safety and heterogeneous treatment effects of diabetes treatments in social risk interventions and develop individualized and intelligent social risk management tools to be integrated into diabetes healthcare. Thanks so much, Dr. Goh. I'm going to turn it over to you now. Thank you very much, Dr. Pack, for the very nice introduction. I'm Serena Goh, assistant professor in the Department of Pharmaceutical Outcomes and Policy at the University of Florida. I'm very excited to have the opportunity to share our research through the ADA webinar, Removing Disparities in Diabetes Care. My presentation today, it's about a tool we are in the process of development called iSmile, individualized intelligent social risk management in type 2 diabetes healthcare. Yes. So before the lecture, I would like to disclose that I received consulting fee from Pfizer outside the presentation and the relevant work. And my research is supported by several federal agencies and pharma foundation. So the story that I'm going to tell today, it's about the development of a social risk management tool for diabetes care. Named iSmile, as I mentioned, the storyline include why we developed iSmile, how we followed the AHRQ's learning health systems framework to build the data capacity, develop the machine learning models, and how we clinically interpreted of our AI machine learning model results in the context of type 2 diabetes care, as well as the public health implications of the iSmile tool. So it has been very well documented that type 2 diabetes and its complications disproportionately affect racial and ethnic minority groups, as well as socioeconomically disadvantaged communities. We all know that social determinants of health are conditions where people live, which have been recognized as the root causes of type 2 diabetes disparities. In fact, SDOH accounts for over 80% of modifiable factors that influence the health outcomes of a person with diabetes. With this context, we think that type 2 diabetes is a public health crisis that must be managed beyond traditional medical care. We have to effectively and efficiently to address patients' unmet social needs in order to ultimately improve health equity and health outcomes in millions of Americans living with diabetes. So the importance of social determinants of health has been very well recognized. Many hospitals and practices have started incorporating social determinants of health surveys in their clinical workflows to screen for patients' unmet social needs. However, there has been an extremely low uptake these existing social risk screening tools. According to data from several healthcare systems, that have adopted SDOH survey or SDOH module in their EHR system, the overall uptake rate was only less than 2%. So why the uptake is that low? We think we may obtain some insights from the survey conducted by American Physicians Foundation last year in 2022. So nearly all physicians indicated their patients' health outcomes are affected by at least one social determinants of health. However, and also the 80% believe that US cannot improve the health outcomes or reduce the healthcare costs without addressing social determinants of health. However, nearly 80% of physicians indicated they had limited time during the patient visit to address social determinants of health. Over 85% indicated sufficient workforce to facilitate the patient navigation. What is more striking that even over half of the physicians reported SDOH challenges make themselves to feel stressed and frustration in their daily work of taking care of patients. So overall, we identified two major challenges in incorporating social risk management in clinical care of type 2 diabetes. First, existing screening tool are not automated, making them very difficult to adapt to the already very busy and overwhelmed clinical workflows. Second, there are very limited and unclear information regarding channels and means to act on social risks, even clinicians can identify. And the fact is that the existing evidence showing that mixed or modest, or even non-effectiveness of several SDOH targeted interventions. So our development of iSmile is to address these above mentioned challenges. We proposed a structured approach to integrate social care into type 2 diabetes healthcare. We think that upstream social determinants of health play a more fundamental role that they can not only shape the downstream medical determinants of health as it's seen in this bloom. More importantly, we believe those blue edges and the green part of the bloom, the social determinants of health present important opportunities for improving not only health, but also improving health equities in type 2 diabetes care. So, as I mentioned, our research followed the HRQ's learning health system framework. Our first task is to build data capacity. We first integrate two layers of social determinants of health data into the EHR database, including neighborhood level or contextual level social determinants of health. We then use NLP to extract the person level social determinants of health or individual level social determinants of health. The second task is to divide machine learning models with the SQH EHR integrated data. We apply prediction and inference models to use for screening the patient's MS social needs and identify the effective target for intervention. And the last task is to put everything together into the ISML platform. So, the current study we conducted use the UF Health EHR data is one site of one Florida plus network. One Florida is a statewide EHR database, including total 19 million residents. Most of them from Florida and about two million were in residence in Georgia and the 9,000 from Alabama. So the mainstream data from Florida Plus is electronic health record. And we conducted the linkage with insurance claims data, best index, some registry data and immunization data. We have a total of 13 data partners. The UF Health EHR data is one of the largest data partners in Florida Plus. So in the first step of building data capacity, we did the spatial temporal linkage of contextual social determinants health information to the EHR database using the historical residential information from the individual patients. And for the, excuse me, extracting the person level social determinants of health information such as the financial constraints, employed information, housing, food information. We use NLP natural language processing to extract the clinical notes and conduct the data harmonization between the unstructured data and the structured data and to harmonize and come up the list of individualized, individual level social determinants information. So this is our study design. We included patients diagnosed with type two diabetes in the UF Health EHR data. The study outcome was hospitalization in the follow-up year. And the index state was selected through a random selection of outpatient visit. We collected all the baseline information, including contextual and the person level social determinants of health prior to the index state. So regarding the model development, we first study about polysocial risk score to use machine learning to generate, develop a prediction model, incorporating both contextual and the person level SDOH information. It was used for identifying patient at high social risk. And the second model development task was about a causal effect estimation or causal inference. We use causal AI modeling such as doubly robust causal forest to estimate the counterfactual effect SDOH means what if we address those actionable SDOH items, how the outcome risk going to be modified for the patients. This is the preliminary result for our polysocial risk score in the UF Health PEP2 cohort. We include about 11,000 patients and the statistics for the polysocial risk score was between 0.62 and 0.65. And we're predicting the three months, six months and the one year follow up hospitalization risk. So a note here for this polysocial risk score, we only include the social determinants health information without any individual level demographic or clinical characteristics. So we put the polysocial risk score in a regression model after adjusting for those individual level demographic and the clinical characteristics. The polysocial risk score explained 22%, 19% and 36% of increased risk was three months, six months and one year hospitalization respectively. In this bar chart, the x-axis is the predicted risk score generated by the polysocial risk score. The y-axis is extra hospitalization risk. So in the top 5%, one in three people would be hospitalized in the top 5% of the polysocial risk score while the POTEM5 dye cells hospitalization rate were 10% or lower, suggesting a relatively good calibration performance of this polysocial risk score. So these are results from the causal estimation of social determinants of health effect. Housing instability has been identified as a key risk factor predicting hospitalization in the polysocial risk score model. So we estimated it's counterfactual effect. What if we address the housing insecurity? The left graph shows patient stratification and heterogeneous effect across the subgroups. Based on our model, patient without insurance would bear a higher risk of hospitalization caused by housing insecurity compared to those who had insurance. On the right hand, it shows the variation of individualized effect of housing instability on the hospitalization risk, which means that if we put intervention targeting on addressing patient's housing insecurity, it will be more effective with some patients compared to others. Then we put everything together into the iSmile. This is a initial user interface for iSmile prototype developed based on a patient case, a female, 53 years old, white, Hispanic, three years of type two diabetes, medical history of diabetes, retinopathy, hypertension, obesity. So we applied to the polysocial risk score. This is the top five important predictors all social determinants of health related. According to the polysocial risk score, the three months hospitalization risk was 92. We categorize into high risk. And then we further break down the total hospitalization risk to social and clinical risk. It turned out social risk explained about one third of the total hospitalization risk. Then we re-rank the five important predictors based on their causal effect, because we all know that the strong predictors that's not necessarily causally related to the outcome. After re-ranking, then the next step, our work would be designed a social care hub to navigate patients to the available intervention resources and programs to address social determinants of health. So in summary, the iSmile information can be used to support the shared decision-making between the care team and patient on personalized intervention strategies to manage their social risks. The project has established the methodological framework and generated the real world evidence for effective social risk management in clinical care for type two diabetes. We believe iSmile has the potential to improve health and equity by integrating social care into healthcare leading to a necessary paradigm shift in the US healthcare delivery. So in the very end, I want to acknowledge our amazingly collaborative supportive study team, Dr. Jiang Bin, expert in data science from Florida, Dr. Shengmen, implementation science expert to lead the implementation of the iSmile, Dr. Donna Hu, endocrinologist focusing on type two diabetes care, Dr. Wu, NLP expert to help us to extract the most important STLH information from HR, Dr. Laurie Bilaylo, health services researcher focusing on health disparities. So that's it, thank you very much. Great, thank you so much, Dr. Guo for a very interesting talk just illustrating how advanced analytic methods can be applied to address disparities in diabetes care. Our final presenter today is Dr. Ananta Adala. Dr. Adala is a pediatric endocrinologist and physician scientist at Stanford University addressing disparities in pediatric type one diabetes management and outcomes. As a physician with a background in pediatric endocrinology epidemiology and behavioral health, she aims to build an evidence-based approach to addressing type one diabetes disparities by systematically evaluating youth, family, provider and system level barriers to optimal diabetes care in youth from low socioeconomic and racial ethnic minority groups. She is funded by the NIDDK K23 award to understand and address disparities in pediatric diabetes technology access and utilization. She also is funded from the Maternal Child Research Institute and Helmsley Charitable Trust. To date, her publications have demonstrated that the disparities in pediatric type one diabetes by socioeconomic status are worsening in the US, provider bias against public insurance is common and public insurance mediated interruptions to diabetes technology adversely impact glycemic outcomes. She has also been leading the efforts to improve justice, equity, diversity and inclusion in research at Stanford University through her leadership at Stanford Pediatrics Advancing Anti-Racism Coalition and as the co-chair of TrialNet's Underrepresented Minorities Outreach Committee. We're so fortunate to have Dr. Adala here. I'm gonna turn it over to her now. Thanks so much. Thank you so much for having me and for that very kind, warm introduction. And I'm going to talk today with the title of Reducing Disparities in Pediatric Diabetes and a Roadmap to Equitable Care. Now I know not everybody is a pediatrician out there but what I hope to actually outline here is the roadmap, one that I think will be applicable irrespective of what field you are talking and what field you'll be working in. And so with that, I will start with my disclosures. I have no conflicts of interest, but as stated earlier I do have research support from the NIDDK, the Maternal Child Health Research Institute at Stanford and the Helmsley Charitable Trust. What we will be talking about today is really to try to understand what are the sort of health factors that really put people at risk for inequitable access to diabetes care. We're gonna talk through some of those social determinants of health and talk about how it directly impacts diabetes outcomes. I'm going to use diabetes technology, again, as an archetype or as an example of how to study and address inequities. And I will also hope to give you some opportunities for what solutions might look like. So how will I do that? I always start with a couple of foundational principles just to ground and orient our conversation around disparities. Then we'll head into the, again, very briefly because I think the state of disparities is quite well detailed. And so I'll try to focus the bulk of my time really on the drivers and solutions so we can think about where action can go next. So starting with the foundational principles, what we have is every time I talk about disparities I have an outright first and foremost conversation in the United States, what the relationship between race, ethnicity and socioeconomic looks like, socioeconomics look like. Part of the reason is because as you can see in this 2021 US census data, where you have on the Y-axis, the mean annual income and then on the X-axis time in years. And just so you know, the gray areas, those longitudinal bars there, that they represent sort of recession so that you can know what the broader trends are. And what you can see is while mean household income has increased over the years, there's a very clear separation by race, ethnicity. So there certainly is a relationship between race, ethnicity and socioeconomics in the United States. But to stop there or to blame everything on socioeconomics alone would be both inaccurate and is in fact quite harmful because there are many studies that have shown that irrespective of an individual socioeconomic status, their race, ethnicity puts them at higher risk for adverse outcomes that is not biologically driven and is in fact entirely driven by the social construct of how we approach race. I take this opportunity to also just remind everyone while we're looking at census data on race, ethnicity, that the way that race, ethnicity was partitioned and created in the United States was somewhat arbitrary and was not based on biological underpinnings. So it's really important that every time I'm talking about disparities by race, ethnicity, I want you to know I'm coming at it from the social construct because that's really what we are choosing to measure even though we've attributed other aspects to it. The second is a really basic definition on what really is a disparity. And this most straightforward definition that I've ever found and I really anchor on this actively when thinking about disparities is disparities are at that intersection of efficacious treatment and inequitable access. So if you have inequitable access of an inefficacious treatment, we might not necessarily call that a disparity and equally if you have an efficacious treatment and everyone is getting it, then again, there's not a disparity there, but it's where those two components overlap that we really want to hone in and think about what equitable distribution means. And then finally, I will be front and center. I always say, I try not to bury the lead, which is what is that roadmap to equitable access and what is that roadmap to equity? And the very first step is recognition of disparities. And I think that the fact that this webinar exists that all of you are attending, these are all indications that there is a social recognition and a cultural recognition now, particularly driven forward in almost a quantum leap, both by the COVID-19 pandemic and America's national call for racial justice. The second is to really understand what is the scope of those inequities? What do they look like? What do they look like in my field? What do they look like in my research area? And the third is to evaluate what are those drivers? Because until then, we can't really think about where the actionable areas are. Next, we take those areas of disparities and we ask the individuals and the communities that are most impacted by these disparities to help us as researchers, as clinicians, as individuals who care, to prioritize what is most important so that we can then intervene and come up with an intervention that is really meant to address those drivers in a stakeholder-driven fashion, which can then finally be disseminated and scaled. Now, we're not quite there on the whole roadmap and that's why we're still walking the path, but I will focus the first portion of this talk on really detailing these three things. Where are we? So the state of disparity is really, I will break it down. While an individual can be marginalized in multiple identities, I'm going to focus specifically on race, ethnicity and socioeconomic status with the lens of diabetes technology. So starting with race, ethnicity, again, as an American social construct, what I am presenting here is 2021 data from the Type 1 Diabetes Exchange that is looking at the mean hemoglobin A1C differences by individuals who are non-Hispanic white in the gray listed first, non-Hispanic black listed second, Latinx or Hispanic listed third, and other or unknown because of the fact that many people do not necessarily readily identify in the existing categories. And what you can see is individuals who self-identify as non-Hispanic black have the highest A1Cs and those who identify as non-Hispanic white have the lowest. And there is this gradation that you can appreciate. Going hand in hand with this, because I did tell you I was going to talk about diabetes technology, is that in the reddish color there is CGM use and in the yellow is pump use. And what you can see in the same paper is they looked at if you were likely to be looking at the non-Hispanic white, non-Hispanic black and Hispanic individuals, you can appreciate that pump use and CGM use as highest in non-Hispanic white individuals. Now there is this other category which includes unknown and it's a little bit catch all and there is some heterogeneity there, but in those individuals who have identified themselves as one racial ethnic group, you can see that there is variations in diabetes technology utilization. Now transitioning to socioeconomics because that is often how our most vulnerable populations get access, it's through insurance as has been discussed. When looking at insulin pumps, I'm going to orient you to this graph because this type of a graph is going to repeat a few times over. And what you have here in the light blue is earlier data, 2010 to 2012, and in the dark blue is 2016 to 2018. This is pediatric data specifically and we looked at about 16,500 youth from the Type 1 Diabetes Exchange. And we gave each of them a socioeconomic quintile. We said, and that was an equal weighted average of the educational level, the highest educational level of either parent, insurance type and annual income. And what we looked and saw was how did those pump use in this first panel change by socioeconomic status? And there are two take home points here. The first is that between the earlier and later time points, pump use increased across all sociodemographic strata, but you can appreciate that there is a strong socioeconomic relationship where the lows of the lowest socioeconomic status had the lowest pump use. We saw something very similar for CGM, many of the individuals on this call will know that a 2010 CGM was not quite the same as a 2018 CGM where by the time we got to 2018, there was often fewer calibrations, greater accuracy and sort of an overall improved user friendliness. But what you can hopefully appreciate again here, the Y-axis changes to CGM use, the X-axis stays the same, but you can appreciate that the increase in diabetes technology use when it comes to CGM was preferentially largest in those at the highest socioeconomic status. And again, coming back to our definition of what a disparity is, there appears to be inefficacious access and there's a large body of data that talks about how the inequitable access is actually going to start resulting in disparities. And the way that we see it here in the same cohort, we did an analysis looking at just the A1C, again, the highest socioeconomic status here and the lowest socioeconomic status there. And then now the Y-axis has changed to hemoglobin A1C. You can, there's sort of two take-home points. The first is that A1C has actually increased across all socioeconomic strata between the earlier and later time point, which is not the direction that we wanna be heading in. But what you can hopefully also appreciate is that rise in A1C is actually the largest in those youth from the lowest socioeconomic status. And what we thought was that, and you can appreciate this from the actual widening of that slope. And when you control for diabetes technology, you see that that interval worsening doesn't act, it actually goes away. Those essentially become parallel lines. So hopefully what I've done for you here is lay some of the groundwork that says, what a disparity is. Think about how we're talking about race, ethnicity, and then the fact that diabetes technology access specifically seems to be one driver of adverse outcomes. And while we're talking about drivers, let's move on to our next area, which is really what are those drivers and what are some solutions for them? I'm going to focus on technology access, public payers as one representation of the structural determinants of health. And then I'm also going to present some data on provider bias. Because when I think about closing the hemoglobin A1C gap that we've observed essentially for many decades now, to me, there's sort of four key aspects, which is addressing provider bias, ensuring technology access as a modifiable risk factor for adverse health outcomes in diabetes, both A1C and psychosocial, addressing the payer coverage gap. And then one of the root cause analyses that we've come to is actually a lack of diversity in clinical trials appears to be a key reason why these system level changes have not been favorable for minoritized individuals. And in order to do that, I'm going to try to do this in a solution on a data-driven fashion. And the first is I will present the gatekeeper study, which is looking at technology access and provider bias. Next, I'll present the 4T study, which is a local Stanford study that aims to give everybody technology, specifically continuous glucose monitoring within the first 30 days of diabetes diagnosis, type 1 diabetes diagnosis. And the third is the BT1D study, which is really looking to build the evidence to address disparities. So starting with the Gatekeeper study, now these data have been published and the goal of the Gatekeeper study was to really address and assess what the role of provider implicit bias is on diabetes technology recommendation. And the way that we did that was through clinical vignettes and ranking exercises and an a priori definition of what it means for a provider to have diabetes bias against public payers for recommending diabetes technology. The entire study was designed and all definitions were finalized before we actually rolled the intervention out because of the nature of trying to study implicit bias, it's really imperative that we're very concrete on the front end. And to kind of cut to the chase, the results that we found was that the longer that a provider was in practice specifically, so not their age, but the longer they were in that system of trying to get healthcare coverage for their patients, doing prior arts and engaging in a system that tends to be pretty challenging as our prior speakers have discussed, the longer they're in that system and longer they're in practice, the more likely they are to have bias against individuals of public insurance. And that's a really helpful thing to know because there's a few things as was discussed, there are changes on the horizon for healthcare policies and the coverage now is much better than it used to be even five years ago. And then second, it does also mean that this is something that is almost programmed and conditioned is our hypothesis by the way that the healthcare delivery system is. So if we can simplify the healthcare delivery, the payer system, then there's a potential that we won't predispose providers to having implicit bias against public payers or individuals who are publicly insured. Now we'd expanded this using the type one diabetes exchange data. And what we found was a re-demonstration of having insurance mediated bias the longer you were in practice. And then very interestingly, now this was done in the last year. So this is in the moment when many of us have had implicit bias training. Many of us have reached out to try to understand what the nature of disparities are by race ethnicity. And very interestingly, those individuals who could most readily say, I recognize my own racial bias, we're actually at the highest risk for having or sort of behaving with respect to diabetes technology or selecting options with respect to diabetes technology that puts them at risk for racial bias. So the thing I will say is patients are very aware of this. And there are two qualitative studies I will bring in. This one has already been published looking at both adult and pediatric patients. And the patients, they say, I would prefer to be on an insulin pump because I think it would help me control my diabetes better. But because I've been deemed uncontrollable, chances I can't get one. And then other people feeling very judged, I went to ask for a CGM and I'm gonna cut to the bottom here where the provider said, are you too lazy to check your glucoses? And that's why you wanna... So the reason why I bring the quotes in is because it's not just something we're doing without consequence as clinicians, but it's something that also is being noted and heard loud and clear. So what do we do to undo the reach of bias? The first component is absolutely education. I know I said that educating oneself is just because we have had education doesn't necessarily protect us, but that is absolutely a starting step. The next is automation. And I'm gonna present some data from the next study to really say that removing that clinical call, that judgment, that let me... Actually is very helpful in removing implicit biases specifically because they're just that, they're implicit. And then reevaluation of your existing infrastructure to make sure that there is active changes and updates to counter systemic racism. So what do I mean by automation? I'm gonna use the 4T study as an example. Part of why as a disparities researcher, I appreciate the study is because the underlying goal of this study was not actually to address disparities. It's just, there were some principles that were built into it that also addressed it while we were at it. So the goal of the study is actually to show that CGM utilization within or initiation within the first month of type 1 diabetes diagnosis in a pediatric population is feasible and acceptable. And part of how we did that is we started the CGM. We looked at person reported outcomes throughout the study period. And then there was an algorithm to help make sure that there was proactive insulin changes that were happening. For those individuals who were interested and eligible, there was also an exercise component. But this study was designed specifically to address technology access. So what we did was even if insurers or payers didn't cover or there were gaps in coverage, we ensured that there was a year of uninterrupted CGM access which is associated with overall improvements. The second was also just making sure that everybody got that technology. Every single individual youth who was diagnosed with type 1 was approached. There was no judgment on whether the family could or couldn't handle it. And the word judgment I use there intentionally because enrollment into clinical trials is often somewhat mediated by how providers think or clinicians think people are ready. So in order to address these three components and what did we find? The take-home message was that really universal and uninterrupted access to CGM, it did help all individuals. So I'm gonna show you two paired bar graphs that essentially one looks at by insurance status and the other looks by race, ethnicity. And what we found is the Y axis here is the percentage of youth who are meeting criteria. The X axis is the month since type 1 diabetes diagnosis. The gray bars are our historical because we compared to an older cohort and then the blue bars are our study pilot population. Those in the, lots to clarify here. So those in the darker here are those individuals who hold inherent privilege either because they are non-Hispanic or are privately insured. Those are in the lighter bars, those in the light gray and the light blue are either Hispanic or publicly insured. And what you can see is I'm gonna kind of narrow you into the 12-month data here, is that individuals who had public insurance by 12 months or identified as Hispanic by 12 months did better than their counterparts. But as you are, as the astute eye in here notices, there is an improvement, but not an entire elimination. So what we saw was that everybody had an improvement in their overall glycemic outcomes, which is really counter to what I presented to you regarding the type 1 diabetes exchange data, but it didn't result in a complete narrowing and closing of the gaps, which really makes sense because technology access is just one component of what it means to have overall improvement. The next study that I'm going to take you to is called, it's called Building the Evidence to Address Disparities in Type 1 Diabetes. This is a very early pilot that is funded both by the NIDDK as part of my K23 and through the Structural Racism Grant that the Stanford Maternal Child Research Institute has given me. And the goal of this was to actually flip provider clinician bias and discrimination and understand in a detailed and targeted fashion what the family's perception is and what their experience is. The next was to think about how this impacts diabetes technology access. And the third is to just demonstrate that inclusion into clinical trials is feasible for minoritized population. The broad study structure is that there were surveys, focus groups, and advisory boards that we engage these parents of kids who are less than 12 years old. And the first really exciting message is that we found that by hiring a linguistically and culturally congruent research coordinator, we were able to have recruitment goals that were just, I'm gonna call them like a superhero. So we use the superhero because I'm a pediatrician first and foremost. So we actually exceeded our recruitment goals by quite a bit. So we over-recruited for our survey, for our interviews, and for our focus groups. The next thing was we had a very diverse cohort. 100% of our youth, of our parents were publicly insured. 69% made less than 50K, which in the Bay Area, which is where I am, is a very, very modest annual income. 72% identified as Hispanic or Latino, and 60% were Spanish speaking. And we had about 96% survey completion rate and very, very rich focus group and advisory board engagement. So this slide is just really to let you know that it's very doable, particularly when we adopt principles of research equity. The findings were very interesting in that those individuals who were Hispanic or Latinx and preferred to speak English rather than Spanish had a very unique and distinct pattern of how they reported their surveys. The surveys looked across four domains. When it came to discrimination, those individuals who identified as Hispanic or Latinx who preferred English were far more likely to endorse discrimination than any other subgroup. And in fact, those who identified as Hispanic and said Spanish was preferred actually endorsed the least. Childhood trauma, which was measured by the ACEs, tended to be highest than the non-Hispanic English preferred individuals. And interestingly, technology acceptance in psychosocial states and health was lowest in those individuals who identified as Hispanic with English as their preference as compared to those individuals. And for example, who were Hispanic and Spanish is preferred endorsed the highest. So this is very important because we tend to overgeneralize by either language or by ethnicity, but in fact, there's a lot more heterogeneity that's worth us knowing. These data will be presented soon in more numerical detail at the upcoming ADA sessions. And then the other thing from the focus groups that was really fascinating was many individuals would start with saying, I don't know that I can call it discrimination, but... And then they would detail very clearly an experience that many of us would understand to be discrimination. And these occurred when you look on the social ecological model across the entire spectrum of what it means to engage in healthcare. For two closing kind of quotes, as far as what happens, these participants said they judge you in all honesty because I noticed a complete shift in the way I was treated when I got out of my pajamas and dressed in the way that I normally dress. Then all of a sudden, a whole nother level of communication and respect comes into play. And importantly, as a solution, they identified the community as a key solution. Now, we coded our groups in whatever language they spoke, but it is translated for those of you who do not speak Spanish that I do think the community would help a whole lot because I don't know anyone who has diabetes at all and I don't know where I can get the information from. So what I've started here is we've talked through the first three steps and what we're really working on now is about taking from the evaluation of the drivers of those disparities to really prioritizing those stakeholder intervention, which is really where our next steps are moving forward. So with that, I'm going to take a moment to really acknowledge my funders, but also the fantastic team that really helps all of this research happen. And I will come off of screen share and transition over to the question and answer session. Thank you, everyone. Wonderful. Thank you so much, Dr. Adala, for a very interesting talk about such an important topic and also providing a wonderful framework for all of us to understand how to approach the understanding of disparities. And thank you to all of our presenters. So right now, we've got about nine minutes left for the Q&A session. So I'm going to start by just reading some of the questions. The first question is for Dr. Cromer. And the question is from Yanbo Zheng asking about first, you know, the Medicare and why analysis within elders. And also if you could just share some ideas about what you think the causes are with the inequity in medication use. Absolutely. So to the first question about why Medicare and why an older population, this topic has been studied in many, many commercial insurance databases, but we prefer to use Medicare because it's a little more universal in the U.S. that everyone, once they're over 65, almost everyone goes on to Medicare. And so there's a good amount of diversity there, especially socioeconomic diversity. In terms of isolating to age greater than 65, we did that because while the majority of people on Medicare are there because they're age greater than 65, there's a small number of people who are there for other reasons like end-stage kidney disease, ALS, or certain disabilities. And so we wanted to target a kind of more homogeneous population that represents classic Medicare. In terms of the reasons behind these disparities, I think some of my other panelists did an excellent job covering some of these. You know, frequently racial and socioeconomic disparities are blamed on socioeconomic status, but as Dr. Adala mentioned, that's, you know, just one very small piece of the many reasons that there are racial and ethnic disparities. And in our study in particular, we adjusted for race and ethnicity and socioeconomic status in the same model. So they have independent, you know, contributions to this disparity. And after that, you know, we've heard concerns about provider bias, implicit bias, and kind of the gatekeeper effect. I think that's certainly a factor. I think their, you know, language is often a factor that if you are worried, for example, for a GLP-1, there's a lot of counseling that has to happen when you start. You know, we started a low dose, we're gonna go up slowly. We're gonna make sure you're not having any side effects. You may not see effects immediately, but once we get to a good dose, then you'll see them. And if you are pressed for time using an interpreter, or if you're just not comfortable with how that conversation is gonna go in another language, then maybe, you know, implicit bias, you end up not prescribing that medicine for reasons like that. So I think there's many, many factors that go into it, but those are just a few. Wonderful. Thank you. Thank you for the response. The next question is for Dr. DeJong. Are Part D plans transparent about what they cover and what the costs are? And also, if you can just share any final thoughts for this talk. Yeah, I think that's a wonderful question. I think they are transparent. The problem is, you know, with so many different individual Part D plans all over the country, it can be really hard for providers to, you know, providers or patients to kind of go to the right website and look up what tier their drug is and what the expected co-pays will be. There are websites, like there's one called eNavi, E-N-A-V-V-I, where providers can go to the website and kind of quickly look up their patient's Part D plan and what tier the various SGLT2s are, for example. And some medical records have integrated information like that. I think it's gonna get easier with the IRA policy changes because it's really simplifying a lot of aspects of Part D coverage. And so you don't have to worry quite so much about which specific phase your patient is in. They'll pay the deductible and then they'll be in the coverage phase. And when they hit $2,000, they'll kind of cap out at out-of-pocket costs. Great, thank you for the clarification and explanation. I'm gonna turn it over to Dr. Cromer to moderate any final questions for the last speakers. Yeah, so for Dr. Guo, there were a number of questions about the polysocial risk score. I'm gonna give them to you one by one. And the first was, when developing that score, could you consider the effects of age? Because age, you know, elders may have more disadvantaged social factors, but they also have higher risk of hospitalization because of their age. I know you mentioned that you adjusted for age in some of your models, but can you comment on how it went into your polysocial score? Yeah, thank you very much. That's a great question. So in the development of the polysocial risk score, we conducted a series of sensitivity analysis. The base model only include a contextual level and the individual level of social determinants of health. Actually, for the prediction utility, after we added age, the overall prediction utility didn't really change that much. And that's our final decision, and it's to remove it from the model because it really, yeah, from the perspective of the polysocial risk score, it didn't improve the utility for, but after we developed the score, we put it in a logistic regression, like adjusting for age is, so the question addressed age, older patients have a higher risk of hospitalization and they may have different portfolio of social determinants of health, and we adjusted for age, risk ethnicity, and wide range of clinical characteristics to examine how much the polysocial risk score can explain the additional risk on the top of those risk factors. And then one other question about the score. Someone asked, are all disadvantaged social factors associated with positive coefficients, so increased risk of hospitalization, or were any of them protective? Or any of them protective? So, we include about, finally, we include about 200, 200 contextual liable factors, yeah. And I don't remember the exact number, and about 16 individual social determinants of health, and our final model really didn't include anything that we think is incorporate biological reasons of causing the hospitalization. And because from the perspective of machine learning models, the linear model more explained by the association between the predictor and outcome, and the tree-based model, it's like more complex interaction accounted for. So we have top, we listed the top 25 SDOH predicting the final health outcomes, and then apply the causal inference method to estimate the causal effect. Great. And I don't actually see any questions for Dr. Adala, but I'm very curious about the gatekeeper effect. And aside from, you know, improving our own education, knowing about our own implicit bias, do you have any recommendations for how we can get around that? Like, are any plans to test like prompts for physicians who, you know, we may see disparities in their prescribing, for example? Yeah, as a clinician, probably the last thing I wanna see and as another prompt, right? We tend to ignore those. I actually just had clinic yesterday and dismissed a couple of things I think I was supposed to acknowledge. So actually the suggestion is actually to move away from either the individual living with diabetes, their care system, or actually which includes clinicians to changing the underlying protocols or structures or procedures. So that's part of why those 4T data for me as somebody who's interested in addressing this implicit bias issue were really compelling. That was just a new sort of, it's a pragmatic research study technically. So what it did was it changed all clinical protocols within our clinical care center. And so every single individual who was newly diagnosed with type 1 and followed in our center, they get approached for a continuous glucose monitoring to start within usually the first. And the average time to start is around seven days. And we went from having pretty serious differences by race, ethnicity, by insurance status to getting. So for me, automation is one of the best things that you can do to help combat implicit bias because implicit biases tend to find their hold in what we call stage system two or fast thinking. And we have to be able to think fast. We can't constantly slow down every single second of every single day. But if our systems are primed to help us succeed, and then that way, even those where maybe disparities or biases are not on their radar, it doesn't need to be. The system is primed and has started to shift in a way where everybody gets approached. And that's the cultural norm. That's the protocol norm. So for me, I actually tend to think that's the best way to address it rather than adding any more to us. I'll speak for myself. No more buttons. Wonderful. Thank you so much. I think we're out of time. So I just want to say a big thank you to our panelists, Drs. Cromer, DeJong, Guo, and Adala, to the ADA, Nikki, and Logan, and ADA leadership for helping us put together this webinar. I hope it's been very informative. I've certainly learned a lot. We've learned about the spectrum of patients who are affected from pediatrics to older adults and some of the different dimensions of disparities. And as Dr. Adala said also, just the first step is also to raise awareness. And I hope we've done that through this webinar. So thanks to everyone. Have a wonderful rest of the day.
Video Summary
The webinar focused on addressing disparities in diabetes care, particularly in access to medications, out-of-pocket costs, and the impact of social determinants of health. Dr. Cromer presented research showing disparities in medication use among different racial and ethnic groups. Dr. De Jong discussed the challenges of Medicare Part D and the need for transparency in coverage and cost information. Dr. Goh presented an individualized social risk management tool that aims to address social determinants of health in diabetes care. The presenters highlighted the importance of addressing disparities in healthcare and the role of social determinants of health in improving patient outcomes. They also emphasized the need for stakeholder engagement and community involvement in developing interventions to address disparities. Overall, the webinar provided insights into the challenges and potential solutions for reducing disparities in diabetes care.
Keywords
disparities in diabetes care
access to medications
out-of-pocket costs
social determinants of health
racial and ethnic groups
Medicare Part D
transparency in coverage
cost information
individualized social risk management tool
patient outcomes
stakeholder engagement
community involvement
interventions to address disparities
American Diabetes Association 2451 Crystal Drive, Suite 900, Arlington, VA 22202
1-800-DIABETES
Follow us on
Copyright All rights reserved.
×
Please select your language
1
English