A Key to Improving Outcomes Is Understanding and Addressing Health Care Disparities – Even in Ocular Oncology: 2024 Lorenz Zimmerman Lecture Free

I would like to thank the organizers of this great symposium for inviting me to give this talk. I’ve learned so much sitting here today already and am humbled by the invitation to give this lecture. Today I’d like to talk to you about health care disparities. To give you some of my background, I’m a comprehensive ophthalmologist here in town at Northwestern Medicine, an ocular pathologist, and I coordinate our department’s community outreach programs. Additionally, I’ve spent a day every week for the last 25 years providing eye care at one of our affiliated Federally Qualified Health Centers.

I want to start by saying what this talk is not. This is not a “Hey this is an important subject but it’s not applicable to my everyday practice in oncology and pathology” talk. I hope that by the time we are done, you are convinced that understanding and addressing health care disparities will have an immediate and direct benefit to both your patient outcomes during an encounter and population-level health outcomes. When I first started practicing medicine, I thought the best way to deal with disparities was to walk into an exam room and treat each patient without bias and to the best of my ability. I often asked myself “Is this the best way to address health care disparities?” The more I reflect on this, my answer is in some respects yes, and in others no. The overwhelming evidence is that one must look at disparities differently than we have in the past in order to address them. So how do we view things differently? If there is one concept that I want you to take away from today’s talk, it is the social determinants of health (SDOH) and their impact on disparities. SDOH are comprised of health care access and quality, neighborhood and built environment, social and community context, economic stability, and education access and quality [1] (Fig. 1).

Let’s think about uveal melanoma as an example. With regards to predictors of prognosis, we know that tumor size, cell type, prognostic testing such as gene expression profile, and American Joint Committee on Cancer staging are very good predictors of prognosis. Do SDOH play a role in prognosis, and if so, how much do they matter? To help answer these questions, let’s look at the literature for evidence of how various individual and community factors impact overall health. In 2016, Hood et al. [2] analyzed county level data across the United States to assess four categories and how they impacted two outcomes: quality of life and length of life. The four categories were: social and economic factors (education, employment, income, social support, safety), heath behaviors (diet, smoking, obesity, etc.), clinical care (access to care and quality of care), and physical environment (housing, air and water quality, etc.) [2]. Their study determined that approximately 40% of modifiable determinants of health are due to social and economic factors, 30% due to health behaviors, 20% due to clinical care, and 10% due to physical environment. They concluded that clinicians and medical educators should realize that the “greatest improvements in population health will require addressing the social and economic determinants of health.” In other words, the SDOH are where we need to focus our efforts if we want to have the greatest impact on health outcomes.

We practice medicine at the point of care, and as we’ve seen from the speakers at today’s pathology and oncology symposium, we examine a patient with a disease, diagnose and stage it, and then determine the best treatment. In order to understand SDOH and their impact, we have to change that thinking. We must understand that the SDOH impact this disease before, during, and after our diagnosis. Once we understand that, we can formulate interventions at each of these stages to improve outcomes.

First, we will start by looking back in history to try and understand some of the root causes of disparities as they relate to SDOH. Winston Churchill believed that the farther back you look in history the farther forward you are likely to see. Today, we are going to look at the social determinants and “look back” in history for context. We will use Chicago as our as our case study…our laboratory so to speak. We will first look back at non-oncology eye disease, then look back at ocular oncology diseases, and finally look forward to solutions.

Here is a very brief history of Chicago. It was incorporated in 1837, and about 35 years later a fire decimated much of the city and made almost half of the residents homeless overnight. To some extent, this allowed the city to rebuild with more urban planning and forethought. In 1916, the Great Migration began and over the next 50 years 500,000 Blacks would migrate to the city. Shortly after that, in 1919, a series of race riots began. In 1937, the Chicago Housing Authority (CHA) was incorporated, in part to build housing for this influx of people. In 1943, the first CHA housing project opened. By 1950, most of the CHA residents were Italian immigrants. In 1962, the Robert Taylor homes opened, which at that time was the largest housing project in the country with 4,400 units on the city’s South Side neighborhood. Initially, the affordable housing units were to be dispersed throughout the city, but the politicians at that time pushed back and finally approved a 2-mile long and several block wide area of land for this housing project to be built. The homes were flanked on either side by highways, rail lines, and light industry. By 1975, 65% of CHA residents were Black, with many of the households led by single females. In 1990, the federal government took over CHA citing unlivable living conditions. That same year, Northwestern Ophthalmology began a partnership providing on site eye care at a federally qualified health center, Winfield Moody Health Center, which was adjacent to another large CHA housing project, Cabrini Green. By 2000, 96% of CHA residents were Black or Hispanic.

We’ll now turn to how this history of public housing relates to poverty. Here is a Zip code map of the city of Chicago and suburbs (Chicagoland) with a blue background detailing the percent households below 200% of the federal poverty level (Fig. 2). Light blue represents less poverty in each Zip code, and darker blue more poverty. The thick black lines represent the major corridors of highways, rail lines, and industry. The majority of the CHA housing units are adjacent to these black lines. Decades of policies, biases, and decisions by political leaders led to concentrated areas of poverty, minority residents, highways, railways, industry, and a substantial number of CHA residences. The majority of these areas fall within the red ovals on the map, and are on the city’s South and West Sides. Concentration of these factors in small areas directly contributed to neighborhood-level health care disparities in the region.

An example of this can be found by analyzing disparate levels of pollution and poor air quality which can lead to higher levels of various diseases in residents of these neighborhoods. An analysis by the city of Chicago’s Public Health Department in its 2020 Air Quality and Health Report concluded that, with regards to air pollution and health, “the areas of greatest concern are primarily located on the South and West Sides of the city. In particular, parts of the city bisected by highways with high concentration of industry are overburdened, experiencing high levels of both pollution and vulnerability” [3]. Another example of neighborhood-level disparity is access to healthy food, with Chicago’s highest concentrations of urban food deserts being in these same neighborhoods on the South and West sides [4, 5].

So how does all of this impact disparities in health outcomes? In 2019, researchers at New York University analyzed life expectancy gaps within US cities and found the greatest gap was in Chicago. In Streeterville, an affluent neighborhood near downtown Chicago, the average life expectancy of a resident was 89 years. Just about 8 miles south is another Chicago neighborhood on the city’s South Side called Englewood. It has significantly higher poverty, higher percent minority population, and is within one of the “red ovals” shown in Figure 2. The life expectancy of a resident in this neighborhood was 61 years. This is an almost 30-year life expectancy difference in two Chicago neighborhoods only 8 miles apart, with residents of both neighborhoods adjacent to the same health system. Their conclusion was that your Zip code can impact your health as much, or more than your genetic code [6].

We’ll shift gears and talk about a traditional disease timeline and look at factors that impact outcome. For most diseases, there are intrinsic factors that we are born with, and subsequently things like health behaviors (smoking) and exposures (UV light, pollution) throughout life that can increase likelihood of developing a disease. There typically is a time lag between developing and diagnosing a disease. Once diagnosed, the disease is treated.

When we look at this timeline from the perspectives of traditional clinical care and the SDOH, there are opportunities to impact health outcomes in both areas. In terms of the traditional model of clinical care, we implement early screening programs, encourage healthy behaviors, and continually develop better medications and treatments for a disease. All of these traditional efforts are essential and should continue. Despite these efforts, however, we still find that certain groups of people living in close proximity to one another have poorer health outcomes compared to each other (e.g., life expectancy). Each of these groups are members of the same society with access to the same screening, same diagnostic methods, and same treatment options. This is where analyzing the SDOH is important. When we look at outcomes through the lens of SDOH, we find there are many more opportunities to improve health outcomes by addressing the SDOH. For example, certain at-risk populations have poorer maternal health, higher exposure to air pollution, poorer nutrition, less access to medical care due to lack of insurance, delayed detection of disease, and even undergo different treatments than others with the exact same stage of disease. It is in these areas that we need to target our efforts to have the greatest impact on improving our society’s overall health. This is where we are going to focus on for the remainder of our time.

Let’s use glaucoma as an example. We are taught very early in our ophthalmology training that glaucoma occurs earlier and is more severe in Blacks compared to non-Hispanic White individuals, with almost 6 times the rate of advanced vision loss [7]. At first glance, many people look at this disparity and conclude that this is due to genetics. While there may be a genetic component for some portion of Blacks with glaucoma, a definitive genetic association to explain all of this disparity in disease burden has not been discovered. This makes sense because the Black population is a heterogeneous phenotypic and genotypic group of people.

Let’s look at nongenetic factors such as air pollution. Ambient fine particulate matter, particulate matter less than 2.5 micrometers in diameter (PM2.5), is a significant part of what we call air pollution. PM2.5 is responsible for 85,000–200,000 excess deaths per year in the United States. A study done by Tessum et al. [8] in 2021 demonstrated that PM2.5 exposure was significantly higher in people of color, particularly Blacks and Hispanics. These findings were true in most states, and in both urban and rural areas. At first glance, one could postulate that this difference is due to poverty. The authors subsequently controlled for income and found that while the gap slightly narrowed at higher income levels, the disparities persisted with Blacks and Hispanics having higher PM2.5 exposure at nearly all income levels.

In addition to cardiac, pulmonary and oncologic diseases, PM2.5 can impact glaucoma. We know that there is a direct dose-response relationship between PM2.5 levels and ganglion cell-inner plexiform layer thickness [9], a key indicator of glaucoma progression, so this is not “just genetics” that is leading to Blacks having glaucoma diagnosed earlier and having more severe vision loss. In the case of PM2.5, a key factor that results in higher exposure to this pollutant is society’s historical systemic bias that has led to minority populations being housed in areas of increased density of industry, highways, and gasoline and diesel burning vehicles. Addressing this disparity and SDOH will have an impact on individual and population-level glaucoma incidence and outcomes, independent of other efforts such as new diagnostic and treatment measures.

We are going to change gears and talk about Chicago in terms of diabetes and diabetic eye disease. Using a large regional health data repository called HealthLNK, our group analyzed the prevalence diabetes, prevalence and severity of diabetic retinopathy (DR), and hemoglobin A1c (HgA1c) [10]. HealthLNK contains data from approximately 2.4 million people in Chicagoland and has granular patient level data from the region’s five academic medical centers and the county health system. As expected, we found that patients with Medicaid had more diabetes, more DR, more severe retinopathy, and higher HgA1c compared to those without Medicaid. The prevalence of DR amongst patients with Medicaid was higher than those with other insurance (25.8% vs. 17.6%; p < 0.0001). In terms of severity, the prevalence of proliferative DR in patients with Medicaid was higher than patients with other insurance (5.9% vs. 3.4%; p = 0.0001). Mean HgA1c in Medicaid patients was greater than patients with other insurance (7.7% vs. 7.4%; p < 0.0001). We chose to look at Medicaid because most patients in this program are means tested, and we used this as a surrogate for poverty because HealthLNK does not contain individual income data.

We felt that this small difference in HgA1c likely was not the main driver of significantly higher rates and severity of DR, so we took things a step further to determine if neighborhood, ethnicity, and household income (poverty) play a significant role [11]. Using HealthLNK data, we did a ZIP code analysis of the 2.4 million people to analyze these factors using a type of math very similar to what is called sabermetrics or “baseball math.” To use a baseball analogy, when opposing managers determine where to position their players to maximize the chances of them being exactly where the batter is going to hit the ball, they use historic data that plots where the batter has hit every ball in prior at bats. They subsequently reposition the defense accordingly. In other words, they use historic data to predict future performance. We combine our sabermetrics analysis with our geographic information system mapping to geolocate these disparities.

There are 221 ZIP codes in Chicagoland, and we found 36 “high risk” ZIP codes where there was a high rate of diabetes, high risk of DR, and very low rates of DR diagnosis. In other words, there was a significantly high amount of undiagnosed DR in these areas. Not surprisingly, almost all of the 36 ZIP codes (Fig. 3, high risk neighborhoods indicated by large green dots) fall within the red ovals that encompass the previously described marginalized neighborhoods. Analysis of socioeconomics of these areas reveals high rates of poverty (dark blue background), high percent minority residents, close proximity to highways and industry, and higher pollution. Poverty and percent minority households in these ZIP codes were both significantly higher than the Chicago average (Table 1).

So what do we do with this information? Here’s is where we transition from describing disparities to addressing them. Just like baseball managers position defensive players where the batter is most likely to hit the ball, we used this map to place non-mydriatic cameras in select neighborhoods where they have the highest probability of detecting DR. We can use our data and geographic information system mapping to pinpoint with great accuracy, often within a few square city blocks, the optimal locations to place our cameras in order to detect the most DR. The three black X’s on the map in Figure 3 are where we deployed these cameras.

In our first 300 patients that we screened, a stunning 45% of them had findings of retinopathy, whereas screening in a general population would have an expected rate of 18–20%. Over half of all screens resulted in a recommendation for referral for in person ophthalmic evaluation for suspected DR or other conditions such as glaucoma. This is an example of using big data and mapping techniques to deploy limited resources to the areas with the highest need that will have the greatest impact.

Let’s pivot and turn our attention to ocular oncology. Just like many other diseases, in ocular oncology we have disparities with regards to incidence, age and stage at diagnosis, and in some circumstances different treatments are provided to patients of different ethnicities even with the same stage of cancer. With retinoblastoma, high income countries have very high survival rates, above 95%, and in lower income countries survival rates can be 50% or less [12]. I would be remiss if I didn’t stop and say that I stand here in awe of what many of you in this audience have done to decrease the worldwide disparity in retinoblastoma survival. Your efforts have increased awareness of the disease, decreased the time to diagnosis, brought infrastructure and treatments to areas of the world that were previously not available, and improved survival. Your efforts serve as a model for how a professional society or group of motivated providers can close a disparity gap.

In Brazil, Mattosinho and colleagues [13] analyzed differences in retinoblastoma survival rates and, as expected, found outcomes varied by income level with patients in families with low income having significantly worse survival than higher income families. What I found interesting about this study was their finding that retinoblastoma patients in families having more than one child under 5 years of age at home had a 200% higher increased risk of death compared to those with families with no other small children at home. As a provider, if there was a risk factor that increased mortality by 200%, we would want to know about this at the point of care. While this risk factor cannot be changed, once aware of it, the care team can make additional resources available such as transportation, childcare and others to help decrease mortality. Configuring an electronic health record (EHR) to notify the provider at the point of care or modifying new patient intake questionnaires to capture this type of family information are examples of using “non-clinical” data to decrease disparities in outcomes.

In addition to individual and family socioeconomic factors leading to disparities in outcomes, the actual neighborhood itself has been found to be a risk factor, regardless of a patient’s socioeconomic status. Altamirano-Lamarque and colleagues [14] studied neighborhood-level impact by analyzing the Childhood Opportunity Index (COI). The COI score has a number of factors that can be consolidated into three main elements: education, health and environment, and social and economic factors. By now, these are very familiar sounding terms. They found that children in low CIO scored neighborhoods had increased odds of presenting with more advanced, Group D or E, retinoblastoma. This disparity existed even after controlling for individual social economic factors. In other words, where you live is associated with more advanced stage of disease at presentation.

Analyzing the National Cancer Database for uveal melanoma, retinoblastoma, and conjunctival melanoma, Choudhry and colleagues [15] found that patients with no insurance and those with Medicaid presented with more advanced primary clinical tumor classifications that those with other insurances. These disparities also exist in ocular surface squamous neoplasia (OSSN) with Black patients having a diagnosis at an earlier age and a higher risk of mortality [16]. The finding of increased mortality in Blacks was present even after adjusting for age, biological sex, Charlson/Deyo score, primary payer, income, education, histologic type, and time to definitive treatment.

There are also disparities in the types of treatments that are administered in ocular oncology. Using the Surveillance, Epidemiology, and End Results (SEER) 18 Registries in a multicenter population-based study, Rajeshuni and colleagues [17] studied enucleation rates in retinoblastoma. They found that children from non-White, Hispanic, and lower socioeconomic status were more likely to receive enucleation for this disease, independent of the stage of diagnosis [17]. One year later, the same group found similar findings with uveal melanoma [18].

We will use the remaining time to discuss additional strategies for addressing these disparities. I have to admit that I don’t have any simple answers to solve these disparities, but I will tell you about some of the things that we’ve implemented at Northwestern. Before we do that, I want to reiterate that if the causes are multifactorial then the solutions have to be multifaceted. We will talk about solutions that involve implementing alerts to the provider at the point of care, using big data to discover or identify patients with higher risk of poor outcome, changing medical education to emphasize SDOH and their impact on health outcomes, and encouraging our health care systems to implement solutions.

We can begin by designing EHR workflows that can identify risk factors associated with disparities and notify providers at the point of care. A straightforward example is a point of care alert that pops up when we prescribe medicine with a high copay. If patients can’t access or afford a treatment, this impacts outcomes. We have to go beyond this. We learned that in certain areas, having a second child at home under the age of 5 was associated with a 200% increase in mortality in a patient with retinoblastoma. If this were a lab value, it would be flagged as a “critical result” and prominently displayed in our EHR. Any provider would want to know of a risk factor that doubles mortality. Even though this is not a lab value, providers should be made aware of this or any similar factor, whether or not is related to SDOH, and be given an opportunity to mitigate this risk. Additionally, our EHRs should not solely be a data entry portal. We have fast, real-time algorithms that can look at both SDOH and clinical health variables. These algorithms and things like artificial intelligence should be used to alert providers both at the point of care and at the population health level to identify those at higher risk for poor outcomes. They can also be configured to present providers with options to mitigate these risk factors. Finally, with regards to big data and EHRs, as a profession, we have to make sure that our clinically relevant pathology (histopathologic findings) and oncology (clinical staging, treatments) data are recorded in a structured way and passively flow into our data repositories. This will allow for more rapid and accurate analysis. As we have learned today, in order for the clinical data to be much more meaningful and actionable, it must be combined with SDOH data.

Disparities researchers have indicated that we need to reform some aspects of medical education. Many of us read residency applications, and at some point, the majority of students applying to ophthalmology say something similar to “I want to help the underserved.” Initially, I wondered if this is just due to being young and idealistic. The more I ponder this, I think they know more than I do, because by saying this they are identifying the exact place where we need to be to have the greatest impact on improving health outcomes. As medical educators we need to realize this and act on it in clinical care, research, and medical education. At Northwestern, we have a formal didactic “disparities curriculum” with a group of faculty and residents assigned to design and implement it. This includes a dedicated grand rounds every quarter. Those rounds have the most robust discussions and highest participation rates from medical students, residents, fellows, and attendings. We have structured rotations at community health centers, and our learners are engaged in disparities research and developing SDOH-based solutions.

We have system wide disparity programs in our academic medical center. The Northwestern Summer Scholars program is an initiative where we partner with two high schools. They are located in the South Side and on the West Side; both schools are within those familiar “red ovals” in Figures 2 and 3 outlining neighborhoods with greatest health disparities. The program allows 30 high school students to participate in a 6-week summer program for each of their three summers between freshman and senior years. They learn about medicine, use our surgical simulation labs, attend seminars on how to write a college essay, and have a first-hand look at opportunities for careers in medicine, research, and medical administration. Hopefully some of them will eventually join our workforce, because the data shows that a medical system with diversity in providers leads to better health outcomes [19].

Finally, as ophthalmologists, we have to continually remind our health care systems to recognize that, compared to many other clinical factors, the SDOH have an outsized impact on individual and population health outcomes. Whether it’s designing solutions to prevent vision loss from diabetes or improving outcomes in ocular oncology, during discussions on how to accomplish this, the SDOH must have “a seat at the table.” As I’ve learned from my American Academy of Ophthalmology advocacy experiences, “If you are not at the table, you are on the menu.”

In conclusion, I am sure that when each of us walks in an exam room, closes the door, and treats our patients to the best of our ability without bias or judgment we’re not part of the problem, but I am convinced that when we recognize and address the Social Determinants of Health and their impact, we will all be part of the solution. Thank you very much.

The author has no financial interest to declare.

This research was supported by National Eye Institute grant 1R01EY034444-01 Health Disparities in Utilization, Quality, and Outcomes for Three Common Ocular Conditions (HealthDOC) and an unrestricted grant from Research to Prevent Blindness, New York, NY.

P.J.B.: conceptualization, data gathering and data analysis, writing, editing, and revising the manuscript.

Presented at the 2024 American Academy of Ophthalmology Meeting Chicago, IL, October 19, 2024.