5.3 Mapping Cancer Health Outcomes and Disparities

[Slide: Mapping Cancer Health Outcomes and Disparities]
 
Slide 1
I am Jaclyn Hall, I am faculty with the College of Medicine and the Cancer Center at the University of Florida, and I’m also a Geographer. And yes, a lot of the work of my research team is making and analyzing maps.
 
Slide 2
[image of a whimsical cartoon map]
But we all love maps. Maps are a part of our childhood play and our creative adventures.
 
Slide 3
[image of children examining a globe]
Maps are an important part of how we learn and share information about our world. For cancer population sciences, maps are very important for describing and understanding the disparity and cancer health outcomes of our communities. As a Citizen Scientist, your research team is likely to be using maps from the literature and generating your own maps and public health data. Now let’s talk about how maps are used in public health and what to be aware of when discerning these spatial health data.

Slide 4
[map showing U.S. cancer mortality rates by county using color gradients to indicate higher rates in red and lower rates in blue]
Most public health maps display county-level data, because counties have long been a standard reporting unit. And most county health statistics are presented as choropleth maps – the type of map that displays a color intensity in proportion to a statistical variable. Here we see a choropleth map of county-level cancer mortality created by the Institute of Health Metrics Research and Evaluation at the University of Washington.
 
Instantly we see spatial trends – we see high values in the southern plains, in the Mississippi Valley, and in Appalachia, all places where poverty level is high. Now cancer is a varied disease with cancers impacting different organs, in different people, from different places. But this map is mortality and there are numerous ways that poverty contributes to higher cancer mortality: less access to health care, less cancer screening, higher probability to live in a community with high industrial and environmental pollution, not to mention how higher smoking rates are related to poverty. When I see this map of cancer mortality, I think what I’m seeing is poverty.
 
We should not rely on statistics of combined health outcomes because the most common cancer will dominate any statistical or geographic trend and you’ll lose the ability to see the disparity in other less-common cancers. Different geographic and socio-demographic communities experience unique health risks, so it’s important to look at each individual cancer rather than relying on a map of combined data.
 
Slide 5
[map showing U.S. mortality rates by county from lung, bronchus, and tracheal cancers, with higher rates in red and lower rates in blue]
Now we’ll look at a few specific cancers. Here is lung and bronchus cancer. It does have a similar geographic trend to the all-cancer mortality because about 30% of cancer deaths are related to lung cancer and smoking. But lung cancer and smoking is also related to high poverty. In this map, we see a clustering of high rates around Appalachia which is also related to exposure to environmental contaminants.
 
Slide 6
[map showing Non-Hodgkin’s lymphoma mortality with red shading to indicate higher rates above the Appalachian Mountains]
In this map of Non-Hodgkin’s Lymphoma mortality, we see a different geographic trend with high rates throughout the northern part of the US’s agricultural region. I also want to draw your attention to the stark boundary at the border of the Virginia and Kentucky states. Usually when I see a rate difference along a state boundary, with other health statistics, I would think it was a difference in state-level reporting or definition, but not with cancer data. They’re very highly regulated and highly maintained data. Sometimes we’ll see an actual geologic boundary and that’s what we have here the top of the Appalachian ridge. Some studies suggest that Non-Hodgkin’s Lymphoma incidence is related to exposure to certain herbicides and insecticides. So, the trend we see here is related to the top of the Appalachian Mountains being a geographic barrier to the winds moving east that might be possibly carrying those environmental contaminants.

Slide 7
[map showing U.S. mortality rates by county from bladder cancer with higher rates in red shown in the northeastern region of the country]
In this map of bladder cancer mortality, again, we see a very different geographic trend. Bladder cancer in the US is not common, but when we look specifically at this cancer, we see real geographic disparity that our research teams need in order to ask and answer the right questions. It wasn’t until 2016 that the NIH determined that families drinking well water in the upper Northeast had an elevated risk for bladder cancer due to the possibility of arsenic in the drinking water.
 
Slide 8
[map showing geographic distribution of arsenic levels in surface water, with more red shading to indicate high rates in the northeast]
Here’s a map from the USGS of arsenic in surface water and we can see higher rates in the Northeast.
 
Slide 9
[map showing U.S. mortality rates by county from liver cancer, with red shading for higher rates in south Texas]
In this map of liver cancer our eyes are drawn towards southern Texas. This is because one of the communities in the US which has the highest risk of liver cancer is Hispanics and ethnicity is the main driver of what we see in this map. But I also want to draw your eyes to the north around the area of South Dakota. Here we see a few outliers that go against the main geographic trend. We should also notice these outliers because there’s something to learn there and it’s an important part of the story of cancer burden in your community. For those of you from this region, you will know that these counties are the location of several Native American reservations whose populations are struggling with multiple health concerns, including poverty and lack of access to health care.

Slide 10
[Florida county-level map showing all cancer mortality, 2013-2017, with the highest rates in dark blue and within northern counties]
So when looking at your own state data, be aware of those spatial outliers. Here is my state, the state of Florida. We’re looking at all cancer mortality. In general, the trend in Florida is higher mortality, high rates of poverty and smoking in the north. So higher cancer burden in the north but we see a few geographic outliers.

Slide 11
[Okeechobee County is outlined in yellow within the Florida county-level map showing all cancer mortality, 2013-2017]
First in the south, we’ll look at Okeechobee County, a rural county with high poverty and high smoking rates.
 
Slide 12
[Leon County is outlined in yellow within the Florida county-level map showing all cancer mortality, 2013-2017]
In the north is Leon County which has Tallahassee, a mid-sized city that is our state’s capital and has two large universities. The higher education and income in this urban area is associated with lower cancer mortality.

Slide 13
[Sumter County is outlined in yellow within the Florida county-level map showing all cancer mortality, 2013-2017]
And in the center of the state is Sumter County, in almost all county-level health statistics, Sumter County will look like the healthiest place to live in the US. Sumter County has some of the largest and fastest growing retirement communities in the country, and many of these are upper middle-class golf course retirement communities. Outside of these communities, the rest of the county is rural with the southern half of the county being unpopulated and part of the green swamp. But in the retirement communities, these individuals are older and age is actually the main risk factor for cancer mortality. But health statistics should always be age standardized. Standardization is a set of calculations which will account for the age structure of each county. Sumter County has the nation’s highest percentage of over 65, with over 50% of this county being over 65. Because these retirees are wealthy and well resourced, that creates this lower cancer mortality that we see for Sumter County.

Slide 14
[slide shows the Florida county-level map showing all cancer mortality, 2013-2017]
Now let’s move on and look at, specifically, the legend. This map was generated as a quantile map with 5 categories each with the same number of counties. This is a standard method to display data if you have values with a normal distribution. If we look at the average range of the values in the categories, they are between 15 and 36 points, except for [the darkest blue category, indicating the highest cancer mortality, is outlined] the highest, highest burden, darkest blue category- it spans 260 points. When we see this, we know there’s likely a data outlier. This large span is caused by one outlier:

Slide 15
[Union County is outlined in yellow within the Florida county-level map showing all cancer mortality, 2013-2017]
Union County. When I first noticed this outlier, I researched the county. I learned that this is a lovely forested county, not too different from the other rural counties around it. But the main industry in Union County is corrections. There are several large prisons in this small rural county, one of which has the state medical prison which has a cancer center. So, every inmate from throughout the entire state who has cancer is located to this county, and any death statistic is attributed to this small county. It doesn’t have a large enough population to counteract this inflated numerator. Here is an outlier due to misaligned data, data that come from different institutions. And on most of the maps
 
Slide 16
[Note appears specifying that the mortality is per 100,000 alongside the Florida county-level all cancer mortality map, 2013-2017 with Union County outlined in yellow]
I create,
[Union County is now outlined and filled with grey shading within the Florida county-level all cancer mortality map, 2013-2017]
I note the outlier like this.

Slide 17
[slide shows the Florida county-level map showing incidence of cervical cancer, 2013-2017] Here’s a different map of cervical cancer incidence. Maps are used to show a geographic trend, clustering, spatial outliers, but what if there’s no geographic trend like this map? If a map looks like this, with no obvious geographic story to tell, then I would recommend that maybe these data that don’t need to be presented on a map. People love maps, but a map will take up space on the page and time from a viewer trying to understand what is the geographic story of this map. Sometimes a good old-fashioned table sorted alphabetically or from greatest to least cancer burden is what is appropriate. One of the reasons we don’t see a geographic trend in cervical cancer incidence is the number of rural counties in northern and south Florida, with small populations, which lead to unstable rates.

Slide 18
[slide shows the Florida county-level map showing incidence of cervical cancer, 2013-2017 with Lafayette and Glades Counties outlined in yellow]
Two of the counties have zero cancer incidence, they’re both rural counties with high smoking rates and high poverty, and all the challenges that come with rurality and poverty. Glades and Lafayette counties are two of the smallest counties in Florida; neither has a hospital. Some counties are so small that people may seek major health care outside the county. But then you have Miami-Dade: has the same population as the entire state of Mississippi.

Slide 19
[graph depicting cervical cancer incidence in Miami-Dade county vs state of Florida, 2005-2017. The county and state rates are both stable and similar to each other.]
Here we see cervical cancer incidence from Miami-Dade and we see that from year to year the incidence doesn’t vary widely and it doesn’t vary widely from the state of Florida as a whole.

Slide 20
[graph depicting cervical cancer incidence in Glades county vs state of Florida, 2005-2017. The Glades rate is stable at 0 except for 2006, when it appears to spike at the top of the graph, while the state rate is stable.]
But when we look at Glades County, because it has a small population, only 13,000 people, most years, it has zero incidence of cancer but 1 case in 2006 produces a rate that was 3 times the state rate.

Slide 21
[graph depicting cervical cancer incidence in Bradford county vs state of Florida, 2005-2017. The Bradford rate appears to ping pong from the top of the graph to the bottom while the state rate is stable.]
Here is the annual trend from Bradford County – another rural county in north Florida that will be very similar to rural counties in your state. It shows a rate indicative of small counties, where one year might be very high, one year it’s very low. This is what we call an unstable rate due to small numbers. This is why for most of our statistics coming from our University of Florida Cancer Center, we combine data from three or more years.

Slide 22
So when mapping cancer health outcomes and disparities, pay attention to:
individual cancers and their unique trends,
note any rates following state boundaries- not so much of cancer but for other health outcomes that might be related to the cancer your research team is studying,
look and learn from those geographic outliers and other data issue outliers,
always use age-adjusted rates because all of our counties have different age structures, and ask is there a geographic story and learn about that story because you will be asked about it, and remember that small numbers issue- those unstable rates, and that’s why it’s good to combine years of data.

Slide 23
[slide title: Mapping Cancer Health Outcomes and Disparities]
Understanding how data are displayed in maps can help our research teams to visualize patterns and that helps our cancer centers target interventions, [slide text: allocate resources], all with the goal to reduce the burden of cancer health disparities [slide text: through primary prevention, early detection, delivering cancer care to diverse communities]. Thank you.