Most of our lives have been touched by cancer in some way, probably the reason that in recent years, citizen-led efforts to raise money and awareness have surged in popularity. But the fact is, despite all those R&D dollars, we’re still elephant-groping blind men when it comes to characterizing cancer. We’ve got pieces of the genetic picture (mutations and certain gene variants predispose us to certain types of cancer), and we’ve got some notion of environmental risks (stay away from UV rays, leaching water bottles, and seismic-zone nuclear plants). But putting them all together into a coherent picture continues to confound us — what turns an orderly clump of cells into a metastasizing mess?
This visualization, by the folks at GSAREH, doesn’t have all the answers, but boasts a phenomenal amount of cancer and cancer-related data, giving us a way to mine for patterns and trends across populations. Selected as a one of two award winners in the National Cancer Institute’s 2010 "Enabling Community Use of Data for Cancer Prevention and Control Challenge," it impressed us too, with its user-friendly interface, enormous data set, and malleability: You can browse and navigate ready-made thematic maps, build customized cancer maps, search and manipulate cancer data, and even generate charts based on demographics down to the county level.
The simplest way to dive in is to navigate the pre-made cancer maps using the buttons in the toolbar. Load the "lung cancer" "breast cancer" or "colon cancer" maps and you’ll get an instant view of deaths due to these diseases, county by county, across the entire US. You can also see the geography of anti-cancer screens such as mammograms and proctoscopy, as well as cancer-related indicators like smoking and obesity. Not all the data is available for all counties, resulting in a plethora of white patches, for example, when searching mammograms. But one striking view can be seen here by looking at lung cancer deaths, at left, and smokers on the right.
The similarities are uncanny: In both maps we’ve got twin cancer/smoking hotspots in Nevada, Arizona, and South Dakota, and a deep, dark patch stretching from western Texas on up through the Virginias — a cancer alley that covers a sizable swathe of America’s tobacco-growing regions.
But before we jump to attribute all this lung cancer to smoking, we should pause to consider that Appalachia is also the heart of coal country. And while there’s still much dissonance over direct links between coal mining and lung cancer, researchers have documented an array of environmental and human health effects that come with mountain-top removal: It leaches toxins into streams, contaminating local water supplies; meanwhile, drilling and blasting pollute the air with carbon monoxide, dust, and coal particles, the last of which carries a carcinogenic slush of cadmium, nickel, arsenic, and zinc.
Another interesting feature to explore is the deceptively demure "More…" button, which allows you to layer more datasets atop your original map. Cancer mortality for white men, white women, black men, and black women can be found here, as can the comprehensive Toxic Release Inventory, a publicly available database containing information on toxic chemical releases and waste management activities in the US. Here’s what the layering Toxic Release Inventory on top of Breast Cancer looks like, first the entire US:
…then, a close-up of the Great Lakes region:
Each year any company that produces more than 25,000 pounds or handles more than 10,000 pounds of toxic chemicals must report it to the TRI. So you might expect there to be a strong link between so much poison dumping and cancer. But if there is one, it’s not apparent from this map, where the red triangles appear as often in darkly shaded counties (lots of breast cancer) as in light ones. (Colon- and lung cancer showed similar profiles).
Does that mean that there’s no link between toxic releases and cancer? Or simply that we aren’t looking at the right kinds of cancer? The pre-made maps will provide laypersons (present company included) plenty to explore, but if you really want to start making hypotheses like these and probing more complex questions, you’ll want to make use of this visualization’s powerful embedded tools. The first is a Search tool, launched by clicking the binocular icon. You then select a drawing tool (circle, rectangle, polygon) and a data set, so that when you draw a shape on the map, the relevant data profile will pop up. Here, for example, is what we got by drawing our boundary around Wayne County, Virginia and selecting "Risk Factors and Access to Care" (left). We did the same for the county surrounding Manhattan (right):
For squint-relieving purposes, here are some of the salient differences:
Wayne County New York County
% Reporting No Exercise 36.4 24.5
% Reporting Few Fruits/Vegetables 83.1 71
% w/ Obesity 28.3 15.9
% w/ High Blood Pressure 31.4 23.1
# Primary Care Physicians per 100K Persons 39.3 287.7
You can also use the Search tool in the opposite way: first define your search criteria and let the visualization find the matching geographic regions. For instance, you might want to find out about counties where mammography screening is between 0% and 70%, and where pap smear screening is between 0% and 80%. This level of fine-grained analysis is possible for an entire rolodex of cancers from bone & joint, bladder, and brain to thyroid, uterine, and polycystic ovarian.
Admittedly, most of us won’t be needing to parse information at this granular of a level. Nor will we have the expertise to begin drawing conclusions about cancer and vast repository of lifestyle, demographic, economic, and healthcare statistics, available here.
So, yes, with data-rich dashboards like these, there’s the risk that unwarranted assumptions will be made, and causal links drawn where none in fact exist. On the other hand, this visualization offers a kind of plasticity that will be invaluable in the hands of the cancer research community. For those with an even basic knowledge of disease etiology, tools like this will be invaluable, enabling them to see relationships that were previously invisible. And they’ll know what — and what not — to make of them.
For the rest of us, there’s still plenty to explore in a visualization that ultimately amounts to not one visualization, but a visualization engine that generates infinitely malleable graphics according to your data-parsing pleasure. So, if you can put aside your Sherlock ambitions and regard this not as a culprit finder but as tool for exploring fascinating geographic correlates, forge ahead: Check out the Search tool, the Mapping tool, and the Pie-chart maker too. Learn what you will, and please send us your thoughts! We’ll share them with GSAREH as they continue to look at how geospatial research emerging from academia can help public and private organizations be more responsive to reality on the ground. When that reality is as mystifying an animal as cancer, we’ll need all the help we can get.