Sources of Spatial Health Data

The Three Pillars of Spatial Health Data

Most spatial health analyses, from mapping disease clusters to assessing healthcare access, rely on combining three core types of data:

1. Health Outcome Data (the "Numerator"): This is information about the health event itself. It could be the locations of patients with a specific disease, mortality counts, or the addresses of hospitals and clinics. This data tells us what happened and where.
2. Population & Demographic Data (the "Denominator"): To put health events in context, we need to know about the population at risk. Simply mapping raw case counts can be misleading; a city will always have more cases than a rural town just because more people live there. To calculate meaningful rates (e.g., incidence per 100,000 people), we need population data. This tells us who is affected.
3. Contextual & Environmental Data (the "Covariates"): These are data on other features of the landscape that might influence health, helping us explore the why. This includes the built environment (roads, schools), the natural environment (rivers, elevation, air pollution), and socioeconomic factors.

Let's explore the sources for each of these pillars.

Pillar 1: Health Outcome Data

Health data can come in two main forms: individual-level records and aggregated statistics.

Individual-Level Data: Registries and Surveillance
The most detailed health data comes from records for individual people or events. This often originates from:

• Disease Registries: These are systematic collections of data about people with a specific disease. Cancer registries are a classic example, tracking every new cancer diagnosis within a geographic area.
• Disease Surveillance Systems: Public health departments track reports of notifiable diseases (like influenza, measles, or COVID-19) from labs and hospitals.
• Hospital Admission Records: These can provide data on where patients who are admitted for various conditions live.

A crucial feature of this data is that it usually contains a patient's street address. To use it on a map, this address must be converted into geographic coordinates (latitude and longitude), a process called geocoding, which we will cover in detail later in the course.

The following reading from the CDC's Field Epidemiology Manual explains how investigators collect and use this type of location information.

As you can imagine, this individual-level data is extremely sensitive. Due to patient privacy regulations (like HIPAA in the United States), direct access to raw address-level data is heavily restricted and usually limited to public health officials or researchers who have gone through a rigorous approval process.

Aggregated Health Statistics
Because of privacy concerns, health data is often made publicly available only after it has been aggregated. This means the individual case counts are summed up into geographic units, like counties, ZIP codes, or census tracts. The data is then presented as a total count or a rate for that area.

The map below is a perfect example of aggregated health data.

This map, prepared by the Illinois Department of Public Health, visualizes two types of aggregated health data for colorectal cancer:

• Incidence Rate: The number of new cases per 100,000 people. This is displayed as a choropleth map, where each county is shaded based on its rate category (the blue-shaded polygons).
• Mortality Rate: The number of deaths per 100,000 people. This is shown using proportional symbols, where a larger red circle indicates a higher mortality rate within that county.

You can see from the "Sources" section at the bottom of the map that this data comes from official sources like the National Center for Health Statistics and the Illinois State Cancer Registry. These are exactly the kinds of organizations that collect and disseminate health statistics.

Pillar 2: Population and Demographic Data

To calculate the rates you see in the Illinois map ("per 100,000 population"), the raw number of cancer cases (the numerator) had to be divided by the number of people living in each county (the denominator). The primary source for this population data is a national census.

The Role of the Census
Government census agencies (like the U.S. Census Bureau) are tasked with counting the entire population and gathering key demographic information. This data is essential for spatial epidemiology. The census provides two critical components:

1. Demographic Data: Population counts, broken down by age, sex, race, ethnicity, income level, housing status, and more.
2. Census Geographies: A nested hierarchy of administrative boundaries for which the demographic data is tabulated. In the U.S., this hierarchy includes states, counties, census tracts, block groups, and blocks. These boundary files are themselves a critical type of vector polygon data.

The CDC's Field Epi Manual also highlights the importance of census data for defining the population at risk.

Pillar 3: Contextual and Environmental Data

The third pillar involves a vast range of other spatial data that provides context. The Illinois map, for instance, includes the locations of Federally Qualified Health Centers (the black dots). An analyst could use this to begin exploring questions about healthcare access.

Sources for this type of data are incredibly diverse and can include:

• Government Agencies:
  • Environmental Data: The Environmental Protection Agency (EPA) provides data on pollution, Superfund sites, and water quality. The US Geological Survey (USGS) provides data on elevation, land cover, and hydrology.
  • Infrastructure Data: Departments of Transportation maintain data on road networks.
• Crowd-Sourced Data: Platforms like OpenStreetMap provide detailed, free global data on roads, buildings, points of interest (like restaurants and parks), and much more, all contributed by volunteers.
• Remotely Sensed Data: Satellite imagery can be processed to derive information about land use, vegetation, impervious surfaces (like parking lots), or even standing water, which is relevant for vector-borne diseases like Zika. The CDC Field Manual provides a great example of this in the section Providing Context with Supplemental Data.

Exploring Data Portals

Finding the right dataset often means searching through data clearinghouses or portals. The U.S. Centers for Disease Control and Prevention (CDC) maintains a page that is an excellent starting point for finding public health-related spatial data.

This exercise should give you a concrete appreciation for the breadth of data available and help you recognize the names of major data providers that you will encounter throughout your career.

Conclusion

In this lesson, we've organized the vast world of spatial health data into three main pillars: health outcomes (numerators), population demographics (denominators), and contextual factors (covariates).

• Key Takeaways:
  • Health data is found either as restricted, individual-level patient records (which require geocoding) or as publicly available aggregated statistics for geographic areas like counties.
  • Population and demographic data, which is essential for calculating rates, primarily comes from national census bureaus. They provide both the demographic tables and the corresponding geographic boundary files.
  • Contextual data, which helps explain spatial patterns, is diverse and comes from environmental agencies, infrastructure departments, and open-source projects like OpenStreetMap.
  • Data portals, like those hosted by the CDC, are invaluable starting points for discovering and accessing these various datasets.

You now have a complete conceptual map of spatial data: you know its structure (vector/raster), how it's located (CRS), and where to find it. In our final lesson for this module, we will discuss the different tools we can use to bring all of this together, examining the distinct roles of GUI-based software like QGIS and programming with R in a typical spatial analysis workflow.