How to read the data
Every number on Vellora is one of three things: an estimate of how many people live with a condition, an observed count of an outcome such as deaths, or an association between two measures. This guide shows how to tell them apart and how far each one can be taken.
Three kinds of figure
| Kind | What it tells you | What it cannot tell you | Example on Vellora |
|---|---|---|---|
| Estimated prevalence | The likely share of adults who report being told they have a condition, modeled from a national survey. | Undiagnosed cases, exact counts, or change over time. | High blood pressure, diabetes (CDC PLACES) |
| Observed outcome | Events counted from records, such as deaths from death certificates or new cancers from a registry, as a rate. | Why the rate is what it is, or how many people live with the condition. | Heart disease death rate (CDC), cancer incidence (Texas Cancer Registry) |
| Association | Whether two county measures tend to rise or fall together. | That one causes the other, or anything about individuals. | Correlations on the Research page |
Try any measure below to see which kind it is and how it was made.
- Type of figure
- Modeled estimate
- Age adjustment
- Crude prevalence (PLACES data value type CrdPrv). PLACES also publishes age-adjusted prevalence (AgeAdjPrv), which CDC suggests for comparing counties with different age structures; it is not loaded in this build.
- Who is counted
- Adults aged 18+
- Latest data
- PLACES 2025 release; BRFSS 2023
- Uncertainty
- 95% confidence interval as published by CDC (method changed beginning with the 2023 release)
- Comparing over time
- CDC states the model does not support tracking change over time. Each release is its own series and is never charted as a trend.
A risk factor, not a measure of cardiovascular or kidney disease itself.
Full definition in MethodologyReading a map
Explore and the cardiovascular research page colour each county by one measure in one period.
- Colours are relative. Counties are split into up to five quantile groups with roughly equal numbers of counties. A dark county is high compared with other Texas counties in that map, not high by any clinical standard, and the same shade means different values on different maps.
- Hatched means withheld. A suppressed county has no published value, usually because its population is very small. It is never zero.
- Read the legend. The legend gives the class break values, the unit and how many counties reported. Check whether the unit is a percentage of adults or a rate per 100,000.
- Big counties are not more important on the map. Each county is one shape whatever its population, so a large, sparsely populated county draws the eye more than a small, dense one.
Comparing two counties
Every PLACES estimate comes with a 95% confidence interval. Pick two counties and watch how the intervals change the reading.
Loading CDC county data…
Crude versus age-adjusted
The PLACES measures on Vellora are crude prevalence: the share of all adults, with no adjustment for age. Conditions such as coronary heart disease become more common with age, so a county with many older residents can show a higher crude figure without its residents being at higher risk at any given age.
The heart disease death rate is age-adjusted to the 2000 U.S. standard population, which removes differences that come only from age structure. Never compare a crude figure with an age-adjusted one, or a percentage of adults with a rate per 100,000.
Reading an association
The Research page reports how strongly two county measures move together (Pearson and Spearman correlation), how many counties were left out because a value was missing, and a p-value adjusted for testing several pairs at once.
- A strong correlation does not mean one measure causes the other. Both may follow age, income, rurality or access to care.
- It is an ecological comparison: it describes counties, not people.
- PLACES measures are produced by related models from the same survey, which can make them look more alike than the underlying conditions are.
See an example: high blood pressure and heart disease deaths
Three numbers people often misread
- Confidence intervals
- The range after a value (for example, 95% CI 8.1 – 9.9) shows the uncertainty in the estimate, as published by its source. Wider ranges mean less certainty. Poverty uses the Census Bureau’s 90% margin of error instead. Overlapping or non-overlapping ranges alone do not establish whether a difference between counties is meaningful.
- Percentage points
- The gap between two percentages is measured in percentage points. A change from 9% to 12% is 3 percentage points, which is a 33% relative increase. Vellora labels which of the two it shows.
- Modelled estimates
- CDC PLACES values are model-based small-area estimates: CDC combines national BRFSS survey responses with county population data in a statistical model rather than counting every resident. They describe likely county levels, and CDC states the model does not support tracking change over time.
Before you draw a conclusion
- Is this an estimate, an observed count, or an association?
- Which year of data, and which population (all adults, adults 35+)?
- Is it crude or age-adjusted, and are you comparing like with like?
- How wide are the confidence intervals?
- Were any counties withheld, and could that change the picture?
- What else differs between the places you are comparing?