4.6: Public Health Statistics

Public health uses data -- including a lot of numerical data -- to understand what's happening at the level of population health. Public health relies on a number of different types of statistics that are routinely collected, such as births, deaths, hospitalizations, and rates of disease. All of these are discussed below.

Health statistics are analyzed by epidemiologists and biostatisticians (think of it as "math for the life sciences") to (a) describe what's happening in a population and (b) understand the relationships among different exposures and the diseases or illnesses associated with them.

Vital statistics, prevalence, incidence, life expectancies -- all these are examples of health statistics.

Like all sciences, epidemiology has to deal with a degree of uncertainty. That isn't because science isn't working -- it's because it is the nature of science to be constantly evolving. When I was a kid, the idea that earthquakes were due to movement of huge plates on the earth (plate tectonics) was just one theory -- now it is the dominant explanation for why earthquakes arise. When I was a kid, we thought ulcers were caused by stress and if you had an ulcer, you were told to drink milk and rest more. Now we know that stomach ulcers are caused by bacteria and you take an antibiotic, not milk, to cure it. Knowledge is expanding through a slow accumulation of many different studies that clarify the answer to scientific questions over time. In the Covid-19 epidemic, you may remember, there have been a number of scientific discoveries -- and also some scientific mistakes. For example, it took a very long time for epidemiologists to confirm that covid could spread via aerosols, not just droplets. (Both aerosols and droplets are small water drops in the air that contain the virus, but they behave very differently. Droplets can only travel about 6 feet and they fairly quickly fall to the ground. Aerosols are much smaller and can stay suspended in air for a longer time and can move around with air currents. Masks are ideal for protecting against the transmission of covid through droplets. Ventilation is probably our best tool for protecting against the transmission of covid through aerosols -- although masks also help.) Uncertainty is a natural part of the scientific process.

Statistical analysis often produces a result with a "margin of error." For example, in elections, you might see that Candidate A is leading over Candidate B by a margin of "two to five percentage points." That's a margin of error -- they know the real number is somewhere between two and five, but they can't quite pin it down. Still, it provides a useful snapshot of what's happening in the election campaign. Similarly, in public health, we might have a "margin of error" around the measurement of how often an exposure is associated with a disease. For example, in the US, people who smoke tobacco die an average of 10 years younger than people who never smoked tobacco. That 10 years isn't a precise number -- there's an association between tobacco and premature death that falls in a margin of error around 10 years. When the "margin of error" is all above zero, we can still be certain that the exposure to tobacco (in this case) is truly related to the health outcome (premature death). When the "margin of error" includes the number 0, then we can't be sure. For example, in that election example, if we said that Candidate A is ahead of Candidate B by 3 percentage points, give or take 4 points, then our margin of error is -1 to 7% (taking that 3% and looking at the numbers 4 points below and 4 points above.) If Candidate A is -1% "ahead" then they aren't really ahead! It would be more accurate to say the two Candidates right now are tied, and we can't tell who is ahead, though it seems Candidate A might be taking the lead.

Below, we discuss various types of health statistics that public health uses most often.

OPTIONAL Video defining epidemiology

The epidemiologist in this video speaks rather fast, but he does an excellent job of explaining what epidemiology is, in just four minutes with useful diagrams. You can turn on captions (the CC symbol on the bottom bar of the video) if you like. (It may take a moment to load, because I set it to start about 2 minutes into the video.) I appreciate how he breaks down epidemiology's focus on examining the connection between an exposure to a germ or a condition and the health outcome that it might produce.