Understanding & Interpreting Population Health Measures
Updated: Nov 15, 2022
Population health measures give us the information we need to assess and improve community health. They help us answer the questions we have about communities by quantifying health and well-being outcomes, community conditions, and other factors. For example, if we are curious about the overall health of our communities, we might look at infant mortality rate, proportion of people living in poverty, and median household income—these measures help paint a picture of community health.
Many common population health measures either a) summarize a distribution, like means and medians, or b) characterize a part of a distribution, like ratios, proportions, and rates. Here, “distribution” refers to a set of data from which we can generate information, for example survey responses, test scores, daily recorded temperatures, and case counts. We use population health measures—specific measures of the distribution—to make sense of data. Learning to interpret community health measures requires a foundational understanding of what the different measure types indicate and how they can be used. Here, we summarize common types of population health measures—many of which are included on IP3 | Assess— and give examples of how they are applied. It’s important to interpret community health measures accurately, in order to glean key takeaways and implications for future work to advance community well-being.
Summarizing a Distribution: Measures of Central Location and Spread
Measures of central location and spread are useful in summarizing a distribution of observations for continuous variables, which can have an unlimited number of values between the lowest and highest points of measurement. Measures of central location like mode, mean, and median describe where the peak—the center—of observations is located. Measures of spread or dispersion, like standard deviation, percentile, quartile, and range, describe the dispersion or variation of values. Measures of central location and spread help us understand the contours of our distribution.
Mean (Central Location)
Use in Practice
The mean, arithmetic mean, or average is the value that is closest to all other values in a distribution (the arithmetic center). Means are calculated by adding all observed values in a distribution, and dividing the sum by the total number of observations.
When data are normally distributed, (the distribution is symmetrical and has the classic bell-shaped curve), the mean is our best measure of central location. The mean is considered the center of gravity of the distribution. In the example of Particulate Matter 2.5 Level, our distribution is thousands of individual observations of ambient concentration of PM2.5 throughout a year. To make sense of it, we take the mean and interpret it as the center of all observations in the distribution.
Poor Mental Health Days - Age-adjusted average number of reported mentally unhealthy days per month
Particulate Matter 2.5 Level - Average annual ambient concentrations of PM 2.5 in micrograms per cubic meter
Median (Central Location)
Use in Practice
The median (also the 50th percentile) is the middle of a set of data put into rank order. The median is equal to a single observed value (if there are an odd number of observations) or the average of two observations (if an even number of observations).
When data are asymmetrical or skewed, median is our choice measure of central location, as it is less affected by outliers or extreme values than arithmetic mean. We can interpret medians as the number that divides a distribution in half, with half of all observations being higher than the median and half being lower. In the example of Median Age, half of the population is older and half the population is younger than the value.
Median Age - Median age (in years) for the population
Median Household Income - Median household income (in U.S. Dollars) for the population
Use in Practice
Percentiles divide a distribution into 100 equal parts such that each percentile contains 1% of total observations. The Pth percentile (P ranges from 0 to 100) is equal to the value that P% of observations fall at or below. In other terms, a percentile gives us the percentage of observations at or below it.
Percentiles can be used to help characterize various parts of a distribution. For example, Inequality - 80/20 Ratio examines two percentiles—the 20th and the 80th percentiles—which are equal to the incomes for the top 20% and bottom 20% of households, respectively.
Inequality - 80/20 Ratio - Ratio of household income at the 80th percentile to income at the 20th percentile, representing the disparity in incomes between the top 20% and the bottom 20% of households
Absolute Upward Mobility - Expected income by percentile rank for children whose parents are at the 25th percentile of the national income distribution
Beyond the measures described above there are other measures of central location—mode, midpoint, geometric mean–and spread–range, quartiles, standard deviation, standard error—that may be important to consider in your data and measurement work. (Learn more.)
Standard deviation is an important measure of spread and is core to the scoring methodology on IP3 | Assess. Standard deviation tells us how widely or tightly observations are distributed from the mean. IP3 | Assess uses standard deviations to calculate a Z-score for any given geography in the platform, which tells us how many standard deviations you are away from the mean. (Read more!)
Characterizing Part of a Distribution: Frequency Measures
Frequency measures characterize part of a distribution to another part of the distribution, or to the entire distribution. Frequency measures include ratios, proportions, and rates. In contrast to measures of central location and spread, which can be used to characterize continuous variables, frequency measures are used to summarize categorical variables. They tell us how often a type of observation occurs.
Use in Practice
Ratios are calculated by dividing one variable by another. The output represents the relative magnitude of the two variables. Ratios are often normalized such that either the numerator or the denominator equal one.
Ratios are used to compare two values, regardless of whether their denominators are related. This allows for an “apples-to-oranges” kind of comparison. In the example, Student-Teacher Ratio, the number of total public school students is compared to the number of Full Time Equivalent (FTE) teachers. The ratio is given in terms of the number of students to teachers, i.e., X number of students to 1 FTE teacher.
Student-Teacher Ratio - Ratio of public school students to FTE classroom teachers
Gender Pay Gap - Ratio of women's median earnings to men's median earnings for all full-time, year-round workers
Use in Practice
Proportions compare a part of a distribution to the whole. They are a type of ratio–one in which the numerator is included in the denominator. To calculate, divide the observations of interest (part) by the total number of observations (whole). Proportions can be expressed as decimals, fractions, or percentages.
If ratios allow for an “apples-to-oranges” comparison, proportions allow for an “apples-to-all-fruit” kind of comparison, as the numerator must be included in the denominator. In the example Current Smoking, the number of adults who have smoked 100 or more cigarettes in their lifetime and currently smoke everyday is compared to the total number of adults; the product is multiplied by 100 and presented as a percentage.
Current Smoking - Percentage of adults aged 18 years and older who report having smoked 100 or more cigarettes in their lifetime, and currently smoke every day or some days
Vacant Housing - Percentage of residential addresses that are vacant
Use in Practice
A rate is a measure of frequency with which an event occurs in a defined population over a specified period of time. Rates are calculated by dividing the total number of events (e.g., diagnoses, deaths, incidents) during a defined time period (e.g., 1 year, 5 years) by the population. Rates are typically normalized to a standard population size, such as 1,000 or 100,000.
In population health, we often use rates to describe disease or mortality frequency (i.e. incidence and mortality rates). Rates allow us to compare disease frequency across geographies, groups, and time periods. In the example of Chlamydia, the count of new chlamydia cases over a five year period is divided by the population exposed and normalized to a population of 100,000.
Chlamydia - Number of new chlamydia cases per 100,000 population
Infant Deaths - Number of deaths among infants (less than one year of age) per 1,000 live births
Other Types of Measures
Besides measures of central location and spread and frequency measures, you’re likely to encounter a variety of other measures, including aggregate measures, summary measures, measures of disparity, and measures of association.
Weaving Data with Story and Experience
Population health measures are one piece of the puzzle in using data to drive community health improvement. There are many sources of data and different ways of knowing that are critical to consider as well. At IP3, we recommend using multiple types of data and information, including qualitative data and storytelling, to weave together a portrait of your community that is shared and iterated upon by the public, and particularly by priority populations. We also encourage engaging community residents with lived experience and asking about the extent to which the population health data resonates with what life is really like in a community.
Stratifying measures by demographic groups, like race, ethnicity, gender, age and other population groups can expose important disparities. By looking at a single measure for the population as a whole, and also for groups within the population, we can identify population groups to prioritize in order to improve health equity (see below, where we explore infant deaths, the number of deaths among infants (less than one year of age) per 1,000 live births.)
Population Health Measures for Community Action
Our choice of measure when exploring population health data is important—it shapes how we see the world and plan for community health improvement. For example, we recently helped a client select an ACES (Adverse Childhood Experiences) measure as part of their assessment work. We zeroed in on two ACEs measures: Mean number of ACEs reported by adults, and Proportion of adults who report 4 or more ACES. The mean tells us on average how many ACEs an adult experienced, while the proportion gives us a sense for how large the group with severe exposure to ACEs in a given community is. This client wanted to prioritize severe ACEs in their assessment, so we selected the proportion for their framework.
The types of measures selected when examining a community shape how we see a community and give a directive for taking action. Selecting and interpreting appropriate population health measures is imperative to using data to inform and drive strategies to advance equitable well-being. IP3 | Assess is our web-based data solution to community assessment and supports changemakers exploring population health data by making hundreds of indicators immediately available for any given geographic area or community. If you're interested in learning more, please schedule a call with our team!