CI*Rank: Frequently Asked Questions

Why Compare Ranks Instead of Rates?

Ranks are based on the age-adjusted rates of the corresponding measures. This website provides plots and tables for both ranks and rates. After selecting Database and Data Retrieval Options, the Results will show the plot and table of Confidence Intervals (CI) for Ranks.

Example: California County-level Female Breast Cancer Incidence, 2006-2010

Confidence Intervals for Rankings
Ranking Based on Age-Adjusted US Cancer Incidence Rates for California - ranks graph
Confidence Intervals for Age-Adjusted Rates
Ranking Based on Age-Adjusted US Cancer Incidence Rates for California - rates graph

The red dots in the plots show the rankings and rates of female breast cancer age-adjusted incidence for the counties in California in 2006-2010, and the blue lines represent the confidence intervals of the rankings and rates. Since rankings for the counties are consecutive integers (e.g., 1, 2, ...), the red dots form a straight line. Glenn County, with an annual average population of 13,842, was ranked No. 37, but the confidence interval (2, 57) overlaps with most other counties.

View the full results for the ranks of female breast cancer incidence in California. By clicking the “CI for Rates” tab on the results page, the plot and table of confidence intervals for rates will appear.

There are two points to pay attention to:

  1. In the “CI for Ranks” plot, the red dots form a straight line because rankings for the counties are consecutive integers.
    • Because age-adjusted rates for consecutively ranked counties may differ by any amount, the red dots that show the observed rates in the “CI for Rates” plot are no longer a straight line.
    • The order of the counties in the rate plot remains the same as in the rank plot because they are ordered by rates from the highest to the lowest.
  2. The variability of a county’s rate depends on the population size in the county, but the variability of a county’s rank not only depends on the variability of that county’s rate, but also on the variability of the closely ranked counties’ rates. For example
    • Los Angeles County has an annual population of 4.9 million women, so the CI for the rate is very narrow, between 114.1 and 119.3, but the CI for the rank is quite broad, between 29 and 43.
    • While Los Angeles County’s rate is stable, the rank of Los Angeles depends not only on the variability of the Los Angeles County rate, but also of the variability of closely ranked counties. Those counties (Siskiyou, Humboldt, Glenn, Kern, and Yuba) all have very small populations. Because those counties' rates are quite variable, the ranks will change considerably due to the random variability of the group of counties.
Los Angeles and Closely Ranked Counties
Female Breast Cancer Incidence Rates, 2006-2010, with 95% CI
California County Rank Lower CI
for Rank
Upper CI
for Rank
Rate Lower CI
for Rate
Upper CI
for Rate
Population
Siskiyou 34 5 57 118.0 83.2 146.2 22,508
Humboldt 35 11 55 117.0 96.9 138.0 66,504
Los Angeles 36 29 43 116.6 114.1 119.3 4,946,293
Glenn 37 2 57 116.5 68.4 164.2 13,842
Kern 38 25 50 115.8 106.5 125.4 394,645
Yuba 39 7 57 115.5 81.4 143.2 35,252

What Questions Can CI*Rank Answer?

The following hypothetical scenarios are provided to demonstrate possible use cases for CI*Rank.

  • Scenario 1: Erie County, New York, which is located by Lake Erie and includes the city of Buffalo, has a population of 919,000 according to the 2010 Census. The County Department of Health is investigating the mortality rates of screenable cancers, i.e. cancers in colon and rectum, female breast, and cervix, in order to design and implement screening programs
  • Scenario 2: The Pennsylvania Department of Health is promoting an intervention program to prevent and control coronary heart disease across the state. Someone in the department uses the CI*Rank website to look for information. The question is: which counties have the highest coronary heart disease mortality rates?
  • Scenario 3: The State of Kentucky has a total of 120 counties, and most of these counties are in sparsely populated rural areas. Instead of county-level rankings, a staff member from the State Department for Public Health is looking for rankings of lung cancer mortality rates at the level of Area Development District (ADD).

Scenario 1

Erie County, New York, which is located by Lake Erie and includes the city of Buffalo, has a population of 919,000 according to the 2010 Census. The County Department of Health is investigating the mortality rates of screenable cancers, i.e., cancers in colon and rectum, female breast, and cervix, in order to design and implement screening programs. A staff member from the department looks at Erie County’s mortality rates of the three cancers and the corresponding rankings in the state. Here are some questions she asks:

  1. Where does Erie County rank for the rate of female breast cancer mortality in New York State in the most recent period?
  2. Where does Erie County rank for the mortality rates of colon and rectum cancer and cervical cancer?

Where does Erie County rank for the rate of female breast cancer mortality in New York State in the most recent period?

Step 1: Select Measure

  1. Select the Database: US Mortality 1990-2014 Adjusted Rates
  2. Select Geographic Area: Show Ranks by County for New York

database selection: US Mortality by county


Step 2: Data Retrieval Options

  1. Select options for Years 2006-2010, All Races, Female, All Ages
  2. Pick Breast Cancer as the Cause of Death.
  3. Change the Confidence Intervals setting to “Individual” from the default of "Simultaneous."

To choose which type of confidence interval to use, select "individual" if you are interested in a specific geographic region. If you are looking at the overall rankings across many geographic regions within a state or the whole country, use the default of "simultaneous." Because the staff member is looking at ranks and rates for Erie County specifically, "individual" confidence intervals are used in this scenario.

data retrieval options: New York state, breast cancer cause of death, individual confidence interval setting.

Step 3: Examine the Results

View the graph and table of CI for both Ranks and Rates for female breast cancer mortality rates in New York:

  • Erie County is ranked No. 1 for female breast cancer mortality rate during 2006 and 2010. The 95% CI for the ranking is between 1 and 9.
  • The age-adjusted mortality rate was 26.98 deaths per 100,000 women, with the 95% CI between 25.149 and 28.849 deaths.

Step 4: Determine where Erie County ranks for the mortality rates of colon and rectum cancer and cervical cancer.

Repeating the 3 steps in Question 1 for colon and rectum cancer (select both sexes) and then for cervical cancer (female only), she finds that

  • For colon and rectum cancer mortality
    • Erie County is ranked No. 38 during 2006 and 2010. The 95% CI for the ranking is between 27 and 51.
    • The age-adjusted mortality rate is 15.272 deaths per 100,000 men and women, with the 95% CI between 14.274 and 16.287 deaths.
    • View the results for colon and rectal cancer mortality rates in New York.
  • For cervical cancer mortality
    • Erie County is ranked No. 19 with a 95% CI of 12 to 22.
    • The age-adjusted mortality rate is 1.651 deaths per 100,000 women, with the 95% CI between 1.182 and 2.149 deaths.
    • View the results for 5-year cervical cancer mortality rates in New York.
    • Counties with fewer than 10 cervical cancer deaths are suppressed. As longer periods are considered, more cervical cancer deaths are included and fewer data are suppressed. It is generally advised to consider longer time periods in ranking small geographic areas so that few data are suppressed.

    • View the results for 10-year cervical cancer mortality rates in New York.
      • Erie County is ranked No. 29 with a 95% CI of 20 to 33.
      • The age-adjusted mortality rate is 1.809 deaths per 100,000 women, with the 95% CI between 1.453 and 2.185 deaths.
    • View the results for 15-year cervical cancer mortality rates in New York.
      • Erie County is ranked No. 41 with a 95% CI of 29 to 48.
      • The age-adjusted mortality rate is 1.953 deaths per 100,000 women, with the 95% CI between 1.662 and 2.265 deaths.

Step 5: Using the Results

With the data available on the CI*Rank website, Erie County decides to allocate more resources to promote breast cancer screening across the county.

Scenario 2

The Pennsylvania Department of Health is promoting an intervention program to prevent and control coronary heart disease across the state. Someone in the department uses the CI*Rank website to look for information. The question is: which counties have the highest coronary heart disease mortality rates?

Step 1: Database Selection

  1. Select the database, US Mortality 1990-2014 Age-Adjusted Rates.
  2. Select Geographic Area: Show Ranks by County for Pennsylvania.

database selection: US Mortality by county

Step 2: Data Retrieval Options

  1. Select the options for Years 2006-2010, All Races, Both Sexes, All Ages.
  2. Choose Coronary Heart Disease from the list of Other Causes of Death.
  3. Use the default Confidence Intervals setting of “Simultaneous."

To choose which type of confidence interval to use, select "individual" if you are interested in a specific geographic region. If you are looking at the overall rankings across many geographic regions within a state or the whole country, use the default of "simultaneous." Because the person is interested in the rankings of coronary heart disease across all the counties in the state, the default of 'Simultaneous' Confidence Interval is selected to show the overall picture of the disease burden in the state.

data retrieval options: Pennsylvania,coronary heart disease cause of death, simultaneous confidence interval setting.

Step 3: Examine the Results

View the graph and table of CI for Ranks for Pennsylvania's 67 counties.

The state uses these rankings to decide which counties need the intervention and to allocate scarce resources.

Scenario 3

The State of Kentucky has a total of 120 counties, and most of these counties are in sparsely populated rural areas. County-level rankings are quite variable because of the small population sizes. View the Lung cancer mortality rankings for Kentucky by county.

Instead of county-level rankings, a staff member from the State Department for Public Health is looking for rankings of lung cancer mortality rates at the level of Area Development District (ADD). There are a total of 15 ADDs in the state of Kentucky, each an aggregation of several counties. By aggregating counties into ADDs, the population sizes are increased and the ranking will be much more meaningful.

Because Kentucky ADDs were not in the original CI*Rank database, the person requested we add a special region. We added the Kentucky ADDs to the CI*Rank database, and now it’s available to any users who are interested in that special region.

To Select a Special Region

After selecting to view results by "Special Region" on the database selection form, a list of special regions currently available in CI*Rank is provided.

Kentucky ADD's

View the Results

Lung cancer mortality rankings by Area Development Districts (ADDs) in Kentucky

Because this person is looking at rankings across all ADDs, the default of simultaneous confidence intervals is used in the results. If this person was interested in one specific ADD in comparison to the others, then individual confidence intervals would be used.

To Request a Special Region be Added to CI*Rank

Contact information is provided in the "?" link next to the "Special Region" option on the database selection form.

Request a special region to be added to CI*Rank

How Meaningful Are Ranks in Different Scenarios?

CI*Rank will help to answer the following questions:

  • How meaningful are the ranks?
  • When are the ranks really different from each other?

The answers depend on the:

  • Size of the population of the geographic areas
  • Number of years of data included, and
  • The relative frequency of the cause of death or cancer type under consideration.

These four examples show various scenarios in ranking rates.

Ranking US Mortality by State, 2010, All Causes

The first graph shows the confidence intervals (CIs) of ranking state-level mortality rates for all causes of death for the year 2010. Displayed here are the states ranked as the top 25 with the highest mortality rates. The red dots are the mortality ranks and the vertical blue lines show the 95% CI. Because the total number of deaths in each state is high, the CIs are narrow except for the few states with sparse population (e.g., Alaska, Wyoming, etc.). View the full results in CI*Rank.

A graph displaying US mortality by state, 2010, all causes.

 

Ranking US Mortality by State, 2010, All Cancers

The second graph shows CIs of ranks for cancer mortality rates for all states and DC in 2010. Again, only the states ranked as the top 25 are displayed here. Because cancer deaths only account for a small proportion of all deaths, we can see the CIs of ranks are much wider compared to the previous graph. View the full results in CI*Rank.

A graph displaying US mortality by state, 2010, all cancers.

Ranking Incidence in New Jersey by County (2006-2010), All Cancers

The third graph shows the county-level rate ranking within a state. Shown here are all cancer incidence rate ranks for all New Jersey counties. To protect confidentiality of the county-level data, a minimum of 3 years is required for data retrieval. Shown here are all cancer incidence age-adjusted rate ranks for all New Jersey counties. Five years of data (2006-2010) are aggregated, and the CIs for large counties are very small compared to the less populated counties.  For example, Essex County has an annual average population of 780,947, and the CI for the ranking is (17, 20). Salem County, with an annual average population of 65,968, has a CI of (2, 14) for the rank. View the full results in CI*Rank.

A graph displaying incidence by county, New Jersey (2006-2010), all cancers.

Ranking Breast Cancer Mortality in New Jersey by County, (2006-2010), Black Females

In the fourth graph, we further focus on the deaths of a specific cancer (breast cancer) among a subpopulation (black females) in New Jersey counties. Unlike the previous example where the population size was driving the width of the CI, this example is showing that due to the small number of deaths (most counties have less than 100 black female breast cancer deaths during the 5-year period), the CIs are wide for all counties. The CIs almost all overlap with each other, indicating that ranking does not make much sense in this situation. Data are suppressed in five counties (Cape May, Hounterdon, Salem, Sussex, and Warren) with less than 10 deaths.  The CIs almost all overlap with each other, indicating that ranking does not make much sense in this situation. In situations like this, ranks should not be done to compare the geographic regions of interest. View the full results in CI*Rank.

A graph displaying breast cancer mortality, New Jersey (2006-2010), black females by county.

CI*Rank

Last Updated: 02 Jan, 2020