Guidance for Comparing States’ Cancer Data

Note: For additional information on data interpretation, please refer to the USCS Technical Notes—Interpreting the Data.

Careful consideration is needed in interpreting and comparing rankings of state cancer rates.

A natural reaction of some readers when looking at figures that rank their state’s cancer rates is to seek explanations as to why their state has higher incidence or death rates for some cancers than other states or than the national average. For example, some may be alarmed that exposure to environmental carcinogens may be responsible when in fact there are several other more likely explanations. The following points should be kept in mind when interpreting or comparing these rankings—

Differences Among Racial and Ethnic Populations

Some cancers have different cancer rates for different racial and ethnic populations. For example, breast cancer incidence rates are usually higher in white women than in women of other racial and ethnic populations, and prostate cancer incidence rates are higher in black men. Therefore, when comparing cancer rates across states, the racial makeup of the state’s population should be taken into account and may be determined through the statistical adjustment of rates by race and ethnicity. However, presentation of rates for specific racial and ethnic populations may be preferable and is more easily understood by a lay audience.

Variations in Populations and Health Behaviors

Some differences in cancer rates among states may be explained by differences in known risk factors among the populations of those states. For example, one finds higher rates of lung cancer and other tobacco-associated cancers in states with higher rates of smoking. Although environmental carcinogens are responsible for some cancer cases, a majority of cases appear to be related to lifestyle factors such as smoking, and geographic variations in cancer rates are thought largely to reflect variations in these lifestyle factors.

Variations in Medical Care

Variations among states in medical care factors may also result in differences in cancer rates. In states where higher percentages of the population participate in cancer screening, more cancers will be diagnosed. Screening leads to earlier detection of tumors that have a better prognosis and may at times find tumors that grow so slowly that they would not otherwise be recognized in a person’s lifetime. Therefore, the cancer incidence rate without additional information only tells part of the story.

Influence of Aging on Cancer Rates

The likelihood of being diagnosed with cancer increases steadily with age. These rates have been adjusted for age so that states can be compared without concern that differences in their rates result from differences in the age distribution of their populations. However, this adjustment may be imperfect if the relationship between age and cancer risk is not the same for all states.

Measuring Burden

The importance of cancer as a public health problem in a state is more a function of the absolute rate of cancer rather than the state’s relative ranking in incidence or mortality. For example, Utah has proportionately fewer smokers than other states and also has the lowest lung cancer incidence rate of any state. Nevertheless, in Utah lung cancer kills more people than any other cancer, a fact that might be overlooked if one focused only on its low ranking in incidence compared with other states. Also, the true burden of cancer on the health care system and economy of a state is determined by the number of people diagnosed with or the number of people dying of cancer and not by the age-adjusted cancer rate. Therefore, the observation that the cancer rate in one state appears high compared with other states may obscure the fact that the absolute number of cases is not large.

Completeness of Cancer Incidence Data

States contribute cancer incidence data to these ranking figures if their registries collected 90% or more of the cancers. Because states vary in their completeness above 90%, rankings may vary to a minor extent because of differences in reporting completeness.

Random Factors and Cancer Rates

Even if registries were able to collect 100% of diagnosed cancer cases, there would still be some uncertainty in computed cancer rates because many factors contribute to the incidence and death rate in any given year or state, and some factors exhibit random behavior. Chance plays a role in determining if and when cancer develops in an individual, whether that cancer is detected, whether the information is entered into the cancer registry, and whether that cancer progresses and leads to death. For these reasons, the reported rates are expected to vary from year to year within a state even in the absence of a general trend. Caution is warranted, therefore, when examining cancer rates for a single year, and especially when the rates are based on a relatively small number of cases.

Confidence Intervals

A 95% confidence interval for the rate is an interval that is expected to contain the true underlying rate 95% of the time. Confidence intervals around the observed state age-adjusted rates are available to help with interpreting the results. Because of the variation in the population sizes and number of reported cases and deaths across states, there is more uncertainty in the incidence and death rates for some states compared with others. The confidence intervals provide a measure of the variability in the rates and some perspective for making state-specific comparisons. It should be noted, however, that using overlapping confidence intervals to conclude that rates are not significantly different is not recommended. This is a conservative test because it fails to detect significant differences more often than does standard statistical hypothesis testing.

Public Health Importance

Another consideration when comparing differences between rates is their public health importance. For some rates in this report, numerators and denominators are large and the standard errors are small with the result that some statistically significant differences may be so small as to lack importance for decisions related to population-based public health programs.