How to Read an Output Table from the WISH Cancer Module
Here is an example of an output table from the WISH Cancer Module:
1. Understand rows, columns and labels
In the example table, the rows are labeled according to the regions of residence: Southern, Southeastern, Northeastern, Western, and Northern. The columns are labeled Cancer Incidence, Population, Cancer Incidence Rate and 95% Confidence Interval.
The column headings indicate the table provides four kinds of information related to age-adjusted rates for cancer in Wisconsin for years 2004-2008. Under "Detail Information," Column 1 shows the number of cancer cases (frequency) for each region selected for the query. Column 2 shows the population denominator for each region, used in the calculation of the rate. Column 3 shows the age-adjusted cancer incidence rate, the selected measure for the query. Column 4 shows the 95% confidence interval, which is provided for all rates shown in the Cancer Module.
2. Read the statistics
Looking at the row labeled "Southern," we see that in the Southern Region in the five-year period 2004-2008:
3. Interpret numbers carefully and with caution before reaching conclusions
Descriptive statistics seem straightforward, but it is always advisable to consider all the relevant data before reaching a conclusion. For example, the table above shows the Northern Region has the fewest cancer cases; however, this should be interpreted in view of the population for that region, which is the smallest of all five regions. The age-adjusted rates show the Northern Region does not have the lowest age-adjusted rate, but rather a rate higher than rates for the Southern and Western regions. Reviewing the confidence intervals, the Northern Region's rate is significantly higher than rates for the Southern and Western regions. This means that although the Northern Region had the lowest number of reported cancers during 2004-2008, this region had the third highest age-adjusted rate of cancer incidence.
Purpose of age-adjusted rates. Cancer occurs more frequently with increasing age, so a population with a larger proportion of elderly will have more cancers than a younger population of the same size, assuming all other contributing factors are the same. To make meaningful comparisons, the age distributions of the populations are weighted to one standard population, in this case the 2000 United States population. This direct age-adjustment method removes the bias due to age differences between populations. By convention, cancer registries in the U.S. currently adjust incidence and mortality rates to the 2000 U.S. population. It should be remembered that the age-adjusted rate is a hypothetical number: it is the rate that would occur if the population of interest had the same age distribution as the U.S. standard population.
How to interpret confidence intervals. Confidence intervals (95%) are included with rates in WISH tables to facilitate comparisons. The range between the lower and upper confidence interval limits defines where the "true" age-adjusted rate would be with 95 percent probability. A narrow confidence interval implies that the rate has been more precisely estimated, whereas a wider confidence interval implies less certainty that the calculated rate is the true rate.
Comparing two confidence intervals can serve as a preliminary test of the statistical significance of differences. Generally, when the confidence interval for the area of interest does not overlap with the confidence interval for the comparison area, we say that the difference between the two areas is statistically significant; i.e., the difference between the two rates is more than would be expected by random variation or chance. However, if we are making many comparisons, we may still find statistically significant differences just by chance. In fact, with a 95% confidence interval, we expect that 5% of the comparisons will show a statistically significant difference by chance. Thus, with 72 counties and 26 cancer sites, we might see as many as 93 instances where the rate for a county is statistically significantly different from the state rate just by chance.
Purpose of the query. Finally, interpret the statistics based on the purpose of your query and how the information will be used. If hospital administrators are making plans for cancer treatment programs, the number of cancers and the crude rate for their county would be relevant. If researchers are interested in the cancer incidence surveillance for environmental risk factors, age-adjusted rates would be more useful.
Last Revised: February 12, 2014