# Using Age at Diagnosis in Selection and Table Tabs

In a Limited-Duration Prevalence session, care should be taken if selecting or displaying cases by age at diagnosis, especially if percents are being calculated. In prevalence, age at diagnosis is not linked to the population data (age at prevalence is), and therefore, populations will not be affected by the use of this variable. When using this variable, the responsibility is left to you to understand the population at risk for any given statistic. By definition, some statistics may be zero while others may be incomplete.

For example, age at diagnosis could be used to calculate the prevalence of childhood cancers. Suppose that you want to calculate 24-year limited-duration prevalence of childhood cancer. For this example, we will define childhood cancer as any cancer diagnosed at an age less than 20 years old. In this case, all estimated prevalence statistics will be for people with a prevalence age less than 44 years. Someone diagnosed with cancer at an age less than 20 within the previous 24 years could not be over 43 years old at the prevalence date. The populations will not be automatically subsetted to account for this.

Consider the results calculated for this example for each age group. If calculating 24-year limited-duration prevalence for childhood cancer (age at diagnosis is less than 20), then the following statements are true:

Age at prevalence is 0-23:

• The entire childhood of people ages 0-23 (at prevalence) was during the limited-duration period.

• Therefore, all cases are counted; this is essentially complete prevalence. Counts, populations, and percents will match results for all ages combined; that is, as if no selection were made using age at diagnosis.

Age at prevalence is 24-43:

• People aged 23-43 on the prevalence date became 20 years old at some point in the duration period. They could have been diagnosed with a childhood cancer, or they could have been diagnosed with a non-childhood cancer, if they were aged 20 or older at diagnosis.

• Therefore, the counts, but not the populations, are lower than if no selection had been made using age at diagnosis.

Age at prevalence is 44 or older:

• By definition, people aged 44 or older could not have been diagnosed with a childhood cancer during the duration period. That is, no one diagnosed with a childhood cancer in the past 24 years could now be age 44 or older.

• Prevalence estimates for these cells are zero by definition.

• The populations for these cells are the total number of people aged 44 or older on the prevalence date. That is, the populations would be the same if no selection had been made using age at diagnosis.

The concepts described above are particularly important when evaluating table cells for age groups that overlap. For example, the cells for age at prevalence equal to 40-44 and all ages combined include people who could not be counted by definition as well as people who could. These cells could be easily misinterpreted.