Time-dependent Cohort Summary Statistics in Atlas

The Atlas Characterizations option allows you to create cohort-level descriptive summary statistics.

Time-dependent statistics (e.g. Demographics Age) are based on the first time a person enters the cohort. This time is defined using Cohort Entry Events when creating a cohort in Cohort Definitions.

In some cases (e.g. GWAS), cohorts can be created either using Inclusion Criteria or Cohort Entry Events in Cohort Definitions. However, cohorts for time-dependent summary statistics in Characterizations must be created using Cohort Entry Events.

Correct time-dependent summary statistics using Cohort Entry Events

Below, we have defined a cases cohort for e_xcessive earwax_ using Cohort Entry Events. The first time a person enters this cohort is when e_xcessive earwax_ is diagnosed for the first time in a person’s life.

We have also defined a controls cohort using Cohort Entry Events. The first time a person enters this cohort is the earliest time when they have any health record (Any Visit) in FinnGen.

In Characterizations, the cases cohort will have the blue age distribution plot showing that the first time a person is diagnosed with e_xcessive earwax_ is usually between 60 and 85 years with an average age of 65 years. The controls cohort will have a much younger age distribution with an average age of 25 years.

Incorrect time-dependent summary statistics using Inclusion Criteria

The summary statistics will change if we define the cases cohort using Inclusion Criteria after first adding all persons into the cohort using Cohort Entry Events.

With the settings above, the summary statistics for the cases cohort is also based on the first time a person has any health record in FinnGen. The summary statistics no longer show the time of the first excessive earwax diagnosis.

Summary

• Time-dependent summary statistics (e.g. Demographics Age) depend on Cohort Entry Events to correctly define when a person enters the cohort

• Cohorts for summary statistics that are not time-dependent (e.g. Demographics Gender) can be defined either using Cohort Entry Events or Inclusion Criteria