Cohort tables

Cohort table

A cohort is a set of people that fulfill a certain set of criteria for a period of time.

In omopgenerics we defined the cohort_table class that allows us to represent individuals in a cohort.

A cohort_table is created using the newCohortTable() function that is defined by:

A cohort table.
A cohort set.
A cohort attrition.

Let’s start by creating a cdm reference with just two people.

person <- tibble(
  person_id = c(1,2),
  gender_concept_id = 0, year_of_birth = 1990,
  race_concept_id = 0, ethnicity_concept_id = 0
)
observation_period <- dplyr::tibble(
  observation_period_id = c(1,2), person_id = c(1,2),
  observation_period_start_date = as.Date("2000-01-01"),
  observation_period_end_date = as.Date("2025-12-31"),
  period_type_concept_id = 0
)
cdm <- cdmFromTables(
  tables = list(
    "person" = person,
    "observation_period" = observation_period
  ),
  cdmName = "example_cdm"
)
cdm
#> 
#> ── # OMOP CDM reference (local) of example_cdm ─────────────────────────────────
#> • omop tables: person, observation_period
#> • cohort tables: -
#> • achilles tables: -
#> • other tables: -

Now let’s say one of these people have a clinical event of interest, we can include them in a cohort table which can then be used as part of an analysis.

cohort <- tibble(
  cohort_definition_id = 1, subject_id = 1,
  cohort_start_date = as.Date("2020-01-01"),
  cohort_end_date = as.Date("2020-01-10")
)
cdm <- insertTable(cdm = cdm, name = "cohort", table = cohort)
cdm$cohort <- newCohortTable(cdm$cohort)

The cohort table will be associated with settings and attrition. As we didn’t specify these in newCohortTable() above they will have been automatically populated. You can access the cohort set of a cohort table using the function settings()

settings(cdm$cohort)
#> # A tibble: 1 × 2
#>   cohort_definition_id cohort_name
#>                  <int> <chr>      
#> 1                    1 cohort_1

Meanwhile, you can access the cohort attrition of a cohort table using the function attrition()

attrition(cdm$cohort)
#> # A tibble: 1 × 7
#>   cohort_definition_id number_records number_subjects reason_id reason          
#>                  <int>          <int>           <int>     <int> <chr>           
#> 1                    1              1               1         1 Initial qualify…
#> # ℹ 2 more variables: excluded_records <int>, excluded_subjects <int>

Cohort attrition table is also used to compute the number of counts that each cohort (ie from the last row of the attrition). It can be seen with the function cohortCount().

cohortCount(cdm$cohort)
#> # A tibble: 1 × 3
#>   cohort_definition_id number_records number_subjects
#>                  <int>          <int>           <int>
#> 1                    1              1               1

Note that because the cohort count is taken from the last row of attrition, if we make changes to a cohort we should then update attrition as we go. We can do this

cdm$cohort <- cdm$cohort |>
  filter(cohort_start_date == as.Date("2019-01-01")) %>% 
  recordCohortAttrition("Require cohort start January 1st 2019")
attrition(cdm$cohort)
#> # A tibble: 2 × 7
#>   cohort_definition_id number_records number_subjects reason_id reason          
#>                  <int>          <int>           <int>     <int> <chr>           
#> 1                    1              1               1         1 Initial qualify…
#> 2                    1              0               0         2 Require cohort …
#> # ℹ 2 more variables: excluded_records <int>, excluded_subjects <int>
cohortCount(cdm$cohort)
#> # A tibble: 1 × 3
#>   cohort_definition_id number_records number_subjects
#>                  <int>          <int>           <int>
#> 1                    1              0               0

An additional, optional, attribute keeps track of the concepts used to create the cohort. In this example we do not have a codelist associated with our cohort.

cohortCodelist(cdm$cohort, cohortId = 1, type = "index event")
#> Warning: No codelists found for the specified cohorts
#> 
#> ── 0 codelists ─────────────────────────────────────────────────────────────────

We could though associate our cohort with a codelist

cdm$cohort <- newCohortTable(cdm$cohort, 
                             cohortCodelistRef = dplyr::tibble(
                                 cohort_definition_id = c(1,1),
                                 codelist_name =c("disease X", "disease X"),
                                 concept_id = c(101,102),
                                 type = "index event"
                               ))
cohortCodelist(cdm$cohort, cohortId = 1, type = "index event")
#> 
#> ── 1 codelist ──────────────────────────────────────────────────────────────────
#> 
#> - disease X (2 codes)

Each one of the elements that define a cohort table have to fulfill certain criteria.

Cohort Set

A cohort set must be a table with:

Lower case column names.
At least cohort_definition_id, cohort_name columns (cohortColumns("cohort_set")).
cohort_name it must contain unique cohort names (currently they are cased to snake case).
cohort_definition_id it must contain unique cohort ids, all the ids present in table must be present in the cohort set and the same ids must be present in cohort attrition.

Cohort Attrition

A cohort attrition must be a table with:

Lower case column names.
At least cohort_definition_id, number_records, number_subjects, reason_id, reason, excluded_records, excluded_subjects columns (cohortColumns("cohort_attrition")).
cohort_definition_id it must contain cohort ids, all the ids present in table must be present in the cohort attrition and the same ids must be present in cohort set.
There must exist unique pairs of cohort_definition_id and reason_id.

Cohort Codelist

A cohort codelist must be a table with:

Lower case column names.
At least cohort_definition_id, codelist_name, concept_id, type columns (cohortColumns("cohort_codelist")).
cohort_definition_id it must contain cohort ids, all the ids present in table must be present in the cohort attrition and the same ids must be present in cohort set.
type must be one of “index event”, “inclusion criteria”, and “exit criteria”

Cohort Table

A cohort table must be a table with:

It comes from a cdm_reference (extracted via cdm$cohort).
It has the same source than this cdm_reference.
Lower case column names.
At least cohort_definition_id, subject_id, cohort_start_date, cohort_end_date columns (cohortColumns("cohort")).
There is no record with NA value in the required columns.
There is no record with cohort_start_date after cohort_end_date.
There is no overlap between records. A person can be in a cohort several times (several records with the same subject_id). But it can’t enter (cohort_start_date) the cohort again before leaving it (cohort_end_date). So an individual can’t be simultaneously more than once in the same cohort. This rule is applied at the cohort_definition_id level, so records with different cohort_definition_id can overlap.
All the time between cohort_start_date and cohort_end_date (both included) the individual must be in observation.