Skip to contents

Summarise treatments

Source: vignettes/summarise_treatments.Rmd
summarise_treatments.Rmd

Introduction

After creating a study cohort, for example of some specific condition of interest, we may be interested in describing the treatments received by the individuals within it. Here we show how such a summary can be obtained.

Create mock table

We will use mock data contained in the package throughout the vignette. Let’s modify cohort tables cohort1 and cohort2 in our mock dataset, so the first table includes 3 cohorts of health conditions (our study cohorts), while the second contains three are of treatments they could receive.

library(DrugUtilisation)

cdm <- mockDrugUtilisation(numberIndividual = 200)

new_cohort_set <- settings(cdm$cohort1) |>
  dplyr::arrange(cohort_definition_id) |>
  dplyr::mutate(cohort_name = c("asthma", "bronchitis", "pneumonia"))

cdm$cohort1 <- cdm$cohort1 |>
  omopgenerics::newCohortTable(cohortSetRef = new_cohort_set)

new_cohort_set <- settings(cdm$cohort2) |>
  dplyr::arrange(cohort_definition_id) |>
  dplyr::mutate(cohort_name = c("albuterol", "fluticasone", "montelukast"))

cdm$cohort2 <- cdm$cohort2 |>
  omopgenerics::newCohortTable(cohortSetRef = new_cohort_set)

Notice that cohort1 is a cohort table with three cohorts representing three different conditions:

settings(cdm$cohort1)
#> # A tibble: 3 × 2
#>   cohort_definition_id cohort_name
#>                  <int> <chr>      
#> 1                    1 asthma     
#> 2                    2 bronchitis 
#> 3                    3 pneumonia

And cohort2 is a cohort table with three different treatment cohorts:

settings(cdm$cohort2)
#> # A tibble: 3 × 2
#>   cohort_definition_id cohort_name
#>                  <int> <chr>      
#> 1                    1 albuterol  
#> 2                    2 fluticasone
#> 3                    3 montelukast

Summarise treatment

The summariseTreatment() function produces a summary of the treatment received by our study cohorts. There are three mandatory arguments:

  1. cohort: cohort from the cdm object.
  2. treatmentCohortName: name of the treatment cohort table.
  3. window: a list specifying the time windows during which treatments should be summarised.

See an example of its usage below, where we use summariseTreatment() to summarise treatments defined in cohort2 in the target cohorts defined in cohort1.

summariseTreatment(
  cohort = cdm$cohort1,
  treatmentCohortName = c("cohort2"),
  window = list(c(0, 0), c(1, 30))
)

strata parameter

We can also stratify our cohort and calculate the estimates within each strata group by using the strata parameter.

cdm[["cohort1"]] <- cdm[["cohort1"]] |>
  PatientProfiles::addSex() |>
  PatientProfiles::addAge(ageGroup = list("<40" = c(0, 39), ">=40" = c(40, 150)))

results <- summariseTreatment(
  cohort = cdm$cohort1,
  treatmentCohortName = c("cohort2"),
  window = list(c(0, 0)),
  treatmentCohortId = 1,
  strata = list("sex", "age_group")
)

Notice that we have also used the treatmentCohortId parameter to specify that we only want to explore albuterol (which corresponds to the cohort id 1 in our cohort table) across the cohorts defined in cohort1.

other parameters

The summariseTreatment() functions also has other input parameters which can be tuned:

  • cohortId: to restrict the analysis to a particular cohort definition id in the target cohort.
  • indexDate: what column to use as the index date to start the analysis. By default we use cohort_start_date, but any other date column, such as cohort_end_date, can be specified instead.
  • censorDate: whether to end the analysis at any specific date. Otherwise we will follow the individuals until end of their respective observation period.
  • mutuallyExclusive: by default set to FALSE, this will consider the treatments separately, so an individual can belong to different treatment groups at the same time (i.e. if they are treated with multiple drugs). Therefore, for each target cohort, we could have a sum of percentages of all treatment drugs greater than 100%. If set to TRUE, non-overlapping treatment groups will be assessed (with multiple drugs in some of those if needed), so that all individuals will belong to only one of them, and the percentages will add up to a 100.
result_not_mutually_exc <- summariseTreatment(
  cohort = cdm$cohort1,
  treatmentCohortName = c("cohort2"),
  window = list(c(0, 0))
)
#> ℹ Intersect with medications table (cohort2)
#> ℹ Summarising medications.

result_mutually_exc <- summariseTreatment(
  cohort = cdm$cohort1,
  treatmentCohortName = c("cohort2"),
  window = list(c(0, 0)),
  mutuallyExclusive = TRUE
)
#> ℹ Intersect with medications table (cohort2)
#> ℹ Summarising medications.

tableTreatment(result = result_not_mutually_exc)
CDM name
DUS MOCK
Treatment Estimate name
Cohort name
asthma bronchitis pneumonia
Medication on index date
albuterol N (%) 5 (7.58 %) 6 (11.76 %) 7 (8.43 %)
fluticasone N (%) 10 (15.15 %) 5 (9.80 %) 9 (10.84 %)
montelukast N (%) 6 (9.09 %) 5 (9.80 %) 2 (2.41 %)
untreated N (%) 45 (68.18 %) 35 (68.63 %) 65 (78.31 %)
not in observation N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) 
tableTreatment(result = result_mutually_exc)
CDM name
DUS MOCK
Treatment Estimate name
Cohort name
asthma bronchitis pneumonia
Medication on index date
albuterol N (%) 5 (7.58 %) 6 (11.76 %) 7 (8.43 %)
fluticasone N (%) 10 (15.15 %) 5 (9.80 %) 9 (10.84 %)
montelukast N (%) 6 (9.09 %) 5 (9.80 %) 2 (2.41 %)
albuterol and fluticasone N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)
albuterol and montelukast N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)
fluticasone and montelukast N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)
albuterol and fluticasone and montelukast N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)
untreated N (%) 45 (68.18 %) 35 (68.63 %) 65 (78.31 %)
not in observation N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)

In this case, as no individual was given more than one of the treatment drugs, the percentages are the same. However, we can easily see how these analyses would provide different results in other contexts.

Visualise results

The package includes table and plot functions to help visualise the results from summariseTreatment(), like we have just used to show the results above.

Tables

The tableTreatment() function generates a table in gt, flextable, or tibble format from the summarised_result produced by summariseTreatment(). This function has customisation options to format the table according to user preferences.

tableTreatment(result = results)
CDM name
DUS MOCK
Sex Age group Treatment Estimate name
Cohort name
asthma bronchitis pneumonia
Medication on index date
overall overall albuterol N (%) 5 (7.58 %) 6 (11.76 %) 7 (8.43 %)
untreated N (%) 61 (92.42 %) 45 (88.24 %) 76 (91.57 %)
not in observation N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)
Female overall albuterol N (%) 2 (6.90 %) 3 (11.11 %) 2 (5.13 %)
untreated N (%) 27 (93.10 %) 24 (88.89 %) 37 (94.87 %)
not in observation N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)
Male overall albuterol N (%) 3 (8.11 %) 3 (12.50 %) 5 (11.36 %)
untreated N (%) 34 (91.89 %) 21 (87.50 %) 39 (88.64 %)
not in observation N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)
overall <40 albuterol N (%) 5 (9.80 %) 6 (13.64 %) 4 (6.67 %)
untreated N (%) 46 (90.20 %) 38 (86.36 %) 56 (93.33 %)
not in observation N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)
>=40 albuterol N (%) 0 (0.00 %) 0 (0.00 %) 3 (13.04 %)
untreated N (%) 15 (100.00 %) 7 (100.00 %) 20 (86.96 %)
not in observation N (%) 0 (0.00 %) 0 (0.00 %) 0 (0.00 %)

Plots

The plotTreatment() function creates a bar plot showing the percentage of treated and untreated in each cohort, stratum, and time-window. This function offers customization options for colors, faceting, and handling of strata.

plotTreatment(
  result = results,
  facet =  sex + age_group ~ window_name + cohort_name,
  colour = "variable_level"
)