TVTmatching

The TVTmatching is an implement of a matching algorithm with time-varying treatment. Each case is assign to one or several controls such that their time “at risk of the treatment” matches.

Installation

You can install the development version of TVTmatching like so:

remotes::intall_github("jaromilfrossard/TVTmatching")

Example

You can explore the control candidates for each case using:

library(TVTmatching)
data("tvtdata")
tvt_sum <- tvt_summary_cases(tvtdata,
                             date_statrisk = stos,
                             date_treatna = treatna,
                             date_eos = eos,
                             state_eos = eos01, 
                             by = fct)
tvt_sum
#> # A tibble: 4 × 10
#>   id_row time_a…¹ time_…² state…³ state…⁴ fct   data_c…⁵ n_con…⁶ state…⁷ state…⁸
#>    <int>    <dbl>   <dbl>   <int>   <int> <chr> <list>     <int>   <int>   <int>
#> 1      1      366     731       1       1 A     <tibble>       2       1       1
#> 2      2      273     579       1       1 B     <tibble>       2       1       1
#> 3      3      122     821       1       0 A     <tibble>       3       1       2
#> 4      4       62     396       1       0 B     <tibble>       3       1       2
#> # … with abbreviated variable names ¹​time_atrisk, ²​time_eos, ³​state_treat,
#> #   ⁴​state_eos, ⁵​data_control, ⁶​n_control, ⁷​state_eos_0_control,
#> #   ⁸​state_eos_1_control

You can match using dates:

set.seed(42)
tvt_match<- tvt_matching_date(data = tvtdata,
                              date_statrisk = stos,
                              date_treatna = treatna,
                              date_eos = eos,
                              state_eos = eos01)
tvt_match
#> # A tibble: 8 × 5
#>   group id_row treat_tv  time_eos_tv state_eos_tv
#>   <int>  <int> <chr>           <dbl>        <int>
#> 1     1      1 treatment         365            1
#> 2     1      5 control           396            0
#> 3     2      2 treatment         306            1
#> 4     2      8 control           154            0
#> 5     3      3 treatment         699            0
#> 6     3      1 control           244            0
#> 7     4      4 treatment         334            0
#> 8     4      1 control           304            0

If you have data in time you can use:

set.seed(42)
tvtdata_num <- tvt_date2numeric(data = tvtdata,
                                date_statrisk = stos,
                                date_treatna = treatna,
                                date_eos = eos,
                                state_eos = eos01)
tvt_match2 <- tvt_matching_num(data = tvtdata_num,
                               time_atrisk = time_atrisk,
                               time_eos = time_eos,
                               state_eos = state_eos,
                               state_treat = state_treat)
tvt_match2
#> # A tibble: 8 × 5
#>   group id_row treat_tv  time_eos_tv state_eos_tv
#>   <int>  <int> <chr>           <dbl>        <int>
#> 1     1      1 treatment         365            1
#> 2     1      5 control           396            0
#> 3     2      2 treatment         306            1
#> 4     2      8 control           154            0
#> 5     3      3 treatment         699            0
#> 6     3      1 control           244            0
#> 7     4      4 treatment         334            0
#> 8     4      1 control           304            0