Add support for model_get_n() for tidycrr model objects

[ddsjoberg]

I can try to add this support at some time if you're in a busy season :)

remotes::install_github("MSKCC-Epi-Bio/tidycmprsk")
library(tidycmprsk)

crr(Surv(ttdeath, death_cr) ~ age + trt, trial) %>%
  gtsummary::tbl_regression(exponentiate = TRUE) %>%
  add_nevent() # number of events (first competing event)

[larmarange]

Dear @ddsjoberg

Can you please check #144 Does it fix the problem? Would you have some other tests to check that results are as expected?

library(tidycmprsk)
library(broom.helpers)
library(gtsummary)
#> 
#> Attachement du package : 'gtsummary'
#> Les objets suivants sont masqués depuis 'package:broom.helpers':
#> 
#>     all_continuous, all_contrasts
#> L'objet suivant est masqué depuis 'package:tidycmprsk':
#> 
#>     trial

mod <- crr(Surv(ttdeath, death_cr) ~ age + trt, tidycmprsk::trial)
#> 11 cases omitted due to missing values

mod %>% model_get_n()
#> # A tibble: 4 x 4
#>   term        n_obs n_event exposure
#>   <chr>       <dbl>   <dbl>    <dbl>
#> 1 (Intercept)   189     154    3762.
#> 2 age           189     154    3762.
#> 3 trtDrug B      98      83    1888.
#> 4 trtDrug A      91      71    1874.

mod %>%
  tbl_regression(exponentiate = TRUE) %>%
  add_nevent() %>%
  as_kable()

Characteristic	Event N	HR	95% CI	p-value
Age	154	1.01	0.99, 1.03	0.6
Chemotherapy Treatment	154
Drug A
Drug B		1.52	0.88, 2.62	0.13

^{Created on 2022-02-01 by the reprex package (v2.0.1)}

[larmarange]

Hmmmm I'm not sure about the results. For Coxph models, the number of events include all type of events.

If I understand properly crr(), you want only the number of events defined by failure option? Can you clarify?

[larmarange]

I have updated the PR to consider, as the number of events, only status defined by failcode. Is it the expected behaviour?

library(tidycmprsk)
library(broom.helpers)
library(gtsummary)
#> 
#> Attachement du package : 'gtsummary'
#> Les objets suivants sont masqués depuis 'package:broom.helpers':
#> 
#>     all_continuous, all_contrasts
#> L'objet suivant est masqué depuis 'package:tidycmprsk':
#> 
#>     trial

mod <- crr(Surv(ttdeath, death_cr) ~ age + trt, tidycmprsk::trial)
#> 11 cases omitted due to missing values

mod %>% model_get_n()
#> # A tibble: 4 x 4
#>   term        n_obs n_event exposure
#>   <chr>       <dbl>   <dbl>    <dbl>
#> 1 (Intercept)   189      52    3762.
#> 2 age           189      52    3762.
#> 3 trtDrug B      98      31    1888.
#> 4 trtDrug A      91      21    1874.

mod %>%
  tbl_regression(exponentiate = TRUE) %>%
  add_nevent() %>%
  as_kable()

Characteristic	Event N	HR	95% CI	p-value
Age	52	1.01	0.99, 1.03	0.6
Chemotherapy Treatment	52
Drug A
Drug B		1.52	0.88, 2.62	0.13

^{Created on 2022-02-01 by the reprex package (v2.0.1)}

[larmarange]

Side question: should we support tidycuminc models?

[ddsjoberg]

Everything looks flawless! Thank you so much for adding!

[ddsjoberg]

Side question: should we support tidycuminc models?

What did you have in mind? These estimates are not regression models; they are more like the KM results from survival::survfit().

I also wrote a gtsummary-like function for them and added them to the tidycmprsk package.
https://mskcc-epi-bio.github.io/tidycmprsk/reference/add_cuminc.html

[larmarange]

Thanks for the feedback. It's merged now.

No worry, I didn't look into the details of tidycuminc. If there is no need, that's OK with me.