Carry the last observed confounder value forward through follow-up

The censoring model needs a value on every follow-up row. This function fills each plain confounder column of a trial panel. It carries the last observed value forward within each person-trial.

Usage

tteenrollment_fill_followup_confounders(trial)

Arguments

trial

A TTEEnrollment object at trial level.

Value

The TTEEnrollment object, invisibly.

Details

The carry starts from the entry-window snapshot, so the first row of a person-trial is never left NA. $s1_impute_confounders() supplies that snapshot value for a person-trial with no observed value before follow-up.

The function reads trial$design for every column name. It fills only the confounders that carry a .tte_entry__ snapshot column. A panel with no snapshot column at all comes back untouched.

The function sorts trial$data by id_var, then tstart_var, then tstop_var. It modifies trial$data by reference. It appends "fill_followup" to trial$steps_completed once. A second call appends no second token and changes no value.

The function stops when a filled row would still hold NA. That happens when the snapshot is NA and no earlier row of the same person-trial holds a value.

See also

tteenrollment_fill_aggregates() for the counts the fill changes.

Other tte_methods: tteenrollment_combined_combine(), tteenrollment_fill_aggregates(), tteenrollment_fill_summary(), tteenrollment_irr_combine(), tteenrollment_rates_combine(), tteenrollment_rbind()

Examples

d <- data.table::data.table(
  enrollment_person_trial_id = rep(c("a", "b"), each = 3L),
  tstart = rep(c(0L, 4L, 8L), 2L),
  tstop = rep(c(4L, 8L, 12L), 2L),
  exposed = rep(c(TRUE, FALSE), each = 3L),
  event = 0L,
  income = c(NA, "mid", NA, NA, NA, "high"),
  .tte_entry__income = rep(c("low", "mid"), each = 3L)
)
design <- TTEDesign$new(
  treatment_var = "exposed",
  outcome_vars = "event",
  confounder_vars = "income",
  follow_up_time = 12L
)
trial <- TTEEnrollment$new(d, design)
tteenrollment_fill_followup_confounders(trial)
trial$data$income
#> [1] "low"  "mid"  "mid"  "mid"  "mid"  "high"