Build a temporal filtering specification

Creates an omop_temporal_spec object that defines how events are filtered relative to a cohort index date or calendar dates. The spec can combine index-relative windows, calendar date ranges, and event selection (first/last N) and deterministic minimum-gap collapsing.

Usage

omop.temporal(
  index_window = NULL,
  calendar = NULL,
  event_select = NULL,
  min_gap = NULL
)

Arguments

index_window

Named list with start and end (integer days relative to the cohort index date). Negative values denote time before the index date.

calendar

Named list with start and end (character ISO 8601 dates, e.g. "2020-01-01").

event_select

Named list with order ("first" or "last"), n (integer; number of events), and optional by = "grain" (default) or by = "concept". Concept mode keeps the first/last N independently for each concept within a cohort episode when index_window is present, otherwise within each person.

min_gap

Positive integer days, or a named list with days, optional by = "concept" or "grain", and optional keep = "first" or "last". Adjacent events no more than days apart form one chain. The normalized policy defaults to concept-specific chains represented by their first event.

Value

An omop_temporal_spec object (a list with class c("omop_temporal_spec", "list")).

See also

ds.omop.plan.events, omop.date_handling

Examples

if (FALSE) { # \dontrun{
# Events within 1 year before each index episode; keep its first 3 events
temporal <- omop.temporal(
  index_window = list(start = -365, end = 0),
  event_select = list(order = "first", n = 3)
)
plan <- ds.omop.plan.events(plan, "conditions",
  "condition_occurrence", temporal = temporal)
} # }