Add 2026 MHT exposure variables to a person-week skeleton

Derives menopausal hormone therapy (MHT) exposure from Swedish prescription registry (LMED) data, using the definitions pinned for the 2026 MHT study. `add_lmed_v20250909()` categorises dispensed product names into MHT groups. It corrects the dispensed durations of products whose recorded defined daily doses are unreliable. It derives the approach-based treatment variables. It then builds the `rd_approach*` exposure variables on top of them.

Usage

add_lmed_v20250909(skeleton, lmed, verbose = TRUE)

Arguments

skeleton

A `data.table` person-week skeleton with an integer `id` column and a character `isoyearweek` column of the form `"YYYY-WW"`. Modified by reference; see `Value`.

lmed

A `data.table` of dispensed prescriptions. The function reads exactly four columns:

`lopnr` - person identifier, matched against `skeleton$id`
`produkt` - product name
`edatum` - `Date` of dispensing
`fddd` - numeric dispensed duration in days

`lmed` is NOT modified. The function first subsets `lmed` to the identifiers present in `skeleton`, which allocates a new `data.table`. Every subsequent `:=` then writes to that internal subset. The caller's own `lmed` is left untouched.

verbose

Logical. If `TRUE` (the default), report progress with `message()`. Set to `FALSE` for a silent run.

Value

An internal logical flag, NOT the skeleton. You MUST NOT assign the result: `skel <- add_lmed_v20250909(skel, lmed)` would overwrite your skeleton with that flag. Call the function for its effect, then carry on using the object you passed in.

`skeleton` is MODIFIED BY REFERENCE. The caller's own object gains one logical column per MHT product category (`A1` ... `I2`), the character columns `approach1`, `approach2` and `approach3`, and the eight exposure columns `rd_approach1_single`, `rd_approach1_multiple`, `rd_approach2_single`, `rd_approach2_multiple`, `rd_approach3_single`, `rd_approach3_multiple`, `rd_approach3b_single` and `rd_approach3b_multiple`. If you still need the input skeleton unchanged, pass `data.table::copy(skeleton)`.

`lmed` is NOT modified. Only `skeleton` is modified in place.

Details

`add_lmed_v20250909()` performs several steps:

Restricts the LMED data to individuals present in `skeleton`
Categorises products into MHT groups (`A1` ... `I2`) from product names
Applies duration fixes for specific products (IUDs, minimum doses)
Flags the product categories in force in each person-week
Derives the approach variables, bridging gaps of up to four weeks
Builds the `rd_approach*` exposure variables, applying the `previous` / `exclude` and minimum-duration rules

Note

This entry point differs from the earlier one in four ways. It flags the product categories with a `data.table::foverlaps()` interval join instead of a nested loop over categories and ISO weeks. It reads the LMED identifier from `lopnr`. It takes a `verbose` argument. And it additionally computes the `rd_approach*` exposure variables, which the earlier entry point does not produce.

Prescription rows whose dispensed interval is missing or inverted, which a missing or negative `fddd` produces, are dropped with a warning.

Examples

library(data.table)

# copy() the skeleton: it is modified in place. `lmed` is not modified,
# but copy() it too so the shipped fixture is never altered by accident.
skeleton <- copy(fake_skeleton_mht)
lmed <- copy(fake_lmed_2026)

# the fixture deliberately holds one negative-duration row, which warns
suppressWarnings(add_lmed_v20250909(skeleton, lmed, verbose = FALSE))
#> [1] FALSE

# `skeleton` itself now carries the derived columns
skeleton[, .N, keyby = .(rd_approach1_single)]
#> Key: <rd_approach1_single>
#>    rd_approach1_single     N
#>                 <char> <int>
#> 1:             exclude  2560
#> 2:   local_or_none_mht  2196
#> 3:            previous   140
#> 4:        systemic_mht   720