claimpivot: Pivot a record of secondary diagnoses into long format

In rplzzz/AHRQcomorb: Convert diagnosis codes to comorbidity lists

Pivot a record of secondary diagnoses into long format

Description

Pivot a record into long format, keeping paired diagnosis and POA fields together in the new format.

Usage

claimpivot(
  tbl,
  idcols,
  diagstem = "ICD_DGNS_CD",
  poastem = "CLM_POA_IND_SW",
  poahandle = poa_cms,
  poaconv = TRUE,
  nahandle = na.pass
)

Arguments

tbl	Table of data in wide format
idcols	Name of the columns used as identifiers. If more than one, they will be joined into a single synthetic key called 'id' (using mergekeys) because comorb only accepts one key column.
diagstem	Character stem for the names of the diagnosis fields (default is 'ICD_DGNS_CD').
poastem	Character stem for the names of the POA fields (default is 'CLM_POA_IND_SW')
poahandle	Function for converting strings in POA fields to POA indicators (see details).
poaconv	Flag indicating whether to convert POA values, or just leave them as-is.
nahandle	What to do with POA values that don't match either the true or false strings. Default is to leave them as NA in the output.

Details

Often a record for which we want to find the comorbidities will be organized with one row per record, and the diagnoses and POA indicators as fields within the row. This function pivots such tables into a long format with one row per diagnosis, and fields for the identifier, diagnosis, and POA flag. (The difference between this and an ordinary pivot is the pairing of the diagnosis and POA fields.)

The diagnosis and POA fields are identified by a stem, followed by a number. For example, "ICD_DGNS_CD13" for the 13th diagnosis code and "CLM_POA_IND_SW13" for the corresponding POA flag. In the output, each diagnosis will be matched up with its POA, and the ID will be replicated for each row.

If the POA column is of type character, it will be converted to an integer 0/1 value. This behavior can be suppressed by setting poaconv to FALSE. By default, the CMS convention of using 'Y' for true will be used, with all other values deemed to be false. Other conventions can be used by supplying a function as the poahandle argument. Several such functions for commonly encountered conventions are built in (poa_functions), or a custom function can be supplied.

Entries that decode to neither true nor false will be assigned the value NA. Optionally, the function can raise an error if this happens (pass na.fail in the nahandle parameter.) This checking will also catch diagnosis code fields for which no corresponding POA code was passed.

rplzzz/AHRQcomorb documentation built on Aug. 10, 2024, 8:12 a.m.