naaccr

Summary

The naaccr R package enables researchers to easily read and begin analyzing cancer incidence records stored in the North American Association of Central Cancer Registries (NAACCR) file format.

Usage

naaccr focuses on two tasks: arranging the records and preparing the fields for analysis.

Records

The naaccr_record class defines objects which store cancer incidence records. It inherits from data.frame, and for now only makes sure a dataset has a standard set of columns. While naaccr_record has a singular-sounding name, it can contain multiple records as rows.

The read_naaccr function creates a naaccr_record object from a NAACCR-formatted file.

record_file <- system.file(
  "extdata/synthetic-naaccr-18-abstract.txt",
  package = "naaccr"
)
record_lines <- readLines(record_file)
## Marital status and race fields
cat(substr(record_lines[1:5], 206, 216), sep = "\n")
#> 30188888888
#> 40188888888
#> 20188888888
#> 20188888888
#> 30188888888
library(naaccr)

records <- read_naaccr(record_file, version = 18)
records[1:5, c("maritalStatusAtDx", "race1", "race2", "race3")]
#>   maritalStatusAtDx race1                      race2                      race3
#> 1         separated white no further race documented no further race documented
#> 2          divorced white no further race documented no further race documented
#> 3           married white no further race documented no further race documented
#> 4           married white no further race documented no further race documented
#> 5         separated white no further race documented no further race documented

By default, read_naaccr reads all fields defined in a format. For example, the NAACCR 18 format used above has 791 fields. Rarely would an analysis need even 100 fields. By specifying which fields to keep, one can improve time and memory efficiency.

dim(records)
#> [1]  20 867
records_slim <- read_naaccr(
  input       = record_file,
  version     = 18,
  keep_fields = c("ageAtDiagnosis", "countyAtDx", "primarySite")
)
dim(records_slim)
#> [1] 20  3

Like with most classes, one can create a new naaccr_record object with the function of the same name. The result will have the given columns.

nr <- naaccr_record(
  primarySite = "C010",
  dateOfBirth = "19450521"
)
nr[, c("primarySite", "dateOfBirth")]
#>   primarySite dateOfBirth
#> 1        C010  1945-05-21

The as.naaccr_record function can transform an existing data frame. It does require any existing columns to use NAACCR’s XML names.

prefab <- data.frame(
  ageAtDiagnosis = c(1, 120, 999),
  race1          = c("01", "02", "88")
)
converted <- as.naaccr_record(prefab)
converted[, c("ageAtDiagnosis", "race1")]
#>   ageAtDiagnosis                      race1
#> 1              1                      white
#> 2            120                      black
#> 3             NA no further race documented

Code translation

The NAACCR format uses similar schemes for a lot of fields, and the naaccr package includes functions to help translate them.

naaccr_boolean translates “yes/no” fields. By default, it assumes "0" stands for “no”, and "1" stands for “yes.”

naaccr_boolean(c("0", "1", "2"))
#> [1] FALSE  TRUE    NA

Some fields use "1" for FALSE and "2" for TRUE. Use the false_value parameter to work with these.

naaccr_boolean(c("0", "1", "2"), false_value = "1")
#> [1]    NA FALSE  TRUE

Categorical fields

The naaccr_factor function translates values using a specific field’s category codes.

naaccr_factor(c("01", "31", "65"), "primaryPayerAtDx")
#> [1] not insured Medicaid    TRICARE    
#> 16 Levels: not insured self-pay insurance, NOS ... Indian/Public Health Service

Some fields have multiple codes explaining why an actual value isn’t known. By default, they’ll all be converted to NA so they can propagate that information in R. But the reasons can be useful, so naaccr_factor and naaccr_record both have a keep_unknown parameter.

naaccr_factor(c("1", "9"), field = "sex")
#> [1] male <NA>
#> 6 Levels: male female other transsexual, NOS ... transsexual, natal female
naaccr_factor(c("1", "9"), field = "sex", keep_unknown = TRUE)
#> [1] male    unknown
#> 7 Levels: male female other transsexual, NOS ... unknown
naaccr_record(sex = c("1", "9"), race1 = c("01", "99"), keep_unknown = TRUE)
#>       sex   race1
#> 1    male   white
#> 2 unknown unknown

Numeric with special missing

Some fields contain primarily continuous or count data but also use special codes. One name for this type of code is a “sentinel value.” The split_sentineled function splits these fields in two.

rnp <- split_sentineled(c(10, 20, 90, 95, 99, NA), "regionalNodesPositive")
rnp
#>   regionalNodesPositive regionalNodesPositiveFlag
#> 1                    10                      <NA>
#> 2                    20                      <NA>
#> 3                    NA                     >= 90
#> 4                    NA       positive aspiration
#> 5                    NA                   unknown
#> 6                    NA                      <NA>

Building

library(devtools)

deps <- packageDescription("naaccr", fields = c("Depends", "Imports", "Suggests"))
deps <- Filter(function(x) any(!is.na(x)), deps)
dep_names <- lapply(deps, function(x) devtools::parse_deps(x)[["name"]])
dep_names <- sort(unlist(dep_names))
dep_list <- paste0("- `", dep_names, "`", collapse = "\n")

To build the naaccr package, you’ll need the following R packages:

To document, build, and test the package, run the build.R script with the package’s root as the working directory.

Project files

First, know this project fills two roles:

  1. Creating a package to work with NAACCR data in R.
  2. Collecting the data needed to process NAACCR files in plain-text and machine-readable formats.
naaccr/
├ R/                  # R files to create the package objects
├ data-raw/           # Plain-text data files and scripts for processing them
│ ├ code-labels/      # Mappings of codes to understandable labels
│ ├ sentinel-labels/  # Mappings of sentinel values to understandable labels
│ └ record-formats/   # Tables defining each NAACCR file format
├ external/           # Downloaded files and scripts to create files in `data-raw`
├ inst/
│ └ extdata/          # Data files for examples in the documentation
└ tests/              # tests and data using the `testthat` package

Files in external only need to be updated or run when NAACCR publishes a new or revised format. In that case, refer to the comments in the .R scripts in that directory for where to download the new files.

Think of these scripts as handy tools for generating data-raw files. Some cleaning of their output may be required.

To run create-record-format-files.R, you’ll need to create an account for the SEER API from the National Cancer Institute’s Surveillance, Epidemiology and End Results (SEER) program. Store the API key as an environment variable named SEER_API_KEY.