REDCapLintR

[wibeasley]

@thomasnwilson suggested that we develop a function/package that can lint REDCap dictionaries and return a markdown/html report. These are the initial rules we thought of while waiting in airport.

---

REDCapLintR: Tool for REDCap Dictionary Good Practices

Working title: "REDCrapR" or "CrapR" or "Moving from REDCrap to REDCap"

• [ ] Rule: no variable should end in "_v\d" (eg, _v2 or _v3) Opinion: variables in a sequence should have a meaningful name that clearly communicates its position in the sequence. Examples of bad behavior: age, age_v2, and age_v2_v2 Suggested fix: Rename variables to age_baseline, age_discharge, age_followup

• [ ] Smell: at least 10% of text variables should have validation

• [ ] Smell: at least 20% of variables should be non-text, like dropdowns or sliders

• [ ] All piped values should originate from variables, events, or smart variables that are currently in the dictionary. Check that all these still exist among a combined list of variables, events, & smart variables. regex: \[[a-z][a-z0-9_-]*\]

• [ ] All embedded variables should originate from variables, events, or smart variables that are currently in the dictionary. Check that all these still exist among a combined list of variables, events, & smart variables. regex: \{[a-z][a-z0-9_-]*\}

• [ ] Rule: all date variables should have the same format within the project. Don't mix & match dmy and mdy.

• [ ] All forms/instruments should be mapped to at least one event

  • exception: the instrument ends with "*_retired"

• [ ] Rule: any variable with something like "phone" in the variable name, field label or field note should have a phone validation. Tokens include

  • phone
  • mobile
  • cell
  • contact number

• [ ] Rule: any variable with something like "number" in the variable name, field label or field note should have a integer or numeric validation. Tokens include

  • number
  • age
  • count

• [ ] Rule: any variable with something like "zip code" in the variable name, field label or field note should have a zip code validation. Tokens include

  • zip
  • zip_code
  • zipcode

• [ ] Rule: any variable with something like T/F, Y/N in the variable name, field label or field note should have a "1" for true/yes/on and "0" for false/no/off. Tokens include (case insensitive):

  • t/f
  • true/false
  • y/n
  • yes/no
  • on/off

• Rule: male & female consistently coded as 1/0, 1/2, or 8507/8532 (for OMOP)

  • m/f
  • male/female

• [ ] ?? can we expand this to tri-state variables like yes/no/maybe or yes/no/null ??

• [ ] Rule: multiple choice responses options are coded as integers (instead of letters)

[genevaneva]

"REDCrapR" 😆