Clinician Scale Measures

End-to-end scripts for harmonizing Canvas and REDCap survey exports, producing analysis-ready data, descriptive tables, and quick pre/post visuals for the DETECT clinician scale measures workstream.

Inputs

• Canvas pre/post survey exports named exactly as the files in data/canvas/.
• REDCap control export plus data dictionary in data/redcap/.
• data/csm_likert_numeric_data.csv – numeric summary of ABEM item means used by the visualization script.

Repo layout

• data/ – raw inputs and derived CSVs (Canvas, REDCap, and harmonized files).
• tables/ – Word exports produced by tables.r.
• visualizations/ – saved plot exports.
• data_harmonization.r – trims Canvas metadata, aligns variable names to REDCap, recodes factors, and writes data/clinican_scale_measures_harmonized_analysis_data.csv.
• tables.r – builds Table 1 (sociodemographics) and Table 2 (ABEM item means) from the harmonized data.
• visualizations.r – reads data/csm_likert_numeric_data.csv plus the harmonized data; saves pre/post mean dumbbell and response distribution plots.
• data_maps.r – shared mappings for variable names, factor levels, site champions, and descriptive ABEM item labels.
• renv* / renv.lock – dependency lockfile and project-local library managed by renv.

Workflow

1. Place the Canvas and REDCap exports in data/canvas/ and data/redcap/ with the existing filenames.
2. Run data_harmonization.r to:
   • remove Canvas metadata and quiz scoring fields,
   • rename Canvas columns to REDCap variable names and dummy-code multi-select race responses,
   • recode REDCap numeric codes to labeled factors and tag site champions,
   • stack Canvas pre, Canvas post, and REDCap control into data/clinican_scale_measures_harmonized_analysis_data.csv.
3. Run tables.r to enforce factor levels and write tables/table_1.docx (sociodemographics) and tables/table_2.docx (ABEM pre/post means).
4. Run visualizations.r to generate:
   • visualizations/likert_answer_mean_changes.png (Canvas pre vs post mean dumbbell plot using data/csm_likert_numeric_data.csv).
   • visualizations/likert_answer_distributions.png (faceted pre/post response distributions from the harmonized data).

Running the scripts

From R:

install.packages("renv") # if not already installed
renv::restore()          # install locked package versions
source("data_harmonization.r")
source("tables.r")
source("visualizations.r")

From the shell:

Rscript data_harmonization.r
Rscript tables.r
Rscript visualizations.r

Outputs

• data/clinican_scale_measures_harmonized_analysis_data.csv – stacked Canvas pre/post and REDCap control records with aligned variable names and readable labels.
• tables/table_1.docx – sociodemographics.
• tables/table_2.docx – ABEM item means with Canvas post minus pre deltas.
• visualizations/likert_answer_mean_changes.png – Canvas pre vs post mean dumbbell plot based on data/csm_likert_numeric_data.csv.
• visualizations/likert_answer_distributions.png – saved histogram facets comparing pre/post response distributions by item.

Notes

• Canvas columns are renamed using redcap_field_labels_pre/post and race text is dummy-coded via canvas_to_redcap_race_map.
• REDCap numeric codes are recoded to labeled factors for readability; site champions are tagged via site_champions.
• Table 1 drops race from the main sociodemographic table and handles it separately to account for multi-select responses.