Cohort-report generates a report for each variable in an input file.

Consider the following extract from a study's project.yaml:

actions:

  generate_study_population:
    run: cohortextractor:latest generate_cohort
    outputs:
      highly_sensitive:
        cohort: output/input.csv

  generate_report:
    run: cohort-report:v3.0.0 output/input.csv
    needs: [generate_study_population]
    config:
      variable_types:
          age: int
          sex: categorical
          ethnicity: categorical
          bmi: float
          diabetes: binary
          chronic_liver_disease: binary
          imd: categorical
          region: categorical
          stp: categorical
          rural_urban: categorical
          prior_covid_date: date
      output_path: output/cohort_reports_outputs
    outputs:
      moderately_sensitive:
        reports: output/cohort_reports_outputs/descriptives_input.html

The generate_report action generates a report that contains a table and a chart for each variable in the input file. The table contains unsafe statistics and should be checked thoroughly before it is released. The chart contains statistics that have been made safe. Cells in the underlying frequency table have been redacted, if:

• They contain less than 10 units
• They contain greater than 90% of the total number of units

Notice the run and config properties.

The run property passes an input file to a named version of cohort-report. In this case, it passes output/input.csv to v3.0.0 of cohort-report. The config property passes configuration to cohort-report; for more information, see Configuration.

Notice that the report is called descriptives_[the name of the input file, without the extension].html. It is saved to the output_path; for more information, see Configuration.

output_path, which defaults to cohort_reports_outputs. Save the outputs to the given path. If the given path does not exist, then it is created.

variable_types, which is required for .csv and .csv.gz input files. Cast the given variables to the given types. Supported types:

• binary
• categorical
• date
• float
• int

The run property can pass multiple input files to a named version of cohort-report. For example:

actions:
  # 3.0.0.
  generate_report:
    run: cohort-report:v3.0.0 output/input_2021-01-01.csv output/input_2021-02-01.csv
    # 3.0.0.

However, if one or more input files are .csv or .csv.gz input files, then variable_types is required; this will cast the given variables to the given types in all input files. It will fail if an input file does not have the given variables.

Please see DEVELOPERS.md.