Shiny Apps

This vignette shows how to render interactive plots and tabular results in Shiny using the visOmopResults package and functions that build on it. Specifically, we will demonstrate:

Set up

Load packages and mock data.

library(shiny)
library(bslib)
library(sortable)
library(shinyWidgets)
library(gt)
library(DT)
library(reactable)
library(plotly)
library(dplyr)
library(visOmopResults)
library(IncidencePrevalence)
library(CohortCharacteristics)
library(shinycssloaders)

# Mock results in visOmopResults
data <- visOmopResults::data

# Remove global options (just in case we have them from previous work)
setGlobalPlotOptions(style = NULL, type = NULL)
setGlobalTableOptions(style = NULL, type = NULL)

Mock data

We will use 3 different mock results, which are:

Shiny App: User Inteface

The Shiny app has three panels, one for each result. All allow filtering by sex strata and provide panel-specific visualization options:

- Baseline characteristics: shows a gt table with controls for headers, groups, and hidden columns..

- Large Scale characteristics: renders as a datatable or reactable, with options to group or hide columns.

- Incidence: displays a static ggplot or interactive plotly plot, with options for colouring, faceting, and ribbons.

Example UI Code

ui <- bslib::page_navbar(
  title = "visOmopResults for Shiny",
  window_title = "visOmopResults • Shiny",
  collapsible = TRUE,
  # Baseline Characteristics (GT table)
  bslib::nav_panel(
    title = "Baseline Characteristics",
    icon = icon("users-gear"),
    bslib::layout_sidebar(
      sidebar = bslib::sidebar(
        title = "Filters",
        shinyWidgets::pickerInput(
          inputId = "summarise_characteristics_sex",
          label = "Sex",
          choices = c("overall", "Male", "Female"),
          selected = "overall",
          multiple = TRUE
        ),
        width = 320,
        position = "left",
        open = TRUE
      ),
      bslib::card(
        full_screen = TRUE,
        bslib::card_header("Table layout"),
        bslib::layout_sidebar(
          sidebar = bslib::sidebar(
            title = "Arrange columns",
            sortable::bucket_list(
              header = NULL,
              group_name = "col-buckets",
              orientation = "horizontal",
              add_rank_list(
                text = "None",
                labels = c("variable_name", "variable_level", "estimate_name"),
                input_id = "summarise_characteristics_table_none"
              ),
              add_rank_list(
                text = "Header",
                labels = c("sex"),
                input_id = "summarise_characteristics_table_header"
              ),
              add_rank_list(
                text = "Group columns",
                labels = c("cdm_name", "cohort_name"),
                input_id = "summarise_characteristics_table_group_column"
              ),
              add_rank_list(
                text = "Hide",
                labels = "table_name",
                input_id = "summarise_characteristics_table_hide"
              )
            ),
            position = "right",
            width = 400,
            open = FALSE
          ),
          # GT output
          gt::gt_output("summarise_characteristics_table") |>
            shinycssloaders::withSpinner(type = 4)
        )
      )
    )
  ),
  # Large Scale Characterisation (DT / reactable)
  bslib::nav_panel(
    title = "Large Scale Characterisation",
    icon = icon("table"),
    bslib::layout_sidebar(
      sidebar = bslib::sidebar(
        # title = "Display options",
        shinyWidgets::pickerInput(
          inputId = "large_scale_sex",
          label = "Sex",
          choices = c("overall", "Male", "Female"),
          selected = "overall",
          multiple = TRUE
        ),
        radioButtons(
          "large_engine",
          "Renderer",
          choices = c("DT", "reactable"),
          inline = TRUE
        ),
        sortable::bucket_list(
          header = NULL,
          group_name = "col-buckets",
          orientation = "horizontal",
          add_rank_list(
            text = "None",
            labels = c("variable_name", "variable_level", "estimate_name"),
            input_id = "large_scale_none"
          ),
          add_rank_list(
            text = "Group columns",
            labels = c("cdm_name", "cohort_name"),
            input_id = "large_scale_group_column"
          ),
          add_rank_list(
            text = "Hide",
            labels = character(),
            input_id = "large_scale_hide"
          )
        ),
        width = 320
      ),
      bslib::card(
        full_screen = TRUE,
        bslib::card_header("Cohort characteristics (large-scale)"),
        conditionalPanel(
          "input.large_engine == 'DT'",
          DTOutput("large_dt") |> shinycssloaders::withSpinner(type = 4)
        ),
        conditionalPanel(
          "input.large_engine == 'reactable'",
          reactableOutput("large_reactable") |> shinycssloaders::withSpinner(type = 4)
        )
      )
    )
  ),
  # Incidence (ggplot → plotly)
  bslib::nav_panel(
    title = "Incidence",
    icon = icon("chart-line"),
    bslib::layout_sidebar(
      sidebar = bslib::sidebar(
        title = "Plot options",
        shinyWidgets::pickerInput(
          "incidence_sex",
          "Sex strata",
          choices = c("overall", "Male", "Female"),
          selected = "overall",
          multiple = TRUE
        ),
        shinyWidgets::pickerInput(
          inputId = "facet",
          label = "Facet",
          selected = "sex",
          multiple = TRUE,
          choices = c("cdm_name", "incidence_start_date", "sex", "outcome_cohort_name"),
        ),
        shinyWidgets::pickerInput(
          inputId = "colour",
          label = "Colour",
          selected = "outcome_cohort_name",
          multiple = TRUE,
          choices = c("cdm_name", "incidence_start_date",  "sex", "outcome_cohort_name")
        ),
        checkboxInput("inc_ribbon", "Show ribbon (CI)", TRUE),
        checkboxInput("interactive", "Interactive Plot", TRUE),
        width = 320
      ),
      bslib::card(
        full_screen = TRUE,
        bslib::card_header("Incidence over time"),
        uiOutput("incidence_plot", height = "520px") |> shinycssloaders::withSpinner(type = 4)
      )
    )
  )
)

Shiny App: Server

1) Baseline characteristics

The server filters results by the selected sex and creates a gt table using the tableCharacteristics() function from the CohortCharacteristics package. This function is built on visOmopResults, which ensures consistent styling and supports arguments to define headers, group columns, and hide columns.

If you have your own <summarised_result> table, which don’t has a dedicated table function, you can instead use visOmopTable() to generate a gt table in Shiny. This allows you to group estimates and configure header, group, and hidden column options in a similar way.

2) Large Scale characteristics

These results are not in <summarised_result> format, as shown below:

data$large_scale_characteristics
#> # A tibble: 952 × 8
#>    cdm_name      cohort_name sex   variable_name variable_level concept_id count
#>    <chr>         <chr>       <chr> <chr>         <chr>          <chr>      <int>
#>  1 my_duckdb_da… denominato… over… Acute allerg… -inf to -366   4084167      113
#>  2 my_duckdb_da… denominato… over… Acute bacter… -inf to -366   4294548      607
#>  3 my_duckdb_da… denominato… over… Acute bronch… -inf to -366   260139      2303
#>  4 my_duckdb_da… denominato… over… Acute cholec… -inf to -366   198809        29
#>  5 my_duckdb_da… denominato… over… Acute viral … -inf to -366   4112343     2388
#>  6 my_duckdb_da… denominato… over… Alzheimer's … -inf to -366   378419        15
#>  7 my_duckdb_da… denominato… over… Anemia        -inf to -366   439777        73
#>  8 my_duckdb_da… denominato… over… Angiodysplas… -inf to -366   4310024      281
#>  9 my_duckdb_da… denominato… over… Appendicitis  -inf to -366   440448       125
#> 10 my_duckdb_da… denominato… over… Atopic derma… -inf to -366   133834        54
#> # ℹ 942 more rows
#> # ℹ 1 more variable: percentage <dbl>

In this case, we use visTable() to generate tables as either a datatable or a reactable, depending on the user’s choice in the UI. The table type is specified with the type argument.

For both table types, we pass the UI-selected columns to groupColumn and hide. We do not generate a header for this result, as it would require restructuring the estimates into a single “estimate_value” column.

The look and behaviour of the tables can be customised through the style argument. Available options can be explored with:

In this vignette, we modify the datatable style in the server code so filters appear at the top of the table instead of the default bottom.

3) Incidence

For incidence results, we use the plotIncidence() function from the IncidencePrevalence package. This function creates a ggplot object, which can be rendered as a static plot with plotOutput or as an interactive plot with plotlyOutput. Users can also select which columns to use for colouring and faceting, and whether to display confidence interval ribbons.

For other results—whether <summarised_reuslt> class or not—you can generate plots in a similar way by using the plotting functions available in visOmopResults.

Example Server Code

Note: Both CohortCharacteristics and IncidencePrevalence functions for plotting and tabulation are built on visOmopResults, which means they share a consistent interface and style.

server <- function(input, output, session) {
  # Baseline (GT)
  output$summarise_characteristics_table <- gt::render_gt({
    data$summarised_characteristics |>
      # filter results by sex
      filterStrata(sex %in% input$summarise_characteristics_sex) |>
      # create GT table
      CohortCharacteristics::tableCharacteristics(
        header = input$summarise_characteristics_table_header,
        groupColumn = input$summarise_characteristics_table_group_column,
        hide = input$summarise_characteristics_table_hide,
        type = "gt"
      )
  })

  # Large scale characteristics
  getLargeScaleResults <- reactive({
    data$large_scale_characteristics |>
      filter(.data$sex %in% input$large_scale_sex)
  })
  # To render as DT
  output$large_dt <- renderDT({
    getLargeScaleResults() |>
      visTable(
        hide = input$large_scale_hide,
        groupColumn = input$large_scale_group_column,
        type = "datatable",
        style = list(
          filter = "top",
          searchHighlight = TRUE,
          rownames = FALSE
        )
      )
  })
  # To render as reactable
  output$large_reactable <- reactable::renderReactable({
    getLargeScaleResults() |>
      visTable(
        hide = input$large_scale_hide,
        groupColumn = input$large_scale_group_column,
        type = "reactable",
        style = "default"
      )
  })

  # Incidence
  getIncidencePlot <- reactive({
    data$incidence |>
      filterStrata(sex %in% input$incidence_sex) |>
      plotIncidence(
        colour = input$colour,
        facet = input$facet,
        ribbon = input$inc_ribbon
      ) +
      theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1))
  })
  output$incidence_plot <- renderUI({
    plt <- getIncidencePlot()
    if (input$interactive) {
      ggplotly(plt)
    } else {
      renderPlot(plt)
    }
  })
}

Run App

To run the Shiny app, copy the code chunks provided in this vignette into a script named app.R, and add the following line at the end:

shinyApp(ui, server)

You can find the complete code run the ShinyApp here.