---
title: "Quarto reports"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{a03_quarto_reports}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse   = TRUE,
  comment    = "#>",
  warning    = FALSE,
  message    = FALSE,
  out.width  = "95%",  # figures occupy ~95% of document width
  out.height = "auto",
  dpi        = 320,    # ensure figure quality
  fig.width  = 6,      # default aspect ratio (can be overridden per-figure)
  fig.height = 3
)
options(rmarkdown.html_vignette.check_title = FALSE)
```

**visOmopResults** supports table and plot creation functions that can be used to generate publication-ready figures and tables. As an example, in this vignette we show how to create tables and figures for a DARWIN-EU study report.

# Quarto document settings

In the header of the Quarto document, we can set a title to identify the study and the document. Additionally, we can point to a Word document containing the template styles for DARWIN reports. The DOCX template can be found [here](https://github.com/darwin-eu/visOmopResults/tree/main/inst), together with the Quarto script used to generate the report.

A Quarto header for producing a Word report might look like this:

```yaml
---
title: "DARWIN-EU PX-CY-Z: Tables and Figures"
format:
  docx:
    reference-doc: inst/darwinReportRef.docx
    fig-cap-location: top
execute:
  echo: false
  message: false
  warning: false
---
```

Next, we use the first R chunk to load packages, load data, and set options/variables used across the script. The results we use are obtained from mock data created with the `IncidencePrevalence` and `CohortCharacteristics` packages, which we have stored in the *[inst]*(https://github.com/darwin-eu/visOmopResults/tree/main/inst) folder of the packge.

```{r, setup, echo=TRUE, eval=FALSE}
# Load necessary packages ----
library(visOmopResults)
library(IncidencePrevalence)
library(CohortCharacteristics)
library(dplyr)
library(tidyr)
library(ggplot2)

# Load mock results stored in the package ----
data <- visOmopResults::data

# Global options ----
knitr::opts_chunk$set(
  out.width  = "95%",  # figures occupy ~95% of document width
  out.height = "auto",
  dpi        = 320,    # ensure figure quality
  fig.width  = 6,      # default aspect ratio (can be overridden per-figure)
  fig.height = 3,
  results    = "asis"  # enable Markdown produced via cat() inside chunks
)

# DARWIN style for visOmopResults plots and tables.
style <- "darwin"
tableType <- "flextable"
plotType  <- "ggplot"
setGlobalPlotOptions(style = style, type = plotType) 
setGlobalTableOptions(style = style, type = tableType)
```

To use the Calibri font family for plots you will need to have the font installed, and for **Windows** registered in R graphics. See vignette on styles.

```{r, echo=FALSE}
# Load necessary packages ----
library(visOmopResults)
library(IncidencePrevalence)
library(CohortCharacteristics)
library(dplyr)
library(tidyr)
library(ggplot2)

# Load mock results stored in the package ----
reportData <- system.file("mockReportData.RData", package = "visOmopResults")
load(reportData)  # loads an object named `data` containing mock results

# DARWIN style for visOmopResults plots and tables.
style <- "darwin"
tableType <- "flextable"
plotType  <- "ggplot"
setGlobalPlotOptions(style = style, type = plotType) 
setGlobalTableOptions(style = style, type = tableType)
```

For tables we choose the `"flextable"` type because it transfers best to Word documents.
To use the Calibri font family for plots you will need to have the font installed. See vignette on styles.

# Characterisation results

The report starts with a table showing baseline characteristics of the denominator cohort in our mock study.

```{r}
data$summarised_characteristics |>
  dplyr::filter(variable_name != "Sex") |>
  tableCharacteristics(
    header = c("sex"),
    hide   = c("cdm_name", "cohort_name", "table_name"),
    type   = tableType,
    style = style
  )
```

If the default order of variable names provided by `tableCharacteristics()` does not match the study requirements, we can create a customised table with this package.

```{r}
data$summarised_characteristics |>
  dplyr::filter(variable_name != "Sex") |>
  dplyr::mutate(
    variable_name = customiseText(
      variable_name, 
      custom = c(
        "Comorbidities"  = "Comorbidities flag -inf to 0", 
        "Comedications"  = "Comedications flag -180 to 0"
      )
    ),
    variable_level = customiseText(
      variable_level, 
      custom = c("HIV" = "Hiv")
    )
  ) |>
  visOmopTable(
    header = c("sex"),
    estimateName = c(
      "N (%)"               = "<count> (<percentage>%)",
      "N"                   = "<count>",
      "Median [Q25 - Q75]"  = "<median> [<q25> - <q75>]",
      "Mean (SD)"           = "<mean> (<sd>)",
      "Range"               = "<min> to <max>"
    ),
    factor = list(
      "sex" = c("overall", "Male", "Female"),
      "variable_name" = c(
        "Number records", "Number subjects", "Age", "Days in cohort", "Prior observation",
        "Future observation", "Cohort start date", "Cohort end date",
        "Comedications", "Comorbidities"
      ),
      "variable_level" = c(NA, "Asthma", "Depression", "HIV", "Opioids", "Antidiabetes")
    ),
    hide = c("cdm_name", "cohort_name")
  )
```

Additionally, we can show the number of people in the overall cohort and stratified by sex. We use the `CohortCharacteristics` plotting function and then apply the DARWIN style and rotate the axis labels.

```{r}
data$summarised_characteristics |>
  dplyr::filter(variable_name %in% c("Number records")) |>
  plotCharacteristics(colour = "sex") +
  themeVisOmop(style = style) +
  coord_flip()
```

# Incidence results

For the incidence results, we want to plot incidence over time by sex group. We can do that with the plotting functions in **IncidencePrevalence**. We add the `"darwin"` style and rotate x-axis labels for readability.

```{r}
data$incidence |>
  dplyr::filter(strata_name == "sex") |>
  plotIncidence(colour = "sex", facet = "sex", ribbon = TRUE) +
  themeVisOmop(style = style) +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1))
```

# Additional results

Sometimes a study includes an analysis not yet supported by the OMOP Tidy R packages. In that case, we might or might not convert our result to a `summarised_result`, either way, there will be no package-specific table or plot function.

For these cases, we can use **visOmopResults** to create styled tables and plots that match the rest of the report.

As an example, we have mock results for a measurement change after an intervention. The tibble looks like this:

```{r, message=TRUE}
data$measurement_change
```

First, we display results in a table. We pivot to the standard columns `estimate_name`, `estimate_type`, and `estimate_value`; this lets us format estimates and use the `header` argument in the `visTable()` function.

```{r}
data$measurement_change |> 
  tidyr::pivot_longer(
    cols      = c("median", "min", "max", "q25", "q75"),
    names_to  = "estimate_name",
    values_to = "estimate_value"
  ) |>
  dplyr::mutate(
    estimate_type = "numeric",
    estimate_value = as.character(estimate_value),
    variable_name  = customiseText(variable_name),
    sex            = customiseText(sex)
  ) |>
  visTable(
    header = "sex",
    estimateName = c(
      "Median [Q25 - Q75]" = "<median> [<q25> - <q75>]",
      "Range"              = "<min> to <max>"
    ),
    hide   = c("cohort_name", "estimate_type"),
    rename = c("Estimate" = "estimate_name", "Variable" = "variable_name")
  )
```

Additionally, we can visualise values before and after in a boxplot:

```{r}
data$measurement_change |>  
  dplyr::filter(variable_name %in% c("value_before", "value_after")) |>
  dplyr::mutate(
    variable_name = customiseText(variable_name),
    sex           = customiseText(sex)
  ) |>
  boxPlot(x = "variable_name", facet = "sex", colour = "variable_name") +
  theme(axis.text.x = element_blank(), axis.ticks.x = element_blank()) +
  xlab("")
```

Note that we did not specify `type` or `style` in the **visOmopResults** functions above because we set these globally at the beginning of the document.

# Captions and footers

## In Markdown

In the reference Word document, styles are defined for figure legends (`Caption`) and footers (`FigureFooter`). These can be used in Quarto as follows:

```markdown
:::{custom-style="Caption"}
**Table 1:** Baseline population characteristics.
:::
```

## In an R code chunk

We might want to create captions and footers inside an R code chunk (e.g., when generating multiple tables/figures in a loop, or when programmatically numbering them). In those cases, we can use `cat()` within a chunk:

```{r, eval=FALSE}
num_table <- 1
cat(paste0(
  ':::{custom-style="Caption"}\n',
  '**Table ', num_table, ':** Baseline population characteristics.\n',
  ':::\n'
))
num_table <- num_table + 1
```

It’s important that we use the chunk option `results = 'asis'` so that the content produced by `cat()` is correctly interpreted by Quarto/Pandoc as Markdown.

# Report example

You can access an example Quarto script and the reference DOCX document [here](https://github.com/darwin-eu/visOmopResults/tree/main/inst).