--- title: "Enrollment Diagrams" output: rmarkdown::html_vignette vignette: > %\VignetteIndexEntry{Enrollment Diagrams} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{r, include = FALSE} ## Use ragg for better font rendering if available if (requireNamespace("ragg", quietly = TRUE)) { knitr::opts_chunk$set( dev = "ragg_png", fig.retina = 1, collapse = TRUE, comment = "#>", message = FALSE, warning = FALSE, out.width = "100%", dpi = 150 ) } else { knitr::opts_chunk$set( collapse = TRUE, comment = "#>", message = FALSE, warning = FALSE, out.width = "100%", dpi = 150 ) } ## Dynamic figure sizing: queue_flow() stashes recommended dimensions from ## recdims(), and the opts_hook on the NEXT chunk (with use_rec_dims = TRUE) ## applies them before knitr opens the graphics device. Plots render via ragg ## (dev = "ragg_png" set above) and knitr captures them natively. .flow_dims <- new.env(parent = emptyenv()) .flow_dims$width <- NULL .flow_dims$height <- NULL knitr::opts_hooks$set(use_rec_dims = function(options) { if (isTRUE(options$use_rec_dims)) { if (!is.null(.flow_dims$width)) options$fig.width <- .flow_dims$width if (!is.null(.flow_dims$height)) options$fig.height <- .flow_dims$height .flow_dims$width <- NULL .flow_dims$height <- NULL } options }) ## Call at the end of a flow-creation chunk to stash dimensions for the next chunk. queue_flow <- function(flow, ...) { ## Measure on the same device family that renders the figures (ragg, set ## via dev = "ragg_png" above) so that non-default fonts---whose metrics ## differ between devices---are sized consistently and the canvas is not ## cropped. Falls back to recdims()'s default pdf measurement otherwise. md <- if (requireNamespace("ragg", quietly = TRUE)) { function() { tf <- tempfile(fileext = ".png") ragg::agg_png(tf, width = 10, height = 10, units = "in", res = 150) tf } } else NULL dims <- selecta::recdims(flow, ..., .measure_dev = md) .flow_dims$width <- dims["width"] .flow_dims$height <- dims["height"] invisible(flow) } ``` Enrollment and flow diagrams provide a complete, transparent account of participant disposition throughout a clinical study. Reporting guidelines developed by the EQUATOR Network—including CONSORT for randomized trials, STROBE for observational cohorts, and STARD for diagnostic accuracy studies—mandate structured diagrams that trace participant flow from initial assessment through final analysis, documenting every exclusion with counts and reasons. In `selecta`, enrollment diagrams are built around the following core functions: | Function | Purpose | |:---------|:--------| | `enroll()` | Establish the starting cohort from data or a manual count | | `allocate()` / `stratify()` | Split into randomized arms (CONSORT) or strata (STROBE) | | `assess()` | Record receipt of a test or procedure (STARD) | Thus, the enrollment pipeline adheres to the following basic structure: ```{r, eval = FALSE} enroll(data, id) |> phase("Enrollment") |> exclude(label, criterion, reasons) |> phase("Allocation") |> allocate(column) |> endpoint(label) |> flowchart() ``` where `enroll()` establishes the starting cohort (from data or a manual count), pipeline functions define exclusion criteria and structural elements, and `flowchart()` renders the final diagram. This vignette demonstrates the package's capabilities across three EQUATOR guidelines using both data-driven and manual construction modes. > *n.b.:* To ensure correct font rendering and figure sizing, the diagrams below are displayed using a vignette-only helper function (`queue_flow()`) that applies recommended dimensions from `recdims()` via the [`ragg`](https://ragg.r-lib.org/) graphics device, with the standard output function applied afterwards (`flowchart()`). In practice, replace this `queue_flow()`/`flowchart()` workflow with a call to `flowsave()` for equivalent printed results: > > ```{r, eval = FALSE} > flowsave(flow, "consort.pdf") > flowsave(flow, "consort.png", dpi = 300) > ``` > > Using `flowsave()` ensures that the figure dimensions are always large enough to accommodate the diagram content, and it is the preferred method for saving flow diagram outputs in `selecta`. --- # Preliminaries The examples in this vignette use the built-in datasets included with `selecta`: ```{r setup} library(selecta) library(data.table) data(selectaex0) data(selectaex2) data(selectaex3) data(selectaex6) ``` Each dataset simulates a clinical study with columns for patient identifiers, eligibility flags, treatment assignments, and follow-up outcomes. The numeric suffix indicates the number of treatment arms: `selectaex0` represents an observational cohort (no randomization), while `selectaex2`, `selectaex3`, and `selectaex6` represent two-, three-, and six-arm randomized trials, respectively. --- # Operating Modes The package supports two operating modes: | Mode | Entry Point | Counts | Cohort Extraction | |:-----|:------------|:-------|:------------------| | Data | `enroll(data, id)` | Computed from data | Available via `cohort()` | | Manual | `enroll(n = 500)` | Supplied by user | Not available | In **data mode**, participant counts and exclusion reasons are computed directly from the dataset. This approach is reproducible, auditable, and enables downstream cohort extraction with `cohort()`. In **manual mode**, the analyst supplies all counts explicitly—useful for diagrams constructed from published summary statistics or when the source data are not available in R. --- # CONSORT — Randomized Controlled Trials The CONSORT (Consolidated Standards of Reporting Trials) statement requires a flow diagram showing the number of participants at each stage of a randomized trial: enrollment, allocation, follow-up, and analysis. The diagram must report exclusion counts with reasons at each stage and the final number analyzed per arm. ## **Example 1:** Data-Driven Two-Arm Trial The most common CONSORT diagram involves a two-arm parallel trial. In data mode, counts are derived automatically from the dataset: ```{r} example1 <- enroll(selectaex2, id = "patient_id") |> phase("Screening") |> exclude("Duplicate records", criterion = is_duplicate == TRUE, included_label = "Unique records") |> exclude("Failed eligibility", criterion = eligible == FALSE, reasons = "exclusion_reason", included_label = "Eligible cohort") |> phase("Allocation") |> allocate("treatment") |> phase("Follow-up") |> exclude("Discontinued", criterion = discontinued == TRUE, reasons = "discontinuation_reason") |> phase("Analysis") |> endpoint("Analysis cohort") ``` ```{r, echo = FALSE} queue_flow(example1) ``` ```{r, use_rec_dims = TRUE, echo = TRUE} flowchart(example1) ``` Each `exclude()` call filters the dataset according to the supplied expression, and the resulting counts populate the diagram automatically. The `reasons` argument accepts either a column name (for data-driven sub-reason counts) or a named numeric vector (for manual specification). The `included_label` argument adds a labeled count box below the exclusion, showing the number remaining after that step. The `allocate()` function splits the flow into parallel arms based on the named column. For a two-arm trial, arms are positioned symmetrically about the center axis with exclusion side boxes to the left and right. ## **Example 2:** Data-Driven Three-Arm Trial Trials with three or more arms follow the same syntax. The layout automatically adapts to accommodate additional columns: ```{r} example2 <- enroll(selectaex3, id = "patient_id") |> phase("Screening") |> exclude("Duplicate records", criterion = is_duplicate == TRUE, included_label = "Unique records") |> exclude("Failed eligibility", criterion = eligible == FALSE, reasons = "exclusion_reason", included_label = "Eligible cohort") |> phase("Allocation") |> allocate("treatment") |> phase("Follow-up") |> exclude("Discontinued", criterion = discontinued == TRUE, reasons = "discontinuation_reason") |> phase("Analysis") |> endpoint("Analysis cohort") ``` ```{r, echo = FALSE} queue_flow(example2) ``` ```{r, use_rec_dims = TRUE, echo = TRUE} flowchart(example2) ``` ## **Example 3:** Manual Mode When source data are unavailable, all counts can be supplied directly. Manual mode is particularly useful for reproducing published diagrams or constructing diagrams from summary tables: ```{r} example3 <- enroll(n = 1200, label = "Assessed for eligibility") |> phase("Enrollment") |> exclude("Excluded", n = 300, reasons = c("Not meeting criteria" = 160, "Declined to participate" = 90, "Other reasons" = 50), included_label = "Eligible cohort") |> phase("Allocation") |> allocate(labels = c("Drug A", "Placebo"), n = c(450, 450)) |> phase("Follow-up") |> exclude("Lost to follow-up", n = c(20, 20)) |> exclude("Discontinued intervention", n = c(15, 15)) |> phase("Analysis") |> endpoint("Analyzed") ``` ```{r, echo = FALSE} queue_flow(example3) ``` ```{r, use_rec_dims = TRUE, echo = TRUE} flowchart(example3) ``` In manual mode, `allocate()` requires explicit `labels` and `n` arguments rather than a column name. When `exclude()` is called after allocation, the `n` argument accepts a vector with one value per arm. ## **Example 4:** Count-First Display Mode The `count_first` parameter reformats all boxes to place the bold count before the label—*e.g.,* "450 Drug A" rather than "Drug A (*n* = 450)": ```{r, echo = FALSE} queue_flow(example3, count_first = TRUE) ``` ```{r, use_rec_dims = TRUE, echo = TRUE} flowchart(example3, count_first = TRUE) ``` --- # STROBE — Observational Cohort Studies The STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement covers cohort, case-control, and cross-sectional studies. Unlike CONSORT, observational studies do not involve randomization; instead, participants are stratified by exposure or another grouping variable. The `stratify()` function replaces `allocate()` in this context, using the more general term appropriate to non-randomized designs. ## **Example 5:** Single-Arm Cohort Before introducing stratification, consider the simplest observational diagram: a single cohort carried through eligibility and follow-up without any grouping. In data mode, `selectaex0` provides an observational dataset with no treatment arms: ```{r} example5 <- enroll(selectaex0, id = "patient_id") |> phase("Enrollment") |> exclude("Ineligible", criterion = eligible == FALSE, reasons = "exclusion_reason", included_label = "Eligible cohort") |> phase("Follow-up") |> exclude("Lost to follow-up", criterion = lost_to_followup == TRUE, reasons = "followup_loss_reason") |> phase("Analysis") |> endpoint("Analysis cohort") ``` ```{r, echo = FALSE} queue_flow(example5) ``` ```{r, use_rec_dims = TRUE, echo = TRUE} flowchart(example5) ``` With no `allocate()` or `stratify()` call, the diagram remains a single vertical column and each exclusion is drawn as a side box. This is the minimal building block from which all other layouts extend. ## **Example 6:** Exposure-Stratified Cohort The following diagram depicts a registry-based observational cohort stratified by exposure level, with per-arm exclusion labels: ```{r} example6 <- enroll(n = 3860, label = "Registry patients") |> phase("Enrollment") |> exclude("Excluded", n = 420, reasons = c("Missing exposure data" = 210, "Prior treatment" = 130, "Withdrew consent" = 80), included_label = "Eligible cohort") |> phase("Stratification") |> stratify(labels = c("Low exposure", "Medium exposure", "High exposure"), n = c(1200, 1300, 940), label = "Exposure level") |> phase("Follow-up") |> exclude("Lost to follow-up", n = c(60, 75, 45)) |> exclude("Discontinued intervention", n = c(20, 15, 30)) |> phase("Analysis") |> endpoint("Analysis cohort") ``` ```{r, echo = FALSE} queue_flow(example6) ``` ```{r, use_rec_dims = TRUE, echo = TRUE} flowchart(example6) ``` The `stratify()` function is the guideline-agnostic generalization of `allocate()`. In fact, `allocate()` is implemented as a thin wrapper around `stratify()` with a default label of "Randomized." Both produce identical diagram structures; the distinction is semantic, reflecting whether the arm assignment was randomized or observational. --- # STARD — Diagnostic Accuracy Studies The STARD (Standards for Reporting of Diagnostic Accuracy Studies) flow diagram tracks participants through index test administration, reference standard evaluation, and final diagnostic classification. Two features distinguish STARD diagrams from CONSORT: inverted exclusion labels (*e.g.,* "Did not receive index test" rather than "Received index test") and terminal cross-classification of results. ## **Example 7:** Index Test and Reference Standard The `assess()` function provides the inverted label semantics required by STARD. Given a label such as "Index test," it automatically generates the side box label "Did not receive index test" and the count box label "Received index test": ```{r} example7 <- enroll(n = 500, label = "Potentially eligible patients") |> phase("Enrollment") |> exclude("Excluded", n = 40, reasons = c("Refused" = 25, "Not meeting criteria" = 15)) |> phase("Index") |> assess("Index test", not_received = 22, reasons = c("Refused" = 12, "Contraindicated" = 10)) |> phase("Reference") |> assess("Reference standard", not_received = 18, reasons = c("Lost to follow-up" = 10, "Inconclusive" = 8)) |> phase("Results") |> stratify(labels = c("Index test positive", "Index test negative"), n = c(180, 240), label = "Index test result") |> endpoint("Final diagnosis", breakdown = list( c("Target condition +" = 160, "Target condition -" = 20), c("Target condition +" = 15, "Target condition -" = 225) )) ``` ```{r, echo = FALSE} queue_flow(example7) ``` ```{r, use_rec_dims = TRUE, echo = TRUE} flowchart(example7) ``` The `endpoint()` function accepts a `breakdown` argument to display sub-classifications within the terminal box (or boxes). A single named numeric vector itemizes one terminal box; a list of named numeric vectors (one per arm) itemizes each per-arm box after a split, as in Example 6 above. In default `grid` outputs, these are rendered in a smaller italic font to visually distinguish them from the main count. For STARD diagrams, this is how the final target-condition breakdown is shown beneath each index-test result. --- # Cohort Extraction In data mode, the `cohort()` function returns the dataset remaining after all exclusion criteria have been applied, enabling a seamless transition from diagram construction to statistical analysis: ```{r} final_data <- cohort(example1) dim(final_data) ``` When arms are present, `cohort()` returns the combined dataset by default. Per-arm datasets are available via `split = TRUE` or by specifying a single arm: ```{r} arm_data <- cohort(example1, split = TRUE) vapply(arm_data, nrow, integer(1L)) ``` The `cohorts()` function returns stage-by-stage snapshots of the dataset at each exclusion step. Each element is a list with `included`, `excluded`, `n_included`, and `n_excluded`, allowing inspection of either the participant counts or the underlying datasets at each step: ```{r} snapshots <- cohorts(example1) names(snapshots) ``` A specific stage can be accessed by name, with counts and datasets available as named elements: ```{r} snapshots[["Failed eligibility"]]$n_included snapshots[["Failed eligibility"]]$n_excluded ``` --- # Inspecting the Diagram Structure Before rendering, the computed graph can be inspected programmatically. The `print()` method provides a text summary of the pipeline: ```{r} print(example1) ``` The `summary()` method returns a tabular representation of every node in the diagram: ```{r} summary(example1) ``` The `recdims()` function returns the recommended figure dimensions (in inches) without rendering: ```{r} recdims(example1) ``` --- # Saving to File The `flowsave()` function saves the diagram to a file (PDF, PNG, SVG, or TIFF) with auto-computed dimensions: ```{r, eval = FALSE} flowsave(example1, "consort_2arm.pdf") flowsave(example1, "consort_2arm.png", dpi = 300) ``` Explicit dimensions override the automatic calculation: ```{r, eval = FALSE} flowsave(example1, "consort_2arm.pdf", width = 10, height = 12) ``` All visual parameters accepted by `flowchart()` are also accepted by `flowsave()`: ```{r, eval = FALSE} flowsave(example1, "consort_2arm_cf.pdf", count_first = TRUE, cex = 1.0, cex_side = 0.8) ``` --- # Further Reading - [Systematic Reviews](systematic_reviews.html): PRISMA and MOOSE diagrams with top-level source convergence - [Split-and-Recombine Diagrams](split_recombine.html): Hybrid topologies for screening validation and exposure classification - [Advanced Workflows](advanced_workflows.html): Factorial (nested-split) designs and hierarchical exclusion reasons - [Graphviz Export](graphviz_export.html): DOT output for Graphviz/DiagrammeR rendering