🔍 Discover →📚 Learn →🛠 Build →📢 Share →🤖 AI & ML

Shiny — NHS Quickstart

🧪 R · Interactive dashboards · SQL Server/Parquet · NHS-friendly hosting

Why Shiny in the NHS

If your team already uses R for stats and reporting, Shiny converts existing analysis into interactive apps with minimal extra code. It’s ideal for internal dashboards where you want auditable R scripts, quick iteration, and secure, behind-SSO access.

Great for: Clinician-Researcher · BI Analyst (R-heavy) · Data Scientist.

---

⚙️ 10-minute install

In the R console:

install.packages(c("shiny","DBI","odbc","ggplot2","dotenv","readr","dplyr","arrow"))

• For SQL Server access, install the Microsoft ODBC Driver 18 for SQL Server for your OS.
• Keep secrets out of code; use .Renviron or the dotenv package locally and a secret store in production.

---

🚀 “Hello NHS” Shiny app (choose a data source)

A) Parquet/CSV (safer pattern)

Folder layout

shiny-nhs/
  .env
  app.R
  data/
    kpi.parquet   # or kpi.csv

.env (local only — never commit)

DATA_PATH=data/kpi.parquet

app.R

library(shiny)
library(ggplot2)
library(dplyr)
library(dotenv)
library(arrow)   # for Parquet

load_dot_env()
data_path <- Sys.getenv("DATA_PATH", unset = "data/kpi.parquet")

df <- arrow::read_parquet(data_path)

ui <- fluidPage(
  titlePanel("NHS KPI Dashboard (Shiny)"),
  selectInput("top_n", "Top N practices", choices = c(5,10,20), selected = 10),
  plotOutput("barPlot")
)

server <- function(input, output, session) {
  output$barPlot <- renderPlot({
    df |>
      arrange(desc(total_appointments)) |>
      head(as.numeric(input$top_n)) |>
      ggplot(aes(x = reorder(practice_id, total_appointments),
                 y = total_appointments)) +
      geom_col(fill = "#005EB8") +
      coord_flip() +
      labs(x = "Practice", y = "Appointments (30d)",
           title = "Appointments by Practice") +
      theme_minimal(base_size = 12)
  })
}

shinyApp(ui, server)

Run in R:

shiny::runApp("shiny-nhs")

B) SQL Server (direct)

Folder layout

shiny-nhs/
  .Renviron
  app.R

.Renviron (local; or use dotenv)

SQLSERVER_SERVER=YOURSERVER
SQLSERVER_DATABASE=NHS_Analytics

app.R

library(shiny)
library(DBI)
library(odbc)
library(ggplot2)
library(dplyr)

con <- dbConnect(odbc::odbc(),
  Driver = "ODBC Driver 18 for SQL Server",
  Server = Sys.getenv("SQLSERVER_SERVER"),
  Database = Sys.getenv("SQLSERVER_DATABASE"),
  Trusted_Connection = "Yes",
  Encrypt = "Yes",
  TrustServerCertificate = "Yes"   # set per Trust policy
)

onStop(function() dbDisconnect(con))

df <- dbGetQuery(con, "SELECT practice_id,total_appointments FROM dbo.vw_PracticeKPI")

ui <- fluidPage(
  titlePanel("NHS KPI Dashboard (Shiny)"),
  plotOutput("barPlot")
)

server <- function(input, output, session) {
  output$barPlot <- renderPlot({
    ggplot(df, aes(x = reorder(practice_id, total_appointments),
                   y = total_appointments)) +
      geom_col(fill = "#005EB8") +
      coord_flip() +
      theme_minimal(base_size = 12)
  })
}

shinyApp(ui, server)

Prefer read-only accounts and views; for production, consider extracting to Parquet and pointing the app at files to minimise live DB access.

C) DuckDB (local analytics)

Folder layout

shiny-nhs/
  app.R
  data/nhs.duckdb

app.R

library(shiny)
library(DBI)
library(duckdb)
library(dplyr)
library(ggplot2)

con <- dbConnect(duckdb::duckdb(), dbdir = "data/nhs.duckdb", read_only = TRUE)
onStop(function() dbDisconnect(con, shutdown = TRUE))

df <- dbGetQuery(con, "SELECT practice_id, COUNT(*) AS total_appointments
                       FROM appointments
                       GROUP BY practice_id")

ui <- fluidPage(
  h3("NHS KPI (DuckDB)"),
  plotOutput("chart")
)

server <- function(input, output, session){
  output$chart <- renderPlot({
    ggplot(df, aes(reorder(practice_id, total_appointments), total_appointments)) +
      geom_col(fill = "#005EB8") + coord_flip() +
      labs(x="Practice", y="Appointments")
  })
}

shinyApp(ui, server)

---

🧰 NHS-ready add-ons

• Branding: apply NHS.UK palette via bslib or use nhsukdown.
• Auth/SSO: place Shiny behind a reverse proxy (IIS/NGINX) integrated with Entra ID / NHS Login.
• Caching: cache expensive queries with memoise or precompute to Parquet on a schedule.
• Validation: add a lightweight checks page (row counts, freshness) rendered from a validation script.
• Containerisation: package with Docker for consistent deploys (then run on App Runner / Azure Container Apps).

---

🚢 Deploy options

• Posit Connect (formerly RStudio Connect) — easiest publishing, SSO integration options.
• Shiny Server Open Source — single box; put behind SSO at the edge.
• Docker + Cloud — containerise and deploy to AWS App Runner or Azure Container Apps; serve over HTTPS; restrict by IP/SSO.
• Intranet — host on a VM, fronted by IIS/NGINX with basic auth or integrated auth.

Document owner, refresh cadence, and datasets in the app README.

---

🔒 IG & safety checklist

• Use synthetic/de-identified sample data for examples.
• Keep secrets in environment variables; never commit .Renviron/.env.
• Apply suppression for small numbers before plotting.
• Log access via reverse proxy and keep an audit trail of releases.
• Minimise live DB access; prefer read-only extracts for dashboards.

---

📏 Measuring impact

• Latency: data refresh → app reflects change.
• Stability: uptime and error rate.
• Quality: validation pass rate (row counts, freshness checks).
• Adoption: weekly active users; stakeholder satisfaction.

---

🔗 See also

See also: R · SQL · DuckDB · Dash · Docker · AWS · Azure · Secrets & .env

What’s next?

You’ve completed the Learn — Shiny stage. Keep momentum:

Continue to BuildJump to my persona