Automated Parameterized Reports for Epidemic Scenario Outputs

Weekly briefings that write themselves — Quarto parameters, scheduled rendering, and professional PDF/HTML output. Skill 3 of 20.

business skills

reporting

Quarto

R Markdown

automation

Author

Jong-Hoon Kim

Published

April 24, 2026

1 The business problem

Every Monday morning, your client’s epidemiologist needs a briefing document: current outbreak status, 14-day forecast, and comparison to last week. You could spend 90 minutes copying numbers into PowerPoint, or you could have a script generate a polished PDF in 30 seconds.

Parameterized reports (1) solve this. A single Quarto document (2) becomes a template; parameters (location, date range, scenario) are injected at render time. The output can be HTML for dashboards or PDF for official briefings — from the same source (3).

2 Anatomy of a parameterized report

A parameterized Quarto document declares its inputs in the YAML header:

---
title: "Epidemic Status Report"
params:
  location: "district_a"
  report_date: "2024-03-15"
  forecast_days: 14
  include_uncertainty: true
---

Inside the document, params$location behaves like any R variable. Render it with custom values from the command line:

quarto render report.qmd \
  -P location:district_b \
  -P report_date:2024-04-01 \
  --output reports/district_b_2024-04-01.html

Or from R:

quarto::quarto_render(
  "report.qmd",
  execute_params = list(location = "district_b",
                        report_date = "2024-04-01"),
  output_file = "reports/district_b_2024-04-01.html"
)

3 Building the report template

Below is the core content of the report template — demonstrating how parameters drive the output.

library(ggplot2)
library(dplyr)

# Simulate what the report would receive from its parameters
# (In the real template, these come from params$...)
location     <- "district_a"
report_date  <- as.Date("2024-04-01")
forecast_days <- 14

# Pull data from database (simulated here)
set.seed(as.integer(report_date))
n_hist <- 60
dates  <- seq(report_date - n_hist + 1, report_date, by = "day")

# Simulated historical case counts
peak_day   <- round(n_hist * 0.6)
peak_cases <- 180
lambda_vec <- peak_cases * exp(-0.5 * ((seq_len(n_hist) - peak_day) / 15)^2)
cases_hist <- rpois(n_hist, pmax(lambda_vec, 2))

df_hist <- data.frame(date = dates, cases = cases_hist, type = "Observed")

# Simple exponential smoothing forecast
last_n   <- tail(cases_hist, 7)
alpha    <- 0.3
smoothed <- numeric(length(last_n))
smoothed[1] <- last_n[1]
for (i in 2:length(last_n)) {
  smoothed[i] <- alpha * last_n[i] + (1 - alpha) * smoothed[i - 1]
}
base     <- tail(smoothed, 1)
trend    <- (tail(smoothed, 1) - smoothed[1]) / (length(smoothed) - 1)
# Clip declining trend
trend    <- min(trend, 0)

fcst_dates  <- seq(report_date + 1, by = "day", length.out = forecast_days)
fcst_median <- pmax(1, base + trend * seq_len(forecast_days))
fcst_lo     <- pmax(0, fcst_median * 0.6)
fcst_hi     <- fcst_median * 1.5

df_fcst <- data.frame(
  date   = fcst_dates,
  median = fcst_median,
  lo     = fcst_lo,
  hi     = fcst_hi
)

# Key numbers for the executive summary table
week_total  <- sum(tail(cases_hist, 7))
prev_total  <- sum(cases_hist[(n_hist - 13):(n_hist - 7)])
change_pct  <- round((week_total - prev_total) / prev_total * 100, 1)
peak_fcst   <- round(max(fcst_median))
total_fcst  <- round(sum(fcst_median))

summary_tbl <- data.frame(
  Metric = c("Cases this week", "Change vs prior week",
             "Peak forecast (next 14d)", "Total forecast (next 14d)"),
  Value  = c(format(week_total, big.mark = ","),
             paste0(ifelse(change_pct > 0, "+", ""), change_pct, "%"),
             format(peak_fcst, big.mark = ","),
             format(total_fcst, big.mark = ","))
)

knitr::kable(summary_tbl, caption = paste("Summary for", location,
             "as of", format(report_date, "%B %d, %Y")))

Summary for district_a as of April 01, 2024
Metric	Value
Cases this week	482
Change vs prior week	-43.8%
Peak forecast (next 14d)	63
Total forecast (next 14d)	689

ggplot() +
  geom_col(data = df_hist, aes(x = date, y = cases),
           fill = "grey65", width = 0.8) +
  geom_ribbon(data = df_fcst,
              aes(x = date, ymin = lo, ymax = hi),
              fill = "steelblue", alpha = 0.3) +
  geom_line(data = df_fcst, aes(x = date, y = median),
            colour = "steelblue", linewidth = 1.2) +
  geom_vline(xintercept = report_date, linetype = "dashed",
             colour = "grey30") +
  annotate("text", x = report_date + 1,
           y = max(cases_hist) * 0.92,
           label = "← history | forecast →",
           hjust = 0, size = 3.2, colour = "grey30") +
  labs(x = NULL, y = "Daily cases",
       title  = paste("Epidemic Status:", location),
       subtitle = paste("Report date:", format(report_date, "%B %d, %Y"))) +
  theme_minimal(base_size = 13)

Epidemic curve with 14-day forecast. This chart is the core visual in the automated client report — the same template generates it for any location and date.

4 Generating multiple reports at once

Loop over all client locations and dates:

# generate_reports.R
library(quarto)

locations    <- c("district_a", "district_b", "district_c")
report_date  <- Sys.Date()

for (loc in locations) {
  out_file <- sprintf("reports/%s_%s.html", loc,
                      format(report_date, "%Y-%m-%d"))
  quarto_render(
    "report_template.qmd",
    execute_params = list(location     = loc,
                          report_date  = as.character(report_date),
                          forecast_days = 14),
    output_file = out_file
  )
  message("Generated: ", out_file)
}

Schedule this with cron (Linux) or Task Scheduler (Windows) to run every Monday at 6 AM, and reports land in a shared folder before the 9 AM meeting.

5 Professional formatting tips

Use gt or kableExtra for tables — they render correctly in both HTML and PDF.
Set format: pdf in the YAML for official documents; format: html for web delivery.
Add a params-driven title block: title: "Epidemic Status Report — ?meta:params.location".
For branded output, add a custom SCSS theme (HTML) or LaTeX template (PDF) once and reuse across all reports.

6 References

Xie Y, Allaire JJ, Grolemund G. R markdown: The definitive guide [Internet]. Chapman; Hall/CRC; 2018. Available from: https://bookdown.org/yihui/rmarkdown/

Allaire JJ, Teague C, Scheidegger C, Xie Y, Dervieux C. Quarto. 2022. doi:10.5281/zenodo.5960048

Knuth DE. Literate programming. The Computer Journal. 1984;27(2):97–111. doi:10.1093/comjnl/27.2.97