13  Chapter 13 — Investigating an Outbreak

Applied Epidemiology in Aquaculture — Part V: Detection & Response

Where this sits — Respond (tertiary prevention). When detection (Chapter 12) raises the alarm, the response begins. This chapter is the capstone of the Respond phase: reading the epidemic curve, running a structured investigation, and estimating where and when exposure happened — using every skill from the book.

13.1 Overview

An outbreak investigation answers four questions under time pressure: what is happening, where, in whom, and why — and then what stops it. The central tool is the epidemic curve, a simple plot of cases over time that encodes the disease’s behaviour. This chapter builds and reads epidemic curves for our farms, extracts the growth rate, back-estimates the exposure window, and lays out the structured steps of a field investigation. It pulls together frequency (Ch 4), association (Ch 5), causal reasoning (Ch 9) and transmission (Ch 10) into one workflow.

We reuse weekly_monitoring.csv (the three cohorts) and regional_srs_weekly.csv (the regional outbreak).

13.2 13.1 Preparation

pacman::p_load(rio, here, tidyverse, incidence2)
mon <- import(here("data", "weekly_monitoring.csv"))
reg <- import(here("data", "regional_srs_weekly.csv"))

13.3 13.2 The epidemic curve

An epidemic curve plots new cases (or deaths) by time of onset. Its shape is diagnostic:

ggplot(mon, aes(week, deaths, fill = species)) +
  geom_col() + facet_wrap(~species, scales = "free_y") +
  labs(x = "Week", y = "Deaths")
  • Point / common-source curves rise and fall sharply around a single exposure. The shrimp AHPND cohort — deaths peaking at week 3, days after stocking — has this signature: a common source (contaminated post-larvae or pond conditions) hitting everyone at once.
  • Propagated curves climb over successive, widening waves as infection passes fish to fish. The salmon SRS cohort — a slower rise to a week-8 peak — is propagated transmission, each generation seeding the next.

Telling the two apart is the first inference an epidemic curve offers, and it immediately narrows the causes worth investigating.

13.4 13.3 Reading the growth rate

The rising limb carries quantitative information. Fit a log-linear model to the growth phase to get the epidemic growth rate and doubling time:

rise <- reg |> filter(week >= 70, week <= 77)
m <- lm(log(cases) ~ week, data = rise)
r  <- coef(m)[["week"]]          # 0.164 per week
dt <- log(2) / r                 # 4.2 weeks  (~30 days)

The regional outbreak doubles about every 4 weeks. Contrast that with the within-population transmission measured experimentally for P. salmonis — a doubling time of about 4 days and R₀ ≈ 4.2 (Mardones et al., in review). The two scales differ by roughly sevenfold, and the gap is itself informative: fast within-cage spread, slower between-farm spread, which is why control operates at both scales (Chapter 14). Growth rate also connects to Chapter 10 — a faster-growing outbreak needs a higher herd-immunity threshold to arrest.

13.5 13.4 Estimating the exposure window

For a point-source outbreak, the curve plus a known incubation (or latent) period back-estimates when exposure occurred: subtract the incubation period from the peak (and from the first and last cases) to bracket the exposure window. For the shrimp AHPND crash peaking at week 3, with onset typically within days of a Vibrio challenge, exposure points squarely at stocking or the immediate post-stocking period — directing the investigation to post-larvae source and pond preparation rather than later husbandry.

13.6 13.5 The structured investigation

A field investigation follows a disciplined sequence (condensed from the classic steps):

  1. Confirm the outbreak is real (rule out a surveillance artefact or lab error — Chapter 6) and verify the diagnosis.
  2. Define a case — explicit, testable criteria (e.g. gross lesions + qPCR below a stated Ct).
  3. Count and find cases — build the line list.
  4. Describe by time, place, animal — the epidemic curve (time), mapping (place, Chapter 14), and attack rates by group (animal).
  5. Generate hypotheses about source and risk factors.
  6. Test them with the measures of association from Chapter 5, adjusting for confounding (Chapter 9).
  7. Implement control and monitor whether the curve turns.
  8. Communicate and document — the Safeguard feedback loop.

Control follows the transmission logic: reduce the number of susceptibles (fallowing, depopulation), cut transmission (biosecurity, density, lice control), and shorten the infectious period (early, correct treatment — Chapter 11). For SRS specifically, the evidence favours immediate depopulation of infected farms, synchronised fallowing, and coordinated neighbourhood action (Lam et al., 2025; Price et al., 2017) — a reminder that for a propagated, area-scale disease, the response reaches beyond the affected cage.

13.7 13.6 Interpretation on the farm

The epidemic curve is the most information-dense object in an investigation: its shape separates common-source from propagated, its slope gives the doubling time, and its position in time points back to exposure. Reading it well turns a chaotic mortality event into a structured problem with a short list of testable causes and a matched set of controls. The discipline — confirm, define, describe, test, control, document — is what keeps a response evidence-based under pressure, and its final step feeds the Safeguard loop so the next outbreak is better anticipated.

13.8 13.7 Exercises

Try it yourself (13.1). Plot the epidemic curve for each cohort. Classify each as point-source or propagated, and justify your call from the shape.

13.2 Fit the log-linear growth model to the regional outbreak’s rising limb and report the doubling time. How would a shorter doubling time change your control urgency?

13.3 The shrimp AHPND curve peaks at week 3. If the incubation period is about 1 week, when was the likely exposure, and which husbandry step does that implicate?

13.4 Write an explicit case definition for an SRS outbreak investigation, using a diagnostic criterion from Chapter 6 and a clinical sign. Why does a vague case definition undermine the whole investigation?

13.5 For a propagated SRS outbreak in one cage of a dense neighbourhood, list one control action at each of the three transmission levers (susceptibles, transmission, infectious period).

13.9 Summary & what’s next

  • The epidemic curve distinguishes point-source (sharp, e.g. AHPND) from propagated (successive waves, e.g. SRS) outbreaks.
  • Its rising limb gives the growth rate and doubling time (here ~4 weeks regionally vs ~4 days within a population).
  • Curve + incubation period back-estimates the exposure window.
  • A structured investigation (confirm → define → describe → test → control → document) applies the whole book, and control follows transmission logic.

In Chapter 14 — Spatial and Network Spread we extend Respond and Anticipate to the landscape: mapping disease, detecting spatial clusters, and tracing spread through movement and connectivity networks.

13.10 References & further reading

  • Mardones FO, Córdova-Bührle F, Diethelm-Varela B, et al. P. salmonis transmission: serial interval, growth rate and R₀. In review.
  • Lam CT, et al. Simulated regional control strategies for SRS. Prev Vet Med. 2025;243:106595.
  • Price D, et al. Fallowing and carry-over of SRS risk. Aquaculture. 2017;473:400–406.
  • Dohoo I, Martin W, Stryhn H. Veterinary Epidemiologic Research. 2nd ed. VER Inc.; 2009.
  • Thrusfield M, Christley R. Veterinary Epidemiology. 4th ed. Wiley-Blackwell; 2018. — outbreak investigation.

Files for this chapter: none new (reuses weekly_monitoring.csv and regional_srs_weekly.csv).