1 Chapter 1 — Why Epidemiology Belongs on a Fish Farm

Applied Epidemiology in Aquaculture — Part I: Foundations

Where this sits — Foundations. This book is organised around five phases of disease management — Anticipate → Prevent → Detect → Respond → Safeguard (the five levels of prevention; Mardones 2020). Chapters 1–3 are the enabling layer beneath all five: the mindset, the tools, and the data skills every phase relies on.

1.1 A tale of two farms

Picture two farms you will get to know very well over the course of this book.

The first is Aysén Salmon Co., a company growing Atlantic salmon in the cold fjords of southern Chile. Their fish start life in freshwater tanks — increasingly in recirculating aquaculture systems (RAS) — and are later transferred to sea cages to grow out. Their veterinarians spend their days worrying about Piscirickettsia salmonis (the cause of SRS, or salmon rickettsial septicaemia), Bacterial Kidney Disease (BKD), and Caligus sea lice.

The second is Lago Verde Tilapia, a warm-water operation raising Nile tilapia in earthen ponds and lake cages. Their headaches are different: Streptococcus outbreaks that flare when the water warms, Tilapia Lake Virus (TiLV), and motile Aeromonas septicaemia.

These two farms look nothing alike — different species, different water, different temperatures, different diseases. And yet, as you will see, the questions their health teams ask are almost identical:

How many of my fish are actually sick, and is that number going up or down?
Which cages, ponds, or batches are worse, and why?
Is this new feed (or vaccine, or water source) actually helping?
If an outbreak starts, where did it come from and how do I stop it spreading?

Epidemiology is the discipline that answers those questions with evidence instead of hunches. This book teaches you to do it, and to do it in R — a free tool that turns a messy spreadsheet of mortality records into a clear answer. Both farms will appear in every chapter, so you always see each idea in a cold-water marine context and a warm-water freshwater one.

1.2 What epidemiology actually is

A classic definition, adapted from Thrusfield’s veterinary epidemiology text, runs like this:

Epidemiology is the study of disease in populations, and of the factors that determine its occurrence.

The word in bold is the whole game. A clinician looks at one sick fish — or one moribund fish pulled from a cage — and asks “what is wrong with this animal and how do I treat it?” An epidemiologist steps back and asks “what is happening across this whole population of fish, and why?”

That shift in scale changes everything. Veterinary epidemiology also explicitly cares about productivity, not just disease — slow growth, poor feed conversion, and downgraded harvests are all legitimate outcomes to study. And because we can almost never examine every single fish, epidemiology is fundamentally about making inferences: looking at a sample and reasoning carefully about the population it came from.

1.2.1 Two ways of seeing the same cage

Here is the difference in one image. Clinical medicine sees only the animals that are visibly unwell. Epidemiology insists on counting everyone — the sick and the healthy — because you cannot interpret 50 dead fish without knowing whether they came from a cage of 500 or a cage of 50,000.

Key idea — the denominator. A count of cases on its own is almost meaningless. “We lost 200 fish this week” could be a catastrophe or a rounding error. The number that gives it meaning is the population at risk sitting underneath it. Most of this book is, in one way or another, about choosing the right denominator.

1.3 Three flavours of epidemiology

Throughout the book we will move between three complementary styles of epidemiological thinking. It helps to name them now.

Descriptive epidemiology organises and summarises what is happening — describing disease and its determinants by animal, place, and time. When the Lago Verde team plots tilapia mortality week by week across their ponds, that is descriptive epidemiology. It quantifies the burden of disease and generates hypotheses.

Analytic (or quantitative) epidemiology tests those hypotheses. It investigates the relationship between an exposure (a suspected risk factor) and an outcome (usually disease), using study designs and statistical analysis to judge how strong and how trustworthy the link is. When Aysén Salmon Co. asks “are cages stocked in river water getting more BKD than cages on RAS water?”, that is analytic epidemiology — and we will answer exactly that question with real numbers in Chapter 5.

Theoretical epidemiology uses mathematical models to predict how disease spreads through a population and to test control strategies on a computer before trying them in the water. We touch on its core intuitions in the outbreak chapter (Chapter 9), particularly the idea of a transmission threshold.

You do not have to choose one. A good investigation usually describes first, then analyses, and occasionally models.

1.4 The epidemiological triad, in water

Disease is never just about the bug. The classic epidemiological triad holds that an outcome emerges from the interaction of three things:

the agent (the pathogen — P. salmonis, Streptococcus, TiLV, sea lice),
the host (the fish, with its species, age, size, immune status, genetics),
and the environment (water temperature, oxygen, salinity, stocking density, water source).

This framing is especially powerful in aquaculture because the environment is so dominant. A Streptococcus strain that causes almost nothing at 24 °C can devastate a tilapia pond at 30 °C. The agent did not change; the environment tipped the balance. Whenever you investigate a problem on either of our farms, train yourself to ask about all three corners of the triad rather than reaching straight for an antibiotic.

1.4.1 Measuring how nasty an agent is

Not every fish that meets a pathogen gets infected; not every infected fish gets sick; not every sick fish dies. Epidemiology gives us three tidy ratios to separate these stages, all of which depend on host factors:

Infectivity — the agent’s ability to infect: infectivity = (number infected ÷ number susceptible) × 100
Pathogenicity — its ability to cause clinical disease: pathogenicity = (number with clinical disease ÷ number exposed) × 100
Virulence — the severity of disease, measured by the case fatality rate: virulence = (number that died ÷ number with disease) × 100

Let’s put numbers to it. Suppose P. salmonis is introduced into a sea cage at Aysén Salmon Co. holding 10,000 susceptible salmon. Over the following weeks, 6,000 become infected, 1,500 develop clinical SRS, and 600 of those die.

Infectivity = 6,000 ÷ 10,000 × 100 = 60%
Pathogenicity = 1,500 ÷ 10,000 × 100 = 15%
Virulence (case fatality) = 600 ÷ 1,500 × 100 = 40%

Now the warm-water parallel. A Streptococcus agalactiae incident hits a Lago Verde pond of 8,000 susceptible tilapia: 5,600 become infected, 2,000 develop clinical signs, and 400 die.

Infectivity = 5,600 ÷ 8,000 × 100 = 70%
Pathogenicity = 2,000 ÷ 8,000 × 100 = 25%
Virulence = 400 ÷ 2,000 × 100 = 20%

Read those side by side and a story emerges: the tilapia Streptococcus spread more readily and made more fish sick (higher infectivity and pathogenicity), but the salmon Piscirickettsia killed a larger share of the fish it sickened (higher virulence). Same three numbers, two very different management problems.

1.5 Your very first R calculation

You don’t need to install anything to read this section — we set up R and RStudio properly, from absolute zero, in Chapter 2. But because this is an applied book, let’s see what that salmon calculation looks like as R code. Don’t worry about memorising the syntax; just notice how readable it is.

# Aysén Salmon Co. — a P. salmonis incident in one sea cage
# The '<-' arrow stores a value in a named object so we can reuse it.

susceptible <- 10000   # fish at risk in the cage
infected    <- 6000    # fish that became infected
diseased    <- 1500    # fish that developed clinical SRS
died        <- 600     # fish that died of SRS

# The '#' marks a comment — R ignores everything after it.
# A '/' divides, and '* 100' turns a proportion into a percentage.

infectivity   <- infected  / susceptible * 100
pathogenicity <- diseased  / susceptible * 100
virulence     <- died      / diseased    * 100

# Print the three results to the screen:
infectivity     # 60
pathogenicity   # 15
virulence       # 40

That is genuinely all there is to it: name some numbers, do arithmetic on the names, and look at the answers. Every analysis in this book is built from pieces this small. By the end of Chapter 2 you will be running code like this yourself; by Chapter 4 you will be doing it on real datasets from both farms.

Try it yourself (1.1). Re-do the calculation for the Lago Verde tilapia Streptococcus incident (8,000 susceptible; 5,600 infected; 2,000 diseased; 400 died). You should recover 70%, 25%, and 20%. Even if you can only do it on paper or in a spreadsheet for now, the goal is to feel comfortable with what each ratio is dividing by.

1.6 One Health: your farm is not an island

Fish health does not stop at the pond bank or the cage net. The One Health perspective recognises that the health of animals, people, and the environment are linked — and aquaculture sits squarely in the middle of that web.

Consider just a few connections that will matter on our farms:

Antimicrobials. Treating an SRS outbreak with antibiotics can, over time, select for resistant bacteria — a problem that reaches well beyond the cage and into human and environmental health. Good epidemiology (knowing when treatment is truly justified) is one of the strongest tools for using these drugs responsibly.
Shared water. Lago Verde’s lake is also used by local communities and wild fish. Pathogens, effluent, and escapees move between farmed and wild populations in both directions.
Food safety and trade. Disease status determines whether a country can export its salmon or tilapia at all. Epidemiological surveillance is what underpins those market-access decisions.

You don’t need to solve One Health to be a good fish health specialist. But keeping it in view explains why we measure carefully, treat sparingly, and document everything — themes that run through the whole book.

1.7 How the rest of the book fits together

Each chapter hands off to the next, building one skill at a time. Here is the road ahead:

Why epidemiology belongs on a fish farm (you are here) — the mindset and vocabulary.
Getting set up with R — install everything and learn the basics, hands-on, with our farm data.
Describing what you see — summarise and visualise data; the statistics you will actually use.
How much disease is there? — prevalence and incidence, the two ways to count disease.
What’s driving it? — risk, odds, relative risk, odds ratios, and attributable fractions.
Choosing a study design — cross-sectional, cohort, and case-control studies and their trade-offs.
How many fish should I sample? — sampling methods and sample-size calculation.
Is it really the cause? — confounding, bias, and reasoning about causation.
Investigating an outbreak — the capstone, pulling every previous skill into one investigation.

A short set of appendices at the end collects an R cheat-sheet, a formula reference, and the datasets.

About the datasets. Starting in Chapter 2, every chapter comes with a small, realistic (simulated) dataset for Aysén Salmon Co. and Lago Verde Tilapia — mortality logs, sampling records, lab results, and so on. They are designed to be downloaded and run alongside the text so you are never just reading about analysis; you are doing it.

1.8 Chapter summary

Epidemiology studies disease in populations, not individuals — and in veterinary work, it also studies productivity.
Because we rarely examine every fish, epidemiology is about inference from a sample to a population.
Always look for the denominator: a count of cases means little without the population at risk beneath it.
The field works in three modes — descriptive, analytic, and theoretical.
Disease arises from the interaction of agent, host, and environment (the triad); in aquaculture the environment is often decisive.
Infectivity, pathogenicity, and virulence are three ratios that separate infection, illness, and death.
The One Health view connects your farm to human and environmental health, and explains why careful measurement matters.

1.8.1 Exercises

1.1 Recompute infectivity, pathogenicity, and virulence for the Lago Verde Streptococcus incident (8,000 susceptible; 5,600 infected; 2,000 diseased; 400 died).

1.2 A neighbouring salmon site reports that out of 12,000 susceptible fish exposed to Caligus, 9,000 became infested, 1,200 showed clinical lesions, and 90 died. Calculate the three metrics. Which is higher here — infectivity or virulence — and what does that suggest about how to manage sea lice compared with SRS?

1.3 For each of the following statements, identify whether it reflects descriptive, analytic, or theoretical epidemiology: (a) “Mortality peaked in week 6 in the eastern ponds.” (b) “Cages on river water had 2.2 times the BKD risk of RAS cages.” (c) “Our model predicts the outbreak will fade once 80% of fish are immune.”

1.4 Pick a disease relevant to your own work and sketch its triad: name a plausible agent factor, a host factor, and an environmental factor that together push fish toward disease.

Next: Chapter 2 — Getting set up with R. We install R and RStudio from scratch, learn to load a farm dataset, and build the handful of skills every later chapter relies on.