# Lesson 3: Infectious Disease, Sanitation, and Hygiene

*Companion-podcast transcript • Sarah & Kiffer*  
*Office Hours episode to listen to after working through the lesson*

---

**Sarah:** Welcome back to Office Hours. I'm Sarah.

**Kiffer:** And I'm Kiffer. This episode goes with Lesson three, Infectious Disease, Sanitation, and Hygiene.

**Sarah:** It's meant for after you've worked through the lesson. We'll add some perspective on how these ideas are used in practice, offer some critique, work through the questions students tend to find thorny with this material, and do some extra worked examples, including a couple that are harder than the ones on the lesson page.

**Kiffer:** Along the way there are a few places where we'll ask you to work something out before we give the answer. When that happens, you'll hear a few seconds of quiet, and you can pause the audio if you'd like more time to think it through.

**Sarah:** We'll take four questions. First, what counts as proof that a particular germ causes a disease, and what should you conclude when Koch's postulates can't be met? Second, how do researchers decide how much of a fall in deaths to credit to clean water, when deaths were already falling for other reasons? Third, why does measles need vaccine coverage close to ninety-five percent, and why can a good average still leave room for outbreaks? And fourth, should school vaccination requirements allow exemptions for reasons of conscience or religion? Kiffer and I see that one differently.

**Kiffer:** Question one. It's tempting to treat Koch's four postulates as a checklist, where a microbe that fails any one item is cleared. Let's test that idea with a made-up case. Imagine a long-term care home where a new skin infection appears among residents. Investigators find a particular bacterium in eighteen of twenty residents with the infection. They also find it on the skin of six of forty healthy staff members. It grows well in pure culture in the lab. Nobody would deliberately infect a healthy person, for obvious ethical reasons, and in the one animal model the team tries, the animals don't get sick.

**Sarah:** So here's the question for everyone. Which of the four postulates does this hypothetical bacterium meet, and should the team conclude that it isn't the cause? Take a few seconds.

*(Pause)*

**Sarah:** I'll try it. Postulate two is met, because it grows in pure culture. Postulate one fails, since it was found in eighteen of twenty cases, and the postulate says all cases. Postulate three fails as well, because the animals stayed healthy, and postulate four can't be checked without postulate three. So by Koch's standard, I'd say the case for this bacterium is weak.

**Kiffer:** Your scoring is right. The conclusion is where I'd slow down. Look at the two groups side by side. Eighteen of twenty is ninety percent of the residents with the infection. Six of forty is fifteen percent of the healthy staff. That's a large difference, and the postulates on their own don't use it at all.

**Sarah:** And there are ordinary reasons a real cause could still miss postulate one. A swab can miss an organism that's there, and a resident who had already started antibiotics might test negative.

**Kiffer:** Exactly. The healthy staff who carry it are a familiar problem too. The lesson's example is Helicobacter pylori, which about half the world's population carries, with disease in only a minority. And an animal model can fail simply because that species isn't susceptible. So failing the classical postulates tells you the evidence is incomplete, and the organism stays on the list of suspects.

**Sarah:** Then what would you look for next, if you were advising the team at this care home?

**Kiffer:** I'd start with the molecular postulates from the lesson, which Stanley Falkow set out in 1988. Compare the strains from residents with the strains from healthy staff. Suppose the residents' strain carries a toxin gene that the staff strain lacks. If knocking out that gene in the lab reduces the damage the bacterium does to skin cells, and putting it back restores the damage, that's a much stronger chain of evidence. I'd also sequence the samples, to see whether the residents share one strain, which would fit spread within the home.

**Sarah:** That's also a good habit for reading the news. When a headline says scientists have found the germ behind some illness, the first question is how often the germ turns up in healthy people, and the second is whether there's a mechanism linking it to the damage.

**Kiffer:** Both questions matter, and the second one is usually where the early reports are thinnest. Finding an organism more often in sick people is a strong lead, and it becomes a causal claim once you can show what the organism actually does.

**Sarah:** The real Helicobacter story has a famous twist on postulate three. Robin Warren and Barry Marshall were arguing that this bacterium caused stomach inflammation, and many specialists doubted them. In 1984, Marshall confirmed by endoscopy that he was free of the bacterium, and a few weeks later he drank a culture of it. About five days after that, he developed symptoms, and a second endoscopy showed severe active gastritis. Warren and Marshall shared the 2005 Nobel Prize in Physiology or Medicine.

**Kiffer:** It's a striking story, and it's worth being careful about what it shows. One person is a thin basis for a causal claim, and it's an experiment that would face hard ethical questions today. Its value came from sitting alongside the evidence Warren and Marshall had already gathered from patients. The general point holds for Semmelweis as well. He had strong numbers from two clinics and no mechanism. Germ theory, and later Koch's framework, supplied the kind of chain that explained why his handwashing worked.

**Sarah:** Question two. The lesson cites Cutler and Miller, who estimated that clean water was responsible for nearly half the fall in total mortality in major American cities in the early twentieth century, three quarters of the fall in infant mortality, and nearly two thirds of the fall in child mortality. I'd like to understand how anyone gets from a falling death curve to a number like that.

**Kiffer:** Let's build it with made-up numbers. Picture a city we'll call Riverton. Its typhoid death rate was forty per hundred thousand people in 1906, and thirty-two per hundred thousand in 1910. At the start of 1911, the city opens a water filtration plant. By 1914, the typhoid death rate is fourteen per hundred thousand. So here's the question. How much of the fall in Riverton's typhoid deaths would you credit to the filtration plant? Take a few seconds.

*(Pause)*

**Sarah:** I'll go first. The rate went from thirty-two before the plant to fourteen after it. Thirty-two minus fourteen is eighteen per hundred thousand. So I'd credit the plant with a fall of eighteen, which is more than half of the earlier rate.

**Kiffer:** Before you settle on that, what was the rate doing in the four years before the plant opened?

**Sarah:** It was already falling. Forty to thirty-two over four years is a drop of eight, or two per hundred thousand each year. Oh, I see the problem. If that trend had simply carried on for another four years, the rate would have reached twenty-four by 1914 with no plant at all. So the plant can only claim the gap between twenty-four and fourteen, which is ten per hundred thousand. I gave filtration credit for eight that were coming anyway.

**Kiffer:** That's the core idea of an interrupted time series. You compare what happened after the change with a projection of the trend from before it. Now let's make it harder. Riverton has a neighbour, another made-up city called Lakeview, that didn't filter its water until much later. Lakeview's typhoid rate was thirty-six in 1906, thirty in 1910, and twenty-one in 1914.

**Sarah:** Okay, let me run the same steps on Lakeview. It fell six over four years, so one and a half per year. Projecting forward four more years gives thirty minus six, which is twenty-four in 1914. The actual rate was twenty-one. So Lakeview ended up three below its own trend, without any filtration.

**Kiffer:** And what does Lakeview's extra drop tell you about the effect of Riverton's plant?

**Sarah:** Something else pushed both cities down in those years. Maybe a regional milk law, maybe a run of mild summers. If it affected Riverton the same way, three of Riverton's ten belong to that, and the plant's share is ten minus three, which is seven per hundred thousand. So of the original eighteen, the filtration plant gets about seven, which is a little under forty percent.

**Kiffer:** That's the answer, along with the assumptions behind it. It assumes the earlier trend would have continued in a straight line, and that the comparison city went through the same other changes. Both can be argued. Cutler and Miller used the different times at which cities adopted filtration and chlorination, which is this same logic applied across many cities. That's a far more careful design than a single before and after comparison, and it still rests on assumptions of this kind.

**Sarah:** It's worth noticing which way my first answer was wrong. Because deaths were already falling, the simple before and after comparison made the plant look better than it was. If deaths had been rising before the plant opened, the same shortcut would have made it look worse.

**Kiffer:** Right, and that's a useful habit for any claim about an intervention. Ask what direction things were moving before it started, because the shortcut always borrows that movement and hands it to the intervention.

**Sarah:** Which is really the McKeown debate in miniature. Several changes happen at once, and each analyst has to decide what the decline would have looked like without one of them.

**Kiffer:** Yes. And sanitation has a further problem, which the lesson describes as invisibility. Its effect shows up most clearly when it fails. In May 2000, in Walkerton, Ontario, heavy rain washed E. coli from a nearby farm field into one of the town's wells. Staff hadn't been properly trained, and the inquiry later found that employees had falsified chlorine tests. Seven people died and about two thousand three hundred became ill.

**Sarah:** So Walkerton is a kind of natural experiment in reverse. It shows what the water system had been preventing every ordinary day before that week.

**Kiffer:** Question three, which is about herd immunity. The lesson defines it as the protection a susceptible person gets when so many people around them are immune that the pathogen can't find an unbroken chain of new hosts. It also says the threshold climbs with contagiousness. There's a simple formula behind that, which goes a little beyond the lesson. The threshold equals one minus one divided by the basic reproduction number, which is the average number of people one case would infect in a population where nobody is immune. For a made-up disease with a basic reproduction number of five, that's one minus one fifth, or eighty percent.

**Sarah:** And the logic is easy to see with that example. Each case would infect five people if everyone were susceptible. If eighty percent are immune, only one in five of those contacts lands on someone susceptible, so each case produces about one new case. That's the break-even point. Above eighty percent immune, the chains start to die out.

**Kiffer:** Exactly. Measles has a basic reproduction number usually given as twelve to eighteen. So what herd immunity threshold does that range give you, using the same formula? Take a few seconds.

*(Pause)*

**Sarah:** One minus one twelfth is about 0.917, so about ninety-two percent. One minus one eighteenth is about 0.944, so about ninety-four percent. Measles needs somewhere between ninety-two and ninety-four percent of people immune.

**Kiffer:** Now add a second idea. The vaccine isn't perfect. The Centers for Disease Control and Prevention puts two doses of MMR at about ninety-seven percent effective against measles. If only ninety-seven percent of vaccinated people end up immune, what share of the population needs two doses to reach those thresholds? Take a few seconds.

*(Pause)*

**Kiffer:** You divide the threshold by the effectiveness. At the low end, 0.917 divided by 0.97 is about 0.945, so about ninety-four and a half percent coverage. At the high end, 0.944 divided by 0.97 is about 0.974, so about ninety-seven percent. That's where the familiar figure of about ninety-five percent comes from, and why it leaves almost no slack.

**Sarah:** And this is where averages mislead people. Suppose, hypothetically, a province reports ninety-two percent coverage overall, but that average combines most schools at ninety-six percent with a minority of communities at around seventy. Measles spreads through the connected groups of people who aren't immune, and a provincial average can hide those groups completely.

**Kiffer:** Canada has lived this recently. A measles outbreak began in October 2024 and kept going. By late October 2025, there were more than five thousand cases across ten provinces and territories. On November 10, 2025, the Pan American Health Organization announced that Canada had lost the measles elimination status it had held since 1998, because transmission had continued for twelve months.

**Sarah:** Which connects to the lesson's distinction between elimination and eradication. Elimination is stopping ongoing local spread while the virus circulates elsewhere, so it can be lost again whenever coverage slips and the virus is brought back in.

**Kiffer:** That brings us to question four, which is how far governments should go to get coverage up. In Ontario, children in school have to be immunized against a list of diseases that includes measles, unless they have an exemption. A medical exemption needs a signed form from a physician or nurse practitioner. Parents can also claim an exemption for reasons of conscience or religion, but first they have to complete an education session with public health and sign a statement before a commissioner for taking affidavits. Sarah, make your case first.

**Sarah:** I'd remove the non-medical exemption. The arithmetic we just did leaves almost no room, and exemptions cluster in particular schools and communities, which is exactly where outbreaks start. There's evidence that removing them works. California passed a law in 2015, which took effect in 2016, ending personal belief exemptions for school entry. In the first year, the share of kindergartners entering school without being up to date on vaccinations fell from about 7.2 percent to about 4.4 percent.

**Kiffer:** That's a real effect, and I take it seriously. I'd point to the rest of that evidence too. Medical exemptions in California rose after the law, and other routes into school took up part of the slack. And the politics can turn against you. New Brunswick introduced a bill that would have required school and daycare vaccination with only medical exemptions, and legislators voted it down in June 2020.

**Sarah:** The lesson's own list of responses includes mandates, incentives, persuasion and structural change, and it says none of them is enough alone. I'd read that as support for keeping a firm requirement in the mix. So is your worry that pushing harder produces a backlash that costs more than it gains?

**Kiffer:** That's part of it. The lesson's history makes the same point. Britain's compulsory Vaccination Act of 1853 was fiercely resisted, and the Anti-Compulsory Vaccination League formed in 1867. I'd keep a non-medical exemption and make it demanding, as Ontario's process does, and put most of the effort into access. That means free vaccine, catch-up clinics in schools, and someone who checks the records at school entry and phones the family.

**Sarah:** But an education session can become a box that families tick. And the children affected by an exemption didn't choose it. A child who gets measles because the school around them fell to seventy percent bears the cost of someone else's decision.

**Kiffer:** That's the strongest form of your argument, and I don't have a full answer to it. Where I'd still hold my ground is that a requirement only helps if the people in the lowest-coverage communities actually comply, and those are the communities where trust is often weakest.

**Sarah:** Then I think we can agree on a practical rule. Start with access and a records check at school entry, because those help everyone. And judge any requirement by what happens to coverage in the lowest-coverage schools, checking whether medical exemptions or other routes rise to fill the gap.

**Kiffer:** I'd sign on to that rule. Let's pull it together with three things to take away.

**Sarah:** First, failing Koch's classical postulates means the evidence is incomplete. Compare cases with healthy people, then look for molecular and genomic evidence before you clear a candidate cause.

**Kiffer:** Second, a fall in deaths after an intervention combines the intervention's effect with everything else that was changing at the time. Compare it with the trend that came before and with a similar place that didn't make the change, and say which assumptions your estimate depends on.

**Sarah:** And third, a herd immunity target depends on how contagious the disease is and how well the vaccine works, and outbreaks follow the pockets of low coverage that a good average can hide.

**Kiffer:** If you'd like more practice, rework the Riverton and Lakeview problem with your own made-up numbers, and compare each result with the trend from before the plant opened.

**Sarah:** Next time, it's Lesson four, Nutrition, Physical Activity, and Sleep, where we turn from infections to the everyday behaviours that shape health over a lifetime.

**Kiffer:** Take care, everyone.

**Sarah:** See you in Lesson four.
