# Lesson 1 — Foundations of Epidemiology (v3 expanded)

*Companion-podcast transcript • Sarah & Kiffer* 
*~5,500 words • ~30 min audio*

---

**Sarah:** Welcome to Office Hours, the companion podcast. I'm Sarah.

**Kiffer:** And I'm Kiffer. This is Lesson 1, Foundations of Epidemiology — really three lessons stitched into one. The history of the discipline, the philosophy of how we know what we know, and what happens when the system designed to produce knowledge breaks down.

**Sarah:** Working definition for someone brand new?

**Kiffer:** Epidemiology is the study of disease patterns, causes, and effects in populations. Not individual patients, but whole groups. Who gets sick, who doesn't, why, when, where. The questions are about prevention, not just treatment.

**Sarah:** And the lesson opens by saying there are two stories running side by side. An optimistic one and a much darker one. You're asked to hold both.

**Kiffer:** The optimistic story is that this discipline helped produce extraordinary improvements in human health. The darker story is about who paid the price. Both are true.

**Sarah:** Okay, section one is the history. Where does it start?

**Kiffer:** Ancient Greece. Hippocrates, around twenty-five hundred years ago. His text Airs, Waters, and Places made the radical claim that disease has natural causes tied to climate, water, geography, the seasons. Before that, the dominant Western explanation was supernatural — the gods are angry. Hippocrates moved the question from the gods to the natural world.

**Sarah:** Once you make that move, you can study disease and intervene on it. He also gave us epidemic and endemic.

**Kiffer:** Epidemic means a disease that visits a community — a surge above what you'd expect. Endemic means a disease that resides in a community, always there at some baseline.

**Sarah:** Then roughly two thousand years of miasma theory — bad air rising up from swamps.

**Kiffer:** Miasma is wrong about the mechanism, but here's the interesting part. Even wrong, it produced enormous public-health benefits. Drain the swamps, build sewers, bring in clean water — all of those reduced disease for the wrong stated reason.

**Sarah:** Then the seventeenth century gives us John Graunt, who taught us to count.

**Kiffer:** Graunt was a London haberdasher. No medical training. In 1662 he published an analysis of the Bills of Mortality — weekly death records London parish clerks had kept since 1532. He found stable ratios of male to female births, seasonal mortality patterns, higher death rates in cities, and he built the first life table.

**Sarah:** And the clerks had been recording deaths for over a hundred years before Graunt picked up the data.

**Kiffer:** That's the first appearance of actor-network theory. Scientific advances depend on networks of people, institutions, technologies. The clerks, the parishes, the printing presses are all part of what made Graunt's work possible.

**Sarah:** Eighteenth century, Jenner and vaccination.

**Kiffer:** In 1796 Edward Jenner showed that inoculating someone with cowpox protected them against smallpox. Smallpox killed roughly thirty percent of those it infected and was eradicated in 1980 — the only human disease ever eradicated.

**Sarah:** And Jenner didn't discover the cowpox-smallpox link from scratch.

**Kiffer:** Dairy workers had observed for decades that milkmaids who'd had cowpox didn't catch smallpox. The lesson flags this as knowledge extraction — the dairymaids made the observation, the physician got the credit, the dairymaids became invisible.

**Sarah:** Then the nineteenth century, where the lesson focuses on four figures.

**Kiffer:** Snow, Semmelweis, Farr, and Nightingale. John Snow first. London, 1854. Cholera outbreak in Soho. Snow doesn't buy the miasma theory. He maps where the victims lived and finds almost all of them had drawn water from one specific pump on Broad Street. He convinces authorities to remove the handle and the outbreak subsides.

**Sarah:** And he compared death rates between two water companies — one drawing from sewage-contaminated Thames, one from cleaner water upstream. The contaminated one had much higher cholera mortality.

**Kiffer:** Snow didn't do it alone. He needed Farr's mortality records, cooperative officials, cartographic tools, and a public already politically primed for sanitary reform.

**Sarah:** Second figure, Ignaz Semmelweis. Vienna, 1840s.

**Kiffer:** Two maternity wards. One run by physicians, one by midwives. The death rate from a postpartum infection called childbed fever is ten to eighteen percent in the physicians' ward versus two to four percent in the midwives'. Semmelweis figures out the physicians are walking from autopsies straight into pelvic exams without washing their hands. He introduces handwashing with chlorinated lime, mortality drops to one or two percent — and his findings get rejected.

**Sarah:** Physicians were offended at the suggestion that their hands could be instruments of death.

**Kiffer:** Semmelweis was eventually committed to an asylum and died there. Having the right answer isn't enough. Institutional resistance can delay life-saving knowledge for decades.

**Sarah:** Third, William Farr.

**Kiffer:** The quiet bureaucrat hero. Farr worked for the Registrar General of England and Wales and built the basic toolkit of vital statistics. He standardized disease classifications so deaths in Manchester could be compared to deaths in London. And he gave us excess mortality — the gap between observed deaths and the deaths you'd have expected. That's what everyone reached for during COVID-19.

**Sarah:** Fourth, Florence Nightingale.

**Kiffer:** Most students know her as a nursing reformer, the lady with the lamp. Fewer know she was a serious statistician. She used a polar area diagram — sometimes called a coxcomb — to show that far more British soldiers in the Crimean War died from preventable infectious disease in unsanitary military hospitals than from battle wounds. One of the first uses of data visualization as policy advocacy.

**Sarah:** What those four have in common is they didn't work from theory alone. They took an existing stream of data and turned it into a public argument that something had to change.

**Kiffer:** Numbers to action. And actor-network theory — developed by Bruno Latour and Michel Callon — names the structure. Breakthroughs aren't produced by individual geniuses. They're produced by networks, including non-human actors like microscopes, printing presses, even the cholera bacterium itself.

**Sarah:** Twentieth century. Are we actually making progress against disease?

**Kiffer:** Kermack and McKendrick in 1927 published the Susceptible-Infected-Recovered model — usually shortened to S I R. It divides a population into three compartments and is the mathematical foundation of every epidemic curve you saw during COVID-19. Around the same time, occupational epidemiology takes shape — coal miners and lung disease, asbestos and mesothelioma.

**Sarah:** After the Second World War, the focus shifts to chronic disease.

**Kiffer:** Two studies every student needs to know. The British Doctors' Study, launched in 1951 by Richard Doll and Austin Bradford Hill, followed over forty thousand British physicians and established the causal link between smoking and lung cancer. And the Framingham Heart Study, started in 1948, has followed an entire community in Massachusetts for over seventy years and gave us the major cardiovascular risk factors. Bradford Hill in 1965 also gave us the criteria for evaluating whether an association is likely to be causal. Still the workhorse framework today.

**Sarah:** And the randomized controlled trial emerges around the same time — investigators randomly assigning people to an intervention or a placebo. The 1948 streptomycin trial for tuberculosis is considered the first modern one.

**Kiffer:** So has all of this produced progress? The lesson leans on Hans Rosling. His book Factfulness argues the world has gotten dramatically better on many measures, and most people have systematically distorted views.

**Sarah:** Headline numbers?

**Kiffer:** Child mortality from forty-three percent in 1800 to under four today. Extreme poverty from about eighty-five percent to under ten. Life expectancy more than doubled. And smallpox, which killed three hundred million people in the twentieth century alone, was eradicated in 1980.

**Sarah:** And Rosling isn't saying the world is fine.

**Kiffer:** No, he's saying the world is both better than most people think and still deeply flawed. If we believe nothing works, we become fatalistic. If we see vaccination and sanitation have produced measurable improvement, we have reason to invest in more. And the actor-network perspective pushes back on the temptation to credit epidemiology alone. The improvements are the product of intersecting networks — science, infrastructure, governance, culture, economics, social movements. Epidemiology is one node in a much larger network.

**Sarah:** Which sets up section three of the historical arc. The shadow history.

**Kiffer:** The optimistic story is true. It's also incomplete. The accumulation of medical knowledge depended in significant part on the suffering and exploitation of enslaved people, colonized populations, and marginalized communities.

**Sarah:** And the lens the lesson uses is Michel Foucault.

**Kiffer:** Foucault wrote about how the modern state knows and governs the bodies of its citizens. Two concepts. Biopower is his term for the power to manage life itself at the level of populations. The medieval king's power was the right to take life or let live. Biopower is the modern state's power to regulate birth rates, mortality, fertility, health, longevity. Biopolitics is the practical side — census-taking, vital statistics, public-health campaigns, quarantine, immigration screening, disease surveillance.

**Sarah:** Which lands on the claim that epidemiology is, at its core, a biopolitical science.

**Kiffer:** It emerged alongside the modern state's need to know, count, and manage populations. That's not inherently sinister. Surveillance saves lives. But Foucault wants us to ask who gets counted, whose deaths are invisible, who benefits, and who controls the categories.

**Sarah:** Then historian Jim Downs gives us the cases. His book is Maladies of Empire.

**Kiffer:** Downs argues the standard origin story, centered on Snow in London, obscures a deeper history. The systematic study of disease patterns didn't begin in 1854. It was developed earlier through the infrastructures of colonial empire, the slave trade, and military campaigns. Plantations, slave ships, and military barracks were the original epidemiological laboratories.

**Sarah:** Walk us through the plantation as laboratory.

**Kiffer:** The conditions mirror an epidemiological study — people in a defined area, systematic surveillance, identifiable exposures, and people with no power to refuse. Plantation physicians developed working theories of yellow fever and cholera. That knowledge travelled to colonizing societies. The enslaved people whose suffering produced the data didn't benefit. And the methods we credit to Snow were used earlier in colonial settings by physicians like James McWilliam and Gavin Milroy. They're absent from standard histories because the populations involved weren't valued. Nightingale's statistical work also drew on data from Britain's colonial and military infrastructure. The lesson is doing the both-and move pretty deliberately.

**Sarah:** Then two twentieth-century case studies that shape the consent rules every modern study operates under. First, Tuskegee.

**Kiffer:** From 1932 to 1972. Forty years. The United States Public Health Service enrolled three hundred ninety-nine African American men with syphilis in Macon County, Alabama. They were told they were being treated for, quote, bad blood. They weren't being treated at all. The researchers wanted to study the natural progression of untreated syphilis. Penicillin became the standard cure in the late 1940s. It was cheap, available, and deliberately withheld.

**Sarah:** Exposed by a whistleblower in 1972.

**Kiffer:** The institutional response was the National Research Act of 1974 and the Belmont Report in 1979, which established three principles — respect for persons, beneficence, and justice. Every research ethics board in North America still operates under those. And the legacy isn't just historical. Research shows African Americans report lower trust in medical institutions, linked to lower clinical-trial participation and elevated vaccine hesitancy during COVID-19.

**Sarah:** Second case study, closer to home. Nutritional experiments on Indigenous children in Canada.

**Kiffer:** From 1942 to 1952, documented by historian Ian Mosby. About a thousand children attending six residential schools across five provinces. Residential schools were institutions designed to forcibly assimilate Indigenous children by removing them from their families. The Truth and Reconciliation Commission characterized the system as cultural genocide.

**Sarah:** And the experiments happened inside that system.

**Kiffer:** Government researchers, knowing the children were already malnourished, divided them into experimental and control groups. Some received vitamin supplements. Others were deliberately kept on deficient diets to serve as controls. In some cases, dental care was withdrawn so researchers could observe disease progression. No consent was sought. In 2023, the Canadian Medical Association formally apologized to Indigenous Peoples for its role in medical racism. We're still in the early stages of institutional reckoning.

**Sarah:** Then the lesson takes one more step. The Black Belt example, which traces a chain from prehistoric geology to contemporary heart disease.

**Kiffer:** Start with the Cretaceous period, a hundred million years ago. A shallow inland sea covered what's now the southeastern United States. Marine organisms fell to the seafloor and compressed into a crescent-shaped band of rich dark soil. The Black Belt, named for the soil. Ideal for cotton. In the nineteenth century, the Black Belt was where enslaved Africans were concentrated in the largest numbers. After the Civil War, those counties retained large Black populations but experienced very little economic development. A landmark study by Kramer and colleagues in 2017 found southern counties with higher concentrations of enslaved people in 1860 experienced significantly slower declines in heart disease mortality in the late twentieth century. The geography of slavery in 1860 is still visible in the geography of heart disease more than a century and a half later.

**Sarah:** And the principle the lesson draws.

**Kiffer:** Health disparities are not natural facts. They're produced by historically specific systems of power, exploitation, and exclusion. Geology created conditions for cotton. Cotton depended on slavery. Slavery created demographic patterns maintained by structural racism. And structural racism produces the poverty, stress, and lack of access that drive disparities today. You can't fix that with a smoking-cessation app. This is not an argument against epidemiology. It's an argument for a more self-aware, ethical, and equitable epidemiology.

**Sarah:** Which sets up the second major section of the lesson. Ways of knowing. If we have to sit with that discomfort, we have to take seriously the question of what counts as knowledge in the first place.

**Kiffer:** This part might be the most important piece of the whole lesson. Every other lesson in the course assumes you have a working answer to the question we're about to ask. When you read a study and decide whether to trust it, you're already operating from some implicit theory about what counts as knowledge.

**Sarah:** The lesson gives you three philosophical pillars. Epistemology, ontology, and axiology.

**Kiffer:** Epistemology asks how we know what we know. Ontology is one step deeper — what's there to know. Is there one objective reality, or multiple realities shaped by culture? Axiology is about values. Should research be value-free? When you choose to measure cholesterol but not housing quality, that's a value-laden choice.

**Sarah:** And the three pillars package together.

**Kiffer:** If you take a position on what's real, that commits you to a position on how you can know it, and that commits you to a stance on what role values should play. Which builds into paradigms — Thomas Kuhn's idea from 1962. Working researchers don't pick their pillars from a menu. They inherit a paradigm where all three are specified.

**Sarah:** The lesson walks through five.

**Kiffer:** Positivism — one objective reality, learned through direct observation, research should be value-free. Post-positivism is positivism with humility. There's an objective reality, but our understanding is always imperfect. Karl Popper's falsificationism — we can never fully prove a theory, only fail to disprove it. Most modern epidemiology operates from a post-positivist stance. Constructivism rejects a single objective reality — multiple realities exist, shaped by culture and context. Critical theory takes seriously that reality is shaped by political and historical forces, and the goal of inquiry is emancipation, not just description. And pragmatism says use whichever method best answers the question.

**Sarah:** And the lesson's key move is that the same public-health question can be studied from any of these paradigms, and each produces a legitimate but different kind of knowledge.

**Kiffer:** Take a question like, why do Indigenous communities have higher rates of diabetes? A positivist study would measure individual risk factors. A constructivist study would explore what diabetes means in the lived experience of community members. A critical-theory study would examine how colonization, displacement, and food insecurity produce metabolic disease. Each is doing legitimate research. They'd lead to very different interventions. And the dominance of post-positivism in epi isn't because it's the best paradigm. It's a historical fact about the society the discipline grew up in.

**Sarah:** Which is the bridge into an honest accounting of Western quantitative science.

**Kiffer:** Positivism came out of the Enlightenment — human reason and systematic observation replacing tradition and religious authority. In public health it produced genuine triumphs. Smallpox eradication. The decline of cholera. Identifying tobacco as a cause of lung cancer. Quantitative methods deliver four things really well — objectivity through standardization, reproducibility, generalizability, and statistical power. Each addresses a weakness of unaided human judgment.

**Sarah:** But quantitative methods also can't see certain things. The lesson gives four critiques.

**Kiffer:** First, reductionism. To measure something, you reduce it to measurable units. A national survey of food insecurity using an eighteen-item questionnaire produces reliable, comparable data across provinces. But it can't capture the experience of an Inuit family for whom food security is inseparable from access to traditional lands and from the intergenerational trauma of forced relocation. The number tells us food insecurity exists. Not what it means.

**Sarah:** Second, the view from nowhere.

**Kiffer:** The philosopher Thomas Nagel's phrase for the aspiration to a perspective free from positionality. Critics argue no such view exists. Donna Haraway in 1988 called the claim to objectivity the god trick — the pretense of seeing everything from nowhere. Her answer isn't relativism. It's accountability. Be transparent about where you stand. Sandra Harding's standpoint theory takes it further. People who occupy marginalized positions often have epistemological advantages because they navigate two worlds. In public-health terms, the people experiencing health inequities often understand the systems producing those inequities better than the researchers studying them from outside.

**Sarah:** Third critique. How quantitative methods can reproduce power structures even when researchers think they're being neutral.

**Kiffer:** Foucault again. The categories science uses — normal versus abnormal, healthy versus sick — are instruments of governance.

**Sarah:** Give us a concrete example.

**Kiffer:** Race as a variable. Epidemiologists routinely include race as a covariate, but race is a social construct, not a biological category. There's no consistent set of genes that defines what we call Black or White or Asian. Racism is real. The lived experience of being racialized is real. But the underlying category is something we made up to organize people. Control for race without theorizing why, and you naturalize health disparities and erase the mechanism. The cause is racism, not race.

**Sarah:** Second example, the deficit model in Indigenous health research.

**Kiffer:** Linda Tuhiwai Smith, in Decolonizing Methodologies, argues that much quantitative research on Indigenous health compares Indigenous populations to non-Indigenous benchmarks. The data may be accurate, but the framing produces a narrative of deficit. It treats Indigenous peoples as problems to be studied rather than communities with their own knowledge.

**Sarah:** And the third is the hierarchy of evidence.

**Kiffer:** The conventional hierarchy puts randomized controlled trials at the top and qualitative research very low. The hierarchy implicitly devalues knowledge that doesn't fit the positivist paradigm. You can't easily randomize people to discrimination. You can't blind a study of cultural belonging. So those questions sit lower not because they matter less, but because the methodology we've decided is best can't easily address them.

**Sarah:** Which is the bridge into the constructive answer. If quantitative science alone can't see context or standpoint, what fills the gap?

**Kiffer:** Qualitative traditions, Indigenous knowledge systems, lived experience. The four big qualitative traditions are phenomenology — what it's like to experience something. Grounded theory — building theory from data, especially valuable when dominant theories were never built with the population in mind. Ethnography — prolonged immersion in a community. And participatory action research, which dissolves the line between researcher and researched. Community members are co-investigators. The purpose isn't just knowledge but change.

**Sarah:** Then Indigenous knowledge systems. The lesson is careful that Indigenous knowledge isn't a single monolithic system.

**Kiffer:** Hundreds of distinct nations across Turtle Island alone. But many share four characteristics. Holistic — health connects physical, mental, emotional, and spiritual dimensions, not separated the way Western biomedicine separates them. Relational — knowledge is produced through relationships with land, Elders, other beings. Land-based — specific places teach specific things. And intergenerational — knowledge is transmitted through oral traditions, ceremonies, mentorship, story.

**Sarah:** Which raises the question of how Indigenous knowledge systems sit alongside Western science.

**Kiffer:** The most influential answer in Canadian public health comes from a Mi'kmaw teaching from Elder Albert Marshall. In Mi'kmaw it's Etuaptmumk. In English, Two-Eyed Seeing. Learning to see from one eye with the strengths of Indigenous knowledges, from the other eye with the strengths of Western knowledges, and using both eyes together for the benefit of all. Notice what that doesn't say. It doesn't say replace Western science. It doesn't say validate Indigenous knowledge by Western standards. It doesn't say blend the two systems into one. It says hold both side by side and respect the integrity of each.

**Sarah:** And beyond formal knowledge systems, lived experience is increasingly recognized as a legitimate form of evidence.

**Kiffer:** People who live with mental illness, who use drugs, who experience homelessness, who navigate systems as racialized minorities, possess knowledge that cannot be obtained any other way. Nothing about us without us. Lived experience is evidence in its own right.

**Sarah:** The methodological vocabulary for combining all of these is mixed methods.

**Kiffer:** Three designs in the Creswell and Plano Clark typology. Convergent — quantitative and qualitative collected simultaneously and merged. Explanatory sequential — quantitative first, then qualitative to explain the findings. And exploratory sequential — qualitative first, then quantitative to test. The research question determines the method.

**Sarah:** But mixed methods done badly is just extraction with extra steps. Which is why two ethical frameworks anchor the section. O C A P and community-based participatory research.

**Kiffer:** O C A P stands for Ownership, Control, Access, and Possession. The four First Nations principles for the governance of First Nations data, developed by what's now the First Nations Information Governance Centre. Ownership — a community collectively owns its cultural knowledge and data. Control — First Nations have the right to control all aspects of research that affect them. Access — they must have access to data about themselves. Possession — physical custody. Where the data actually live. Possession is subtle because once samples leave a community's physical control, ownership becomes a legal and political negotiation across decades. And community-based participatory research isn't a method, it's a set of principles. Community members are equitable partners in all aspects of the research process. O C A P says who the data belong to. Community-based research says how the research should be done.

**Sarah:** Which connects to the third major section of the lesson. Research integrity. What happens when the system designed to produce knowledge breaks down.

**Kiffer:** The lesson surveys five overlapping ways the published record can become unreliable. Outright fraud. Industry-funded manufactured doubt. Honest analytical errors. Replication failure. And the absence of community-led ethical infrastructure.

**Sarah:** Start with fraud.

**Kiffer:** The United States Office of Research Integrity defines research misconduct as fabrication, falsification, or plagiarism. Fabrication is making up data entirely. Falsification is manipulating real data — dropping observations, adjusting timelines. Plagiarism is taking someone else's work and presenting it as your own. Those are the bright-line cases. Rare and prosecutable. The harder problem is a much larger grey zone called questionable research practices. The big one is undisclosed analytic flexibility. In any dataset there are many defensible analytical choices. Covariates, missing data, subgroups, statistical tests. When you make those choices after seeing the data and only report the version that gave you a significant result, that's p-hacking. Andrew Gelman and Eric Loken call it the garden of forking paths. Simmons, Nelson, and Simonsohn showed in 2011 that even with a completely null effect, undisclosed flexibility produced significant results more than sixty percent of the time.

**Sarah:** And the case studies anchor the extremes. Wakefield is the famous one.

**Kiffer:** Andrew Wakefield, British gastroenterologist. In 1998 he and twelve coauthors published a paper in The Lancet — a case series of twelve children, no control group. At a press conference he claimed the combined Measles, Mumps, and Rubella vaccine was linked to autism. Vaccination rates dropped across the United Kingdom and beyond.

**Sarah:** Years later, journalist Brian Deer dug into the original medical records.

**Kiffer:** He found extensive falsification. Timelines between vaccination and symptom onset had been altered. And massive undisclosed conflicts — Wakefield was being paid by lawyers pursuing litigation against vaccine manufacturers, and had patented a competing single-measles vaccine. The Lancet retracted in 2010. Multiple very large studies — a Danish cohort of over six hundred and fifty thousand children, a meta-analysis pooling over one point two million — found no association. But the damage outlived the correction. Measles came back where coverage dropped below herd-immunity thresholds. Retraction does not undo a measles outbreak.

**Sarah:** Wakefield was caught by a journalist. Fujii was caught by statistics.

**Kiffer:** Yoshitaka Fujii, Japanese anesthesiologist. About two hundred papers, mostly small randomized trials. At least one hundred and eighty-three contained fabricated data — one of the largest fabrication cases in biomedical history. John Carlisle examined baseline characteristics across Fujii's trials. Real randomization produces groups that are similar but not identical. Fujii's were too similar. The probability of getting baseline distributions that consistent honestly was less than one in ten to the thirty-third power. Even at their worst, fraud cases are rare. So why is the literature still systematically distorted?

**Sarah:** That's where manufactured doubt comes in.

**Kiffer:** The key reference is a 2010 book called Merchants of Doubt by Naomi Oreskes and Erik Conway. They reconstructed a recurring playbook called the tobacco strategy. The heart is a 1969 internal memo from Brown and Williamson. Doubt is our product. The point was never to win the science — it was to prevent the science from looking settled long enough for regulation to be delayed. In tobacco, the science was clear by the early 1960s. Meaningful regulation took forty more years. Once tobacco built the infrastructure, it was rented to other industries — acid rain, ozone, secondhand smoke, climate change. The same individuals keep showing up across unrelated hazards. That topical promiscuity is the giveaway.

**Sarah:** Why is epidemiology specifically the surface attacked?

**Kiffer:** Because industries can't easily fabricate primary observational data. Cancer registries and large cohorts are run by independent investigators across many institutions. You can't buy that data and rewrite it. But you can exploit the genuine epistemic features of observational research — confounding, ecological fallacy, healthy-user bias. The clearest recent example is opioids. Through the 1990s and 2000s, Purdue Pharma promoted OxyContin for chronic non-cancer pain, claiming the addiction risk was negligible. Their evidence was a one-paragraph 1980 letter to the editor reporting an addiction rate of point zero three percent among hospitalized patients. That letter was used over and over as if it established the safety of opioids prescribed at home for years for chronic pain. Different patient population entirely. Strategic amplification of weak evidence.

**Sarah:** Then there's the middle ground. Papers that pass peer review, are written in good faith, and still mislead.

**Kiffer:** Surgisphere is the canonical example. Early in COVID-19, two high-profile papers came out from a small analytics company. One in The Lancet, reporting hydroxychloroquine was associated with increased mortality. The other in the New England Journal of Medicine. Both claimed a registry of more than ninety-six thousand patients across nearly seven hundred hospitals. When independent researchers tried to verify the data, the structure fell apart. Reported numbers from some countries exceeded official national case counts. Both papers were retracted within weeks.

**Sarah:** And then there's p-hacking, even when researchers act in good faith.

**Kiffer:** An analyst making each defensible choice silently manufactures findings. The researcher doesn't have to know they're doing it. Selective outcome reporting and spin show up in roughly forty percent of trials with non-significant primary outcomes. The toolkit for catching this after the fact is statistical forensics — things like the Granularity-Related Inconsistency of Means test, the G R I M test. If you report a mean for a sample of twenty on an integer scale, only certain values are mathematically possible. A reported mean of three point one seven from twenty observations is impossible.

**Sarah:** Then the replication crisis and the reform movement. The opening shot was a 2005 paper by John Ioannidis with the title, no joke, Why Most Published Research Findings Are False.

**Kiffer:** He argued from first principles that under realistic conditions the majority of published positive findings should be expected to be false. The Reproducibility Project in psychology, published in 2015, attempted to replicate one hundred experiments from top-tier journals. Only about thirty-six percent of replications produced significant findings in the same direction. And epidemiology had its own version. The textbook case is hormone replacement therapy. Through the 1980s and 1990s, observational cohorts suggested hormone replacement therapy substantially reduced cardiovascular risk in postmenopausal women. The Women's Health Initiative, a randomized trial of over sixteen thousand women, was halted early in 2002 when the data showed combined hormone therapy increased heart attacks, strokes, and breast cancer. Opposite of the cohort findings.

**Sarah:** How were the cohort studies wrong?

**Kiffer:** Healthy-user confounding. The women prescribed hormone therapy in that era were wealthier, better educated, more likely to exercise, more likely to use preventive services. Those features themselves protect against cardiovascular disease. The credit was going to the hormone therapy. None of this was fraud. It was the cumulative product of confounding, selective reporting, publication bias, and the structural pressure to publish positive findings.

**Sarah:** The reform movement is structured around specific failures.

**Kiffer:** Preregistration — publicly specifying hypotheses, design, and analysis plan before you collect data. The time-stamp lets the community distinguish confirmatory from exploratory analyses. Data and code sharing so other researchers can check for errors. And registered reports — where peer review happens before data collection. You submit your introduction, methods, and analysis plan. If accepted, the journal commits to publish regardless of whether results support the hypothesis. That removes the incentive to produce significant findings.

**Sarah:** And the lesson closes with two community frameworks. O C A P and E G A P.

**Kiffer:** The foundational North American teaching case is the Havasupai — a community of a few hundred people at the bottom of the Grand Canyon. In 1989 researchers from Arizona State University began collecting blood samples for diabetes research. Over the following decade, the samples were used, without further consent, for studies of schizophrenia, of consanguinity, and of population-genetic migration history. The migration findings supported the Bering Strait theory, which directly contradicts the Havasupai's own origin narratives. The community sued, settled in 2010, and the samples were returned for ceremonial reburial.

**Sarah:** And the Canadian case is the Nuu-chah-nulth blood samples.

**Kiffer:** Eight hundred and eighty samples collected in the early 1980s for arthritis genetics research. When the lead researcher moved institutions, the samples moved with him, and over two decades collaborators around the world used them for unrelated research. None of that had been disclosed to or approved by the Nuu-chah-nulth. That case became central to Chapter 9 of the Tri-Council Policy Statement on the Ethical Conduct of Research Involving Humans.

**Sarah:** And E G A P.

**Kiffer:** Evidence in Governance and Politics — a researcher-led network founded in 2009 around field experiments in political science. The centerpiece is the pre-analysis plan, a time-stamped commitment to every analytic decision before you see the data. Committing in advance converts undisclosed flexibility into either a planned confirmatory analysis or an honestly labeled exploratory one. The anchoring case is LaCour and Green. In December 2014, Science published a paper claiming brief in-person canvassing conversations with gay canvassers durably changed voters' attitudes toward same-sex marriage. Within months it fell apart. Two graduate students tried to design a follow-up and found the survey firm had never run the protocol described. Science retracted in 2015. The decisive feature is that LaCour partially complied with transparency norms, and that partial transparency gave the community a trail.

**Sarah:** And the synthesis.

**Kiffer:** Misconduct, manufactured doubt, analytical errors, and replication failure all look at first glance like individual moral failings. The structural reading is that they're features of an incentive system. Produced by how science is funded, evaluated, and rewarded. Which is why the response can't be calls for individual virtue. Reform has to be structural. Preregistration changes what researchers commit to in advance. Data sharing changes what the community can inspect. Registered reports change how journals decide what to publish. O C A P changes who has authority over data. E G A P changes what counts as transparent design.

**Sarah:** Let me try to pull the whole lesson together. First, epidemiology emerged through the convergence of science, technology, and governance — a discipline that grew up alongside the modern state.

**Kiffer:** Second, progress has been driven by networks of actors and institutions, not by individual geniuses.

**Sarah:** Third, global health has improved dramatically. Child mortality from forty-three percent to under four. Life expectancy more than doubled. Smallpox eradicated. Recognizing this isn't complacency — it's the evidence base for continued investment.

**Kiffer:** Fourth, the same institutions that produced health improvements also produced exploitation. Tuskegee, the Indigenous nutrition experiments, the Black Belt chain from geology to heart disease — these histories shape health, trust, and equity decades and centuries later.

**Sarah:** Fifth, what counts as knowledge is itself a methodological choice. Three pillars — epistemology, ontology, axiology — build into paradigms. Epi is mostly post-positivist, but that's historical, not necessary.

**Kiffer:** Sixth, qualitative traditions, Indigenous knowledge systems, and lived experience are valid ways of knowing in their own right. Two-Eyed Seeing, O C A P, and community-based participatory research make integration ethical instead of extractive.

**Sarah:** And seventh, the published record gets unreliable in five ways. Fraud. Manufactured doubt. Analytical error. Replication failure. The absence of community-led ethical infrastructure. The reforms are concrete and structural — preregistration, data sharing, registered reports, O C A P, E G A P. None is a panacea. Each addresses a specific failure.

**Kiffer:** A critical epidemiology asks not just what causes disease, but who benefits from the knowledge, who is harmed by the research process, and whose priorities shape the agenda. The two stories are not rivals. They're both true. The lesson asks you to hold both.

**Sarah:** Take care of yourselves with this material. If any of it sits heavily, that's an appropriate response. It means you're reading carefully.

**Kiffer:** Life expectancy has roughly doubled. Smallpox is gone. Cholera is treatable. The shadow story is also real, and the same period produced both. An honest reading requires holding both.

**Sarah:** Next lesson we move from history and philosophy to method. How epidemiologists actually do the work, and how the systematic-review tradition tries to keep the literature honest. That's Lesson 2.

**Kiffer:** Before class, work through the interactive module and bring the questions that didn't resolve to the in-class session.

**Sarah:** Thanks for listening. We'll see you in Lesson 2.

**Kiffer:** Take care of yourselves. See you in class.