Hi, I'm Justin Lessler. In this module, I'll be talking about Weighing Evidence and Identifying Causes in Epidemiological Studies. This module will have four major objectives. First, we'll help you learn how to turn general questions about diseases and health into precise hypothesis. Second, it'll help you identify appropriate comparison groups for answering this hypotheses Third, it'll teach you how to describe and calculate basic measures of association including the relative risk and the risk difference. Finally, we will go over basic measures of uncertainty in significance to help you decide if your results are statistically powerful. So, here's some resources that will be helpful in understanding the material in this lecture or going deeper. Epidemiology (Fifth edition) by Leon Gordis, is a classic epidemiological textbook because of most of the concepts discussed here. CDC Self-Study Course number SS1987, particularly lesson three goes over many of the concepts that we will discuss in this module. For those of you interested in learning more about John Snow, who uses a running example throughout this module. There's an excellent piece on his life and work called John Snow, The Father of Epidemiology. Then for those who want to understand more about P-values and Confidence Intervals and Tests of Significance in general. There's a blog on the Public Library of Science website called, Is the p-value pointless. A good piece in understanding Confidence Intervals In the BMJ. On the lecture page, I've provided links to these resources where appropriate. In this first section, we'll talk about how to turn general questions into precise hypotheses. Throughout this module, we'll be following the path of John Snow. John Snow is a famed physician of his day and he was anesthesiologists to Queen Victoria, when she delivered her children. He did these investigations of cholera in London that are recognized to be among the first epidemiological studies and we'll use his work as a running example throughout this module. So, think about if you're John Snow and the types of questions you might ask. You might ask a very general question such as, what causes cholera? So, this is a general question of importance of scientific and public health importance. But it's not specific enough to suggest a particular study or experiment. It doesn't propose a particular mechanism or cause that we can test. So, it might be appropriate for framing your life's work, but not for designing or running a specific study. We can narrow this a bit and make a broad hypothesis. John Snow's hypothesis was that color is transmitted by contaminated water. It's more specific than the other hypothesis, and it's in principle testable. It's specific to a mechanism and cause given the knowledge had at the time, they didn't know about bacteria and infectious diseases, but it's not tied to a particular set of data experiment or setting. So, there's no clear observation that will support or refute this hypothesis. So, this can help to guide investigation, guide the types of studies we might design, but it doesn't help us to support or refute the hypothesis. We can make this hypothesis a bit more specific and say people who drink water with sewage in it are more likely to get cholera than those who drink clean water. This is a clearly testable result of the broad hypothesis about cause that we made before. It can be shown to be false. People could drink water with sewage and be not more likely to get cholera. But it's still not tied to a particular study or observation. So, it can serve as the basis for design of experiments and epidemiological studies, but we have to get even more precise if we want to have a highly specific hypothesis that can be tested. So, finally we can make things more precise and make a hypothesis that is tied to a specific study or observation. For instance, at the time people were starting to look at the water of the Thames in London with microscopes. They realized there were tonnes of creatures in it. Downstream of London, where sewage from the city was in the water compared to upstream of London, where there are less sewage. On the right here we have a picture from the time where people had drawn all of the microscopic animals that they were finding in the water of the Thames, the monster soup in the Thames as they called it. Thinking about the water below London where there's more sewage and the water above London where there's less sewage, and knowing that people in London could get water from above the city or upstream of the city or downstream of the city. John Snow was able to make a more specific highly testable hypothesis. That is people who get their water from the less contaminated part of the Thames upstream of London are less likely to die from cholera than those who get their water from the more contaminated part of the Thames downstream of London. So, here the hypothesis is tied to a specific location study or place. It can be tested by specific observation. So, this can be the basis of a specific study or analysis that supports our broader hypotheses and scientific questions. For now, let's think a little more generally about what makes a good hypothesis. We want our hypothesis to be relevant. So, the answering the hypothesis sheds lights on a larger scientific or epidemiological question. It should be specific and that it postulates and exposure for instance, here water from upstream or downstream of the city, an outcome getting cholera and a relationship. Those who get water from upstream are less likely to get cholera than those who get water from downstream. It should be falsifiable. So, there should be some observation that could make us conclude that the hypothesis is false. So, for instance here if people who get their water from upstream and downstream have the same rates of cholera, then we might conclude the hypothesis is false, and it should be precise if we want it to test it in a study. That is it should elucidate exactly the relationship that will be tested and it's expected results. The problem with bad hypotheses is they can lead to beliefs that can't be changed or interrogated with scientific evidence. So, consider one of the more common hypotheses out there, that's not a very good one vaccines cause autism. This isn't specific to an exposure or mechanism. So, if I say it's not true for one exposure or mechanism, you could always say, "Well, maybe it's true for that other one." It doesn't define a particular outcome that can be falsified. You haven't told me what I need to do to prove you wrong, and no hypothesis is truly scientific, if I can't prove it wrong. So, this cannot be proved or disproved with any particular evidence. We can make a more specific hypothesis that could be. So, we could say, children who have received the MMR vaccine are significantly more likely to have autism than those who do not. This has a clearly defined exposure outcome in relationship, and it can clearly be falsified by a given study or observation, and it has been proven to be false, many times including this one example studied by Jain and colleagues. So, some key points for this section. A good hypothesis is the key to collecting and weighing evidence. Broad scientific questions in general hypotheses can help guide research, but are of limited value when evaluating evidence. Evidence should be collected and evaluated based on a specific hypothesis that is falsifiable by experiment or observation. They should have a specified measurable exposure or comparison groups and have specific measurable outcome. So, for an exercise consider that you've had a family dinner for some holiday. Many of your family members developed diarrhea after the family dinner. Make a specific testable hypothesis, we would help you determine the likely cause of diarrhea after the family dinner.
