Hello! This is module 2, Clinical Trial and Insurance Claims Data. This is lesson 2.1.1 Introduction to Clinical Trials. In this lesson we'll discuss what clinical trials are and how they can be useful in collecting data for a technology assessment. First, let's discuss developing a research question. Assume we want to know more about the cost effectiveness of a once per year rheumatoid arthritis treatment, there's currently multiple time per year treatments that exist, either once per week or once per month. So what we need to do in order to assess the cost effectiveness of this new, once per year treatment, is define the research question. That helps us focus and properly define what it really is we want to know. A good way to do this is to follow the PICO protocol. In other words we need to identify the population, intervention, comparators, and outcomes. So for this example, the population may be adult individuals, male and female, with rheumatoid arthritis, who have already tried some form of therapy. The intervention we are trying to test is the once per year treatment, once per year, versus, let's say, the once per month treatment. So that would be the comparator. The outcomes to measure, it would be some measures of either side effects, improvements in mobility or function, Or possibly, some other type of RA outcome that is objectively measurable. Once we know this type of information, we can proceed to design the clinical trial appropriately. Let's talk about the types of clinical studies. First, there are non experimental, or observational studies. In these studies, we observe both exposures and outcomes. Examples of these are case studies or prospective observational studies where we know that there's two populations receiving an intervention, and we use another part of the population that's not receiving the intervention as a comparator. To do these types of studies, we first identify the populations. That's the one receiving the intervention that we're interested in. And then the one that's receiving what we define as the comparator. Then we observe and record the characteristics, and look for associations. So if we follow our example before, we could identify populations that are receiving the new treatment, the once per year drug, and we can observe how their outcomes improve or not improve over the course of the next year. And we could identify the population that's taking the drug once per month and compare the two. Then see if the population has taken a once per year drug has improved over the course of the year more than the population that has been taking the other drug. Alternatively, we could also try an experimental design. In these types of designs we assign an exposure and observe the outcomes. The difference here is that we assign the exposure, not just observe the exposure that's been already applied or used by the population. So the best example of this type of clinical study is a randomized controlled trial, or RCT. Before performing a trial or conducting a trial, we still need to identify participants. Then rather than just observing them, we place them into a common context. We apply the exposure or intervention, and we observe and evaluate the facts of intervention. So in this case, we randomly assign participants to either receive the once-per-year treatment or the once-per-month treatment, and then we follow them for the year or follow up to observe the outcomes. And we try to maintain a common context so that their amount of follow up to their doctor and the amount of measurement of outcomes is similar between the two groups. The benefit of clinical trials in particular, RCTs or randomized control trials, is that they provide the ability to reduce bias and variability in order to identify the effects of treatment or intervention. Note that bias, which we'll discuss in more detail in a later lesson, affects accuracy. Accuracy is how close to a measured value is the true value. Variability, on the other hand, affects precision. Precision is how close to a measured value or how close measured values are to each other. So, if we think of a target with a bulls eye, Bias would be how close the values are to the intervention. In this case, these observations on the outside would be biased, and these observations would be unbiased, because they're close to the true value. Alternatively, you think of precision with the bullseye, it could go either way. So we could have a lot of precision that's still biased, or we could have a lot of precision that's not biased. So, both of these Are precise. But this one's bias. The same can be true when we observe data and not just random shots at a dart board.