Okay. Hi there. My name is Dr. Fernando Heredia. I'm from Chile. I'm a gynecologic oncologist and work at Universidad de Concepcion. And today we are going to speak a bit about epidemiological statistics. So this is like epidemiology 101 because it's quite a basic, but first of all, epidemiology is a basic science that takes care of health happenings. And we will speak today about some fundamentals of this science. We'll speak about variables, the science of investigation studies, validity or value, when we speak about internal and external value of our studies, and of course bias. So this is our agenda. This is what we're going to talk about today. So as a science, epidemiology has evolved since 1970 until now. At first it was only concentrated on observations of a certain health condition or diseases and their consequences. But nowadays it also dictates primary and secondary prevention policies based on observations and studies. So all of these definitions are based on the health disease process. So nowadays epidemiology is defined as the science that studies the distribution of the health disease phenomena. And it studies the determinants that influence or impact specific populations in order to provide solutions for relevant health problems. So its applications or its objectives is to establish the magnitude, distribution, and determinants of health disease, understand clinical cases, identify new diseases, and elaborate effective interventions, like what we just passed with COVID-19, contribute to the sanitary planification of a community, and develop clinical investigation, and of course, what we're doing today, teaching. Okay, so this is like the core of epidemiology. Other important item in basic epidemiology are variables. And a variable is a characteristic that can fluctuate, and its fluctuation is capable of adopting different values, which are observable and or measurable. Briefly, variables can be classified as statistical variables. And in between those, there are qualitative and quantitative variables. In function of the investigation that we are performing, independent or independent variable, intermediate variables, and confusion variables, and epidemiological variables, which can be classified in time variables, place variable, or individual variables. Okay, so qualitative variables are those who assign qualities in a study population, for example, gender, which can be feminine, masculine, or others, working status, dependent, active, retired, or others, and so on. You can also assign numerical codes to these qualitative variables to facilitate the data collection and further analysis, like in these cases, like gender, masculine, feminine, or no data can be assigned to numericals like one, two, three, the same with working status, socioeconomic characteristics, anything that could be considered a qualitative variable. Quantitative variables are expressed exclusively in a numerical manner, like in amounts, okay, like body weight can be expressed as kilograms or ounces or pounds, body temperature in degrees in Fahrenheit or Celsius, etc. And according to to their role in the investigation, variables can be classified in independent variable, which is the exposed variable. It also could be the protective factor or the risk factor that we are studying, which impacts directly on the dependent variable, also called the result variable. That is what we are observing. Intermediate variable is presented as a consequence of the independent of the independent variable and affects directly the dependent variable. It's not really an independent variable, but it's like, it has sort of similar, similar. How could I say that it's, it's something that's not the independent variable, but it also influences on the dependent Okay. So for example, in the case of cervical cancer, the independent variables could be tumor size, histologic type, nodal metastasis or tumoral grades, and the dependent variable would be relapse. But when we speak about intermediate variables, for example, in a study of myocardic infarction, in which high caloric intake is the independent variable, hypercholesterolemia works as an intermediate variable, since it is very much related with the independent variable, and it also impacts directly on the dependent variable. Hope I've made myself clear. And there is also, to confuse much more, the confusion variable. So the confusion variable is not necessarily related with the epidemiological chain of the dependent variable. Like for instance, in colorectal cancer, if we study red meat consumption as the independent variable, a high socioeconomic status could be considered as a confusion variable. Now, when we see epidemiological variables, we must make ourselves three questions. First, when does this happen in time? Okay, if it's in a certain century or decade, or if it's cyclic, that means over a year, or stational, like vaccinations for the influenza virus, etc. Second, where does it happen to locate in a region, in a country, or in a state, or in a hospital, or in a healthcare center? And the who question, that is to characterize, or to individualize the persons affected by gender, by age, to know which is the person affected by the health problem. We can classify this epidemiological, according to their nature, in quality, in quantitative or qualitative variables. The qualitative variables can be measured nominally, like to classify them to classify them in A, B, C, or ordinarily, to give them rank. Quantitative variables can be discrete, to be able to count them, like one, two, three, or can be continuous, to measure them, and those are expressed in decimals. Like, for instance, there's also an interval scale that has to be described, like, for example, in a thermometer, the value zero does not mean the absence of a characteristic or value, and there is also a ratio scale, in which the value zero is the absence of the characteristic. And sometimes the important issue is the comparison between two different measurements. So, like in weight gain or weight loss. There is also an important item in epidemiological variables that is to define the operationalization of variables. This is a series of procedures or indications to perform measurement of a variable defined conceptually in the study, and it is closely related to the kind of technique or methodology used to collect data and review our literature. In operationalization of variables, we can use various items, like the operative definition of the variable, the nature of the variable, the operative level of the variable. In this case, for instance, the variable metastasis has to be defined as the presence of metastasis confirmed by images or pathology. It has a qualitative nature, and the operative level is defined as present, absent, or not documented. In this other case down there, it's adjuvant radiation that is defined as a total cumulative radiation dose, its nature is quantitative, and its operative level is a unit of measurement of radiation that is the gray or centigrade. Now, when we have to design a study, the investigator must decide in the first place whether he's going to observe or to intervene in the health phenomenon. If there is an intervention in the right side, we will call these the interventional studies, and we call this, and there are quasi-experimental studies and experimental studies. If the individuals studied are not healthy, we call this design a clinical trial, and these clinical trials are like in the tip of the pyramid of evidence quality. That's because they are so hard to do, because these are very, very difficult studies to do. In the other hand, the easiest ones, the ones on the left, are the non-interventional studies, also called observational studies. If there is not a control group, we will call them case studies or case series, and they can be longitudinal or transversal according to how many observations were done in a period of time. If there is a control group, we call them analytic, and amongst them there are the cohort studies, in which we observe the effect of the treatment, and the case control studies, in which we study in the affected or treated individuals if there was any variable which could have determined the occurrence or affectation of the individual. It's also very important down there to describe the temporality of the study, if it is retrospective or prospective or hybrid, but again, it is important to locate the observation in time. I think this is the most important slide of the whole presentation. This dictates how your experiment is going to be. Another important issue of epidemiological studies is their validity or value. Validity starts by generating the protocol in a structured way. Data collection must be based on what we're looking for, and data analysis has to be directed by how the protocol was generated. And finally, the communication of our results is relevant. This will result in the internal value, which means that the results are true for the study group in contrast to the external value, which means that our studies, that our results are extrapolable to a population outside our study group. So, these factors can, the errors, I mean, can be, can negatively affect our internal value, and these are called statistic errors. These errors can be related to the study population in selection of patients during the study design, in the measurements during data collection and analysis, and in the comparability when the control groups are not statistically balanced. So, all of these impact negatively on the internal value. And these errors usually translate in bias, bias in the study design, bias in the data collection, or bias in data analysis. And bias can be defined as systematic or random bias. Random is when repeated measurements, either in the same subject or in different subjects of the same study population, differ in an unpredictable fashion. And systematic error, which is not really random, occurs when measurements vary in a predictable manner. And so, one would tend to over or underestimate the real value of all the repeated measurements. So, there are two concepts that will help us understand between these two kinds of bias, precision and exactitude. These target shooting schemes help us understand. The correct measurement that is both here, precise and exact, is near the center of the bullseye. So, if we imagine that the center of the bullseye is the truth, an excellent study would be precise and exact. Our measurement is in red, and the other measurements are in black. In a randomized bias, we can be quite exact, but we'll be not precise. So, we will get near the truth, but we'll be not precise. In systematic bias, we are very precise, because all measurements are around the same measurement, but we are not precise. So, as our idea is to be precise and exact, any potential bias must be assessed in every study to assure internal and potentially external value. And if we translate this to graphics, we can see how measurements compare with the parameter. If this line here is the truth, to be precise and exact would be this. This would be similar to random bias, and this would be similar to what systematic bias means, okay? These are the different kinds of bias. There are selection bias, and these are systematic errors which occur during the selection or follow-up of the study population, and lead to a wrong conclusion about the study hypothesis. Selection bias could originate in the investigator himself, or be the result of a complex relation in the study populations which were overseen or missed by the investigator. Information bias, on the other hand, are errors introduced during the measurement of the exposure of the events or other co-variables in the study population, and they occur in different weights within the compared groups, and lead to a wrong conclusion about the study hypothesis. I think this is the last slide, so these are our references, and thanks again for your attention.

epidemiological statistics

health happenings

distribution of health disease phenomena

variables in epidemiology

measurement scales for variables