 Good morning everyone. We will be continuing the classes on Epidemological Study Designs. We have so far completed the designs and descriptive epidemiology in detail. So today I will be explaining about analytical epidemiology. So we have covered the hypothesis. So in descriptive epidemiology we have made a hypothesis. So to test the hypothesis or to know whether it is true or false or to reject or accept hypothesis we conduct analytical epidemiology. So let's go into detail about the analytical study design. So in analytical epidemiology we will go to the steps or the types before the steps. We have two types of analytical epidemiological studies. As I mentioned the study remains a case control study and covert study and when these studies were used are also depending upon the nature of the study. So this is basically a second major type. The first one was descriptive and the second major type of epidemiology which actually focus on individual within the population but whereas descriptive epidemiology we are focusing on a group of people or a population. So population, descriptives we were collecting. Now we are going to check the individual perspective and the most important thing is testing the hypothesis rather than formulating it. So in descriptive study we were checking the hypothesis. Now we were formulating the hypothesis but in analytical study actually we are testing the hypothesis and rejecting or accepting it. So the first one is case control study. It is also known as retrospective study because it is going backwards because the cases and controls are already present at the beginning of the study. So the collection of data by asking questions or questionnaires or their previous data their hospital records. So it is a retrospective study but it is also known as straw-hawk study. You can see it as the reverse of cohort COHORT. So it is going backward. So cohort is always a prospective study. So opposite of cohort is straw-hawk. So case control studies also known as straw-hawk study. So question might come it as straw-hawk study. So never get confused with cohort study. Straw-hawk means case control study. So the basic thing in case control studies both exposure and outcome have already occurred before the start of disease. So before that you need to understand what is exposure, what is outcome, what is effect and what is cause. So the exposure and cause are same outcome and effect are same. Suppose we will take another example of few people or a group of people who are eating food from street side and who are eating from home and the chances of food poisoning. So our exposure is eating food from the street side because it has more chance to cause food poisoning. So the exposure or the cause is always eating street food and outcome or the effect is food poisoning. So always an exposure will have an outcome and it is the same as cause and effect. Effect and outcome are the disease. The exposure and cause are the reason for it. So exposure is the particular reason for creating that outcome or a particular disease. So here eating street food is the cause. So in analytical study the fundamental thing is we always keep a comparison group unlike descriptive study. In descriptive study we don't have any comparison group because we are just describing it. We are formulating a hypothesis. So here we have a comparison group that is control study. So the basic thing in case control study both the exposure and outcome already occurred before the start of the disease. If you are going to check the food poison of patients in a city or in a town in a village the exposure and outcome. People who have already consumed street food are there and many of them might have already developed the disease. So the exposure and outcome have already occurred and the study is always going from disease to cause because we are trying to find out whether actually eating street food caused food poison. So that is food poison to its causes. We don't know whether it is just a hypothesis eating food from street side caused food poison. So we want to know whether it is actually true or false. So we will be asking questions to the patients that is people who have food poison whether they had consumed more food from street side or they eaten from home and we will compare it or do analysis and we will find out whether the hypothesis is true or false. So it is going from the effect to cause but whereas in cohort it will be opposite. That will be dealing later. So this is the basic design of a case control study. So there will be cases and controls are already present. Cases means people with the disease control means people without the disease. How we select control will be coming in next slides. So to cases and controls the epidemiologist will ask questions and find out how many of them were exposed to our particular cause that is street food. Case is food poison. So people with food poison will be checking collecting data using questionnaire or any method and find out how many were exposed to street food and how many were not exposed to street food. At the same time controls without the disease also will be collecting data that how many were exposed to the street food or how many are eaten street food and home food. So always we need to have a control group. So our assumption is people who have eaten street food will have more chance to get the disease that is food poison than the people who have eaten less from the street side that is we are keeping the controls for comparison. So cases and controls without the disease to the both groups will be asking questions whether they have exposure or they are not exposed. Exposure is tobacco chewing whether they chew tobacco or not chew tobacco. So our assumption is in cases the exposure factor will be more compared to controls. So that is our assumption that is what we are trying to prove. In the cases the exposure will be more compared to control. So that is why we are keeping a control group because in control group without people without the disease will not be having much exposure compared to the cases. People who have the cause will be having more effect people. There is no effect means cause will be less or less cause less effect or less exposure less outcome. So the basic steps in a case control study. The first step is selection of cases and control then we have to match it then we have to measure the exposure finally analysis and interpretation. So how to select cases and how to select controls. Selection based on the diagnostic or eligibility criteria. We have to keep a diagnostic criteria and we have to keep eligibility criteria and we can take from hospital or general population. Whereas the selection of control control should be from hospital if the cases are from hospital and we can take controls the relatives of the cases are from the general population and the number of controls that is usually the optimal range is one is to one. The ideal ideal is one is one not optimal ideal is one is one. If you are taking one case you should have minimum one control and if there is more control the study will be good and you can have as many as four maximum of four. Beyond four it is not much effective. So at least you need to have one control for every case and while selecting this control you need to follow a step called matching that is every control should be matched based on the age and gender. Suppose if you are taking a patient with cancer that is a male patient with 40 age you need to select a control with the same age and same gender but without the disease. So that is what is known as matching to match age and gender. So control will be matched based on age and gender. So the third step is measurement of exposure. So how do you measure the exposure? Because you have already produced cases and controls. Now you need to ask about their exposure history. So in this case that is if you see the case of oral cancer the exposure is smoking or tobacco chewing. So we have cases and control groups so we will be asking the same questions to cases and controls. The exposure factor how many times you chewed how duration you were using this what were the symptoms. Same questions will be asked to cases and controls. So you can use interviews, questions as past records such as hospital records, employment records can take clinical or lab examinations and you need to take the records. That is why it is known as a retrospective study because we are going backward. So we are asking the past history of cases and controls. So most commonly this is the easiest analytical study because only you need to check the past history past medical history and past exposure history of a person. But one thing is you need to ask the same questions to cases and controls. But the main problem is the investigator should not be knowing the cases and control groups identity. If he knows the cases are cases the people are cases what happens is he might have had a chance to ask more questions and report more actively than the control group. Because control group will not have so much answers because only cases will be having more detailed report of their past history because they are having the disease. Controls is not having the disease. So when you ask the same questions how long you were using the tobacco means they will not say I have not used. So the same questions if we ask the investigator shouldn't be knowing the participants are belong to which group. This process is known as blinding. And the case control study it is very difficult to do the blinding and there will be bias. So blinding and bias will be checking in future classes. So the last piece analysis. So how do you analyze the exposure in groups? So what we are checking here is exposure. So how much exposure did the cases and controls have? So the measure of exposure is done by a method known as odds ratio in case control study. So suppose the same example of food poison the contaminated food or the street food. That is exposure of cause of disease and outcome or effect is food poison. This is our hypothesis. Contaminated street food causes food poison. So we select cases that is food poison patients and we select relatives, neighbors, friends of these people. We have more number of control that is 35 cases and 82. We can have 1 is to 4 ratio. That is 35 to 120 up to 140 controls we can have. But this is a scenario. We have 82 patients. So out of 35 cases 33 patients were exposed. That is 33 patients were taken food from street. In control 82 controls 52 people. 55 people were taken food from street. But you have to see that the number is more. You have to check the exposure rate. In cases just 2 people are non exposed. That is not taken food from street. But in control 27 people are taken. So if you see the ratio in cases the exposure rate is 94 percentage and control rate is 67 percentage. Because the exposure rate in among cases that is vertically you have to see exposed people 33 divided by the cases that is A by A plus C 33 by 35 this is 55 by 82. So you will get 94.2 and 67. So the exposure rate among cases is 94 and exposure rate among control is 67. That is what we are trying to prove. The exposure rate would be very high among cases compared to control because people who are exposed will be having high chances of an outcome. Because people who ate food from street will have more chances of having poison compared to people who have not eaten from street. So don't get confused after seeing this number because here it is almost triple amount of control that we are taking. So only thing we need to check the exposure rate. So in control the exposure rate is 55. So horizontal line is exposure rate. This is exposure this is non exposed. This is cases and this is control. So don't get confused and always see X axis we write cases and controls Phi axis we write exposure and non exposure. So this is 55 by B plus T that is 55 by 82 we get 67. So exposure rate is always high in cases compared to controls. So odds ratio it is very simple. We are trying to find out the odds of exposure among cases and controls. Odds means chances. So what are the chances of exposure among cases and odds of exposure among controls. The formula is very simple. You have to do the cross product. That is A by C divided by B by D. So it will become mathematically A D by B C that is A D by B C. You have to take the cross product. So A D by B C is the odds ratio. It gives the odds of exposure to the cases and controls we will be asking about their previous exposure odds. So this is the strength of association between the risk factor and outcome. Risk factor is eating street food than the home food. Outcome is food poison. So it gives a measure of strength of association. So when we multiply it 33 into 27 by 55 into 2 you get 8.1. So that is the measure of strength of association of eating street food and the food poison. That is exposure and outcome association. So in simple words we can say that street food eaters had a risk of developing food poison 8.1 times than that of restaurant or home eaters. So people who had eaten food from street side will have 8 times chance of getting the disease compared to the non-exposure group. Here I put restaurant eaters. Don't get confused restaurant anywhere or home side because we have only one exposure and one outcome in any study. So our exposure factor will be based on the hypothesis. So our hypothesis was like for this case eating street food could be the cause of having food poison. So street food eating will have an 8.1 times of risk in developing food poison compared to the non-exposed group. So here also I put some different this is some hotel name Sagar and Malabar. So don't get confused with the names only you worry about exposed and non-exposed. So that is all about case control study. So in cohort study the thing is everything changes because cohort study is a forward looking study or a prospective study. So you remember the word Strohok. Cohort is forward looking study. So the backward looking study is case control or it is also known as Strohok study. It is prospective study or longitudinal study with follow up we had seen in the designs or we can say incidence and incidence prevalence will be coming later. So cohort what is a cohort? Cohort is nothing but a group of people who have a common characteristics. So the thing is it is proceeds from cost to effect. If the cost comes first if it is in case of our foot poisoning we will be observing people who are eating from street side and we will wait for the effect. Because at the beginning of the study all the people are without the disease that is the most striking part of cohort study. So the central study both the exposure and outcome have already occurred. But in cohort study not the exposure not the effect has already occurred. So we will wait for the exposure and the effect to happen in future time. So that is why it is known as prospective longitudinal or forward looking study. So you can see that study starts present. Today it starts. So in case of retrospective or case control study we will be asking their past history. What happened in previous time. But this is forward looking study so if study starts today now the population is free of disease. Today nobody is having disease. So what we do is we observe group of people. So what happens is we observe group of people. That is we observe a group of college students and the factor that is causal factor they develop in future time. So many of the students will be eating from street side and many are not eating from street side they are eating mess food. So that group a common group we will take some may be 100 students we will take. So 100 many will be going street side for eating food and many will not be going street side for eating food and we will keep on observing them. So at the beginning of study they are not even having factors that is exposure factor. Over the period they will be becoming two groups that is one is exposed group and this is non exposed group. The same scenario in case control study what is happening reverse because in case control study it started from disease and no disease exposed and non exposed. In control study it is going opposite that is study without disease few will develop the exposure group will go into exposure and few will go into non exposure group. And people who eating food from street develop food poison few may not develop food poison. People who eating from mess that is without exposure factor. Food poison few might not develop food poison but our exposure or hypothesis is to test the cause and effect that is eating street food will cause food poison. So in the last year there was a history of food poison during the month of April May. So this year we are planning to conduct on a group of people or a hostile group. So that will become a perspective or a cohort study. Last month there was a history of food poison for hostile people or a group of students. So we were conducting a study on the people and we were like asking questions that becomes case control study. So same group we can do different approach but in cohort study study will go to the future. It is a forward looking or prospective study. At the beginning of the study there will not be any cases. So it is time consuming and it is very difficult to follow up because we need to follow up the students, follow up the participants. Okay. So this is how it is design case control and cohort study. So there are three types of cohort study. One is prospective cohort study, retrospective prospective and combination. So I can give an example. So this is prospective study. So today we are starting the study. So there is a group of people without any disease. Few are exposed, few are in non-exposed group and some will develop disease, some will develop disease. Same with the non-exposure group. Outcome has not yet occurred at the investigation begins. So this is a very clear cohort study. Very particular or very proper cohort study. So retrospective so we can see that if we started in 2008 don't confuse with the study. If we starting the study in 2008 the population without any disease over a 10 period of time. All were group of people without any smoking habit over a 10 period of time they develop the habit that is exposure group. That is few are started smoking habit and few are not developing smoking habit. Okay the same group of people we are following for 20 years. So this is a prospective cohort study. We are also known as concurrent cohort study. So in 2028 the people with exposure that is people with smoking habit few develop disease and some did not develop disease that is lung cancer. And also a non-exposed group few develop disease and few did not develop disease. So you might have that doubt that how come the non-exposed group the lung cancer would develop. That is what our aim is. You have to find out an association between this exposure this smoking habit and lung cancer. So obviously we are postulating we are hypothesizing that the exposure group will have more number of cases and non-exposure group will have very less number of cases. And obviously it is a multi factor we don't know which factor causing lung cancer. Our assumption is that smoking causes lung cancer. So people with smoking habit will have high chance of lung cancer. That is what we are trying to prove in hypothesis. Because there are 100 people with smoking habit they all will not develop lung cancer. And 100 people without smoking habit some of them might develop lung cancer. But when we compare the smoking and non-smoking the lung cancer development will be very high in smoking group. That is exposure will have a big impact in creating the decision. That is what our hypothesis was. So we are trying to prove the hypothesis. I hope that concept is clear. So this is a different thing that is retrospective cohort study. So what happens in the retrospective cohort study? Don't get confused with retrospective study. That is case control study. Here the study was started in 1998. The study we are starting in 1998. And it ended in 2008. So here we started the study in 2008. And it ended in 2028. We are starting in 2008. Don't think of the present 2020. You are starting in 2008. It goes in 2018 and it lasted in 2028. But whereas in this scenario the study started in 1998 and it lasted only 10 years. And in 2008 it stopped. What happens was, when we start the study we have already two groups of exposure and non-exposure. That is not present in the prospective cohort study. But the similarity is in 1998 the group which developed in exposure and non-exposure group the same group will be traced back till 1988. Because the same group of people will be assessed the same group of data will be assessed from their records. From their hospital records school records or habit records. So why we are doing a retrospective cohort study? Suppose we are doing a prospective study we need to it might take 20 years to complete. Starting in 2008 would take 20 years. But in retrospective cohort study, when we starting the study we are already completed at 10 years gap because people already are in exposure and non-exposure group. So we take the exposure and non-exposure group and we will ask the data or we will collect the data and we will follow the same cohort backward. So we will collect the data and we will make sure that all belongs to a single population. Suppose we are taking college students now they are 25 years or they are working. We will ask their college history and make sure that all belong to 1988 category. Or we will go back 10 years back and we will collect data and make sure that all belong to same age group or same college. Okay, it should have a same cohort design. Because cohort means they should have a common characteristics. If you are taking a college student group means the same group should be forward back 10 years back and we should collect data. So the advantage is we are advanced 10 years i.e. 10 year gap we have reduced. So this study will take only 10 years to complete. So we start at the study in 1998 and we will take 10 years to develop the disease. In 2008 they developed the disease and we will check the exposure. It is a little bit complicated when comparing but in retrospective study we are collecting the population data from the records. From their college records or from the student records. But in prospective cohort study we are starting the study with the population and waiting for them to develop into exposure and non-exposure group. Okay, but here we are starting the study with exposure and non-exposure group and we are collecting the population data from the records. So we are saving 10 years. And there will be in retrospective cohort historical cohort or non-concurrent prospective study. It is like we are collecting data from the history. Okay. So if the study is like this that is we are going from here to here to known as retrospective study and if we conduct the study in combined fashion that is historical cohort present scenario and the future follow up. It will be known as combination of retrospective and prospective study. So basic elements are similarly case control study. We have to select the study subjects. We have to obtaining the exact data on exposure. So we were checking also of exposure. Here we will get exact data on exposure because since we are going forward. It is a prospective study. Then we have to make a comparison group. Then we have to do follow up and analysis. So first thing we have to compare we have to select the population. Depending upon the hypothesis we can get general population or special population and we have to get the data. Second step we have to get the data from records, medical examination that will be in a follow up manner. Not from their past records. We have to do a past records when it is a retrospective cohort study. That is a combination. The same cohort will be checked for its records. The difference from case control study is when we are checking a retrospective cohort they are not developed into an actual case. They are just developed into exposure and non-exposure group. But in case control study actually we are checking the exposure of cases and controls. But here we have exposed and non-exposed group. When we are doing a retrospective cohort study not the cases. Then we have to select a comparison group. We can take internal comparison or external comparison group or with general population. You can take any comparison group. So it will automatically form a comparison group when we follow a population for a period of time. Some will develop the habit, some will not develop the habit. So people will not develop the habit. It will become automatically comparison group. Cohert within the same group. Sometimes it will not be available. So we have to go for an external comparison. Then you have to do the follow up. That is the main problem of cohort study. When you are doing follow up the same population will not be there. Some might have already left out of the city. Some might have gone for some other purpose. So there will be always attrition problem. Attrition research says that loss of people during a period of time. So there will always attrition problem in cohort study. Because maintaining the same sample over a period of 10 years and 20 years is very difficult. So it is very complicated study basically conducting a prospective cohort study. It is very difficult. It is administratively very complicated. It is expensive time consuming. But it is very good study because we will get the exact data. We need not to rely on the past entry or the data of patients. We can relate directly. We can see the patients outcome in front of our eyes. Because we are following up we can check it and we can count the actual risk rather than the odds ratio. Outs ratio is in direct way of checking the risk. So it is periodical medical checkup and hospital records, routine surveillance and questionnaires. In many ways we can obtain the data. So obtaining data is same. But in case control study we will be checking their past data which is already being written. One will be asking their past history. This is like follow up. Because we have seen patients without disease and over a period we will be asking questions. So analysis like here is the main analysis is calculation of relative risk. So the same example with disease and without disease analysis part what we are checking is incidence. So the same table with disease without disease and this is exposure and this is not exposure. It is non-exposure. Because the disease will be on x axis always and the exposure will be on y axis. So for 10,000 people 45 and developed poison and the remaining 9,955 did not develop. Whereas in non-exposure group only 5 developed disease. So the remaining were non-developing disease. So the incidence rate we calculate this will go horizontally that is among the exposure rate. The case control study exposure rate was calculated like vertically A by A plus B. Here it will be A by A plus C that is the presence of cases among the exposed group that is exposed group is A plus C presence of cases that is 45 cases among 10,000 that is 4.5. And the number of cases or the incidence of cases or presence of cases among the non-exposure group. Here it was among the cases and among the controls. In case control study here it was among the exposed group and among the non-exposed group. The difference you need to understand. This is prospect here that was retrospective. In retrospective it was going among the cases and among the controls. Here it is going among the exposed group and among the non-exposed group. So 45 by 10,000 and 5 by 10,000 45 by 10,000 45 by 10,000. So you get 4.5 is the incidence among exposed and 0.5 incidence among non-exposed. So the relative risk formula is incidence of disease among exposed that is horizontal and incidence of disease among non-exposed. So you get 4.5 divided by 0.5 is 9. So it implies that there is a 9 times risk of oral carcinoma in tobacco juice compared to non-juice or 9 times risk of getting food poison among street food compared to home food or restaurant. So analysis is different but the meaning is same. There we calculated the odds of exposure. Here we calculate incidence of disease among exposure and non-exposed group is just opposite. So the meaning is same. It is measuring the strength of association between exposure and outcome or the causal factor and the disease. So basically if relative risk is 1 then we can say that there is no association between this exposure and outcome that is eating street food or have getting developing food poison. There is no association. If it is more than 1 we can say that because of eating street food food poison has happened. If it is 2 we can say that twice risk. So greater the strength of association greater the relative risk greater the strength of association between the factor and the cause and effect. One example of another type of risk is attributable risk or attributable risk. It is also known as risk difference. Calculated by like this incidence among exposed minus incidence among non-exposed divided by incidence among exposed. It will be presented in percentage. So 4.5 minus 0.5 divided by 4.5 we get 88.9 percentage. It says that how much is the attributable power of that suspected factor. If 100 people are eating street food how many of them will develop food poison. So the power of that particular factor. Before relative risk we were seeing compared to the non-exposed group what is the risk of exposed. Here we are saying the power of exposure for the risk difference or what is the attributable risk of that particular causal factor. So we see almost 89 people will develop food poison if he eat food from the street side or if 100 people will have smoking habit. 89 people or 88.9 people will develop oral cancer. So that is the attributable risk or risk difference. And one last thing is population attributable risk. So it is what is the population effect. So it is like incidence of the death in the total population. This is only different thing coming here. So total incidence of disease. So all the disease incidence we have to include here in total population minus incidence due to the non-exposed so total population death minus population death happened due to non-exposed. So what is the incidence in non-exposed group divided by incidence in total population. So it also will give a percentage. It provides an estimate that if the suspected factor is removed from the population how much percentage of people will be saved. Suppose after calculating this we get 40 percentage. That is if we remove the causal factor of remove the street food. If we stop all the street food ventures we can save the total food poisoning disease 40 percentage. So if it is smoking and lung cancer, if we stopped stopping if we stopped smoking habit or if we prevented smoking we can save 40 percentage of the total death. So that is all about population attributable risk. The formula is different incidence of total population minus non-exposed incidence divided by incidence of total population. Here it was incidence of exposed. That exposed to become total population incidence of due to all the disease. So ultimately we get the disease of people with only exposed. So this is a little bit complicated. It gives a population attributable risk. It is mainly helping in preventing programs. If you are doing a mass program on tobacco banning of tobacco counseling that program will have this much effect. If we remove that particular factor from that population this much incidence can be saved. If population attributable risk is 70 percentage for the tobacco and lung cancer we stop smoking we can save 70 percentage of the death due to smoking from that population. That is all about case control and coercity. It was a lengthy session but try to understand the concept both are different but one or the same but has same scenario. One thing is going backward and one thing is going forward. The risk estimation also is just one is measuring the exposure other is measuring the incidence amount exposed to it. Thank you.