 Hello everyone, in this video I am going to discuss about Diagnostic study or Validational study. Most of the health research will be focusing on the risk factors, the diagnosis of the disease or the effectiveness of the treatment. For identifying the risk factor and the treatment, the epidemiological study designs are classified and most common classification is this. But in case of the diagnostic test, it's a unique topic where this epidemiological study designs will not touch. As I said it is unique, they have a different sample size calculation, different type of analysis and different results which we get. Before going to the diagnostic study, we have to understand how are we diagnosing a disease. This is an example for the spot diagnosis. That is if you look at this picture, you can easily identify it is a herpes zoster. So this does not need any more investigation, we can directly treat. We have self labelling, patients based on their symptoms directly go into the internet as their friends and they come out with their own diagnosis. Most of the times it will be threatening them. The third type is based on the presenting complaint, you can arrive at the diagnosis of liver failure. Then we have the place of ruling out the possibility of the presence of the disease, then refinements, then reasoning out thereby diagnosing a disease. Then we have the investigations, various types of investigations which can be broadly classified into the tissue sampling or the histopathologies, the biochemical investigations. Then we have the imaging investigations, then we have endoscopy and we have plenty of these investigations which will help in the diagnosis of the disease. Now where and how are these tests used? Most commonly the departments which use this diagnostic study design is or the validation study design is biochemistry, pathology, microbiology, radiology, public health and clinical subjects. And how do they use it is the clinical diagnosis, either they have certain sets of criteria of symptoms and they can have a clinical diagnosis and they can test its accuracy also. They can develop a screening criteria based on a single question or a set of questions that is questionnaire. They can look at the prognosis of the disease, they can look at the response to the treatment, they can screen for a disease. All these settings need validation of study design. Parameters on which diagnostic test is evaluated is validity or accuracy, reliability or otherwise called as precision, reproducibility, repeatability or consistency. The third one is the economicity or efficiency. Here this dots, the images with dots shows the difference between validity and reliability that is accuracy and precision. This is highly accurate and precise. This is highly accurate but there is low precision. This is highly precise but low accuracy. This has low accuracy and low precision. Validity is the ability of the test to detect what it is supposed to measure. Reliability or repeatability is the ability of the test to produce the same results when it is repeated. We need to understand the type of errors before going into the accuracy calculations. We always have to construct this two-bar-to-table for effective analysis of diagnostic test. The first rule for constructing this two-bar-to-table is always gold standard will be on the top that is say the truth or the best test will be in the top. Then the test result which you have will be on the left side then always yes will be first. Based on this we generate the true negatives that is the test result said no and the truth also no this is called as true negatives. On the other hand the test result says the disease present the gold standard also said the disease is present hence you call it as true positives. On the other hand here you have the test negatives but actually by truth they are positive. So you call this as a false negatives here as the false positives. Now validity or otherwise called as accuracy is measured by this following parameters that is sensitivity, specificity, likelihood ratio, positive and negative, positive predictive value and negative predictive value. Sensitivity can be remembered with the phrase positivity among the deceased and negativity in the health. If the test is sensitive then it is a good screening test. If the test is having good specificity then it can be used as a confirmatory test. You can remember with the fourth letter that is the key to remember. So sensitivity, screening, specificity, confirmatory test. So here we are going to see how we calculate the sensitivity and specificity. As I told sensitivity is the positivity among the deceased. Here they are the deceased, they are the healthy. This is test positives, this is test negatives, this is total. So it is positivity among the deceased. So true positives divided by true positives plus false negatives. This is the positivity in the deceased. Here it is negativity in healthy. So Fp plus Tn, so true negativity will be on the top. So this will be negative by the test, actually not deceased will be down. So this is the sensitivity and specificity. We can calculate positive predictive values and negative predictive values. Positive predictive value is out of the total test positives, how many are actually positives? Out of the test detected positives, how many are actually positives? Same way out of the test detected negatives, how many are actually negatives will be the negative predictive value. You can also have this likelihood ratio. At this stage you can remember it is the measure combining sensitivity and specificity. Basically what this likelihood ratio is, the chance of a person having the disease has how many times changed. That is the likelihood ratio. It can be for the presence of the deceased or absence of the deceased, based on which we have likelihood ratio positive and likelihood ratio negative. We also have this Fagin's normogram, pre-test and post-test probability and the likelihood ratio calculation. Then the next important parameter is the reliability, that is the repeatability. It is affected by variations which can occur due to observers, subjects, instruments or techniques. This can be statistically assessed by estimating the degree of agreement between the two measurements. The level of agreement can be calculated by the Kappa statistics. If the microbiologist diagnosis is the gold standard here and the technician's diagnosis is the test we are validating, then we put the 2 bar 2 table like this. To calculate the Kappa statistic, we need to calculate the observed agreement, that is we need to calculate the true positives and true negatives, true positives plus true negatives divided by total is the observed agreement and we need to calculate the expected agreement that is calculated by row total into column total plus row total into column total divided by grand total into grand total which will yield this value. Now the Kappa statistic is given by the formula or the Kappa coefficient is given by the formula observed minus expected divided by 1 minus expected. So 0.85 minus 0.754 divided by 1 minus 0.754 which gives 38.8 percent rate. Now how to interpret this Kappa value is as it is close to 1 or 100 percentage it is the complete agreement that is the agreement is better if it is equal to 1 it is complete agreement if it is close to 1 it is the better the agreement between the either the two observers or the two subjects or the two techniques or the two instruments or any method of measurements. Usually Kappa up to 25 percentage indicate a mild agreement, 26 to 50 percentage indicate moderately strong agreement, 51 to 75 indicates strong agreement, more than 75 percentage indicates very strong or excellent agreement. This is how we interpret the Kappa statistic value for the reliability. Then the third important parameter is the economicity we need to look at whether the test is safe for the patient, cost effective and available at our location. Then we need to ask is there a real benefit out of this test. And the third question we need to ask is will the result of the test change my management in clinical settings that we need to answer this is the third point about economicity. So far we have discussed about the outcomes of the diagnostic test in terms of categorical variables that is this is present or absent. Suppose when the measurements is based on a continuous scale for example random blood sugar HBO and C for a detecting diabetes militants then we have a continuous scale. In this example we have serum ALT for chronic paranchymal liver disease. The diagnosis is confirmed by liver biopsy. So based on the liver biopsy the disease present and absent are categorized like this and the serum ALT levels are categorized like this. Then we have taken the cutoff level of 20 that is more than 20 means then it has a good sensitivity but the specificity will be very low. At the same time if we increase the value to 100 the sensitivity will be reduced but the specificity will be very good. So we need to have a balance between the sensitivity and specificity. If we put up the sensitivity and specificity for different cutoff level as the cutoff level decreases the sensitivity increases. Sensitivity also increases the specificity decreases. In that case we need to identify a cutoff value with maximum sensitivity and specificity. In this scenarios we use a ROC curve otherwise called as receiver operating characteristic curve. This is a curve with sensitivity on the y axis and 1 minus specificity in the x axis. So this is the reference line and we have plotted the sensitivity and 1 minus specificity for different cutoff values. So we need to look at the value which is close to this top corner. The best sensitivity and specificity values can be identified using this ROC curve. We need to understand how are we calculating the sample size for this validation studies or diagnostic studies. First we need to specify the expected sensitivity of the test and specify the acceptable deviation. The expected sensitivity can be taken from previous studies. We calculate A is equal to sensitivity into 1 minus sensitivity divided by the acceptable deviation. If suppose we are validating ELISA test for HIV infection our rough estimate is that sensitivity would be 95 percentage and we accept a deviation of 3.3 percentage on either side then we need to calculate like this 0.95 into 1 minus 0.95 that is 0.05 divided by 0.03 the whole square. So like this we calculate A, yeah it is given here so we get it as 33. Then we need to calculate the actual sample size from A by the formula A by prevalence. So let us say the expected prevalence of HIV infection in the population is 5 percentage then we need to put up 33 divided by 0.05. So which yield 1060 as the desired sample size for your study. What are all the steps in a study on diagnostic test? We need to clearly define the research question and its background significance. Specify the variable of interest whether it is categorical or continuous then the settings then the gold standard which we are going to compare then the reliability or repeatability then the accuracy or validity which is given by sensitivity specificity predictive values and likelihood ratio then we need to calculate the sample size then we need to make every subject undergo both the diagnostic test gold standard test and we need to arrive at the 2 bar 2 table we need to calculate the accuracy parameters such as sensitivity specificity likelihood ratios predictive values and capacity for reliability. We need to decide the usefulness of the test based on the safety cost availability acceptability and economicity for diagnostic test we have a good reporting guidelines called start criteria it is standards for reporting of diagnostic accuracy studies it has 30 essential items as I mentioned here so we need to look at as a checklist based approach we need to mention all these items in our manuscript in order to make our manuscript better so this start criteria will help we also have one more criteria called tripod criteria that is transparent reporting of a multivariate prediction model for individual prognosis or diagnosis easily accessed and downloadable from EquatorNetwork.org now how to critically evaluate a diagnostic test are the results of the study valid you need to look at the start criteria what are all the results you need to look at the validity and its parameters reliability and its parameters will the results help in the care of my patient you need to look at the safety cost effectiveness availability then you need to look at the real benefit present or not and also we need to ask ourselves that whether the result of the test change my management or not these are my references if you have any doubts and feedback you can contact in this email id or post in the comment section if you like this video please share it to your friends if you haven't subscribed to our channel please subscribe thanks for watching this video