Randy Neil Biostats 03: Biostatistics Summary Step 1 - The Basics USMLE

Specificity = TN / (TN + FP)

Case-control.

It starts with disease/outcome status and looks backward for exposure.

Cohort.

It starts with exposure status and compares outcome risk.

Power = 1 − β

About 68% of values fall within ±1 SD, 95% within ±2 SD, and 99.7% within ±3 SD.

Answer: C) 90%

Specificity = TN / (TN + FP) = 180 / 200 = 90%.

Answer: B) Case-control

The outcome is already known, so the study looks backward for the exposure/risk factor.

Answer: C) 85%

Power = 1 − β = 1 − 0.15 = 0.85.

Alright guys, so let's start getting started. We get about 16 of these questions. These are the most one you're most commonly going to see. So, let's get started in a game when you study this, just kind of go through the questions in your head, kind of watch this, in a sense, and working about over and over and over. So, just kind of watch these. First one says a new test, a diagnosis, UTIs and women's being assessed. The comparison goal standard is positive theorem dipstick, plus urine cultures, ultra-given. So, they're going to give you a chart like this, 2x2 table. They're going to ask you what does a test, sensitivity, specificity, positive-positive value, or negative-picked value. And remember how it goes. Now, here they're asking for specificity. Let's just go back to what we remember. Sensitivity, you circle here, go down. Specificity here, go up. Positive-perfect value goes this way. Negative-perfect value goes, goes that way. So, you should have all the formulas just based on this. Now, with specificity, it's here. You've be circled here and went up. So, on the top, goes 180. On the bottom, 180 plus 20, when we connected it, you take both numbers. So, the specificity in that case is 180 over 200. You cancel out, you get 9 every 10, and it's 90%. If they said sensitivity, we would have said, we went here and go down. So, it's 40 over 40 plus 160, 40 over 200. We start canceling. And then we're looking at about 20%. So, in this case, specificity was 90, answer choice E. The next one says, a five-year study is planned to assess the incidence and ideology of respiratory disease in 600 individuals greater than 50 years of age. Study consists of two groups. One carries for a pet dog, and the other one is second is not. Care for any animals in the household. The onset of respiratory symptoms, cultures in serialogical studies will be performed which the following

best describes is study. So, when we see a question like this, basically they're going to ask, they're going to try to see if you know the difference between case control and cohort. And remember what we said, case control has an A and an O. So, an A and an O. Two different things. It's the most meaning like one person has a disease, and one does not. The other one is a cohort, an an O. So, basically both sets of people start out with out the disease. So, when you look at this problem, they talk about these people that are 50 years old, and then they're going to say, well, we're just looking at the ones who care for a dog, and the ones who do not have animals. So, at the onset of the study, both the 50 year old did not have the disease. They're actually looking for it to happen in the future, given these circumstances. So, in this situation, they're actually looking for a cohort study. Remember, oh, there's no, nobody has to disease initially. Now, since it's like relatives, right? Relatives look the same. So, when you see, again, so just remember when you see cohort, you say relative risk. When you say case control, you need to say odds ratio. You just gotta have that memorized, okay? They look the same. They're relative relative risk. Nobody has to disease initially. It can go ahead and look forward and backwards. Now, when someone has to disease and someone does not, you think odds ratio, case control. And you really only want to look backwards, because we don't care about going forward on that one anymore. Says a new study is to diagnose prostate cancers being evaluated, since the activity of the test is 70% in the specificities nony. In the study, there are 100 patients who truly, truly have you to use. And 200, truly not. How many false negatives are there in the study? So, when you get this, it didn't give you the chart initially, so you gotta create it. So,

when you ever get stuck, you say, what do I know? Well, I can always create a chart. And I know up on top, we always say reality goes there. And we would have said a test goes over here, positive negative, positive negative. Gotta be able to label this. Now, positive, if the test is positive, then the reality is positive. That's a true positive. Test is negative, reality is negative. It's a true negative. So, and everything is based on the test. So, if the test was negative, and reality was positive, that's a false negative. Based on the test, test was positive, and reality was negative. That's a false positive. Now, what does this question says? It says a sensitivity to test is 70%. So, we know the sensitivity was this going down. So, we know here, this is an equal 70% somehow. Specificity is here going up. So, we know that's gotta equal 90%. So, in the study, there's 100 patients who truly have UTIs. So, in reality, 100 patients actually had UTIs. So, this category here is gonna equal 100 patients. And 200 who truly do not. So, in reality, there's 200 people who do not have UTI. So, this is a category that goes here. So, this is how many false negatives are in this study. So, that'd be that number. So, I know that something over the total of 100, something over the total of 100 equals 70%. So, what would that number be? 70, correct? So, if we have 70 over 100, that gives us 70%. But that's not what they ask. They ask me if any false negatives. So, 70 plus the 30 equal 100. So, my answer in this one, how many false negatives? This right here, an inch twice C, number 30. Now, if they would ask false positives, I would have said, oh, I know that 90% of 200 is 180 and 180 plus 20 equals 200. So, if they would ask me how many false positives, I could have just said 20. But you always go back, once you write this table out, I don't care what they give you. You just work

backwards and you can solve this. Now, this one says, basically says, what type of study is being conducted? Being conducted. And so, again, we're always, always look back between cohort and case control for the most part. Everything else is pretty self-explanatory. So, it says, a test is being conducted to determine if older students have a lower score on US-mix step one. A group of students older, a group of older students, greater than 50, you took step one, or compared to a group younger students who took step one. Data shun, below. So, here's a two by two table. And basically, they're saying that, you know, there should be a better example here, meaning more of a disease. But here they're saying just a lower test score, versus a no more test score, older person, versus a younger person. So, basically, we already know that, we already pretty much know the outcome in this. So, if we already know the outcome, someone like say, for example, has disease and someone does not, then that's what we called, a case control. Now, if we didn't already know the outcome in this, not in the situation, but if we didn't know the outcome, we can actually go forward looking or backwards, and that is considered a cohort study. But in this situation, we already know the outcome. So, basically, it's equivalent to saying, someone has disease and someone is not. So, that one would be a case control. Now, this is the same question, but now the question is, what is the odds ratio? Okay? And again, we knew that, because it was said case control would be, we always think odds ratio. You got to have that. You may see a question that just say odds ratio, and then you got to go back and say, oh, that's case control. Now, when we say odds ratio, we got to notice odds ratio, versus relative risk. Think odds, when you're in Vegas, you think odds, so it's one number, over one number. When you say

relative risk, you got to think one number, over two. You got to keep that in mind, odds, your in Vegas, three to one odds, two to one odds, one number, over one number, relative risk, one number, over two. You do that, you get it right. Now, when we read math problems, we read them, top to bottom, let's go right. So you got to read this properly. So it's what is the odds ratio, relative to that at the younger student, of older students having lower test scores? So it's very important to understand they're asking older students, compared to the younger. So who's going to go on top? Who's going to be my thing on top? And that's going to be the older students. So it's the older students and it's odds ratios, one number, over one number, 60 over 200, the younger student, 40 over 160. And that's going to answer to a C. Now, this thing was different, and it basically had relative risk, it said relative risk, it looks something like this. It'd be one number, over two, 60, over 60, plus 200, over 40, over 40, plus 160. Very simple difference between those two. Just remember, odds ratio, one number, over, one number. Relative risk, one number, over two. Okay, you do that, you get it right. All right, the next one. It says that biomarker is being used for detection of certain disease. 500 healthy volunteers and 120 patients with the biomarker are used in the figure below. So changing the cutoff value of the biomarker from point B to point A would most likely result in. So again, they're going to give you something like this, some type of chart. So you're going to move the marker from A to B or B to A. Or they're just going to not give you this and say, well, I'm moving the cutoff from 200 down to 100, or they can say moving up from 100 to 200. So biomarker is going to move it left from right to left or left to right. Okay, now, you have to label this properly. And we said,

over to the left, true negative, over to the right, true positive. Everybody keeps our last name. So on this side, right here, if you keep your last name, that's going to be a false negative. And right here, if you keep your last name, right there, it's going to be false positive. And all we ever care about in these situations are the middle numbers. I don't care about him, and I don't care about him, because he will not change. Okay, for our purposes, he will not change. So it says, changing and cut off from B to A. So if I'm moving this thing from B to A, and remember, I always kind of think you're going to adjust this bottom line. So if I move from B to A, I'm actually going to smash that false negative. So false negative goes down. And if I move it from left to right or from B to A, technically, you think this middle line goes, this way, in my false positive, goes up. Now if it went from low to high, the opposite occurs. I'm smashing the false positive, and increasing the false negative. So they're only looking about, well, it's good. For lower sensitivity, well, what's my think for sensitivity? I'd go back and write my chart of what I know. Reality, test, positive, negative, positive, negative. That's a true positive, negative, negative, it's a true negative. Everything's based on the test. So that's a false negative. Everything's based on the test. So that's a false positive. So sensitivity, we said sensitivity. This is going down. So that's a true positive over true positive plus false negative. So it would be lower sensitivity. So here, the only thing that relates to this, that's all I care about, and this would be the false negative. Now false negative goes down. But so if that number goes down, the whole sensitivity actually would go up. So it's not looking like it's A. More true negatives? But wait a second, true negatives were out here. And again, we do

not care about the true negatives or true positives in this scenario. So those things do not change for us. More false negatives. But we just talked about this. If it goes right to left, it smashes the false negatives. So that's actually going to be incorrect. Higher, negative predictive value. Now, again, we went back to that thing. And we said sensitivity goes this way. Specificity that way. Positive predictive value negative predictive value. So the formula for negative predictive value is true negative over true negative plus false negative. So the false negatives in the scenario went down. So if that goes down, my negative predictive value actually would go up. So that could be definitely an answer to choice. So I'm going to put a little mark next to that. Or higher positive predictive value. Positive predictive value was this guy going that way. So the formula is true positive over true positive plus false positive. That's positive. That's positive predictive value. Now, the only thing I care about is false positive. And he actually went up. So if the false positive goes up, the whole number out here, actually, you would go down. So it's not that guy. So the answer in this scenario is higher, negative predictive value. Great question. Remember, all you care about is the middle. You got to label him correctly. Everybody keeps their last name. That's how you know which side to put him one. You're going to slide him to the left or slide him to the right. You're going to smash this guy or smash that guy. And if he smash the left, the right goes up. He smash the right. The left goes up. That's all you care about right there. The following graph shows distribution values from a healthy group of volunteer and disease people. Points A through E represent various points for determining distinctions between these people. What cut off point would determine sensitivity of

1. Well, those are formula for sensitivity. That again, go back to our chart. Reality, test, positive, negative, positive, negative. True, positive, true, negative. Everything's based on the test. It's a false negative. This one's a false, positive. So sensitivity is this guy. So my formula is true positive over true positive plus false, negative. Remember, the bottom is two things. Whatever thing I did, connect the line. You're right, both of them down there for the bottom. All right. So to make this 100%, that means I got to get this guy. Actually, he's got to become basically a zero. So if I notice he's a zero, my sensitivity will be 100%. So where in this chart would I find the false negative is being zero? Well, it's just labeled the chart. We know that's a true negative. This is a true positive. And then all I care about is in the center. I got to label them correctly. That's a false negative. And this is a false positive. So wherever it would make him zero, work and I smash this guy and make the false negative a zero, I slid him all the way right here. We see. Now if they say what is the specificity? Well, I would say, oh, that's specificity. True negative, a returinegative, plus false positive. Work, I make the false positive zero. Work, I make him zero. Oh, if I smash him this way. And then so the specificity would be choice E. All right. So what will happen to the sensitivity and specificity of a test when the markers are moved from the blue curve to the red curve? So from the blue to the red or pink, whatever you see here. And again, you see the scenario all I care about is the inside. I don't care anything about this. Trinegative, I mean, trinegative, true positive. So here, I dropped my line. So on this side, it was actually the false negative. And on this side, it was actually the false positive. But going from the blue to the pink, the area under

this actually went down and went this way. So it's going from here smaller. So the false negatives got smaller. It went down and in. The false positives got smaller. So now all I care about, I got to plug this back into my formulas. And so all I'm looking at is for a sensitivity and specificity. Back to what I know. Positive, negative, positive, negative reality. Test. True positive, true negative. Everything's based on the test. It's a false negative. This one is a false positive. So sensitivity goes this way, specificity goes that way. So sensitivity, positive, plus false negative. And specificity is true negative. Right here, going up, true negative, plus false positive. So on both these scenarios, and both these situations, the false negative and false positive went down. So the false negative goes down. So if he goes down, this whole number gets bigger. So sensitivity goes up. False positive. He went down. So if that number goes down, the whole number goes up. So in this situation, when we went from this, the blue to the pink, the area of the curve got smaller for both of those guys. Smaller. And both those get smaller. It's a higher sensitivity in a higher specificity. All I go back to our formulas, you got to know this. But all I care about in this situation is the center. I don't care about him. I don't care about him. Checking blood pressure to health fair would be an example of what type of prevention. This can come up pretty easily. Just make sure you know these first two. Primary prevention is some like immunizations. And then secondary prevention is something like when you're trying to screen for this stuff, such as checking blood pressure would be a scenario where you'd use secondary prevention. So kind of know these. It's kind of read through them. If you get a chance, primary again, education, immunization before the thing happens. Secondary screen

for disease early to reduce the impact. Those are your two main ones. Basically, make sure you know those. Okay. And let me see here. We talked about the case fatality. Again, when I say case, when I say case, great case rate, all I care about is the case they're talking about. So it says the table shows distribution of spinal cord injuries and death, what is the case fatality for falls. So case fatality for falls. So I don't care anything else except falls. So case fatality. Here's a number of fatal ones. Here's a number of total. Case fatality falls. How many falls total? 20. How many of those were actually fatal? Four. So case fatality for falls is four out of 20. Okay. Just make sure you understand that they're staying within that category. They're not asking about anything about this number, this number, any of that stuff. Case fatality falls. Four out of 20. A new instrument is purchased by the hospital, it checks serum levels of some type of X. The published value for the standard is 40. The technology runs with testlempacians and gets readings of 70, 60, 70, 70, 70, 70, 70, and 75. Respectively, what can we conclude about this instrument? All right. So they're asking about two things. They're asking about accuracy and precision. Now, this is a bull's eye. You know, accuracy. If they're going to give you something that says, the disarmant accuracy, they've got to give you a gold standard. So with accuracy, you've got to have some type of marker or gold standard. Some type of reference to know if you're hitting the bull's eye. So when a bull's eye, the gold standard is going to be the center of the target. So if you hit around this, obviously, if you're getting close to the mark, you're accurate. Okay. The larger you're close to the mark, you're accurate. Now, with precision, you're just in the same area. That's all. You just want to make sure you're pretty

tight with that. So with accuracy, you've got to have some type of gold standard measured yourself by with precision. You know, you're just kind of being the same area. So in a situation like this, what the standard is, 40, yet this guy's hitting these readings of 70, 70, 70, 70, and 75. So he's pretty tight in this region. So this guy's pretty precise. So when you reference him next to what the gold standard is, he's not very accurate. So in this situation, he's precise, but not accurate. Answer choice B. All right, new study showed that the mean HTO level of a non-diabetic patient was 42, and that the mean HTO level in diabetic was 35. Probably this was due to chance, was just 0.05. Okay. There's also a 15% probability of the coconut that there is no difference in the HTO measurement when their reality was one. So first thing, because what is a P value of the study? What's P value mean? Basically, you gotta think a P value just says, what is the chance of it happening, happening by chance. Okay. And in this situation, actually gave it to you. Probability of this was due to chance was 0.05, or changed that 5%. And remember, for a study B actually a good study, it's got to be either point or less. Okay, so 0.0. Four, is that a good study? Yes, because it's less than 0.05, is 0.06 a good study. We're gonna say no, because the rule says, it's got to be 0.05 or less, okay, or less, not higher. What is the power of the study? Well, let's get back to the whole null hypothesis stuff. So when we draw out our null hypothesis, we still stick with our reality over here. I always like to do just the test. Stick with test, but just whatever situation it is, okay. Now, we got to label it correctly. 1 and an 0, 1 and an 0, and an 0, and it's oh, and no, is the null, okay, that's the null hypothesis. So no meaning there is no association, okay. So let's fill out our chart of what

we know. So if the test says there is an association, but in reality there is no association, that's called an alpha error, okay. Alpha error type 1. Again, the test says there is an association, but in reality there was not one, that's an alpha error. Now, the test for situation says there is no association, but in reality there was one, that's called a beta, okay. Beta or type 2 error, okay. And it's all based on words. Gotta know how to first write this out, one and a 1 and a 0, reality in the test. And then you gotta put this into words, fancy on you gotta know this in words. So again, if the test says there was no association, but in reality there was one, beta error, okay. Beta error. If there is an association where there was not alpha error. Now, how do I find this one right here, whether it says, well, yeah, there is an association where reality there is one. How do I find that? It's one minus beta, also known as power. If you can write out this, if you can draw out this, you can answer any question they're gonna ask you pretty much on a step while it comes to the null hypothesis, which you gotta know what it means in words, okay. So it says, what does the power of the study? So we wanna know what this box is right here. So how do I find that? Well, let's see what they gave me. They said that there is, there's also a 15% probability of concluding that there is no difference in the H to E, I'll measure it when and where out, there is one. Again, 15% probability of the concluding that there is no difference. We're gonna try to conclude, the issue of measurement when there is one in reality. So there's no difference when I read out. There was one, there was a 15% probability of that happening. Well, what do I gotta do? I take one, that was beta, this is the point one five. So one minus the point one five, it's gonna be point eight five. And that's actually

gonna be my power, okay. Answer twice C, what you gotta be able to write this box out, okay. And then this is the most common question, just like we did it right here. This is the most common question they're gonna ask that you can understand that this in words actually means the beta box, okay. And so one minus that, that's your answer, okay. Now, the prevalence, the prevalence of prostate cancer compared to two groups of men and the falling down was obtained, baseless data was relative risk, okay. By the way, relative risk. And when we read math problems, read them, top to bottom left to right, they'll wanna prostate cancer in men who had no children, compared to men who had children. So we gotta compare men who had no children to the ones who did. So who's gonna go on top, the men who had no children. Now, it's relative risk. Now, we said odds ratio, we go one number over, one number. If I said relative risk, it's one number over two. Okay, so relative risk, I was able to move, so no children, so that was gonna be 80 over 80 plus 920 because it's one number over two, because it's relative risk. Compared to the men who had who had children, to 20 over 220 plus 1, 2, 8, 0, okay. Now, again, it's at odds ratio, I would have just went 80, and odds ratio in this scenario, I would've been 80 over 920, over 220 over 1, 2, 8, 0, okay. But since it's relative risk, one number over two, over 1, number over two, gotta know it. All right, so then if we kind of wittle that down to 220 over 1500, I mean, we can kind of, do the math here, hopefully they'll give you a lot easier scenario, or 220, you know, you can cancel out a lot of stuff. So if we did do that with 500, it's 3, 2, again, if I did by 10, 24, 44, almost looking, you know, looking close to about 50, roughly 55%, all right, very good. Scatter diagram, chose a correlation between alcohol consumption and test scores.

All right, so you're gonna get a scenario like this, and either the scenario is gonna be going like this way, and you just gotta look at the dots and draw whatever law and you think we've met most match where this thing's going, okay. So in this situation, it's that way. It feels like this, it would be that way. Now, if it's going in this way, it's like, as you drink more alcohol, okay, as you drink more alcohol, your test score is gonna go down, okay. So this is gonna be a negative association, okay? I'm just using negative one, because basically it's basically saying that if I go over one, I can also go up one, it's a one to one ratio, the line looks like it's one to one, I go over one up one, it's equal to negative one. In a situation like this, the more I study, say the more I study, the higher my test score, again, over one up one, or up one, over one, it's a one to one ratio, and then you can kind of go from there now. In this situation for this answer, it would have been negative one, but you gotta be able to understand it, it feels negative point two, this low, would be a little bit bigger. It feels a positive point two, is slow, would be a little bit flatter. So for right now, just understand, negative association, positive association, and then understand the slopes can be a little bit less if it's one of these numbers. All right, last one, it says a study of 200 patients, 200 patients in hospital, patients hospital, I patient, blah, blah, blah, with complications related to pneumonia, and you show their serum cholesterol level is normally distributed with a immunatutan, standard deviation of 15, basalin the study, how many patients would you expect to have cholesterol greater than 240? All right, so it's a standard deviation stuff. So when in doubt, it's kind of right out what you know. So there's a standard deviation curve. It says, there's 200 people in

the study, pneumonia is a zero-hustrel, they mean a 210, so we know the mean is 210. Standard deviation is 15. Now, we know one standard deviation is 68%. You know, two standard deviations is 95%. What? Three standard deviations, what is 99.7%. All right, so you basically have to have this memorized. Now, they love this one, that's 95%. Just because it's a nice number, they know you can use it the 95 a lot. Here's your need to use 68, which got into this 168, 228, 25. Now, an standard deviation of 15. So basically, so what they're saying is, if they come out 15, 15 points, 15 points going down, it's gonna be 195, and that 15 points going up, it's gonna be 225. And basically, that's same between 195 and 225. There should be, that's one standard deviation, there's 68% of the people should be in this in between here. Now, the question on this one says, based on the study, how many patients, how many patients would you expect to have cholesterol greater than 240? Okay, so somehow out here, I gotta get to 240. Okay, well, I can see what they're doing here. So if one standard deviation was adding 15, so another standard deviation, another deviation out would be adding 15. So, they're basically looking at two standard deviation. So I'm adding 15, and that takes me to 240. I could minus 15, and that's gonna take me to 180. So what that's saying is, now in here, between all this, I should have, 95% of the people should be within this range. So again, back to the question, so how many patients would you expect to have cholesterol greater than 240? So really, all we're looking at in this whole problem is the number of people that are out here, going this way. So let's look at what we know. We know 95, 200 people in here, 95% are contained in here. So if I had 200, and if I did 95, okay, so I know 90% is 180, 200, so that should be about 190. So 190 people are in the center

here. So that leaves how many, that leaves 10 that aren't. So 10 should make up the outer edges. But again, they were gonna, they would try to trick you on this, and you know, they went everybody by the Bible in this 10, but that's not the answer because the question is how many patients would you expect to have cholesterol greater than 240? So if there's 10 left, I gotta put five on this side, so the answer was, the grade in 240 would be five, not to 10. So again, when you study this, obviously you just wanna go through these, just real fast and know the concept. This is pretty much the meatmittators of what you'll see on step one. Just kinda go through these, and I think you'll be well on the test. So hope to help guys, we'll see you later.