How to Interpret and Use a Relative Risk and an Odds Ratio

hi Terry Shane you fell for UAV school of medicine odds ratios and relative risks are the main measures of association and observational studies in this video I will discuss how to interpret what they mean and importantly how to apply them in the care of our patients so the outcome measures and observational studies are a relative risk or risk ratio in cohort studies and an odds ratio in case control studies let's look at each of these individually so relative risk or risk ratio is just that it's a ratio and then the numerator we have the incidence or rate of development of disease in the persons who are exposed divided by the incidence or rate of developing disease and those who are not exposed and a relative risk or risk ratio expresses how many times more or less likely an exposed person is to develop the outcome relative to an unexposed person and the key here is relative as we'll see in a minute how do we interpret this well if our relative risk is one there's no risk of the outcome because the numerator and denominator exactly the same the relative risk is greater than one that means it's increased risk of the outcome and exposed persons the relative risk is less than 1 that means reduce risk of the outcome and exposed persons so let's see how we interpret a relative risk in this case I want you to interpret what a relative risk of 1.36 means pause the video look through these answers choose the one you think is correct restart the video and see what my answer is so let's see how you did what did you choose well a relative risk goes along with a cohort study and cohort studies look at risk of disease so anything here that has exposure in it can be thrown out immediately next the relative risk is 1.36 which is greater than 1 so that means risk is increased so I have one answer that says risk was reduced so I can get rid of it completely so now I'm down to two answers and this is a useful formula to figure out this answer the relative risk increase or relative risk reduction is the absolute difference between 1 - relative risk and then you multiply it by 100 and so because my relative risk is 1 point 3 6 I know I have to calculate a relative risk increase and when I run my numbers through I know my answer should be number five or 36% so relative risk of 1.36 means risk of disease is 33 excuse me 36% greater and exposed persons relative to unexposed persons so remember this formula it's something you'll need both for your step exams etc but also importantly in clinical practice so let's do two more examples see if you really understand this so the first one as we want you to interpret a relative risk of 0.8 pause the video come up with your answer restart to see how I answered it so what did you get so what I got is a relative risk of 0.8 means that risk of the outcome and expose group is reduced by 20% or occurred 20% less relative to the unexposed group remember that one minus relative risk is 0.2 or 20% how do you interpret this one a relative risk of 3.3 pause the video we start to see what my answer is so what did you get well what I got was the risk of the outcome and exposed group is increased by 200 30% relative to the unexposed group an alternative interpretation is that the risk of the outcome is 3.3 times greater in the exposed group relative to the unexposed group so again using that formula from the previous slide will give you this 230 percent but one thing you have to keep in mind is I keep using the term relative risk and it is all relative so relative risk of three means you're three times more likely to develop the outcome if you're exposed and if you're not but are you impressed by this number well I'm impressed if the patient starts out at the 30 percent baseline risk and goes up to 90 percent but I'm not so impressive they start out at at point one percent and only go to point three percent so the key here is you have to assess baseline risk of your patient then apply the relative risk to get their final risk of developing an outcome if they're exposed to something this is how we use this information in clinical care it's not that big of a deal if I can interpret a relative risk an odds ratio or calculate it I have to be able to use it to help take care of my patient and the next slide will show an example of how this is done so this is the Nurses Health Study or the data from it I've deleted out some of the extraneous information to make this a little clearer and let's focus on this multivariate adjusted relative risk of women who used estrogen and progesterone compared to non-users so first thing I want to ask you is how do you interpret this relative risk pause the video for a second come up with your answer then restart to see what I put so how I interpret is at the risk of major cardiovascular disease was 61% less and users of estrogen and progesterone compared to non-users same formula from before 1 minus relative risk but let's say I have a woman sitting in the office who I do a Framingham risk calculator on her it's determined that she has a 20% chance of having major cardiovascular disease and I'm trying to decide whether I want to use estrogen and progesterone or not so now what I want you to do is take her baseline 20% risk and use this information from the Nurses Health Study to figure out if she used estrogen progesterone what is her final risk of having cardiovascular disease pause the video restart to see how I answered it so what did you get this is a complex problem but a real one of decision-making in clinical practice and so what I need to do is multiply her baseline risk by the relative risk and I get what her final risk is in her final risk will be about 8% based on the Nurses Health Study if you use estrogen and progestin and that makes sense it reduced her risk because the relative risk was less than one so this is how I would use an observational study to help me take care of a patient I would take their baseline risk and this Framingham risk calculator is a prediction rule so it's best to use something like a prediction rule to figure out baseline risk and multiply it by that of the risk found in the study to get your patients final risk that's clinically how we use this information so what about an odds ratio often physicians interpret it the same as a relative risk of technically it's quite different and an odds ratio again is a ratio and it's a ratio of the odds of exposure and those with disease or cases divided by the odds of exposure and those without disease or control so remember in a case control study we start with cases who have disease and controls who don't go backwards in time for exposure so an odds ratio actually estimates the risk of exposure different than a relative risk but often physicians interpret it in the same way which isn't technically correct an odds ratio is how many times more likely are the odds of finding an explosion someone with disease compared to finding that explosion somewhere without disease and an odds ratio one means there's no change in the frequency of exposure because the numerator denominator the same odds ratio greater than one means there's increased risk of freak or increased frequency of exposure among cases an odds ratio less than 1 means there's decreased frequency of exposure among cases again physicians often use this as an assessment of risk of disease I guess it's okay to think about it that way for clinical purposes but it's not absolutely technically correct so when should I be impressed by a relative risk or nods ratio well I'm going to give you some rules of thumb as suggested by the authors of the EBM Bible and in a randomized control trial because it's the least prone to bias you would be satisfied with fairly small increases or decreases in relative risk but Nicole Watts study which is more prone to bias than the relative risk the number needs to be bigger my relative risk needs to be at least 3 or greater for minor adverse events and things that are significant adverse events I'll ratchet this back and a cohort study they suggest an odds ratio greater than 4 to be impressed with and that's because case control studies are even at greater risk of bias than cohort studies which are at greater risk of bias the randomized control trials again these are just sort of rules of thumb the bigger the number the more impressed you are now when should I believe the relative risk our odds ratio in a study and professor lesser wig who is an epidemiologist based in Australia made this suggest first all you have to compare an unadjusted or raw relative risk or odds ratio and often you might have to make this calculation yourself although sometimes the authors of a study will report it and you want to compare that to the odds ratio or relative risk that was adjusted for at least one known confounder now if this adjustment produces a large decline in the odds ratio relative risk be very suspicious that this is a spurious Association and not real on the other hand if adjustment increases the relative risk or odds ratio or it remains stable you can be confident that this is a valid Association and what I mean adjustment increases or you know comparing yours to the adjusted one is what I'm talking about in these two statements so let's see a real-life example of this this again is the results from the Nurses Health Study and I first have to calculate my unadjusted a raw relative risk because both of these reported here adjusted and the way you do that is just divide or first calculate the incidence of major cardiovascular disease in the women who used estrogen it's just 8 divided by this 27 161 and I divide that by the incidence or probability of developing the outcome and women who are never exposed is just 431 divided by this number when you do that you get zero point 208 when I compare that to age-adjusted that's not too different I'm fairly satisfied at this point of that this might be a real Association but look what happens when I compare it to the multivariate adjusted huge changes in risk so a huge reduction in the estimate of risk and the reason I say reduction is even though this number is greater than this remember 1 minus relative risk this is actually smaller reduction so I'm really worried that this is a spurious Association and not real and I made a previous video about residual confounding that gives some reasons why this might be a spurious Association you might want to look at that this video has helped you understand how to interpret an odds ratio in a relative risk and how to apply it in patient care remember if you have any questions you can contact me through the course website or through the contact me section in my blog have a great day

Channel: Terry Shaneyfelt

Views: 306,319

Rating: 4.9314713 out of 5

Keywords: relativerisk, oddsration, observationalstudy

Id: FZzm3-RRlI4

Length: 11min 0sec (660 seconds)

Published: Wed Feb 20 2013