USMLE World and Step 1 Correlation Data

[RapplixGmed]

Hey guys, I was curious about Usmle World and Step 1 correlation so I went ahead and collected my own data and correlation.

(See attached PDF for proper formatting)

Purpose: Many students are interested in knowing the correlation between USMLE World (UW) qbank scores and what they would get on the step 1 (S1). At the moment there is only one online tool that supposedly does this but it sucks so I went ahead and got my own data.

Method: Since 2004, IMGs have been on a thread in the prep4usmle website regarding their S1 results. To date there are 993 posts. Between 1/7/2010 and 4/6/2011, 35 people made posts that contained both their S1 and UW. I included only posts that stated that these UW scores were from first time random passes or posts that did not comment on the method of going through UW.

Results:
UW S1
Min 55 210
Mean 71 240
Max 87 269
Std. Dev 7.52 14.41




Uploaded with ImageShack.us
Regression Line is solid line. 95% prediction band is dotted lines.

Slope: 1.455 ± 0.2170
Y-intercept: 136.9 ± 15.49

R2 = 0.58

Pearson’s p < 0.01

Conclusion: There is a linear correlation between UW and S1 between the step 1 range of 210 and 269. UW can estimate S1 with an 95% prediction interval of about ±20 points.




Hope this helps!

---

Members don't see this ad.

 

[Rollo]

Good work!

 

[ArcGurren]

At the very least you'll get all those biostats questions right

 

[MsKrispyKreme]

Outstanding.

 

[Mr hawkings]

Wow.

 

[RapplixGmed]

At the very least you'll get all those biostats questions right

LOL. I just thought that people debate this topic way too much and decided to put an end to it with some real data in a parody of an abstract.

Another reason is I think the clinical review site way overestimates your expected score in the upper range. I see people posting about >75% on UW all the time but I almost never seen anybody score >270 like what clin rev predicts.

 

[Encephalectomy]

I think that most/all of the predictors get fairly inaccurate at the higher ranges (probably anything above 255/260). IMO At that point you are getting the vast majority of the questions right and how well you do probably depends on you not making mistakes on test day and on getting a little bit lucky on your question distribution (i.e. getting hard questions that you know). So I don't put much stock in predictors above that except to say that you are or aren't in that range.

 

[Bob101]

Very helpful OP, thanks.

 

[pullandbear]

so a 240 is roughly around 70% correct on UW? Not sure if i'm reading this right.

 

[RapplixGmed]

so a 240 is roughly around 70% correct on UW? Not sure if i'm reading this right.

Yeah, according to this model, it would be around 70.85%. You got it!

Another thing is that most IMGs have very short subscriptions of UW that they take right before they do the exam. Your score correlation would probably be more accurate if you used your average from the last 7 blocks in the model.

 

[raynor123456]

Thanks for your hard work. FYI, your results are pretty much completely consistent with this score calculator: http://www DOT clinicalreview DOT com/solutions/resources/usmle-score-calculator.html

edit: just noticed what you said about the "upper range". I compared it to mid range scores You're right in that regard.

 

[hyrule]

Seems off from the standard

UW% x2.4 + 84 formula

in that one, a 65 = 240

Many people had success with this formula... but it didn't work with outliers (those who started UW at 70% and ended up averaging 80%+)... and I'm sure there are wayy to many variables to account for as well

 

[fahimaz7]

Seems off from the standard

UW% x2.4 + 84 formula

in that one, a 65 = 240

Many people had success with this formula... but it didn't work with outliers (those who started UW at 70% and ended up averaging 80%+)... and I'm sure there are wayy to many variables to account for as well

So what did those guys get then? My first pass will be over 70% in UW.....

 

[da8s0859q]

I included only posts that stated that these UW scores were from first time random passes or posts that did not comment on the method of going through UW.

For the posts which didn't specify, I'm awfully curious as to whether the people doing UW were doing random blocks on the first/only pass. Some of the faculty at my school who are a little more in touch with the whole Step prep thing have suggested that even a consistent mid-60 percentage in UW the first time around, with random blocks, lands many people a score in the 240 territory. Guess that's not too far off.

Awesome thread OP.

 

[iA-MD2013]

How is it possible that after 1 month of studying, my UWorld scores are going down?!!? Most of my studying has been from these practice questions, so I guess it is possible that all of these questions are different so there's no way to improve when I'm trying to learn like this...but still! It doesn't feel good.

I started off in the high 60s and have now settled in the low 60s.

Edit: this may be the wrong thread for this post, but I needed to vent somewhere

 

[RapplixGmed]

Seems off from the standard

UW% x2.4 + 84 formula

in that one, a 65 = 240

Many people had success with this formula... but it didn't work with outliers (those who started UW at 70% and ended up averaging 80%+)... and I'm sure there are wayy to many variables to account for as well

Hyrule,

I have heard of this formula too. The thing is that I don't know where it comes from or what data exists to support it. It is also a very very old formula and there is evidence to suggest that the step 1 has gotten harder in recent years. The nbme exams have certainly been changed to a much harder curve and I think that has happened to the real thing.

I took special care to take data points that were from the last year to capture the difficulty of the most recent tests.

Thanks for your comments. I'm hoping to poll my class after this year's step 1 exam to see if the correlation holds. I may post that data when I get it in a few months. It would be very helpful if some of you guys do that at your schools too and let us know the results.

 

[fahimaz7]

Hyrule,

I have heard of this formula too. The thing is that I don't know where it comes from or what data exists to support it. It is also a very very old formula and there is evidence to suggest that the step 1 has gotten harder in recent years. The nbme exams have certainly been changed to a much harder curve and I think that has happened to the real thing.

I took special care to take data points that were from the last year to capture the difficulty of the most recent tests.

Thanks for your comments. I'm hoping to poll my class after this year's step 1 exam to see if the correlation holds. I may post that data when I get it in a few months. It would be very helpful if some of you guys do that at your schools too and let us know the results.

Good luck getting your class to tell you what their % correct was, or their actual step 1 score is. This stuff is guarded like a recurrent hx of neisseria.

 

[Encephalectomy]

Good luck getting your class to tell you what their % correct was, or their actual step 1 score is. This stuff is guarded like a recurrent hx of neisseria.

ah, but a history of recurrent neisseria means that lots of sharing has been going on.

You could make it anonymous, just throw up a google spreadsheet or something. Would be at least as reliable as anything you find about people's scores on the interwebs

 

[hyrule]

wonder if we could extrapolate based on UW itself if we make the assumption that a 50%ile in UW = median on the USMLE.. which is 220-something.

 

[Encephalectomy]

only problem is that a fair amount of people (myself included) do Uworld twice, which makes the percentiles not particularly worthwhile

 

[fahimaz7]

only problem is that a fair amount of people (myself included) do Uworld twice, which makes the percentiles not particularly worthwhile

Yup. But somehow the averages are still in the 50-60% range.

 

[MossPoh]

only problem is that a fair amount of people (myself included) do Uworld twice, which makes the percentiles not particularly worthwhile

I strongly suspect people put too much stock in what they remember when they do questions a second time. There are some random tricky ones that you remember being tricky or may remember the answer to, but there are many that you answer because you know the concept the second time as well.

I also don't believe as many people go through uworld twice as it seems. I only talked to one person from last year's class (out of over 25 I discussed this with) that actually got through it more than once and that wasn't all the way.

It will be my second time through doing it, but I also haven't looked at the questions or done it for over 2 months when all is said and done and the questions answered were over a longer period than that. I can't even remember quiz questions, so I'm not too worried.