Average USMLE scores

[wanabdoc]

I cannot find any info on this site for the following...

What is the average USMLE scores from caribbean schools (preferribly the big 4) as compared to US schools. Is it the same, higher, lower? I am talking about students who match into a residency.

[sammy123]

answer: lower

[wanabdoc]

Originally Posted by sammy123

answer: lower

You say lower, does that mean slightly lower or a lot lower? (I think the average was around 210-215 according to the NRMP.

[BrendaB_MD]

To my knowledge, none of the schools publish their average scores; however, some of them publish their pass rates. If you assume the scores have a normal distribution, you can estimate the mean score from the pass rate. Scores are negatively skewed so the estimate is just an approximation, but it will get you in the ballpark.

For example, we can start with the numbers for US students to test the calculation. For US students, the mean is 216, the sd is 16 and the pass mark is 185. Then, if the scores are normally distributed, we would expect a 97% pass rate which is not too far off (I believe the pass rate for US students is something like 95%). So, it looks like we need to adjust the calculation by 2% because the scores are not perfectly normal (a pretty minor adjustment).

Pass rate* predicted mean USMLE
97 216
94 210
89 205
82 200
--------------------------------------------
* unadjusted (you can adjust by 2% or use your own fudge factor)

You can do the calculations yourself using the NORMDIST function in Excel.

With respect to the adjustment for skewness, the calculation is probably much better if you calculate the difference. For example, a difference in pass rate of 3% corresponds roughly to a difference in mean USMLE score of about 6. Thus, we might expect that SGU scores are, on average, about 6 points less than the average US scores (possibly better than some of the lower US schools).

[stateofequilibrium]

I would be very interested in hearing the average scores from various carib schools if they ever release them. Though just from personal knowledge of people who took it, and the people they know: You'll hear of many people breaking good scores in the 90+ range. But you will also hear of quite a number of people failing or barely passing by the skin of their teeth or just happy they broke 200.

[wanabdoc]

I was looking through the NRMP results and it seems that family medicine independent applicants had an average of 199 for matched applicants, and 220 for matched internal medicine, 212 for OB/GYN, and 211 for pediatrics. Through my convoluted logic, does this mean the average is around 208-210 at carribbean schools? Is this about right for those of you with an unofficial idea?

[stephew]

sgu had done it in the past at least once. I think. anyway worry more about the pass rates. that is a better indicator od the school. the median is more an indicator of the individual (rough guide only)

[wanabdoc]

Originally Posted by stephew

anyway worry more about the pass rates. that is a better indicator od the school.

The school can't make the students pass or become motivated, it may be a better indicator or the schools student selection process though.

[Narcan]

After taking Step 1 and 2. I think it has more to do with your study habits, than anything the school does.

[BrendaB_MD]

Originally Posted by wanabdoc

The school can't make the students pass or become motivated, it may be a better indicator or the schools student selection process though.

This is a longstanding question on valueMD:

When comparing two schools, can you conclude that teaching effectiveness is better at one if the USMLE pass rate is higher?

The short answer is, not necessarily.

If a statistician were to try and determine whether attending a particular schools has an effect on USMLE, they would proceed as follows:

You would create a statistical model that incorporates the known factors that effect USMLE scores along with the study variable that you wish to test, that is, the supposed school effect. The major factors that are known to effect USMLE are MCAT scores and gpa. To disentangle the effects of MCAT and GPA from school, you would create a regression model that includes all the known effectors along with the new one (school):

USMLE = b0 + b1*MCAT + b2*gpa +b3*school

You would code the school variable as a binary variable. For example, if you were comparing SGU and Ross you would code as follows
school = 1 if person attends SGU
school = 0 if person attends Ross

The coefficient, b3, would measure the effect of attending SGU (i.e. the difference in USMLE score) relative to Ross after adjusting for MCAT and USMLE.** You would test the null hypothesis:
H0: b3 = 0 (doesn't matter what school you attend) vs
H1: b3 <> 0 (does matter).

Because MCAT and GPA are correlated with school (sgu is more selective) it is impossible to determine whether one school has a postive impact on USMLE relative to another without adjusting for those factors. Thus, it is incorrect to make a statement such as, "The pass rate at school X is 5% higher than school Y. Thus, you are likely to get a higher USMLE score if you attend school X."

While we know that MCAT and GPA are predictors of USMLE it is harder to make a strong case for a school effect. The curricula and mode of delivery is virtually identical at each school. They all use the same textbooks. It is hard to understand how sitting in a 300 person lecture hall on Dominica differs from sitting in a 300 person lecture hall in Grenada. And faculty seem to drift around between schools. Thus, I am somewhat skeptical that one would find a school effect even if it were possible to get the data.

The only case where you can tell something conclusive from avg data would be if you had a school with really low entrance stats that had a higher USMLE pass rate, then you could say that this suggests that the school made a difference. For example, if St. James (low entrance stats) had a higher USMLE pass rate than SGU, we could pretty safely conclude that the school had an effect. But, as it is, pass rates correlate with entrance stats so it is impossible to separate the effects of admission selectivity from the effect of the school.

So, should you just choose the cheapest school? Absolutely not. While the USMLE is important, there is more to choosing a medical school than maximizing the USMLE/dollar ratio.

**As an aside, you could construct a similar model using pass rate instead of USMLE as the dependent variable (except you would have to use a logistic regression model); however, the same general reasoning applies.