NRMP match algorithm foolproof?

[username456789]

So, for example, a person who ranked program A as 3rd and was ranked by A as 30th, may lose out to a person ranked 40th by A if the person at 40th ranked A as #1 while the person ranked 30th was ranked by their #4 choice as 2nd in their ROL especially if program A has a relatively high number of spots to fill.

This is 100% false, and I can't believe you still don't understand why.

Sorry for any confusion.

No, I really don't think that you are.

---

Members don't see this ad.

 

[dienekes88]

When I said iterations, I was referring to the algorithm. There may be more integrative ways of implementing the algorithm, but a series of iterations is the most intuitive way to do it. I actually wrote a program to simulate the match and it would usually take 100 iterations to reach a stable solution.

God only knows who the NRMP has staffing their phones, I would not rely on them for details about the match algorithm.



I'm trying I figure out what you just said and fighting the urge to quote Billy Madison.

Program A

#30: Applicant 1
#40: Applicant 2

Program X:
#2: Applicant 1

Applicant 1
#3: Program A
#4: Program X

Applicant 2:
#1: Program A

I believe this is the situation you're describing. In NO event would Applicant 1 match to Program X and Applicant 2 match to Program A, which I believe is what you now think will happen. Either they both go to A, 1 goes to A and 2 does not, or neither does.

As I said, I've actually implemented a version of the algorithm and run simulations of it. Take my word on this: you are confused.

I may also be confused, but I was under the impression that the Gale-Shapley Marriage problem (including the house officer/hospital variation) can produce more than one stable solution. This discussion may be overly technical and truly doesn't apply to the vast majority of applicants, though.

The current model: resident-preference over hospital-preference creates a different solution than the converse both of which are stable but optimized in a different way. However, even though a solution is deemed to be stable and generally favored towards resident-preference, it doesn't guarantee that every match is resident optimized. It just guarantees that one party was optimized. This could be my misunderstanding, though.

Moreover, the house officer/hospital variation increases the complexity of the problem such that it doesn't fulfill the simple markets assertions. There exists the possibility in which neither side is optimized though stable matchings are created. The example of 8 applicants and a few programs is a simple market and is a good approximation of the complex NRMP market, but it doesn't completely capture the intricacy of the problem (more so at the margins). For the vast majority of applicants, though, it's all that is necessary to understand the matching algorithm.

Having said all this, the research that is out there states unequivocally that applicants should rank the programs according to their true preference as any sort of "gaming" only affects their match negatively.

Also, Kongakut is wrong.

 

[deleted103644]

I may also be confused, but I was under the impression that the Gale-Shapley Marriage problem (including the house officer/hospital variation) can produce more than one stable solution. This discussion may be overly technical and truly doesn't apply to the vast majority of applicants, though.

I believe the Gale Shapley algorithm is exactly what is used in the main residency match (didn't know its name before your post but that's what I implemented). The solutions are stable in the sense that at least one party in a pair will be less happy with any alternative matching. Optimality is tricky to define, so depending on your definition it probably isn't an optimal solution, but it is stable and unique*.

*ignoring the couples match.

The current model: resident-preference over hospital-preference creates a different solution than the converse both of which are stable but optimized in a different way. However, even though a solution is deemed to be stable and generally favored towards resident-preference, it doesn't guarantee that every match is resident optimized. It just guarantees that one party was optimized. This could be my misunderstanding, though.

Optimal needs a formal definition. It ensures that there is no case where a program would prefer a specific applicant and that applicant would in turn favor that program over the actual outcome.

What's optimal for applicants? That everyone gets as close to their first choice as possible? Or that the most desirable applicants get their pick and the rest get scraps? There are many ways it could be defined (and by defined I mean a formula to calculate a score for the overall match process which can be minimized/maximized).

Moreover, the house officer/hospital variation increases the complexity of the problem such that it doesn't fulfill the simple markets assertions. There exists the possibility in which neither side is optimized though stable matchings are created. The example of 8 applicants and a few programs is a simple market and is a good approximation of the complex NRMP market, but it doesn't completely capture the intricacy of the problem (more so at the margins). For the vast majority of applicants, though, it's all that is necessary to understand the matching algorithm.

Having said all this, the research that is out there states unequivocally that applicants should rank the programs according to their true preference as any sort of "gaming" only affects their match negatively.

Not familiar with house officer variant.

You don't need many examples to find problems, just the right examples.

The biggest problem with the match algorithm is the couples match, since it can undo the stability of the main algorithm. (Using a similar approach to the main algorithm, you can get caught in cycles although I think there may still be a unique solution.) I'm not sure exactly how it is implemented so I can't say much for sure, but I believe the order in which they run couples matters for the outcome so they made a decision to run them last (resulting in slightly worse outcomes for couples).

The only time rank list strategy matters is if you can a) convincingly communicate your preferences to a program and b) the program wants to minimize the final matching position on its rank list. In this scenario, you might want to bump up your rank of a certain program in exchange for them ranking you higher, but it requires a large amount of trust by both parties and a program willing to overlook your utility as a resident in favor of their final matching position's utility as an indicator of their competitiveness. (In other words, take a lower caliber resident now to convince higher caliber residents later of your desirability). Again, this involves a huge amount of trust since both parties are best served if they can lie to the other party convincingly.

Also, Kongakut is wrong.

Indeed.

 

[dienekes88]

I believe the Gale Shapley algorithm is exactly what is used in the main residency match (didn't know its name before your post but that's what I implemented). The solutions are stable in the sense that at least one party in a pair will be less happy with any alternative matching. Optimality is tricky to define, so depending on your definition it probably isn't an optimal solution, but it is stable and unique*.

*ignoring the couples match.

Optimal needs a formal definition. It ensures that there is no case where a program would prefer a specific applicant and that applicant would in turn favor that program over the actual outcome.

What's optimal for applicants? That everyone gets as close to their first choice as possible? Or that the most desirable applicants get their pick and the rest get scraps? There are many ways it could be defined (and by defined I mean a formula to calculate a score for the overall match process which can be minimized/maximized).

Agree 100%. What should be defined as optimal in the case of multiple stable answers? I believe it might be possible to optimize for fewest unmatched, lowest summation of matched ranks, etc. From that, you could develop a fitness score for each unique stable solution, using your correct definition of "stable."

Not familiar with house officer variant.

The standard problem is one of marriage, and all the men rank and propose to all women; on the flip side, all women rank all the men. The house officer/hospital variant is the NRMP match in which not all candidates are able to apply to every program due to limitations of time and money. Moreover, there is the added complexity of the couples match.

As an aside, the only way I could envision the couples match working is by having each pair fill multiple spots within one locale, though I will admit that I'm not even sure how the rank lists are filled out in a couples match.

You don't need many examples to find problems, just the right examples.

The biggest problem with the match algorithm is the couples match, since it can undo the stability of the main algorithm. (Using a similar approach to the main algorithm, you can get caught in cycles although I think there may still be a unique solution.) I'm not sure exactly how it is implemented so I can't say much for sure, but I believe the order in which they run couples matters for the outcome so they made a decision to run them last (resulting in slightly worse outcomes for couples).

Yes.

I think it's not just the couples. I have a pet hypothesis (without adequate CS background to test it) that more than a single unique stable match may arise depending on the order in which candidates begin to be preliminarily matched. If this were my system, I would probably run the algorithm multiple times and compare based on a fitness score as discussed above.

The only time rank list strategy matters is if you can a) convincingly communicate your preferences to a program and b) the program wants to minimize the final matching position on its rank list. In this scenario, you might want to bump up your rank of a certain program in exchange for them ranking you higher, but it requires a large amount of trust by both parties and a program willing to overlook your utility as a resident in favor of their final matching position's utility as an indicator of their competitiveness. (In other words, take a lower caliber resident now to convince higher caliber residents later of your desirability). Again, this involves a huge amount of trust since both parties are best served if they can lie to the other party convincingly.

Indeed.

This goes down the road of game theory and the prisoner's dilemma. In a single iteration of the match, e.g. applicants not planning on fellowship, then the impetus to be truthful is entirely dependent on the character of the person. However, in the case of repetitive games, e.g. every single residency program, there is a very large penalty for being dishonest: the reputation of lying to applicants. The applicant who will likely go on to fellowship is also compelled to be honest, because telling multiple programs that they are "ranked #1" can result in a bad reputation especially in the case of a small field. The solution that both sides have figured out is to be effusive (calls, e-mails, letters, etc.) without making promises... resulting in everyone thinking that the other is lying. Then I suppose the question is, how do programs react if you don't act effusive, i.e. no follow ups?

Generally, I think we're on the same page.

 

[deleted103644]

Agree 100%. What should be defined as optimal in the case of multiple stable answers? I believe it might be possible to optimize for fewest unmatched, lowest summation of matched ranks, etc. From that, you could develop a fitness score for each unique stable solution, using your correct definition of "stable."



The standard problem is one of marriage, and all the men rank and propose to all women; on the flip side, all women rank all the men. The house officer/hospital variant is the NRMP match in which not all candidates are able to apply to every program due to limitations of time and money. Moreover, there is the added complexity of the couples match.

As an aside, the only way I could envision the couples match working is by having each pair fill multiple spots within one locale, though I will admit that I'm not even sure how the rank lists are filled out in a couples match.



Yes.

I think it's not just the couples. I have a pet hypothesis (without adequate CS background to test it) that more than a single unique stable match may arise depending on the order in which candidates begin to be preliminarily matched. If this were my system, I would probably run the algorithm multiple times and compare based on a fitness score as discussed above.



This goes down the road of game theory and the prisoner's dilemma. In a single iteration of the match, e.g. applicants not planning on fellowship, then the impetus to be truthful is entirely dependent on the character of the person. However, in the case of repetitive games, e.g. every single residency program, there is a very large penalty for being dishonest: the reputation of lying to applicants. The applicant who will likely go on to fellowship is also compelled to be honest, because telling multiple programs that they are "ranked #1" can result in a bad reputation especially in the case of a small field. The solution that both sides have figured out is to be effusive (calls, e-mails, letters, etc.) without making promises... resulting in everyone thinking that the other is lying. Then I suppose the question is, how do programs react if you don't act effusive, i.e. no follow ups?

Generally, I think we're on the same page.

Yup, we're on the same page, but you're wrong about the simple match algorithm (ie no couples) having the potential for getting caught in a cycle and having no unique solution.

In fact, the way I first thought the couples match worked, there would be no cycles or instability there either. (I thought it was the same as primary/supplemental lists where you first match one partner then the other partner using a list based upon the first match. This would work, but places the interests of the first person over the second and the couple overall).

With the couples match, you can set up a scenario with two couples where they keep switching places as you go through the match algorithm, you can't do this with the individual match.

I haven't tried very hard to prove this or see if anyone else already has, but it "feels" right and I've tried running many simulations with the same rank lists and going through the applicants in random orders and 1) there is always a unique solution and 2) for a given set of lists, it's always the same.

I ran these simulations with thousands of applicants and programs and it was still the case, so I think my intuition is most likely right.

 

[deleted103644]

For better or worse, I did not contact any programs post-interview (not even thank you notes) except my #1 advanced program.

I matched to my #2 advanced program and my #1 prelim.

We'd all be better off if we disarmed and gave up the BS post-interview communication, but I'll admit what I did was a gamble. We'd be better off if no one sent thank you notes, etc, the problem is you don't want to be seen as the odd man out.

 

[NotAProgDirector]

I haven't tried very hard to prove this or see if anyone else already has, but it "feels" right and I've tried running many simulations with the same rank lists and going through the applicants in random orders and 1) there is always a unique solution and 2) for a given set of lists, it's always the same.

No need to wonder about it. It's been proven: http://en.wikipedia.org/wiki/Stable_marriage_problem

Note that the match without Prelims and without Couples is completely mathematically identical to the Stable Marriage Problem. Since in the match not all applicants are interviewed by all programs, and the number of spots and applicants are imbalanced, there will be unmatched applicants and unfilled programs at the end. But, without prelims and couples, there will always be one answer and it will be the same.

Note the section on "Optimality of the solution". This clear shows how the answer could vary, depending on how you define the "optimal" solution. Since the match uses an applicant driven process, in the example on the wikipedia page, if ABC were residents and XYZ were the programs, the solution would be the one were ABC get their first choice and XYZ get their third choice.

Adding couples changes the problem in a dramatic way. The problem now becomes "NP Complete" instead, which means that there is an answer, but you often can't come up with a simple algorithm that runs in a reasonable amount of time to find it. Here's why:

Think about what happens when the match is run. Let's say that your list is being processed, somewhere in the middle of the batch (i.e. half of the people have already been processed and have been "tentitively matched". The system tries to put you in your first choice. Turns out that your first choice is full, and the last person matched there is ranked #20. Sadly, you're #21, so you're out of luck. On to your next rank. Without couples, there would be no way in the future for a spot to open for you -- the only way someone could get bumped out would be if they were ranked higher than #20, and bump the person at #20 out (which wouldn't help you at all).

But, let's say that later on in the match process someone ranked at #5 is couple's matching with someone in radiology, and that person get's bumped out of their tentitive rad's match. Because of the couple's match, both spots are freed up. So, all of a sudden there now is a spot open at your top rank which you should get -- but you've already been processed. The standard algorithm only looks down your match list, since (without couples) it's impossible for something up your rank list to become open.

So, now you'd have to run through all of the applicants and tentitively match them, and then run through them all again to see if spots have opened higher on rank lists (and of course everytime you change something, you have to start all over again to see if other matches need to move). This is why couples make the match much more complicated. It also does create a situation where the order that you process applicants might result in a different answer. In order to minimize this, I expect the NRMP processes the couples at the end -- that minimizes the amount of disruptions. Note that being run last doesn't in any way hurt your chances to match -- you'll simply bump someone else out.

Note that this was studied, and the effect of running the match in a "different order" made a minimal difference -- using 5 years of data with 12000+ matches, the largest difference was somewhere around 4-5 spots, and in most cases was lower than that.

 

[drizzt3117]

Here's the paper on it and a detailed explanation if you are willing to read 49 pages.

http://www.bepress.com/cgi/viewcontent.cgi?article=1016&context=wilson

 

[username456789]

For better or worse, I did not contact any programs post-interview (not even thank you notes) except my #1 advanced program.

I matched to my #2 advanced program and my #1 prelim.

We'd all be better off if we disarmed and gave up the BS post-interview communication, but I'll admit what I did was a gamble. We'd be better off if no one sent thank you notes, etc, the problem is you don't want to be seen as the odd man out.

I sent zero post-interview communication . . . no thank you notes, no thank you emails, nothing. I had originally planned to send a few real thank you notes via snail mail (which is why I didn't fire off a bunch of email thank yous) and then got lazy I guess, and by that point it seemed kinda silly to send really late email thank yous. So I just never sent anything.

I got my #1 Derm program and my #1 TY program.

Thankfully, things worked out just fine for me.

 

[dienekes88]

First, I want to say that I appreciate all the well thought out responses. Very interesting discussion. I always enjoyed math.

No need to wonder about it. It's been proven: http://en.wikipedia.org/wiki/Stable_marriage_problem

Note that the match without Prelims and without Couples is completely mathematically identical to the Stable Marriage Problem. Since in the match not all applicants are interviewed by all programs, and the number of spots and applicants are imbalanced, there will be unmatched applicants and unfilled programs at the end. But, without prelims and couples, there will always be one answer and it will be the same.

I can see the unmatched/unfilled programs being the same regardless of order in the Match without prelims or couples.

However:

Roth AE, Peranson E. The effects of the change in the NRMP Matching Algorithm. JAMA 1997; 278: 729 - 732.

1. At least one stable matching can always be found, no matter what Rank Order Lists are submitted.
2. The set of stable matchings always contains a "program optimal" stable matching, and an "applicant optimal" stable matching.
3. The program optimal stable matching is produced by a "program proposing" algorithm, in which residency programs make offers to applicants, starting from the top of each program's rank order list, and allowing applicants to hold at any point in the algorithm the most preferred offer among those so far received. The applicant optimal stable matching is produced by an analogous "applicant proposing" algorithm.
4. The same applicants are unmatched and the same positions are unfilled at every stable matching.
5. When the applicant-proposing algorithm is used, no applicant can possibly improve his match by submitting a Rank Order List (ROL) different from his true preferences. No parallel assertion can be made about residency programs offering more than one position.

When the NRMP match variations are present, none of these five assertions are correct in all cases.Markets with match variations can be constructed for which no stable matchings exist or no optimal stable matchings for either side exist (even when stable matchings exist); therefore, no algorithm always produces an optimal stable matching for either side of the market. Furthermore, different stable matchings may have different applicants unmatched and positions unfilled, and no algorithm always guarantees applicants that stating their true

preferences is an optimal strategy.

This is from the "background" section of the paper. I may be misunderstanding this, though. I've looked through one of the papers that is quoted, and couples matching makes it quite a complex problem:

Aldershof B, Carducci OM. Stable matchings with couples. Discrete applied mathematics 1996; 68: 203 - 207.

3. Conclusion
Allowing couples to express preferences over pairs of positions has improved the usefulness of the NRMP to couples at the cost of some certainty in the results. Particularly troublesome is the possibility of different stable matchings yielding different sets of positions filled and of students assigned.

Note the section on "Optimality of the solution". This clear shows how the answer could vary, depending on how you define the "optimal" solution. Since the match uses an applicant driven process, in the example on the wikipedia page, if ABC were residents and XYZ were the programs, the solution would be the one were ABC get their first choice and XYZ get their third choice.

Adding couples changes the problem in a dramatic way. The problem now becomes "NP Complete" instead, which means that there is an answer, but you often can't come up with a simple algorithm that runs in a reasonable amount of time to find it. Here's why:

Think about what happens when the match is run. Let's say that your list is being processed, somewhere in the middle of the batch (i.e. half of the people have already been processed and have been "tentitively matched". The system tries to put you in your first choice. Turns out that your first choice is full, and the last person matched there is ranked #20. Sadly, you're #21, so you're out of luck. On to your next rank. Without couples, there would be no way in the future for a spot to open for you -- the only way someone could get bumped out would be if they were ranked higher than #20, and bump the person at #20 out (which wouldn't help you at all).

But, let's say that later on in the match process someone ranked at #5 is couple's matching with someone in radiology, and that person get's bumped out of their tentitive rad's match. Because of the couple's match, both spots are freed up. So, all of a sudden there now is a spot open at your top rank which you should get -- but you've already been processed. The standard algorithm only looks down your match list, since (without couples) it's impossible for something up your rank list to become open.

So, now you'd have to run through all of the applicants and tentitively match them, and then run through them all again to see if spots have opened higher on rank lists (and of course everytime you change something, you have to start all over again to see if other matches need to move). This is why couples make the match much more complicated. It also does create a situation where the order that you process applicants might result in a different answer. In order to minimize this, I expect the NRMP processes the couples at the end -- that minimizes the amount of disruptions. Note that being run last doesn't in any way hurt your chances to match -- you'll simply bump someone else out.

Note that this was studied, and the effect of running the match in a "different order" made a minimal difference -- using 5 years of data with 12000+ matches, the largest difference was somewhere around 4-5 spots, and in most cases was lower than that.

Where was that study published? I'd like to read it. I honestly have nothing better to do with the rest of my year: I'm just preparing manuscripts and exercising...

Here's the paper on it and a detailed explanation if you are willing to read 49 pages.

http://www.bepress.com/cgi/viewcontent.cgi?article=1016&context=wilson

I'll have to read this.

 

[NotAProgDirector]

Where was that study published? I'd like to read it. I honestly have nothing better to do with the rest of my year: I'm just preparing manuscripts and exercising...

It's the same article posted above. I took some liberties with quoting the numbers, but as you'll see the effect was very small (0.1% of applicants)

 

[dienekes88]

It's the same article posted above. I took some liberties with quoting the numbers, but as you'll see the effect was very small (0.1% of applicants)

Great. Thanks!

 

[shivasHeroLike]

I strongly feel that there should be more transparency to this process. It should be known where you ranked on the program's list--since they have access to ours.

 

[deleted103644]

I strongly feel that there should be more transparency to this process. It should be known where you ranked on the program's list--since they have access to ours.

They don't have access to ours. They only know where we match, not how we ranked them. (Assuming "they" = programs.)

 

[shivasHeroLike]

They don't have access to ours. They only know where we match, not how we ranked them. (Assuming "they" = programs.)

oops, you're right. pardon the ignorance

 

[neurosurg777]

This thread now has two different discussions in it, sorry for that.



There is another thread about this recently. The bolded part is probably incorrect in most cases. Some fields are very competitive. In those, you have to have step scores >250 (or insert whatever high cutoff you want) to get an interview. In that case, scores may really not matter anymore -- as a resident with a 270 is not really any better than one with a 250. Both did very well on the exam, and both are likely to ultimately pass their specialty board exams. However, in lower competitive fields there is likely to be a larger spread in board scores. In that case, PD's are very likely to rank someone with better scores (or, perhaps better described as "a better performance in medical school") than those with lower scores. Personality is great, but I want someone who will pass the exams at the end (and knows their basic stuff coming in). I'd rather have a resident who did well in medical school but is a bit of a "dud" personality-wise, than one who has a great personality but didn't do well in medical school. Each PD will set the "weight" of these criteria individually, so it's hard to make sweeping generalizations. But I think that the statement that "everyone is equal at the interview" is mostly untrue.



Sounds like you're doing the right things. I doubt that being a paramedic will make much of a difference, I would only do that if you want to and in addition to what you're already doing. What you really want is some clinical experience at your major teaching hospital -- continue your research, and beg/plead for some way to get some clinical experience there.



Nothing short of "I promise you will match with us" is a guarantee of matching, and even that really isn't (since it's only as good as the word of the PD). I told exactly one student this year that they were guaranteed to match with us -- their spouse is here with us, and I basically told them not to bother interviewing anywhere that would be lower than us. Should they match somewhere else (because they ranked that program higher than us), no hard feelings. I ranked them #1.

Can you contact this program? Sure, and if you had a good relationship with them you certainly could. Unfortunately, you're likely going to be told "It was a really competitive year, and if we had gone down a bit further on our rank list you would have matched with us", which isn't really going to help you. But there isn't much downside to asking, as long as you're polite and especially if you had a relationship with them more than just interviewing.

Sorry for creating two conversations in the thread. Thank you for all your advice. If I have any other questions I will PM you so that we don't have more than one conversation in the thread. Thanks again.

 

[Law2Doc]

I strongly feel that there should be more transparency to this process. It should be known where you ranked on the program's list--since they have access to ours.

It's a real bad idea. It becomes an employer employee problem if the employee knows they weren't really wanted, or that the other employees were better regarded from day one, etc. You are looking at it from an applicant perspective, but you need to look at it as if you were a resident, who finds out that all of your co residents were ranked to match while you clearly were thrown into the list as an afterthought. The way it exists now, everyone can save face and the program can just say "what a great class -- we are excited with who we got!" if it were transparent, nobody could make such grandiose statements and there would always be suspicion.

 

[Kongakut]

This is 100% false, and I can't believe you still don't understand why. No, I really don't think that you are.

.

This was from talking with the NRMP. It was the scenario she gave me. I didn't make it up.

Please, don't be rude. I feel very bad about it. The info I gave was from talking to a PD about it at length which was how he thought the algorithm worked. I trusted him, period, and didn't check the facts. I won't do that again.

I said my peace and then left the discussion so I wouldn't cause any more confusion, but you had to go and put in the last word and make me feel even worse. Nice.

 

[b-real]

.

This was from talking with the NRMP. It was the scenario she gave me. I didn't make it up.

Please, don't be rude. I feel very bad about it. The info I gave was from talking to a PD about it at length which was how he thought the algorithm worked. I trusted him, period, and didn't check the facts. I won't do that again.

I said my peace and then left the discussion so I wouldn't cause any more confusion, but you had to go and put in the last word and make me feel even worse. Nice.

When you're continually wrong, expect to be corrected. And be a big girl about it, rather than being frustratingly obstinate.

 

[Kongakut]

When you're continually wrong, expect to be corrected. And be a big girl about it, rather than being frustratingly obstinate.

 

[Valvool]

Again, I'm sorry you are in the situation you are in and I wish you the best of luck but stop blaming the system and look at yourself. How do you really look on paper? How did you really interview. In my experience the people who are so sure that they interviewed "definitely better than the people that were there on the same day as me" are either cocky, overbearing or have some other personality flaw that turns people off.

I'm curious as well. I was told at numerous places "You will definitely match at one of your top two choices" by the people interviewing me. On paper I am very strong--higher than avg step scores, lots of honors in third year, etc. So my guess is it was something that was off during interviews. I can't say that people always love me, but historically I have been told that I interview well. I was even told that once this past October following an interview that was four on one--a panel interview where four people interviewed me at once. We left the room laughing and one of them smiled at me and said, "You interview very well."

I ended up with nine interviews, I guess I should have taken a few more. But still, I was told I was a top candidate and I did not expect to have to scramble. Because there were no unfilled EM spots, I had to scramble into a specialty that I had not planned on doing. I have wondered what happened, why I did not match, but I haven't felt the need to expend any energy trying to find the answers because I don't plan on reapplying and interviewing again. This process has turned me off; it did not work for me for some reason and it was a huge expense and time sink. If that were not the case I might try again, but that is the case. My plan is to stick with anesthesiology; I feel I can be happy doing any number of things and I hope this works for me.

 

[shivasHeroLike]

It's a real bad idea. It becomes an employer employee problem if the employee knows they weren't really wanted, or that the other employees were better regarded from day one, etc. You are looking at it from an applicant perspective, but you need to look at it as if you were a resident, who finds out that all of your co residents were ranked to match while you clearly were thrown into the list as an afterthought. The way it exists now, everyone can save face and the program can just say "what a great class -- we are excited with who we got!" if it were transparent, nobody could make such grandiose statements and there would always be suspicion.

yeah i see what you're saying.
Why not send a list to each applicant where they ranked on the progam's list that they applied to, except from the program they got in. That way you don't know how you did in your own program which maintains the relationship and also gives you perspective on how well you did with other programs. I can see this being helpful for those that didn't Match that needed some Answers.

 

[deleted103644]

yeah i see what you're saying.
Why not send a list to each applicant where they ranked on the progam's list that they applied to, except from the program they got in. That way you don't know how you did in your own program which maintains the relationship and also gives you perspective on how well you did with other programs. I can see this being helpful for those that didn't Match that needed some Answers.

Because that information would be shared and compiled and used to rank programs, leading programs to rank not in order of preference but in order of most likely to attend to game the pseudo-rankings.

You don't need that information, it would just encourage shenanigans. It might be helpful to medical schools, but only if the information could be kept confidential from any associated residency programs (unlikely).

 

[Chakrabs]

yeah i see what you're saying.
Why not send a list to each applicant where they ranked on the progam's list that they applied to, except from the program they got in. That way you don't know how you did in your own program which maintains the relationship and also gives you perspective on how well you did with other programs. I can see this being helpful for those that didn't Match that needed some Answers.

And this information helps you how exactly? You match and you find out you were in the top-10 in every program you applied. Great! You match, and you see that you're 40 or worse everywhere, well not so great. Were you ranked that low at the place you matched? Are you the black sheep that they're not all too thrilled to be taking? Not so good.

Now lets say you go unmatched. No matter where you were ranked, how helpful is that information going to be for you. Without any feedback from the programs you have no idea if its because you interviewed poorly, didnt have a good applications, board score was inadequate, or just werent academic enough and had applied only to academic programs and should have thrown in a few community programs. It could be a combination of the above, and you'd have no clue if you had a serious deficit (fixable or not) or were a perfectly fine candidate and just fell through the cracks. As mentioned above by AProgDirector, you're unlikely to get the kind of necessary in-depth feedback beyond generalizations.

TL;DR - as nice as it would be to know where you were ranked, its not going to be much help.

 

[Sol Rosenberg]

I strongly feel that there should be more transparency to this process. It should be known where you ranked on the program's list--since they have access to ours.

I don't agree with this for the reasons mentioned by previous posters, but it would be nice to have access to the master list (that PDs have,) and/or find out who matched at the places you interviewed. My only reason for this is curiosity. I met people along the trail and wonder where a bunch of them ended up (but I don't know them well enough to have their email, phone number, etc.) Similarly, there are some programs that I ranked above the program where I matched and I'd like to know who ended up matching there. Yes, there are always informal internet-based lists, but these are usually incomplete. I don't see why it is a big deal to provide this to applicants since, if I'm not mistaken, all PDs get the master list for their specialty anyway.

 

[15MD]

I don't agree with this for the reasons mentioned by previous posters, but it would be nice to have access to the master list (that PDs have,) and/or find out who matched at the places you interviewed. My only reason for this is curiosity. I met people along the trail and wonder where a bunch of them ended up (but I don't know them well enough to have their email, phone number, etc.) Similarly, there are some programs that I ranked above the program where I matched and I'd like to know who ended up matching there. Yes, there are always informal internet-based lists, but these are usually incomplete. I don't see why it is a big deal to provide this to applicants since, if I'm not mistaken, all PDs get the master list for their specialty anyway.

I see both sides of this argument but isn't it pretty easy to look up a program's new residents in a few months when they update their website?

 

[Ronin786]

I see both sides of this argument but isn't it pretty easy to look up a program's new residents in a few months when they update their website?

Lots of programs dont put their resident lists online.

 

[Chakrabs]

I don't think PDs get a masterlist for their specialties. Its my understanding that they only see where the people on their rank list matched. For some of the smaller specialties - neurosurgery, plastics, urology etc. there are master-lists that are compiled on specialty forums like here on Student doctor or a Urologymatch.com.

Here are some:
Derm
Plastics
Urology
Neurosurgery

 

[Sol Rosenberg]

I don't think PDs get a masterlist for their specialties. Its my understanding that they only see where the people on their rank list matched. For some of the smaller specialties - neurosurgery, plastics, urology etc. there are master-lists that are compiled on specialty forums like here on Student doctor or a Urologymatch.com.

Here are some:
Derm
Plastics
Urology
Neurosurgery

Urology, for sure, sends out a master list to PDs (I've seen one.) Not 100% sure about the NRMP match, but I just figured that it was the same. Like I said in my original post here, "Yes, there are always informal internet-based lists, but these are usually incomplete," which is why it would be useful to have the master list that PDs get.

 

[deleted103644]

Urology, for sure, sends out a master list to PDs (I've seen one.) Not 100% sure about the NRMP match, but I just figured that it was the same. Like I said in my original post here, "Yes, there are always informal internet-based lists, but these are usually incomplete," which is why it would be useful to have the master list that PDs get.

The NRMP does not give out a master list. PDs only find out where people on their rank lists matched.