reasonable to avoid fellowship?

[radrad5]

hi all

starting to re-think fellowship route. may just pursue tele for these reasons, not ranked in order

1. i'm absurdly over being a trainee
2. opportunity cost of fellowship year
3. make hay while sun shines - job market is red hot rn - reimbursement will probably continue to decline
4. i'm bullish on AI and think our lifespan as human rads is probably in the 10-20 year range, so even just one extra year of attending salary may be more impactful than one would think
5. flexibility of tele - i can live wherever
6. i hate procedures
7. i enjoy reading ED / acute / inpt studies more than outpt exams

is this route too short-sighted?

are there any non-fellowship trained rad that can comment on their experience?

any rad who went straight to tele can comment on their experience?

thanks much!

---

Members don't see this ad.

 

[Mo991]

Not sure about all the reasons, but if AI is on top of your list, having a fellowship will help you keep a Job in the future

 

[maxxor]

Yes it’s short sighted.

The market won’t always be good. When it’s bad, you’re competing with the people who did fellowships.

Do a non acgme fellowship and moonlight a lot. That will make the year suck less.

 

[2brads]

Do the fellowship. The future is looking more and more like subspecialized reads. Would you rather do fellowship now or when the market turns and you’re older and less used to the academic environment?

 

[MadRadLad]

Why not just pick a chill fellowship where you can moonlight as maxxor says? Without a fellowship you are bottom of the totem pole if the market turns again.

If you really don't care you can do tele but you will be somehow even more bottom of the totem pole and take a paycut on the $/RVU side of things.

Also you are unlikely to be hired by competitive practices without a fellowship, although this can vary by region. Clinicians are pushing more and more for subspecialist reads. It's a marketing thing.

 

[guytakingboards]

1. i'm absurdly over being a trainee

Understandable but there are a few considerations to be aware about. As a fellow, you are often treated by attendings with far more respect and collegiality than as a resident. A lot of that resident angst you really don't deal with as a fellow. I enjoyed fellowship year probably the most of any year of my radiology career. It was super chill. Which leads me to my next point:

First year attendinghood is no walk in the park. As chill fellowship was, being a newbie attending was really unpleasant at times. That adjustment to finally signing cases, building speed, etc... And that's beyond troubles of a joining a big, onsite PP like it sounds like you won't be doing. I had to drive all over town, gladhand and kiss ass like no one's business and go back to reading all this **** I avoided during fellowship year. I personally wouldn't recommend people be in a rush to get to year 1 attendinghood.

1. opportunity cost of fellowship year
2. make hay while sun shines - job market is red hot rn - reimbursement will probably continue to decline
3. i'm bullish on AI and think our lifespan as human rads is probably in the 10-20 year range, so even just one extra year of attending salary may be more impactful than one would think

As many people have said, the opportunity cost can be fairly minimized with some external moonlighting. The job market is red hot now, but many groups would just as well hire you now (while a resident) for two years later. You could get the job now and go do fellowship anyway.

I wouldn't price in reimbursement declines into your math; that just is what it is. Same thing with AI. No one can predict with any certainty when AI will start costing radiologists jobs.

1. flexibility of tele - i can live wherever
2. i hate procedures

I think this will limit your career earnings as much as anything. The better tele gigs will get you up in the range of low $30's/wRVU. The crappy ones will be mid-high $20's/wRVU. Flip side, a PP group will solid contracts can be anywhere from $40's-50's/wRVU. A buddy of mine signed for a small PP in the Midwest to make in $60's-$70's/wRVU. You will give up a considerable amount in your career going the pure tele path.

1. i enjoy reading ED / acute / inpt studies more than outpt exams

Most young rads coming out of training prefer more acute stuff. It gets old. I'm a few years out of training and I'm starting to find the call case mix pretty mind numbing. Slogging through a ton of negatives with the occasional positive stroke or appy.

I'd say probably >80% of the people in my section with more than 20 years in radiology have switched to/lean heavy OP preference now. I didn't understand it at the beginning, but it makes alot of sense to me now. Chill days.... are nice. A "miss" on an outpatient degen spine MRI usually doesn't cause dire consequences. Flip side: when you pound high-acuity cases and you miss something, you open yourself up to a lot of liability. Even beyond liability, the first couple of misses I had where the patient had a bad outcome, I took that really hard.

Side point: I derive a lot of professional satisfaction from being fellowship trained and good at my sub-specialty.

Basically, go to fellowship.

 

[pathogeN7]

Is it possible in Radiology to do a few years as an attending, then go back for a fellowship, as is the case in some other specialties?

 

[2brads]

It’s possible. But it’s not easy going from attending salary to fellowship salary. Hours and vacay is also probably going to be worse. Plus, you’ve presumably started on some sort of partnership track that will be derailed. And then, if you have settled down, you may be geographically constrained or be living away from home for a year.

So yeah, you can do it. Why the hell you’d want to do it is beyond me though.

 

[SirGecko]

I’m milmed so hardly any of us get to go to fellowship straight out of the gate. I have known several radiologist without fellowship training who got out and didn’t have trouble finding a job but they had been practicing for years before separating from the military. If you have a long proven track record fellowship is less essential but it offers a bit more flexibility in future jobs than going straight to just tell would. Most of the younger people I know that are getting out choose to do fellowship. (It’s often one of the reasons they get out actually)

I owe the Navy a ton of time so my ability to do fellowship is limited by the Navy but if they start offering more spots I would likely apply. If I still haven’t been to fellowship after practicing for 12 years and I’m retiring from the Navy? Might be a different story because I could be trying for more of a part time gig. Who knows?

Long story short: I you should probably go to fellowship rather than pigeon hole yourself into exclusive tele. It’s not the end of the world to not have fellowship but you need experience to make up for it.

 

[vm26]

hi all

starting to re-think fellowship route. may just pursue tele for these reasons, not ranked in order

1. i'm absurdly over being a trainee
2. opportunity cost of fellowship year
3. make hay while sun shines - job market is red hot rn - reimbursement will probably continue to decline
4. i'm bullish on AI and think our lifespan as human rads is probably in the 10-20 year range, so even just one extra year of attending salary may be more impactful than one would think
5. flexibility of tele - i can live wherever
6. i hate procedures
7. i enjoy reading ED / acute / inpt studies more than outpt exams

is this route too short-sighted?

are there any non-fellowship trained rad that can comment on their experience?

any rad who went straight to tele can comment on their experience?

thanks much!

Great post/discussion...Current job market is pretty unique (disclaimer I have only been out of training for about 10 years)...shortage of rads, demand for sub-specialization yet need for general rad skills, dire need for overnight rads coverage, private equity infiltration, decreasing reimbursements, increasing volume/demand/ and case complexity of imaging 24/7, midlevel encroachment, AI, inflation and increased cost of just about everything.

With this said, a part of me what's to say take the money ASAP if you can, while a part of me would encourage you to do a fellowship. Aside from the subspecialty title, you will learn a tremendous amount.

Also Im not sure how much mentorship you would get going straight into tele-rads. I routinely encounter stuff that I am baffled with and having a colleagues available for a consult is extremely important

 

[MouseChair]

AI? Reading a study the other night with a complex trauma, patient with an incidental mass, post op changes in bowel, abscess, RP bleeding etc... computers cannot even identify squiggly ECGs with enough accuracy let alone all this

but otherwise I think I agree... 12 more rotations in any given subspecialty is it really going to make you *that* much better of a radiologist, especially if you wanna do general? unfortunately I feel like we are all pretty risk averse as a group but I think you could get away with doing no fellowship right now, and can attending pay quicker, partnership earlier etc....but I am a resident and don't know market dynamics like others in this thread probably.

 

[MadRadLad]

AI? Reading a study the other night with a complex trauma, patient with an incidental mass, post op changes in bowel, abscess, RP bleeding etc... computers cannot even identify squiggly ECGs with enough accuracy let alone all this

but otherwise I think I agree... 12 more rotations in any given subspecialty is it really going to make you *that* much better of a radiologist, especially if you wanna do general? unfortunately I feel like we are all pretty risk averse as a group but I think you could get away with doing no fellowship right now, and can attending pay quicker, partnership earlier etc....but I am a resident and don't know market dynamics like others in this thread probably.

It's not hard to imagine 12 months in a given subspecialty would make you much better, especially with exposure to more complex cases. You do probably 4-6 months of MSK or neuro total in most residencies unless you do a minifellowship. For some jobs, knowing the ins and outs of fmri or complex MSK oncology are probably not that important but for others they are, which is where the extra training helps.

You can definitely get away without fellowship as long as you understand that the pool of available jobs, while still big, is not what it is is for fellowship trained candidates and if the market dips again you look much worse on paper even if you are a very good radiologist, and you might end up accepting worse terms to save a year overall. You can still moonlight for attending pay in many fellowships which eases the burden.

 

[Neopolymath]

Fellows have great options at my program. Treated very well. Decent lifestyle or worse lifestyle with a great compensatory paybump if it's important to you. Chance to learn something fully like neuro. Be eligible for good jobs in good cities that still have interesting cases. Can be the go-to person for XYZ subject in practice making you a more valuable part of the team.

Seems like a no-brainer but I'm not a burnt out R4.

 

[guytakingboards]

AI? Reading a study the other night with a complex trauma, patient with an incidental mass, post op changes in bowel, abscess, RP bleeding etc... computers cannot even identify squiggly ECGs with enough accuracy let alone all this

but otherwise I think I agree... 12 more rotations in any given subspecialty is it really going to make you *that* much better of a radiologist, especially if you wanna do general? unfortunately I feel like we are all pretty risk averse as a group but I think you could get away with doing no fellowship right now, and can attending pay quicker, partnership earlier etc....but I am a resident and don't know market dynamics like others in this thread probably.

Yes. Fellowship can make you THAT much better of a radiologist. It takes one complex area of radiology and makes it routine as a CT abd/pelv.

As a neurorad, I have as much diagnostic certainty and report quality dictating out a de novo GBM as I do an acute appy. (had one last night on call). That makes ME more comfortable dictating those studies and obviously the clinician reading my report.

Having that niche makes a big difference. My group has an internal deep nights team and they all have different fellowship training. You can always tell when someone is dictating outside their depth. I.e. the MSK guy reading brain tumor MRIs or the Neuro guy reading a shoulder. Imagine being the person who can't read anything complex at a high-level.

Quicker to partnership is a tricky thing. I doubt the partnership market for a fresh residency graduate/non-fellowship trained is that great. The better, more desirable PP gigs definitely want a fellowship trained rad.

EDIT*

Also, define "general". General as in the average case mix a 2nd year resident on-call might see (PF,US,CT) minus procedures is a very different beast than generalist sitting in a hospital seat slinging barium, doing light procedures, reading mammo and reading inpatient/outpatient MRI/NM on top of PF/US/CT. The latter clearly benefits from additional training in *some* area.

 

[Clean7795]

I am surprised whenever residents think a fellowship doesn't add anything.

Are you all going to programs where you feel comfortable putting out high quality reports for wrist MR, rectal MR, head and neck MR, IAC MR, mammo, nucmed, as well as doing US & CT guided biopsies?

Most residents have at least 1 area where they are not comfortable in and could improve. Even if you "ask for help", you should be able to provide some help in return to others, which is difficult if you have no area of expertise. The only time where it doesn't add much is if you plan on doing evening/night teleshifts forever and pushing off complex cases for the morning rad. Most rads don't want this sort of lifestyle though.

I would agree 12 months is probably excessive for most, but that's just how it's structured (you gain some knowledge and provide service x1 year).

 

[MadRadLad]

I am surprised whenever residents think a fellowship doesn't add anything.

Are you all going to programs where you feel comfortable putting out high quality reports for wrist MR, rectal MR, head and neck MR, IAC MR, mammo, nucmed, as well as doing US & CT guided biopsies?

Most residents have at least 1 area where they are not comfortable in and could improve. Even if you "ask for help", you should be able to provide some help in return to others, which is difficult if you have no area of expertise. The only time where it doesn't add much is if you plan on doing evening/night teleshifts forever and pushing off complex cases for the morning rad. Most rads don't want this sort of lifestyle though.

I would agree 12 months is probably excessive for most, but that's just how it's structured (you gain some knowledge and provide service x1 year).

It might just be practice location. If you are in a community hospital where there aren't that many subspecialists the difference in reads won't matter as much.

Some of the non-flagship hospitals in our system are busier than ever and doing more advanced stuff. There is more push from the clinicians there for fellowship trained reads esp in rectal MR and head/neck. Rads at our main site are already too busy to have them read over there, so when we hire we are looking for fellowship grads only - only thing that makes sense in terms of staffing multiple sites.

Imo just shortsighted to skip over 1 year of fellowship after 10+ years of training to settle for worse jobs the rest of your career. Plus, if the midlevel or AI push happens it will give you a more valuable skillset than just generalist work.

 

[GadRads]

I found fellowship helpful. It made be a better radiologist. Also feels good to have solid expertise in an area. If you do skip fellowship, do at least 6 months in a dedicated area. I work with another attending who didn't do a dedicated fellowship after residency, but completed a 1 yr mini-fellowship in a subspecialty. Besides some struggles in yr 1, I don't see any difference between this attending and others now 3-4 years into independent practice.

 

[radrad5]

very helpful insight and advice. thanks to all.

 

[knosway14]

non us radiologist here (europe /spain).

I am working on a private practice, im a a fresh graduate working as a general radiologist, the youngest of a group of 10 radiologist, all of them with an area on experience (neuro, msk, breast, pelvic), we mostly do CT / MRIs… not an expert but seems like having a fellowship its of a lot of help or just being an expert on something. I guess there are general skills for radiologist and advanced skills depending on your area of expertise. One year in an average lifetime is nothing…

of all the things u mentioned im concerned about AI …..

Just for curiosity…. Have you seen AI doing complex studies :

- perianal fistulas.

- endometriosis (deep).

- dynamic hepatic studies (for HcC).

or something very common : spine mri, shoulder, knee … non GD brains….

just curious to know what my fellows American colleagues have got so far with AI…

chest Ct for nodules and mammograms are out of question cuz they are already dominated by AI to the best of my knowledge….

just want to know if you actually believe AI can do rational/ analytical thinking and perhap complex cases.

sorry for my bad English…

 

[SeisK]

non us radiologist here (europe /spain).

I am working on a private practice, im a a fresh graduate working as a general radiologist, the youngest of a group of 10 radiologist, all of them with an area on experience (neuro, msk, breast, pelvic), we mostly do CT / MRIs… not an expert but seems like having a fellowship its of a lot of help or just being an expert on something. I guess there are general skills for radiologist and advanced skills depending on your area of expertise. One year in an average lifetime is nothing…

of all the things u mentioned im concerned about AI …..

Just for curiosity…. Have you seen AI doing complex studies :

- perianal fistulas.

- endometriosis (deep).

- dynamic hepatic studies (for HcC).

or something very common : spine mri, shoulder, knee … non GD brains….

just curious to know what my fellows American colleagues have got so far with AI…

chest Ct for nodules and mammograms are out of question cuz they are already dominated by AI to the best of my knowledge….

just want to know if you actually believe AI can do rational/ analytical thinking and perhap complex cases.

sorry for my bad English…

This probably isn’t the right thread for your questions.

From my understanding even in Europe AI is only a second reader for specific findings. You guys have replaced nodule reads with AI completely without prospective trials? That is very bad practice.

I am not concerned. Radiology is a descriptive specialty and an atomic and pathological context of findings matters.

 

[knosway14]

This probably isn’t the right thread for your questions.

From my understanding even in Europe AI is only a second reader for specific findings. You guys have replaced nodule reads with AI completely without prospective trials? That is very bad practice.

I am not concerned. Radiology is a descriptive specialty and an atomic and pathological context of findings matters.

No we had not replace anything yet, to the best of my knowledge there are studies in which AI perform better than general radiologist on mammograms, use of AI in clinical practice is very unusual in my experience, thats why i wanted to know what kind of use is AI is having in the US.

 

[2brads]

I haven’t seen any AI “in the wild” in my area. I read a post some time ago where a practice was using AI for pulmonary nodules. He said it was frustrating, although I don’t remember exactly what the problems were.

 

[maxxor]

I haven’t seen any AI “in the wild” in my area. I read a post some time ago where a practice was using AI for pulmonary nodules. He said it was frustrating, although I don’t remember exactly what the problems were.

I have experience with the Siemens product that is deployed on the scanner.

It is too sensitive, so every report now has to address the false positive nodules.

 

[2brads]

I have experience with the Siemens product that is deployed on the scanner.

It is too sensitive, so every report now has to address the false positive nodules.

Yes. It’s possible this was the issue and addressing false positives would be more frustrating than just looking for the nodules without AI. Do you get occasional updates to address these issues?

 

[maxxor]

Yes. It’s possible this was the issue and addressing false positives would be more frustrating than just looking for the nodules without AI. Do you get occasional updates to address these issues?

I have not seen any updates separate from cyclical scanner software upgrades. There does not appear to be rapid iteration.

I also used to work for a system that deployed Viz.AI. Im not neuro so I can’t say whether or not the LVO performance is any good; anecdotally, the neurorads complained about it but they weren’t very specific with their complaints.

The CMS NTAP technical payments were really predicated on the built in messaging system which I honestly feel is superfluous if your system has epic chat.

 

[guytakingboards]

I have not seen any updates separate from cyclical scanner software upgrades. There does not appear to be rapid iteration.

I also used to work for a system that deployed Viz.AI. Im not neuro so I can’t say whether or not the LVO performance is any good; anecdotally, the neurorads complained about it but they weren’t very specific with their complaints.

The CMS NTAP technical payments were really predicated on the built in messaging system which I honestly feel is superfluous if your system has epic chat.

My hospitals use RAPID for their AI stroke program. It's fine. Sometimes it does stuff well. Some times it does stuff really poorly.

For LVO's: meh. Granted all stroke CTA's in my group are read by fellowship-trained neurorads but really we humans aren't missing LVO's all that often. Maybe this is more helpful for more general rad-types who don't read H&N CTA's very often but I don't find this feature all that remarkable. This is really for the clinicians to see the AI report before the human gets a call/dictation out.

For hemorrage r/o: dumb feature. it's false positive rate is way too high. any intracranial calcification dings it as positive.

For perfusion: this is where I find it helpful. we had a case the other day where the AI Perfusion detected a distal M2 stroke. It was completely occult on the DWI/ADC but in retrospect you could see it on the MRA. Might have been easy to miss without the perfusion.

All that is to say, my takehome as a neurorad is that stroke AI is no where near actually replacing a human radiologist.

 

[2brads]

Anyone have a list of currently available/approved AI products for imaging? I'm looking forward to the day I can just copy and paste the AI findings for pulmonary nodules. Doesn't sound like we are anywhere close though.

 

[maxxor]

Anyone have a list of currently available/approved AI products for imaging? I'm looking forward to the day I can just copy and paste the AI findings for pulmonary nodules. Doesn't sound like we are anywhere close though.

I think the business case for diagnostic AI has not been demonstrated yet.

There is all sorts of fda approved AI accelerated image acquisition and reconstruction from the scanner makers. GE, Siemens, Canon, and United Digital all have it.

The most impressive tech demo I saw years ago was this company at rsna Radlogics. But they got slapped by the FDA for marketing things that weren’t approved.

They still don’t have the things they were demoing as FDA cleared. Their cleared algorithms are toys.

 

[2brads]

I think the business case for diagnostic AI has not been demonstrated yet.

There is all sorts of fda approved AI accelerated image acquisition and reconstruction from the scanner makers. GE, Siemens, Canon, and United Digital all have it.

The most impressive tech demo I saw years ago was this company at rsna Radlogics. But they got slapped by the FDA for marketing things that weren’t approved.

They still don’t have the things they were demoing as FDA cleared. Their cleared algorithms are toys.

Yeah, I get the impression that they have some cool stuff right now, but it's not approved and they can't just do the tech thing of "move fast and break stuff" because the FDA will stomp on them, which I'm totally fine with.

 

[MouseChair]

can't predict the future but no. lots of posts about AI here, you can read them

if AI can read a complex scan with post op changes, fluid/blood scattered about, new lesions that are either benign and nothing to worry about or possibly malignant and warrants further follow up, and compare scans to prior studies dating back many years while accounting for subtle changes in patient positioning or scan technique across different modalities... then AI can take our jobs. but I suspect that there will be many more casualties and it won't just be radiologists left out in the cold

In fact AI will probably augment radiology. like RAPID software for strokes. it will be radiologists who use AI that will replace radiologists who don't probably.

 

[.Asterion]

currently applying to medical school here so I am about 11 years and many hurdles out from getting into the rads attending job market

would you be concerned about AI if you were going into rads in 5 years? what about 10-15? I thought I wanted to do neurology previously but really like the idea of doing diagnostic medicine and reading/working alongside other radiologists for a living. Ive been learning computer science and starting on some deep learning fundamentals in the last year partially because I think ml domain knowledge could be extremely useful as AI becomes a bigger part of rads.

Hi, was more worried about AI before I started residency. After actually doing radiology residency I’m not worried about AI. I have worked on some AI projects in the past. But no means am I an expert but I do have a good understanding of how it works and I’m not worried.

 

[SeisK]

currently applying to medical school here so I am about 11 years and many hurdles out from getting into the rads attending job market

would you be concerned about AI if you were going into rads in 5 years? what about 10-15? I thought I wanted to do neurology previously but really like the idea of doing diagnostic medicine and reading/working alongside other radiologists for a living. Ive been learning computer science and starting on some deep learning fundamentals in the last year partially because I think ml domain knowledge could be extremely useful as AI becomes a bigger part of rads.

It is impossible to predict the future accurately. I spent some time arguing why AI won’t take over some months ago if you search this forum, but since then I’ve just decided to simplify the argument as the following:

Radiology is an anatomic and descriptive specialty. What that means is that a lot of times the most critical aspect of what we do is piece what is going on in the imaging together into a story that makes sense for the clinician, or sometimes just to describe what we see. Logically linking disparate findings is something AI isn’t even currently working towards being developed on.

A trivial example is a renal ultrasound done to evaluate for an obstructing stone. We’ll often see hydronephrosis but not the stone itself. We have to be sure to describe in the report that this could be a stone or not, but a CT is recommended to evaluate for a stone or some other obstructing cause. You’d think the recommendation would be obvious, but you’d be surprised at just how many hospital, ED, and primary care physicians don’t know this. Even subspecialists sometimes.

A more complex example would be a chest abdomen pelvis CT which demonstrates ascites and an irregular adnexal mass with several rounded 4-6mm solid pulmonary nodules. Have those nodules been there before? If so, for how long in relation to the mass? Does this patient have something else going on that can explain the ascites? The disparate pieces matter for the larger diagnostic picture. Single-word diagnoses like “pneumothorax” exist, and AI in clinical practice can help us catch some we miss sometimes, but it’s honestly very rare that the single-word diagnosis was something the clinician was looking for or something they’d find useful in isolation.

Or another one: there’s a collection in the abdomen for someone without a past medical history. Where is it and where did it come from? Missed ruptured appy? Complicated diverticulitis? Related to Crohn’s? Is the collection even infectious? Does it appear drainable? The synthesized conclusion from other findings in the imaging or in the patient’s history matters. This in addition to all the other **** I mentioned in the other post, which is a lot.

I’m not worried. I’m increasingly worried no AI application will deliver on the promise of making us more efficient. We’re starting to need it.

 

[Steve_Zissou]

Quick related question. Does it matter if the fellowship is non-ACGME (MSK or ER) at all?

 

[maxxor]

Quick related question. Does it matter if the fellowship is non-ACGME (MSK or ER) at all?

The majority of fellowships that do not offer a CAQ are non ACGME.

 

[Steve_Zissou]

The majority of fellowships that do not offer a CAQ are non ACGME.

And CAQ is not really necessary, I take it?

 

[Neopolymath]

I believe one benefit of non-acgme is less moonlighting restrictions.

 

[Steve_Zissou]

I believe one benefit of non-acgme is less moonlighting restrictions.

I guess my question is more, is there any benefit to having ACGME recognition other than maybe for a select few academic jobs?

 

[maxxor]

I guess my question is more, is there any benefit to having ACGME recognition other than maybe for a select few academic jobs?

It’s more related to the relative renown of given fellowship rather than if it is acgme or not.

Many “top” fellowships are not ACGME at all.

 

[guytakingboards]

I guess my question is more, is there any benefit to having ACGME recognition other than maybe for a select few academic jobs?

As mentioned, if you do a CAQ specialty like neuro or peds then the primary certification pathway to CAQ requires an ACGME fellowship. My large multi-specialty group pretty much requires CAQ for neuro. It's a marketing point for some groups that all neuro reads are by CAQ rads.

The benefits of doing an ACGME fellowship in a non-CAQ specialty like MSK, body or ER are much less so. The ACGME rules are there to protect the trainee, by stating the fellow can't be forced/required to act in a non-trainee role. I.e. the ACGME body fellow can't be scutted out to staff another service during the week or be forced to serve as an attending in the evenings/weekends. It does not prevent the fellow from doing it after hours, but even then they couldn't staff their own specialty.

Fellowship year IS a good time to get some experience final signing cases and in general moonlighting to keep your non-specialty skills up. Plus you can make extra money moonlighting. So non-ACGME programs in non-CAQ fields are preferred by many. To my knowledge, there's no added prestige to doing an ACGME body or MSK fellowship vs a non-ACGME fellowship.