For discussion in the broader radiology community, I’d like to share Radiology Partners’ March 9th announcement email regarding their new time-based work unit system.
I was really hoping to get more specific examples of this system after the April 1 go-live, but from what I’ve gathered, the transparency on the system is generally low, and most people really don’t know exactly what’s happening under the hood. If anyone from RP would like to provide concrete examples, I would absolutely love to share them.
I’m not including a small partial table circulating online, as I cannot verify its accuracy. The consensus I have heard consistently is that ER-style CT is down (head CT, c-spine, CTA chest); brain MRI is down; screeners are down, and diagnostic mammo is up; plain films are up.
The official language from RP does not describe the methodology. I’ve heard that exam weighting is different by context (ER, inpatient, outpatient), but I’m not sure how much. This would be more fun/interesting/fruitful if we had a fuller picture. Nonetheless, we can still discuss some consequences of this move regardless of the specifics, because the discussion here does not hinge on whether the actual custom weighting used today is bad/wrong.
RP Colleagues,
As you’ve heard from your leadership, your practice is moving to time-based work units (TBWUs) on April 1, 2026. This unit of measuring work is being implemented across all RP practices to ensure equity and transparency, creating a unified approach to scheduling, cross-practice coverage and compensation.
Note the framing. “Equity and transparency” is the marketing. “Unified approach” is the actual product. A PE-backed aggregator that has always insisted its practices are “locally led” is now standardizing the core work unit across every practice it owns.
I don’t think RP has a buyer, but certainly, finally exerting more operational dominance over its flock would be important to shoring up the argument that it is a single, stable, cohesive business.
Custom units can create new winners and losers overnight, so even the same system would feel different to radiologists if it were something built collaboratively by the team that everyone agreed to and not something forced onto it. Without transparency, such a change requires incredible and ongoing trust.
To inform this change, RP’s Clinical Transformation Team performed an in-depth analysis of a large, diverse dataset of 52M exams across most RP practices and all subspecialties to ensure statistical significance. TBWUs measure work based on the average time required to read an exam.
Fifty-two million is a big number. It also happens to be 52 million exams read by rads who were compensated on RVUs, where many radiologists are incentivized to cherry-pick and crush high-RVU studies.
Radiologists also know that exam type is only part of the puzzle. Context and complexity matter a lot. It’s easy to say that screening mammography and brain MRIs are overvalued, but there is a big difference between a non-contrast stroke protocol brain MRI and a GBM follow-up. In the example table floating on the internet, a non-contrast chest CT was valued higher than a CTA chest. The only way that would make sense is if we are ignoring context and letting healthy young person chest pain CTAs drown out cancer and chronic lung disease noncons.
Any time-based metric derived from this dataset inherits the distortions of the system it’s replacing. The CPT code is a terrible metric for effort, but just moving the slider up and down and down to get people to stop letting non-con chests, thyroid US, and plain films pile up is a crude lever.
It seems as though the exams that linger on a cherry-picked list are the ones that have been bumped by this system, so exam desirability is being tweaked, and the need to address pileups in the backlogs may be a significant weighting factor. But once rads are incentivized under the new system to crush plain films—which, oh by the way, are exactly the studies Mosaic is being positioned to pre-draft—those times will fall. And when RP applies the same methodology to recalibrate, the numbers will presumably move again. This is not going to be a one-time change; it is the beginning of a perpetual motion machine, and the initial settings are based on what’s backlogged or what Mosaic can/can’t do.
TBWUs offer:
Equity, balance and accuracy: Aligns compensation and productivity with actual time, not billing distortions.
In theory, complexity- and effort-based internal RVUs are fine, maybe even the exact right thing to strive for. I’ve written about them before. Some physician-led private practices run custom, zero-sum internal systems where total internal production matches CMS numbers in aggregate, and the group decides how to weight specific work relative to that total. That can work. It works when there’s transparency, governance, and buy-in. It works when the people affected by the weighting are the people doing the weighting.
Implementation and governance are the whole game. They’re also where corporate central planning has the hardest time earning trust. “Locally led” becomes a hollower phrase every time one of these corporate-wide pronouncements lands in the inbox.
Simplicity and transparency: Creates a predictable structure for compensation, enabling equitable comparisons across practices using one standardized measure.
The real transparency test is whether RP publishes the underlying data or at least the full breakdown. Show RP rads the distribution of read times by CPT code across those 52 million exams. Show how ED versus IP versus OP TBWUs were derived to see if even the pretense of context impact is respected. Show the process for future adjustments—who approves them, on what cadence, and with what rad stakeholder input. Publish the change logs.
I don’t know the size of the governance committee, but I strongly suspect they, like the RUC, will provide recommendations and not decisions. That power is almost certainly with the executive leadership, likely the CMO.
A standardized measure that only the standardizer can see inside of is not transparent. It’s just standardized.
Better support for practices: Enables a unified approach to scheduling, cross-practice coverage, compensation, workload balance and flexibility to adapt as payers, reimbursements and external environment shifts.
The need for practices to cover for each other in an insufficiently staffed enterprise would seem to be a key feature here.
Quality care: Creates conditions to balance workloads, allow for cross practice coverage and protect time for more complex exam types.
Reasonable in theory. RVUs aren’t great. Everyone agrees that GI fluoro is undervalued, for example.
If the system actually did this, it might even work. If each CPT code gets an emergency, inpatient, and outpatient weighting, that might even get part of the way there. But the reality is that even different groups and locations have different relative complexity. Level 1 trauma cases will be different on average from Level 3 trauma, even if both types of patients are in the ED. At most three buckets per code is a very coarse instrument in the AI era.
FAQs
Why can’t we stay on RVUs?
They reward reimbursement patterns, and because of how they are created and adjusted they are subject to potential political imbalances.
The RVU system’s flaws are well known, but “they reward reimbursement patterns” is the same thing as saying “they are based in shared reality.”
There is a meaningful difference between a flawed system where everyone in American radiology plays by the same rules and a proprietary system where one corporate owner sets the rules for the subset of the field it happens to own. The latter means the winners and losers are chosen by your boss, and your employer is actively tweaking things to guide your behavior and manipulate your compensation. To be clear, RP is totally within its rights to do this! It’s their business, and after all those group purchases have matured and those 5-year vesting cliffs have come and gone, they’re running it.
But this is a completely different level of trust than custom units for a democratic private practice, which would mean everyone coming together to hash it out and try to fix the failures of the RVU system together (and even that would be extremely contentious in most groups).
How will TBWUs affect physician compensation?
TBWUs more accurately measure physician work, shifting from exam volume to a model that reflects the time required to interpret studies. The intent is to better align the work being performed with productivity, minus the intent to negatively affect compensation. If the new measurement creates unintended shifts, productivity thresholds can be adjusted to maintain balance and consistency.
The announced rollout is reportedly zero-sum or close to it on day one. Presumably, almost all practices transitioning from CMS RVUs to a custom internal work unit would do it this way. Zero-sum is the only politically survivable first step. In the short term, it incentivizes reading the plain films, thyroid US, and non-con chests that have been piling up on the out-of-control worklists.
In the long run, zero-sum is not a promised constraint. There is no guarantee that adjustments will continue to track. As long as the changes are subtle enough, it could even feel like the frog in the slowly warming pot. To confirm this point, see RP’s answer about Mosiac below.
Even if the tables were never changed, one should expect the lowering of some CT and MRI vs radiographs to result in less money to radiologists over time as the growth of MRI and CT outpaces radiography. Zero sum today won’t be zero sum tomorrow.
Of course, plenty of radiologists are salaried, so we shouldn’t pretend that this directly impacts everyone’s compensation in a competitive job market. But whether or not it’s used this way, it remains possible for such a system to be used to clandestinely lower reimbursement. RP’s announcement doesn’t promise any specific communication, transparency, or accountability for future adjustments. No hard conversations with the thousands of radiologists they employ. It could just require updating a table.
A custom system actively creates winners and losers even when the underlying design is defensible. That’s tolerable when the people choosing the design are your partners. It’s harder to swallow when it’s a third party doing it to you.
Will TBWUs change how much I make on night shifts, weekends or high-volume modalities?
Existing compensation structures (productivity thresholds, shift differentials, etc.) still apply. TBWUs provide a more accurate measure of work, but practice-level decisions around comp structure remain.
In order not to lose their entire call team system-wide, they have to play nice.
But this is where the contractor and moonlighting workforce should pay attention. Salaried rads with productivity bonus thresholds are one population. Per-click readers, moonlighters, and the contractor army are a different population.
If the new TBWU table nukes the effective value of neuro CT and a practice’s productivity threshold doesn’t adjust in lockstep, an ED reader may have just signed on for a pay cut without a meeting. And anyone who was enjoying cherry-picking the status quo can also take their talents elsewhere.
Breast imagers are probably another cohort to watch. Diagnostic values are up and screeners are down, but screening moonlighting is a big draw for lots of mammographers. Given the possibility of running breast imagers off and their relative scarcity, it also makes you wonder how far off Mosaic drafts for screeners are (probably months).
If you’re being paid per study, and the studies you enjoy reading are now worth less, you’ll probably moonlight somewhere else.
How do AI tools like Mosaic fit in over time?
As Mosaic continues to deploy, radiologists find it easier to read in less time, and report quality and efficiency improve, RP intends to share the value created by these tools with radiologists both clinically and financially. If an AI tool meaningfully reduces average read time, the associated TBWU value would be adjusted by a smaller percentage so that radiologists generate more in total compensation in the same amount of time. The timing of any adjustment is still under discussion and is currently anticipated between July and September.
Read that paragraph twice.
The sequence is: deploy AI that reduces the time a study takes, observe the reduced time, adjust the TBWU for that study downward—but by less than the time saved, so rads keep “more in total compensation in the same amount of time.” The radiologist’s share of the productivity gain is set by the size of the adjustment. The size of the adjustment is set by RP. The data underlying the adjustment is held by RP.
(While rubber-stamping AI-slop AI-drafts would presumably be economically encouraged, the liability is on the rads, whereas any associated economic benefits are at best split. You can decide if you think that sounds reasonable or not.)
Notice also that plain films—one of the study types Mosaic is reportedly pre-drafting—got a big bump in the initial TBWU table. A time-based metric on a study whose time is about to fall because of a tool the same corporate parent is selling. That bump will presumably not survive recalibration over the long term. “Currently anticipated between July and September” is perhaps corporate-speak for we’re giving you Q2 to enjoy it.
The honest version of this FAQ reads: as Mosaic takes on more of the read, we will reprice the read. The share that rads receive is a policy choice made by us, not a guarantee. That choice will be defended internally with data the rads can’t see. This is the first step toward “AI drafted this for you, we’ll pay you what you’re worth to us.” Anything that still feels like a great deal will be adjusted down as radiology whack-a-mole.
To reiterate: I would argue it is different for a physician-owned practice to change its own internal accounting so its members feel valued and to guide desired behavior (and doing so is extremely challenging). It’s another thing for a third-party employer to change the game and leave you with the promise that, over time, you won’t make less money: that they will capture the benefits of increased efficiency thanks to the implementation of AI, but not so much that you’ll actually make less money than you do today per hour, even if you do make less money than you do today per study. (That may even be a real promise that they’ll stand by. It’s up to readers/radiologists/reality to decide. But even then, what happens long-term to reimbursement if AI is making you more efficient? Consequences have consequences. What will RP do about sharing then?)
We live in a tumultuous radiology market, increasingly full of loosely affiliated teleradiologists. More than ever, even fresh graduates are starting their careers in the tele market. Perhaps we will eventually see massive worklists with automatic assignment based on radiologists’ “skills,” demand, and preferences. Don’t want to read body CT? No problem, we can remove those from your list and adjust your comp. Oh, today we’re in surge pricing for MSK MRI. Maybe we’ll have a generation of rads who will describe their dynamically-priced, AI-assisted, per-study compensation as freedom. Or that may not happen at all; I won’t pretend to know where this is going.
But the first step to further cog-ifying the radiologist within a money-printing machine is to decouple the work they do from the revenue they generate while promising it’s not about compensation.
The first step is a zero-sum, equity-framed, transparency-branded transition to a new internal work unit.
That first step is now done.
