As a follow-up to the two posts about Radiology Partners’ new time-based work units, the first quarterly TBWU adjustment is taking place quickly: July 1. This will incorporate changes related to any efficiency gains for groups on the Mosaic platform. From the May 15 FAQ:
Based on the TBWU guiding principles, our objective is to improve overall physician compensation per shift, per hour, per year. In order to do that, the efficiency gains from any RP tools (such as Mosaic Drafting or Capture) will be shared with radiologists. In the first TBWU review, 80% of the value from efficiency gains will go directly to our radiologists. After that and through the end of 2027, no less than 50% of the value of the efficiency gains will go to our radiologists.
It’s easy to roll your eyes at that corporate phrasing, but we should be fair about what RP is talking about when it comes to splitting efficiency gains. As previously stated, their intention has always been to reduce TBWUs to “partially” capture efficiency improvements, and their guiding principle is to capture an increasing percentage from the delta itself and not from the total work unit. The work unit of course will decrease, but not by a ton, at least not at first. RP estimates that 20% of the improvement will be 2-7% of the work unit. Yes, this means that they bumped plain films for just a fraction of a year before using Mosaic to start siphoning them back down.
Total aggregate work credit (± compensation itself) would only decrease if productivity didn’t increase to account for the saved time. So, you need to kick faster to tread water, but you arguably won’t care because the kicking will be easier.
To reiterate: Since these are time-based work units, what they’re doing is that when the time-to-read goes down thanks to Mosaic, they won’t move it all the way down the full amount; they’ll leave some of the efficiency gain to rads. For this first iteration of the time difference, it’s 20% to RP and 80% to radiologists. At this early time point, one would venture that the time savings are not going to be huge, so the relative generosity of the split is probably mostly in name. Until the end of 2027, they said they won’t take any more than 50% of the improvement. Reimbursement/credit is going down per case in these adjustments; it’s just that they claim they’re going down less than the time savings, and therefore total work units produced by a human won’t be decreased in aggregate over the course of longitudinal work: you’ll just read more cases because it will be so easy, or something along those lines.
How the proposed improved “overall physician compensation per shift, per hour, per year” will survive over the long term if the market softens and payers catch on remains to be seen. Also, we can acknowledge that the time saved for individual radiologists may be very different than the average time saved. Even if an individual rad doesn’t experience efficiency gains, they still pay the efficiency cost.
As the use of similar tools grows across the field, a relative pay differential could also widen relative to other practices that are able to capture similar productivity gains from other products while reaping the majority of the benefits. That would, of course, depend on how expensive those products are for those practices relative to the efficiency gains and who controls the compensation. Who’s to say how it will shake out? Not me. The laws of supply and demand will keep operating, second-order effects are inevitable, and there are a lot of moving parts. The technology may increase efficiency, but what we do with that depends on the contracts. Employers are incentivized to capture those gains and raise the production floor. Private practices are incentivized to maximize revenues for themselves.
Procedural Time
The FAQ contains an amusing doublespeak regarding IR vs “Non-workstation time”:
IR procedures: Given the unavailability of time studies for IR procedures, TBWUs for IR exams equal the RVU value.
Non-workstation time: Diagnostic mammography and CT coronary angiography (exams with a non-workstation component) include additional time beyond image interpretation. Based on data-driven time studies, incremental seconds were added to these exams to more accurately reflect total time spent, resulting in updated TBWU values.
I am not an IR, but I would venture it is indeed possible to know how long various average IR cases take. It may even be easier than adding “data-driven” “incremental seconds” to fluff numbers. As for non-workstation time…what about fluoroscopy? Something tells me those incremental seconds are largely uncaptured. The difference is that coronary CTA and diagnostic mammo are lucrative, growing service lines that need to be supported.
Perhaps the more salient problem with including IR is that truing up TBWUs to account for the time of procedures would take too many RVUs out of the pool and drive DR values down too much in a zero-sum calculation.
An Incomplete Accounting
The reality is, no one has this data but RP. No one is going to be performing an audit, and no one can ensure honesty or integrity when it comes to anything here (or really at any other practice). This is a privately held company, and we shouldn’t pretend like they have any need to, because if the 12-radiologist TBWU Physician Advisory Committee making “recommendations” to the CMO is enough accountability for the RP rank and file, then it’s enough.
But let’s assume that time-based work units are exactly that: work based on time, fairly calculated and revised on a continual basis. They have argued that TBWUs are better and fairer than RVUs, and that may even be true.
At the same time, TBWUs, like RVUs, are still an incomplete accounting of work.
Averages are averages, and not every case is average. Giving you an average work unit for every case does not necessarily equate to fairness over time. The attempt to adjust based on patient context is insufficient when you consider the broad swath of radiology. The positivity rate, the complexity of an exam, the timing of the exam, the relevance of priors, call findings to clinicians, etc—all these things might impact a true weighting system.
For a company that is touting its AI-native platform, one might ask why just adjust for time when there is so much more? Why not go all the way and do a true effort RVU? Because it’s a hassle they don’t need to attempt.
As I suggested previously, the primary motivation for simple time-based accounting is that time changes can be measured before and after, and those magical promised AI-driven productivity gains can be garnished in increasing amounts over time. How would a multimodal AI-derived work unit account for help from AI? This recursive question is, I would venture, much harder to measure than “time went down.” TBWUs may be better than RVUs, but fairness is not the goal per se. Rather than being an ideal way of internally fixing a broken system, they are the ideal way of “sharing efficiency gains” (repricing the work) and driving the behaviors needed to address the most egregious list behaviors/cherry-picking/backlogs.
For radiologists, the only way to “win” is to use those seconds of “average improvement” to read more cases to make up for the decreases in TBWUs. They frame it as “sharing the upside/gains,” and that’s technically true.
Ultimately, those with the power, leverage, and control have a stronger say in who gets to reap the economic benefits of these new technologies.
I can see why some would view this repricing as the beginning of a depressing hamster wheel on turbo mode.