Medical procedure relative value systems have been available for decades. The Medicare program has used the Resource-Based Relative Value Scale system (RBRVS) since the late 1980s as the basis for its physician fee schedule. In theory, this system is to factor both clinical and economic components of delivery of medical care such that two radiologists of similar skill, working at the same pace, should generate similar resource-based relative value units (RVUs) even with materially different caseloads.
This article will portray results from compilation of a large database of different practices, using the year 2003 RBRVS. It will begin with detailed information on one of the practices to illustrate how the data was organized and conclude with statistics from all practices. The perfect benchmark is average daily RVUs, rather than an annual number. Even with the large amount of data used in this analysis, we can only speculate on such a number.
Below are the real-world impediments:
Certain practice members will have administrative duties pertaining to the department and/or their legal entities that are critical to the role of the group but not measurable with RBRVS.
Some older members will cover fewer days and perhaps not take call. Others may work only call hours and not cover days.
The radiologists cannot control referral patterns into the facility where a physician will be available to read but there are not enough cases. Conversely, there are no data that suggest how many hours per day a radiologist can interpret caseload if an unlimited amount of examinations are available to read.
Teaching physicians will have caseloads affected by resident/fellow training.
There will be materially different payor structure, depending on both region and institution. Cash value per examination may dictate the volumes per radiologists based on trade-off between income and time off.
Some physicians in the sample might have either started or left in the 12-month time span. Their compiled RBRVS might imply part-time status when in fact they were productively full time while with the group.
The data was derived from receivable systems. There can be coverage relationships not running through these systems that influence the number of radiologists in the group. Those RVUs will be missing, and the implication is that the study will understate true productivity.
My consulting work requires interface with both regional and national billing companies. They provided the examination statistics used in this study. The statistics are accurate; the practice and physician identities are in code. The data is for all examinations performed by each practice from May 2002 to April 2003. It is organized by physician, then CPT-4 code. Charges were provided, but are unnecessary for the study, though it does provide relative pricing benchmarks. The 2003 RBRVS is available in spreadsheet format from the Centers for Medicare and Medicaid Services (CMS) download Web site (CMS files for download for Medicare payments systems).
A special compilation was requested for a single practice that provides examinations/charges per physician for each day in this time span. This will be used to estimate RVUs per day, the more appropriate benchmark. It was not possible to acquire the daily statistics for every practice in the sample because the sorting matrix is very large. Even though today’s billing systems have powerful computing capacity, such an endeavor would require 365 sorts per physician. The daily information just for one practice was very large and cumbersome to work with; compromises were required to compile the information, as will be noted shortly.
SINGLE PRACTICE STATISTICS
The first model (Table 1) is the summary of information pertaining to client “A07”. This will be followed by physician-specific data on four of the members to illustrate modality-specific production and days of coverage.
The “FTE” is a statistical designation of the status of the physician; this does compromise the calculation. There are 21 radiologists in this practice that interpreted examinations in this 12-month cycle. Any member whose RVU count exceeds 85% of the greater of the mean/median was designated as “1”. The others were designated by the ratio of their RVU count to the greater of mean/median. Another test was performed using each physician’s number of workdays, and the same screening criteria. That resulted in an “FTE” count of 15.15.
The columns showing Exams, RVU, and Fees are from the database compilation per physician. The “RVU/Exam” column offers perspective on caseload configuration per physician. Another table will illustrate this. The higher the number, the greater the proportion of the more time-consuming examinations.
|Table 1. Summary of statistics for practice A07
The “Conversion Factor (Con Fac)” column is the division of the practice fees by the RVU. This offers a perspective on the relative pricing of this specific practice in relation to the others. If the fee schedule was perfectly relative for every examination, then the numbers for each doctor would be identical; variability implies that the procedures are not relatively priced.
Table 2 (page 54) offers a perspective on the examination/RVU configurations of four members of this practice; their code identities will help you locate their corresponding statistics in Table 1:
The “RVU/Exam” column provides insight into the aggregate differences based on modality; the higher the number, the more time-consuming the procedure. The non-70000 series codes pertaining to surgical cases are separated from the diagnostic codes to depict the substantial differences in assigned RVU. At the conclusion of this article there will be a table that mathematically depicts the average RVU/examination for the entire population. Note also that the fee schedules for mammograms and surgical procedures are more conservatively priced based on the conversion factor (fees divided by RVUs). This is typical for most practices.
VJM is clearly the most productive member of this practice, with responsibilities crossing every modality. Note the absence of MRI cases. They are paid through a facility coverage agreement and the billing company has no statistics on this caseload. VJM recorded the second largest number of coverage days in the entire group; it is suspected that he or she is relatively new. Generally, the newest members handle the largest caseload as part of their growth to parity. Note, also, that all four in this sample handle caseload in every modality. Even though this is a busy practice with relatively large volume, there appears to be little subspecialization. This adds flexibility to scheduling both daytime and on-call coverage.
Table 3 (page 57) is the next phase of the process where we seek to establish examinations per day to then estimate daily RVU volumes.
The examination statistics for every day were organized into a large table that included the weekday of each date. Holidays falling on a nonweekend day were also identified, because caseload is definitely affected. The next phase was to examine patterns that implied nighttime coverage on a nonweekend day. Many of these coverage days were evident because those assigned night coverage often have regular hours off. Each of the four physician compilations has columns distinguishing full-day coverage versus on-call. Holiday, Saturday, and Sunday volumes assume on-call coverage. The area of greatest interest is the “Full” column, which portrays the average caseload for daytime coverage. Unfortunately, as stated earlier, the volumes do not include MRI. This poses another problem. A weekday that appears light and was designated as on-call might have been a day that the radiologist was covering the MRI facility. VJM handled the largest caseload per day for the full days.
|Table 2. Modality-specific data by physician.
The middle blocks isolate the full and on-call days. VER logged in the largest full days and VSB covered the most on-call slots. As an aside, this compilation revealed others in this practice that only cover Friday to Sundays (not in this four-physician sample).
The last block completes the math by illustrating examinations by type of weekday. The totals differ slightly from the other tables because this data was derived from a separate database and there was a lot of manual compilation; errors were probably made in the reorganization of the data.
The average daily volume (full nonweekend day) can be used to calculate the daily RBRVS benchmark for these four radiologists as depicted in Table 4, page 57.
The extension of each physician’s average daily examination volume and their specific RVU count per examination provides the RVU per day. The last four columns illustrate the RVU generated from the number of coverage days. The last line is the average of this four-physician population. Unfortunately, the range of RVU/day for this small sampling is rather broad (60.40 to 39.89), but the four members chosen are representative of the 13 statistically classified as full time (the other eight are considered part time).
SUMMARY OF ENTIRE POPULATION
The total sample comprised 15 practices with a very large range of examination volume and physicians per group. The summary of compilation is exhibited in Table 5 on page 58.
All billing companies track collections pertaining to each practice location. The top third of the table identifies the database statistics for the hospital caseloads. The professional fees are listed here to help compute the conversion factor (CF). This is the division of fees by the total RVU. If you were to multiply the practice-specific CF by the RVU per CPT-4 code, you would come up with a relatively priced fee schedule that equals the same charge base. Some practices will use most of the RVU values to construct a schedule (rounding to the nearest whole dollar), except for surgical codes. The application of practice-specific CF to the surgical codes results in very high fees not recognized by payors.
The hospital examination base results in an average RVU/examination of .90 and a conversion factor of $109.66. The year-2003 Medicare CF is $36.7856. The multiple between fee schedule and Medicare ranges between 2.31 and 3.82, with the average at 3.00.
|Table 3. Daily examination statistics by physician.
The middle block of data includes only the examinations and a calculation of the professional RVUs on this base. The charges are not listed because they are global fees. It would be possible to compute the global RVU on this population, resulting in two interesting benchmarks:
The conversion factor, and its relationship to that existing for hospital cases,
The ratio of the professional component to global.
Note that the RVU per examination is lower for the offices. There are few, if any, surgical procedures in this database and the proportion of MRI/CT scans is probably less than hospital caseload.
|Table 4. RVU statistics for full coverage days only.
The merged statistics show combined data for the five practices with outside imaging facilities. It is here that the statistically computed “FTE rad’ is listed immediately to the right of the number of physicians in the group that interpreted cases in this 12-month span. This is an admitted weakness in the modeling because it uses the RVU data to determine full-time equivalency. It would have been better to secure the number of days each physician worked in this 12-month span, a daunting task in light of the volume of sorts per practice. If an individual practice were interested in duplicating this exercise, it would be easier to compute averages per day because each member’s schedule would be readily available. The billing company can supply examination and RVU statistics per day as part of its regular reporting.
|Table 5. Summary of benchmark statistics on total population.
The columns at bottom right are the division of total examinations and RVUs by the “FTE” physician number. The mean and median of this population are at the bottom. Again it needs to be emphasized that there are probably some missing examination and RVU statistics in the population because the billing company does not bill for negotiated facility coverage contracts.
The data reveals some interesting effects of market forces, yet it is not necessarily consistent. Practice A46 is a substantial outlier with examination and RVU volumes very different than the rest. Note the CF for this practice, the lowest of the sample. It does have one of the least complex caseload configurations (fifth lowest at .83). Practice A07 also possesses a relatively low fee schedule and case mix, and the lowest RVU per physician. Conversely, practice A93/A94 has the most aggressive fee schedule, the fifth highest caseload mix (.90), and the second highest RVU per physician. There is weak positive correlation between FTE and RVU average (.055; 1.00 is perfect correlation). The correlation of CF to RVU average is only slightly more direct (.097).
The last table (Table 6) compiled from this database is an examination equivalency model based on average RVUs per examination by modality.
|Table 6. Examination equivalency multiples based on modality RVU/examination.
The table is to be read top to bottom for each modality in the column heading. The modality is designated as a bold “1.00”. All the other numbers in the column are the number of procedures required to generate the same number of RVUs. Example: A radiologist assigned to read diagnostic cases would have to read 4.94 examinations for every 1 CT scan.
There is little question that RVUs per physician is a better measurement of clinical productivity than either examinations or their charges. There is debate whether RVUs can be a foundation for distribution of some or all of the income pool of a practice. It depends on realistic assumptions about caseloads per working day. In the hypothetical scenario where physician A only interpreted CT scans for a day, while physician B interpreted diagnostic examinations, the table implies how many more examinations physician B must read to match RVU generated by physician A. The other issues are burnout and clinical efficiency. What is a realistic number of interpretation hours in a day before stress levels affect health and accuracy? Can both physicians work the same number of hours per day, assuming there is sufficient caseload to interpret on a constant basis?
This study defines protocol for individual practices concerning compilation or RVU data. It is uncertain how much outside caseload is missing from the sample used to produce the benchmark averages. The tables at least offer minimum benchmarks that answer questions about income systems, based on productivity. Many key payors use relatively valued fee schedules to reimburse for interpretation. Knowledge of CF in relation to payor fees and RVU volume per physician can help a group know if they are staffed efficiently to maximize their incomes. Also, if a new coverage opportunity presents itself, a practice can quickly determine staffing needs based on review of RVU volumes determined by multiplying examinations/CPT code by the RVU/CPT code.
James A. Kieffer, MBA, is president, ProForma Financial Group Inc, a consulting service in radiology reimbursement, based in Nashua, NH