Introduction

The long-studied relationship between surgical volumes and outcomes has consistently demonstrated better outcomes among surgeons with higher annual procedure-specific surgical volumes [1,2]. Findings have influenced the Leapfrog Group, a consortium of large employers and insurers, to develop and publish minimum volume standards for surgeons for a number of surgical procedures as a proxy for anticipated better care outcomes [3].

In the absence of the availability of surgeon-specific risk-adjusted mortality and patient-reported outcomes measures – which are critical to informed patient choice – patients seeking high-quality elective surgical care might reasonably want to choose surgeons who perform relatively higher volumes of the procedure that they are considering. In early 2024, the Centers for Medicare and Medicaid Services (CMS) released several publicly available ‘utilization’ datasets that provide 2020 data on individual clinicians and provider- and procedure-specific volumes (and percentile rankings for those volumes) for twelve common procedures, the provider’s primary specialty, and additional information on providers (such as National Provider Identifier (NPI) number, and ZIP Code in which the provider practices) [4].

To help consumers make informed choices regarding their healthcare procedures, we sought to understand: 1) Which specialists perform these procedures; 2) The distribution of volumes across procedures and Dartmouth-Atlas-defined Hospital Referral Regions (HRRs); and 3) The feasibility of using these data to identify providers with high procedure-specific volumes.

Materials and Methods

Data

CMS published publicly available datasets designed to help consumers identify provider-specific volumes and volume percentiles among those conducting twelve common procedures: cataract surgery, colonoscopy, coronary angioplasty and stenting, coronary artery bypass grafting, open hernia repair, minimally invasive hernia repair, hip replacement, knee replacement, mastectomy, pacemaker insertion, prostate resection, and spinal fusion [4].

From that CMS website [4], we obtained: the “National Downloadable File” (DAC_NationalDownloadableFile.csv) that provides National Provider Identifier (NPI)-specific information on primary specialty and practice ZIP Codes; “Utilization Data” (Utilization.csv) that provides NPI-specific information of the count of and volume percentile for each of the twelve procedures in 2020; and “2020 Clinician Utilization Data” (ClinicalUtilization_2020.csv) that provides NPI-specific data on ancillary and support services that were provided by providers listed in the “Utilization Data.” The “2020 Clinician Utilization Data” file reflects the reality that multiple providers who are involved in a procedure might bill for it, including ancillary providers such as anaesthesiologists, radiologists, pathologists, assisting providers, and other professionals, even if they are not the primary proceduralist.

Data preparation

We limited data to NPIs with practice ZIP Codes located in the continental United States, Alaska, and Hawaii. We assigned providers to Dartmouth Atlas-defined HRRs based on practice ZIP Codes [5]. HRRs are aggregations of ZIP Codes of Medicare fee-for-service beneficiaries who sought care and are considered healthcare markets [5]. We also obtained the total number of Medicare enrollees in each HRR from the Dartmouth Atlas website.

In the “National Downloadable File,” 697 NPIs were missing information on primary specialty or practice location; therefore, we used the National Plan and Provider Enumerator System (NPPES) [6] to provide those data.

While the data included a provider-specific ‘volume percentile’ for procedure-specific volumes, we assigned each provider to a quintile of volumes for a specific procedure based on volume counts. Because of clustering of volumes, we sometimes needed to create quartiles or terciles of volumes; in those cases, we report data as quintiles by retaining the lowest and the two highest quintiles. Because CMS did not provide actual volumes if the number of procedures performed was between 1-10, we substituted the number “5” for all such volumes.

Analysis

From the combined “National Downloadable File” and

“Utilization Data” dataset, we calculated the number of providers and procedures that were associated with each primary specialty for each procedure. We used those data and medical practice knowledge to identify which specialty types were likely to include the primary proceduralists; we limited our data analysis to those procedure-specific specialty types. Using the “2020 Clinician Utilization Data” file, we confirmed that none of the ancillary providers were included in the primary proceduralist data.

For each procedure, we calculated mean count for each quintile, the mean CMS-assigned volume percentage, the number of providers (overall, and in the case when multiple primary specialists performed the procedure, the number in each primary specialty) in each volume quintile, and the proportion of providers who provided any of a specific procedure and were in the highest and top two volume quintiles in each HRR.

Because certain specialties might conduct several of the twelve procedures (for example, orthopaedic surgeons might perform hip replacements, knee replacements, and spinal fusions; general surgeons might perform colonoscopies, open hernia repairs, minimally invasive hernia repairs, and mastectomies), we identified proceduralists who performed multiple types of procedures and determined the proportion of providers conducting a single procedure type or multiple procedure types that were in the highest volume quintile for each specialty.

Finally, for each proceduralist type, at the HRR level we calculated the minimum and maximum number of proceduralists, the number of high-volume proceduralists per 100,000 Medicare enrollees, and the minimum and maximum proportion of proceduralists that were in the top or the top two volume quintiles.

As all data used were publicly available, the study did not constitute human subjects research and IRB approval was not required.

Results

Table 1 shows the number of providers performing each of the 12 procedures by specialty, before restricting data on ancillary providers provided in the “2020 Clinician Utilization Data” file. There were 81 primary specialty types assigned to these 12 procedures, including the ancillary providers. We used numbers of providers and the specialty-specific volumes of procedures (Supplemental Online Table 1) to identify the primary specialties for proceduralists performing each type of procedure. The number of primary specialties ranged from one (for cataract surgery, hip replacement, knee replacement, and prostate resection) to three (for pacemaker insertion or repair). The assigned primary specialists constituted between 75.3% (spinal fusion) and 99.1% (prostate resection) of the total number of specialists attributed to each procedure and between 89.1% (colonoscopy) and 99.7% (prostate resection) of the total number of procedures performed.

Table 1:  Number of providers performing each procedure type by specialty. Bold numbers indicate primary specialties.

Limiting further results to the primary specialties, for each procedure, the top part of Table 2 shows the mean volumes and CMS-assigned procedure-specific percentile rankings for each volume-based quintile. The quintiles align with mean volumes and volume percentiles, but the mean number of procedures in each quintile varied considerably (for example in Quintile 5 from 22.06 for open hernia repair to 793.14 for cataract surgery). The bottom part of Table 2 shows the distribution of numbers of providers for each procedure across volume quintiles. At the procedure level, the proportion of providers in the highest 2 quintiles ranged from 39% (prostate resection) to 58% (spinal fusion) and in the highest quintile from 20% (for coronary angioplasty and stenting, open hernia repair, and prostate resection) to 29% (for spinal fusion). At the specialty level, the proportions varied considerably. For example, only 2% of general surgeons were in the highest quintile for colonoscopy while 72% of cardiac electrophysiologists were in the highest quintile for pacemaker insertion or repair.

Table 2:  Mean volumes (and CMS-assigned percentile rankings) for each procedure by volume quintile (top) and distribution of the number of providers by volume quintiles (bottom).  The proportion of procedure-specific providers in the highest and highest two quintiles is provided (bolded numbers indicate the combined proceduralists).

Table 3 provides information on the distribution of specialties when the specialists might have performed more than one of the 12 procedures. For cardiologists, interventional cardiologists, and general surgeons, a smaller number of types of procedures performed was associated with a higher proportion of those specialists being in the highest volume quintile. For orthopaedic surgeons, a higher proportion of specialists were in the highest volume quintile if they performed two of the possible procedures, as opposed to one or all three.

Table 3: Distribution of specialties by the number of providers and the number of procedures they performed by each specialty type. Cardiologists and interventional cardiologists could perform coronary angioplasty and stenting and pacemaker insertion or repair; orthopaedic surgeons could perform hip replacement, knee replacement, and spinal fusion; and general surgeons could perform colonoscopy, open hernia repair, minimally invasive hernia repair, and mastectomy. The percentile of providers and procedures in the highest volume quintile is provided as a function of the number of procedures performed by the specialist.

For each procedure type, at the HRR level, Table 4 shows the minimum and maximum total number of proceduralists, total number of proceduralists per 100,000 Medicare enrolees, and the top two and the top volume-based quintiles; further, the number of HRRs with no providers at all and none in the top two and the top volume-based quintiles is provided. While most HRRs have at least one proceduralist, a substantial proportion of HRRs do not have providers in the top quintile (for example, 33 HRRs do not have a coronary angioplasty and stenting proceduralist, 96 do not have a coronary artery bypass grafting proceduralist, 20 do not have a minimally invasive hernia repair proceduralist, 26 do not have a hip replacement proceduralist, 25 do not have a mastectomy proceduralist, 48 do not have a prostate resection proceduralist, and 37 do not have a spine fusion proceduralist in the top volume quintile). Supplemental Online Table 2 provides data on these variables for each HRR.

Table 4:  For the total number of proceduralists, the number of proceduralists per 100,000 Medicare enrolee, and the highest two and the highest volume-based quintiles, the minimum and maximum number of providers in one of the 306 hospital referral regions (HRRs) and the number of HRRs with no providers.

For each procedure, the Figure shows variation in the number of highest-volume-quintile proceduralists per 100,000 Medicare enrolees at the HRR level. As expected, there are more such proceduralists per-capita for the more common procedures (cataract surgery, colonoscopy, knee replacement, and hip replacement) and fewer for less common procedures (coronary artery bypass grafting, open and minimally invasive hernia repair, and mastectomy). The distribution of high-volume proceduralists varies by procedure as well, with relatively fewer cataract surgery and colonoscopy proceduralists in the Midwest, but relatively more high-volume coronary angioplasty and stenting, knee replacement, pacemaker insertion or repair, and spinal fusion proceduralists in the Midwest.

Figure 1: For each procedure, variation in the number of highest-volume-quintile proceduralists per 100,000 Medicare enrolees at the Hospital Referral Region level.

Discussion

We used publicly available data, generated by CMS to help consumers identify proceduralists with higher volumes for twelve procedures, to identify the types and numbers of specialists who perform those procedures, the volumes of procedures performed, the distribution of procedures and proceduralists across volume quintiles, the geographic distribution of proceduralists, and, for four specialties, the relationship between a specialist performing multiple procedure types and their likelihood of being a high volume proceduralist. We found that only certain specialists should be considered when determining the numbers of procedures performed, that there is substantial variation in numbers of procedures and providers across volume-based quintiles and across HRRs, and that, for general surgeons, cardiologists, and interventional cardiologists, doing more types of procedures is associated with lower likelihood of being in the top quintile for any one of them.

We also found that the data are complex, require some knowledge of how healthcare works, and necessitate a substantial amount of data manipulation. While CMS is to be commended on making data publicly available, the data are not consumer friendly. And, while CMS has incorporated volume percentile data into their “Find & compare providers near you” tool [7], that tool does not provide consumers the opportunity to easily compare volumes or to differentiate who might be primary proceduralists.