A study conducted by the Harvey L Neiman Health Policy Institute (HPI) has identified a large percentage of unidentified interventional radiologists ‘hidden’ within Medicare data. The research has uncovered a previously unforeseen trapdoor, in which self-designated specialty in Medicare claims data have proven to be an “inadequate selection method” resulting in the omittance of practicing interventional radiologists from records.
Mikki D Waid (Harvey L Neiman Health Research Policy Institute, Reston, Virginia) and colleagues reference interventional radiology (IR) as arguably one of the most poorly understood fields of medicine in the introduction to their research. They assert that the overlap in services across radiology has created a “challenge” for identifying interventional radiologists.
The authors note that definitions of IR to date have been “insufficient” for identifying “replicable samples of interventional radiologists for research”. Additionally, they posit that the self-designated physician specialty found in Medicare data—which are typically derived from the Medicare Provider Enrolment, Chain, and Ownership System (PECOS)— have been shown to be “unreliable” and does not accurately represent the true level of self-described interventionists when compared with physicians’ own websites.
“This research does not attempt to make a judgement on which radiologists should be considered interventional radiologists,” Waid and colleagues underline. Rather, their research sought to identify if interventional radiologists are accurately represented in the available data.
The primary sources for their analysis were the 2015–2019 Centres for Medicare and Medicaid Services (CMS) Provider and Payment Utilization and Payment Data Physician and Other Supplier Public Use File (PUF), and 2015–2019 data from Optum’s de-identified Clinformatics Data Mart (CDM) database.
The PUF provides physician-level data including gender, self-designated specialty, state, zip code of practice, and counts and payments by service performed. The CDM database includes full patient-level medical claims data from approximately 20 million members enrolled in commercial and Medicare Advantage plans.
To identify practicing interventional radiologists within PUF and CDM data, Waid et al looked at the number of procedures that were billed as IR-related treatment and rank-ordered radiologists IR-related work as a percent of total-billed work relative value units (wRVUs). Characteristics were analysed at various threshold percentages. The authors also note that external validation used Medicare records with the Society of Interventional Radiology (SIR) membership records.
Their results show that, of the 37,131 unique radiologists, in the Medicare data, 8,010 (21.6% of total unique self-designated diagnostic radiologists and interventional radiologists), 3,572 (9.6%), and 2,060 (5.5%) met the 10%, 50%, and 90% thresholds, respectively. Above a 10% IR-work threshold only 24% of selected practicing interventional radiologists were designated as such in the data; above 50% and 90% thresholds, percentages rose to 42% and 48%, respectively. The average IR-related work (surgical, invasive, and evaluation and management services) reported among practicing interventionists was 45, 84 and 96% of total wRVUs, for the 10, 50 and 90% thresholds. Pulling key takeaways from their results, the authors highlight that in both datasets, most practicing interventional radiologists at all IR-work levels up to 95% were designated as diagnostic radiologists in the source data. They contend that, even at the highest levels of IR-related work effort, no more than 47.6% of interventional radiologists are “appropriately” self-designated in the Medicare sample.
Furthermore, Waid et al identified characteristics that were associated with “greater likelihood” of being a practising interventional radiologist. Among these characteristics were male sex, fewer years since training, practicing in a metropolitan area, and employment within a “smaller” practice. The authors opine that future research exploring the “relationships between workforce, practice, system, and community factors” to understand the fundamental IR workforce dynamics could illuminate “factors that predict patient access to IR services”.
The fluidity of IR in definition and the pervasive overlap of its services has created statistical invisibility for practicing interventional radiologists, the authors salient conclusions assert. “Physicians’ self-designation could vary over time—defined as diagnostic one year and interventional the next. They also do not necessarily always define themselves. Their institution or billing provider may take on this power, particularly in the PECOS system,” the authors state.
“There is an absence of an available gold standard for assigning specialty that reflects both training and medical practice,” Waid and colleagues pose. “Identifying IR physicians in claims data poses a challenge that hinders the progress and comparability of research to broadly characterise and measure the value of this important independent medical specialty.” The authors conclusions demonstrate a need for better representation and definition of IR and call for improved selection method to avoid the omittance of “the majority” of practicing interventional radiologists.
