Editor’s note: In the April issue of Health Affairs, Jean Mitchell reported that “self-referring” urologists, who billed Medicare for both prostate biopsies and the associated surgical pathology services, charged Medicare for more specimens per prostate biopsy than non-self-referring urologists sent to independent pathology providers. Additionally, the regression-adjusted cancer detection rate was higher for men treated by urologists who did not self-refer. “This suggests that financial incentives prompt self-referring urologists to perform prostate biopsies on men who are unlikely to have prostate cancer,” Mitchell wrote. In a separate Health Affairs Blog post published today, Deepak Kapoor and David Penson criticize Mitchell’s methodology and argue that it invalidates her results. Mitchell responds to this critique below. You can hear Mitchell discuss her study here.

Deepak Kapoor and David Penson state that my study, which was published in Health Affairs after rigorous scrutiny including double-blind peer review, is “methodologically flawed and factually inaccurate.” They are mistaken. Below are responses to issues raised in the Kapoor and Penson critique, which either misinterpreted or failed to recognize many important aspects of the study.

Kapoor/Penson Comment 1: 12 cores are the international standard for prostate biopsies.

Response 1: The study objective was to compare the practice patterns of urologists who self-refer to pathology services in which they have a financial interest to those of urologists who do not self-refer.  The study did not evaluate adherence to guidelines. That said, if guidelines determine physician practice patterns, then one would not expect to observe differences associated with financial self-interest. The critique cites three references published in 2009 or 2010 that recommend a 12 core prostate biopsy be performed (Carroll et al. 2009; El-Hakim et al. 2010; Kawachi et al. 2010).   While conducting the study, I reviewed the National Comprehensive Cancer Network (NCCN) 2010 guidelines which recommend that 12 tissue cores be extracted per biopsy procedure (HCPCS code 55700).  If a 10-12 tissue sample biopsy, with each sample placed in a distinct vial was the standard during the time period examined (2005-2007), it is odd that only self-referring urologists appeared to follow this standard.  As noted in the Discussion section of the article, “Guidelines apply equally to both self-referring and non self-referring urologists. There is no reason why a change in guidelines would alter the practice patterns of self-referring urologists alone.”

Kapoor/Penson Comment 2: The author indicates she used a study methodology designed to “investigate emerging market trends,” previously described “in the peer review literature”.  For these precedents, the author cites 3 studies, all written by the same author, and all published in the same non-peer reviewed industry newsletter, which is not indexed in PubMed and is not a serious scientific publication.

Response 2: Insourcing of pathology services by urology groups was an emerging market trend during the time period of the study (2005-2007). The statement that the precedents  cited in my article include three reports by the same newsletter is in error. On the contrary, the cited references (see notes 2, 17-19 in the article) include one report by the Medicare Payment Advisory Committee (MedPAC) and two articles on physician-owned specialty hospitals published in Health Affairs (2003; 2006).  An additional citation was a report by the Center for Health System Change, a highly regarded not-for-profit institution that has been documenting changes in the health care industry that has occurred since 2000.  These four citations are prior examples from the peer reviewed literature that adopted a similar targeted market area design to investigate nascent market trends.  As stated in the article: “The Medicare Payment Advisory Commission, Centers for Medicare and Medicaid Services, and the Center for Studying Health System Change frequently use a similar approach to study nascent market trends. (2,17-19).” (page 742 first paragraph).

For example, the Centers for Medicare and Medicaid Services contracted with RTI International to conduct a study to compare referral patterns, quality, patient satisfaction and community benefits of physician-owned specialty hospitals versus their peer competitors. Both the American Hospital Association and the Physicians’ Hospital Association provided recommendations regarding the market areas to be examined. Site visits were conducted and Medicare claims data for a six month period in 2003 were also analyzed. The final cities included in the analysis were: Dayton, OH; Fresno, CA; Hot Springs, AR; Oklahoma City, OK; Rapid City, SD; and Tucson AZ. Findings from the CMS report were ultimately published in Health Affairs in 2006.

A private company Uropath evidently played an important role in promoting the pathology self-referral business model. Marketing materials dated April 2005 listed names of urology practices in Florida and Texas that had established self-referral arrangements using the Uropath model. During this time period, Uropath held marketing presentations to which they invited prospective clients.  Beneficiaries who resided in counties where Uropath self-referral arrangements were established were included in the study.  (Uropath apparently terminated business activity in 2008).

Kapoor/Penson Comment 3: In addition to using a trade newsletter as a scientific source to justify her study design, she uses the same publication to choose the limited counties included in the analysis. No information was given as to how these counties were chosen a priori, nor was any statistical analysis presented to illustrate why this very small subgroup is a valid cross-section of either her study or control arms.

Response 3: The article reported that self-referral arrangements for pathology services had been established in 13 counties included in the study.  These counties were identified from a variety of sources including reports published in the industry trade newsletter Laboratory Economics, marketing materials prepared by Uropath (the company promoting establishment of self-referral arrangements) dated April 2005, and information reported on the websites of urology groups.  Extensive internet searches and analysis of Medicare claims confirmed that self-referral arrangements involving at least one urology practice had been established before the end of 2007 in each of the 13 counties included in the study.  The counties were also selected to represent different geographic regions. Seven counties without self-referral arrangements were also included as controls to investigate the presence of a contagion effect; that is, whether the presence of self-referring urologists in a county affects practice patterns of non self-referring urologists practicing in the same county.  The selection criteria resulted in a sample of Medicare beneficiaries who resided in a set of geographically dispersed counties throughout the US.

Kapoor/Penson Comment 4:  Hospital practices were arbitrarily excluded. Mitchell arbitrarily excluded from the analysis that hospital based physicians that referred to pathologists using the same tax ID and effectively are in the same group. No rationale for this decision was provided.

Response 4:  Both statements are in error. As stated on page 744 paragraph 1: “About one-third of the biopsies and associated surgical pathology services were performed in hospitals.”  On page 743 the article described the approach to classify cases where the biopsy was performed at a hospital and “the pathology claim number matched the biopsy claim number because the urologist and pathologist were members of the same physician practice plan. Such episodes were therefore classified as not self-referral.” Clearly, these episodes were included in the analysis, not excluded.

Kapoor/Penson Comment 5: The study period and sample sizes are small. Although Medicare data are available for urologists from all regions of the country, she elected to limit her analysis to selected counties. She states that a larger national sample “would probably only include a small percentage of cases treated by self-referring urologists. Thus, any comparisons by self-referral status would likely yield insignificant results.”

Response 5:  The study analyzed data from the time period 2005 thru 2007 for three reasons. First, as noted above, during this time period the “insourcing” of pathology services by urologists was in its infancy but becoming more prevalent as this business model was adopted by large urology group practices.  Second, because insourcing of pathology services by urologists was an emerging market trend during this time period, a national sample would likely have included too small a numerator (i.e. beneficiaries treated by self-referring urologists) to enable statistically meaningful comparisons.  Third, the research project was initiated in 2009 and the study relied on the most recent available Medicare claims data at that time.

Moreover, the sample size is not small.   Over the three year period 36,261 biopsy procedures (HCPCS code 55700) were documented — hardly a small sample.  The approach for identifying men who underwent a prostate biopsy procedure followed a widely cited study by Welch et al. (2007). These researchers analyzed outcomes of 10,429 prostate biopsy procedures (HCPCS code 55700) performed during 1993-2001 among 8,273 men aged 65 years and older who resided in a SEER region.  The authors reported a 32 percent cancer detection rate, consistent with the rates I found among men who had a prostate biopsy procedure performed by a non self-referring urologist (32.4 percent to 35.4 percent).

Kapoor/Penson Comment 6: The author’s most startling finding, that groups with in-house pathology labs had substantially lower diagnosis rates of prostate cancer than other groups, is based on a completely inaccurate methodology and is baseless in fact. She restricted her pool of patients eligible for prostate biopsy to only 5 diagnoses with corresponding International Classification of Disease, Ninth Edition codes (ICD-9), despite there being over 20 diagnosis codes acceptable for prostate biopsy.

Response 6: This criticism is not valid.  The five ICD-9 diagnosis codes listed in the article relate to the office visit with the urologist, not the biopsy procedure. The initial sample was comprised of men who had office visits with a urologist where at least one office visit claim listed one of five ICD-9 codes most frequently associated with prostate biopsy.  Each beneficiary may have had additional office visits with the urologist that listed other diagnosis codes. From this sample, all beneficiaries who underwent a prostate biopsy procedure were selected and analyzed.  The five ICD-9 codes accounted for 81 percent of the diagnoses listed on the biopsy procedure claims, while the remaining 19 percent listed other diagnoses.

Kapoor/Penson Comment 7: She further states, “I examined the ICD-9 diagnosis codes on the corresponding professional or global claims with HCPCS code 88305 submitted to Medicare. If one or more of the claims listed diagnosis code 185, prostate cancer, or 233.4, carcinoma in situ—prostate, I constructed a dichotomous indicator to show if the beneficiary had prostate cancer.”  In fact, there is no absolute requirement that the final pathological diagnosis be used for submission of the Medicare claim; the Common Procedural Terminology (CPT) indicates that CPT code 88305 (level IV – surgical pathology, gross and microscopic examination) is billable for all ICD-9 codes associated with dozens of different CPT code ranges.  Thus, it is highly likely that groups may have billed this code with the reason for the biopsy rather than the result of the biopsy.

Response 7: Critics contend that cancer detection rates as reported are likely biased because it is permissible to list the reason for the biopsy on the pathology claim submitted to Medicare rather than the result of the biopsy.  This contention is not valid.  The rationale for using the diagnosis codes as listed on the pathology claims to identify beneficiaries with prostate cancer was based on Medicare rules (in effect since 2001) for reporting ICD-9-CM codes for diagnostic tests.  CMS requires that,

If the physician has confirmed a diagnosis based on the results of the diagnostic test, the physician interpreting that test should code that diagnosis. The signs and/or symptoms that prompted ordering the test may be reported as additional diagnosis if they are not fully explained or related to the confirmed diagnosis. … Note that physicians are responsible for the accuracy of the information submitted on a bill.

The CMS regulations present a specific example relevant to the study.

Example 1: A surgical specimen is sent to a pathologist with a diagnosis of ‘mole’. The pathologist personally reviews the slides made from the specimen and makes a diagnosis of ‘malignant melanoma’. The pathologist should report a diagnosis of ‘malignant melanoma’ as the primary diagnosis.

Thus, if self-referring urologists misreported the diagnosis on the pathology claims (i.e. reported the reason for the biopsy rather than the result of the biopsy), this would be in violation of Medicare reporting requirements. Moreover, the cancer detection rates found among men treated by non-self-referring urologists (32.4 percent to 35.4 percent) were in accordance with the cancer detection rate of 32 percent reported in the national study conducted by Welch et al. (2007).

Kapoor/Penson Comment 8: The Large Urology Group Practice Association (LUGPA) recently conducted a preliminary review of 8 practices with in-house labs. Their results indicated that with between 2 and 7 years of follow-up, an aggregate of 42,474 prostate biopsies were performed with 16,990 positives or 40 percent.

Response 8: The LUGPA analysis is based on self-reported data from 8 of the 95 LUGPA members. This is a selected sample and there is no mechanism available to verify the data. Their reported 40 percent cancer detection rate is based on 2 to 7 years of follow-up, which cannot be compared to the rates based on a positive diagnosis on pathology claims associated with the initial biopsy procedure.  Including results of follow-up for a period between 2 to 7 years would undoubtedly have resulted in higher cancer detection rates.

Concluding Remarks

The purpose of the study was not to investigate adherence to guidelines, or whether the NCCN or other guidelines are an appropriate standard, or whether significant medical information can be derived from dividing the prostate cores into separately identified (and billable) vials.  The point of the study, and most significant from a policy standpoint, is that financial self-interest as manifested in self-referral of self-generated biopsy tissues specimens was associated with different practice patterns from those exhibited by urologists that did not benefit financially from their referrals. Further, the increase in utilization was linked to lower, not higher, per patient cancer detection rates.

In summary, the policy question is whether the Medicare regulations should accommodate self-referral of services such as surgical pathology as an exception to the federal physician self-referral prohibition. My conclusion is that they should not.  The primary justification for the in-office ancillary services exception was patient convenience, thereby enabling physicians to provide ancillary services (e.g. simple x-rays or clinical lab tests) as part of the office visit. Surgical pathology services are not performed in conjunction with the biopsy procedure. Furthermore, the group practice component of the exception was intended to exclude large multispecialty groups, such as the Cleveland Clinic, where physician members are paid a salary.  The presumption was that referring physicians who belong to large multispecialty group practices would not derive financial benefit from making internal referrals to other physicians within the group.

My study is another example among many studies that demonstrate that financial interest leads to increased utilization. In this case, lower cancer detection rates were also observed. The in-office ancillary services exception has the potential to create conflict between patient welfare and the provider’s financial interest.  Study findings clearly support new legislation to eliminate the in-office ancillary services exception in the federal self-referral prohibition.