Self-Referral of Imaging Does Not Imply Overutilization
Michael Hutchinson, MD, John B. Chawluk, MD, Camilo Gomez, MD, Jack Greenberg, MD, Francis D. Hussey, MD,
William G. Preston, MD, Earl Zimmerman, MD
From the NYU Langone Medical Center, New York, New York (MH); Drexel University, Philadelphia, Pennsylvania (JBC); Alabama Neurological Institute, Birmingham, Alabama
(CG); Drexel University, Philadelphia, Pennsylvania (JG); MRI Centers of New England, Newton, Massachusetts (FDH); UCI School of Medicine, Neuroscience Medical Group,
Laguna Hills, California (WGP); and Albany Medical Center, Albany, New York (EZ).
Acceptance: Received July 17, 2008, and in revised form July 17, 2008. Accepted for publication July 22, 2008.
Correspondence: Address correspondence to Michael Hutchinson, MD, NYU Langone Medical Center, 530 First Avenue, Suite 5A, New York, NY 10016. E-mail: michael.hutchinson@med.nyu.edu.
A B S T R A C T
For several years, some sectors of the specialty of Radiology have complained about the practice of self-referral, where a nonradiologist physician provides and interprets an imaging procedure. It is argued that such practice leads to increased costs since a physician will overutilize technology because of financial incentives. Here we review the literature. The most extensive analysis to date is at odds with the conclusions drawn in the older literature in that it provides little, if any, evidence for overutilization.
We performed our own investigation using a poll study and found no suggestion of overutilization in 33 self-referring neurologists when compared with 900 neurologists who referred imaging to radiologists. The main period of growth in demand for imaging was between 1999 and 2002. Since 2002 there has been a steep decline in the rate of growth, so that it is possible to predict roughly zero growth in MRI utilization by 2009 without any intervention. It is shown why the rise in demand for imaging studies cannot be explained by self-referral, and it is argued that the sudden expansion of demand 8 years ago was caused by simultaneous technological improvements in the 3 major imaging modalities. Finally, it is shown how self-referral may actually reduce costs by facilitating the transfer of care from the hospital and ER to the office.
Introduction
Along with all the bad news, something positive is happening in Medicine: a quiet revolution driven by MRI. It is now possible to make pathological diagnoses in living tissue without using a scalpel and to make clinical decisions that do not involve educated guesswork, which will save lives, improve health, and reduce costs. MRI is a technique of breathtaking range. It is also now relatively inexpensive. A scan today is vastly superior to one from 1990 and yet costs about one-fifth as much after adjusting for inflation.
Yet in spite of this, it is still widely held that MRI is expensive. It should therefore be understood that despite increases since 1999, and despite all that it delivers to modern Medicine, MRI absorbs only three quarters of 1% of all U.S. healthcare spending, about the same as in 1990.1 Moreover, its rate of growth has been in steep decline for several years so that it is possible to predict roughly zero growth by next year. It is also widely held that lacking more strict controls, the United States has more MRI scanners than any other country in the world.
This view is also incorrect. At present, Japan has twice as many MRI scanners per million citizens, and 7 other OECD countries, including Italy, have overtaken the United States in this regard.2
As MRI technology has become more refined, its range has expanded and its applications have become more closely tailored to the increasing complexity and sophistication of specialized care. For this reason, physicians such as cardiologists, orthopedic surgeons, and neurologists, specializing in organ systems, have become increasingly involved in its application and interpretation. These physicians possess a detailed knowledge of anatomy, physiology, and pathology, a background extremely well suited to the complex task of interpreting images.
In addition, their knowledge of the clinical presentation of disease is a great asset for placing the imaging findings in the appropriate context for best patient care. This welcome development has raised concern among certain radiology groups, which have made strenuous efforts over the last decade to restrict imaging to themselves.3-5 Regrettably, radiology management corporations now sell their services to insurance companies to restrict imaging by adding time-consuming and expensive layers of bureaucracy for the ordering physician. Approved studies are directed only to radiologists.
The Self-Referral Scare
The argument for monopolistic control of imaging by radiologists centers on the presumption of overutilization by physicians who refer patients to their own imaging facilities. The Radiology lobby calls this “self-referral,” that is, the practice by which a physician performs a procedure he deems appropriate. If a surgeon recommends surgery, this lobby is silent, but if imaging is self-referred it will be “overutilized,” since physicians purchase scanners for the sole purpose of making money (even though there is nowadays little money to be made). Nevertheless, this extreme view holds that self-referral is by definition unethical.4 The notion that physicians may wish to enhance their ability to practice Medicine is never considered. If physicians had to refer to radiologists, it is argued, fewer studies would be performed and costs would be reduced. A corollary is that the current model, with radiologist as middleman for all imaging studies, is the most economical. This is a concept that deserves careful scrutiny.
80 Copyright ◦C 2008 by the American Society of Neuroimaging
Those who adhere to this viewpoint make repeated reference to a paper from 1990, where Hillman and colleagues analyzed referrals for imaging studies (mainly chest X-rays) by doctors who owned imaging devices, compared with doctors who did not.6 Self-referrers were said to obtain studies at 4 times the rate of doctors referring to radiologists. The number 4 has become familiar to anyone listening to lobbyist David Levin, who has now inflated it to “4 to 8.”5
But even if it were possible to prove that self-referring physicians used to order more chest X-rays, and that this necessarily represented overutilization rather than standard of care—an essentially impossible task—it is reasonable to ask why some radiologists continue to cite studies in a bygone era to explain modern Medicine. For the world has changed in significant ways.
First, self-referral is strictly regulated by the Stark laws, which require that physicians who own imaging equipment must form large groups, and divide any profit equally to avoid direct incentives to self-refer. Second, the technology has changed. Compared to an X-ray that could be performed in seconds in a back office,MRI scans typically take 30 minutes. This means that no more than about 28-32 scans can realistically be done in 1 day.
For a group of 10 physicians, the maximum rate of referral is about 3 scans per physician per day. If a group wanted to order more studies, it would have to purchase another scanner, incur increased costs and risk diminishing returns.
A Direct Look at Self-Referral
The claims of Hillman and Levin simply do not comport with our collective experience. We therefore decided to start with a very simple survey of outside groups, using a straw poll and asking a direct question: how many MRI scans do you actually request?
In 2005, the American Academy of Neurology polled 900 neurologists, selected at random, who refer to radiologists. They ordered a mean of 1.8 imaging studies (MRI and CT) per neurologist per day.
In 2007, we surveyed self-referring neurologists, examining 3 large groups where MRI scans have been performed for several years subject to the Stark laws. None of the authors belongs to these groups or has financial ties with them. The groups were simply asked to add up the number of MRI scans that were selfreferred in the previous month.
In all, there were 33 neurologists in these groups (8, 10, 15). The largest group had 2 MRI scanners. The average number of self-referred scans was 2.0 per neurologist per day (1.8, 2.4, 2.0), about the same rate as neurologists referring to radiologists in 2005. If Levin were correct,5 the rate would be 8-16 scans per neurologist per day.
Furthermore, if self-referring physicians had been tempted to overutilize imaging, the rate would be expected to be close to the theoretical maximum of 3. The fact that it was substantially lower, underscores the lack of evidence for overutilization.
The fact that 2.0 is greater than 1.8 cannot of course be taken to imply that self-referrers order 10% more studies than radiologist-referrers. The point is that Levin’s assertions of 300- 700% increases or 4-8 times the rate are not even remotely consistent. Indeed, a rate of 8-16 scans per physician per day would be far beyond the physical constraints of MR imaging.
The Origin of the Self-Referral Argument
Hillman’s 1990 paper6 is an interesting historical examination of a bygone era. Claims for 4 diagnostic codes were examined, the large majority of these being for upper respiratory symptoms with the chest X-ray as imaging modality. The study design is complex with layers of inclusion and exclusion criteria that give the superficial appearance of thoroughness. However, the paper suffers from a general lack of clarity. First, the use of percentages obscures the fact that the average self-referring physician saw about twice the number of patients as the average radiology-referring physician. This very large discrepancy raises questions as to differences in patient populations and styles of management, yet it was not addressed. Second, large numbers of subspecialty physicians, who might ordinarily be expected to refer to radiologists, were excluded for undisclosed reasons. Instead, the reader is directed to a microfiche that could be obtained after a cash transaction from a P.O. Box at Grand Central Station. Third, there is no explicit acknowledgment that physicians referring to radiologists may have been underutilizing imaging. The number of office visits is missing, which is important since the traditional way of managing an episode was to follow the patient with frequent office visits rather than an X-ray. The costs for these two approaches might have been similar but we cannot know because total costs are not given. We therefore agree with Hillman’s statement that “it is not possible to determine which group of physicians used imaging more appropriately”6—although it is interesting that neither
Hillman nor Levin would ever again profess to any doubt on this issue.3-5 In a later study sponsored by the National Union of Mineworkers (NUM),7 Hillman directly probed for overutilization by not reimbursing self-referred imaging services between 1993 and 1994. Instead of seeing decreased utilization, the rate actually increased 12%, and the NUM revoked its ban on selfreferral. There were no large comprehensive studies of self-referral untilGazelle et al.,8 which is analyzed below. The bulk of studies that do exist, such as those of Mitchell,9 merely document a steady increase in imaging by nonradiologist physicians, and do not make a case for overutilization by the same physicians.
The Gazelle paper8 is a very large, clear, scientifically balanced study on self-referral focusing on modern imaging techniques in the era of Stark-regulated activity. In this paper, the authors correctly criticize prior studies on self-referral.5,6
Eight categories of disease were studied using imaging modalities ranging from MRI to SPECT. After excluding hospital based imaging, the rate of MRI self-referral by neurologists for stroke was found to be only 1.12 times the rate at which neurologists referred to radiologists.
In fact, in 6 out of the 8 disease categories, similar rates of self-referral and radiology referral were found. In 2 categories (chest X-ray for pulmonary disease, and SPECT studies for cardiac disease), however, the relative rates were about 2.5.
Hutchinson et al: Self-Referral of Imaging Does Not Imply Overutilization 81
Despite its size and thoroughness, the Gazelle paper has 3 major design flaws. First, hospital referrals were excluded. This is crucial, since for many doctors it is still standard of care to admit patients to hospitals for “tests,” which almost invariably include imaging studies performed by radiologists. By excluding these common referrals, it is likely that Gazelle seriously underestimates the number of radiology referrals for particular disease categories. The second flaw is that specialists referring imaging studies to other specialists outside their immediate group, but in the same specialty, were nevertheless counted as self-referrers. They should instead have been counted as non-self-referrers. We illustrate this by a hypothetical.
Suppose there are 100 cardiologists and each appropriately sends 1 patient per day for a cardiac study. Fifty cardiologists have nuclear medicine devices, and send all their patients to these devices (self-referral). The other 50 cardiologists send their 50 patients outside their group for nuclear studies (non-self-referral). Twenty five of these 50 patients are sent to radiologists, 25 to cardiologists. The 25 patients sent to other cardiologists are designated as “self-referral” by Gazelle, which means that in total there are 75 “self-referrals” and only 25 radiology referrals. Thus, the relative rate of “self referral” to non-self-referral is 3. If on the other hand the 25 patients are correctly classified as “non-self-referral” then there are 50 self-referrals and 50 non-self-referrals, and the relative rate is 1.
Finally, in the evaluation of coronary disease, imaging can be done in at least 2 ways: cardiac stress echocardiogram, and SPECT. If a cardiology group has a SPECT scanner it can be expected to use it exclusively. Therefore, by focusing only on SPECT referrals rather than total imaging, Gazelle is introducing yet another bias that will suggest overutilization even where none exists. These biases are substantial. When they are corrected, therefore, it seems likely that the numbers will converge to the conclusion that, subject to the Stark laws, physicians who self-refer imaging do so at about the same rate as those who refer to radiologists.
Self-Referral Can Lower Costs and Improve Quality
We now turn from the past to outline a model for the future. At present, a substantial portion of healthcare expenditure goes to hospitalizations and ER visits. We outline here how the Stark laws could spark a revolution in the practice of Medicine, by shifting inpatient and ER care to the outpatient office. We consider three common examples in Neurology where delay can be avoided and expensive ER visits eliminated. This requires rapid integration of both clinical and imaging resources in the same facility.
Transient ischemic attack (TIA) is a warning sign of possible strokes to come. Most patients are sent to the Emergency Room, where a head CT is always performed before admission to the hospital for further tests. These tests invariably include intensive neuroimaging, and the whole process can take 2-3 days at considerable expense.
In an ideal world a TIA would not warrant admission to a hospital. Indeed, the importance of rapid assessment was recently underscored by a study that showed an 80% reduction in the risk of early recurrent stroke through rapid assessment in the outpatient setting.10
With modern MRI the entire workup—including blood tests—can be completed in about 1 hour: MRI of the head, MRA of the blood vessels, and MR cardiogram done at a fraction of the cost and discomfort of hospitalization. The risk factors for further strokes can be quickly identified and modified with medication as an outpatient. We estimate that only 5% of patients would have to be hospitalized after this outpatient work-up, mainly for carotid stenting or carotid surgery. Such procedures could now be performed immediately in same-day surgical suites.
Multiple sclerosis attacks consume resources in a similar fashion. The patient comes to the Emergency Room where a head CT is performed. The patient is then admitted to the hospital for an MRI and several days of high-dose steroids. It would be far more convenient and far less expensive to have an MRI scan immediately and to be treated with intravenous pulsed steroids as an outpatient, and then sent home. Intravenous infusions could then be continued daily as an outpatient in those who are not severely disabled.
Migraine headaches are one of the most frequent causes of ER visits with the statutory head CT and 5-hour wait. As an outpatient visit, a simple exam and history could establish in most cases that no imaging is necessary, or provide if it is. A simple infusion follows, and the patient can be sent home or back to work.
These are commonplace examples of modern, cost-effective care, centered on clinical experts who have ready access to, and knowledge of modern imaging. We estimate that the cost savings from these three diagnoses alone could be as much as several billion dollars a year.
Final Comments
Although still less than 1% of total healthcare expenses, MRI costs have risen since 1999 along with the costs of other imaging modalities. It must first be emphasized that this cannot be explained by abusive patterns of self-referral, since a large majority of studies were performed by radiologists. For Medicare between 2000 and 2004, radiology referrals accounted for almost all of the increase in imaging costs (86%) with 80% of the increases going to hospital based imaging.11
The contrary impression created by Levin et al.3-6,9 is fully explained by the specious use of percentages. That is, by documenting the fact that self-referred scanning has increased in recent years, these lobbyists create the illusion that this is what is driving costs. They then inappropriately equate increases in self-referral to increases in costs, as Gazelle points out.8
Therefore, the likeliest explanation for increased utilization is that advanced imaging techniques are now infusing themselves into every corner of Medicine as the technology has improved. At more or less the same time, in the late 1990s, advances in technology allowed for dramatic improvements in imaging capability. In CT, the introduction of multiple channels allowed angiographic studies of the coronary, pulmonary, and cerebral arteries. This in turn led to the replacement of invasive angiographic and nuclear medicine studies for the evaluation of heart disease, pulmonary embolism, and cerebral aneurysm. In MRI, the introduction of parallel acquisition enabled the imaging of thoracic and abdominal structures for the first time. Finally, the introduction of total body PET scans replaced older and less reliable nuclear medicine studies in the evaluation of metastatic disease. Indeed, the rise of new applications for the 3 new imaging modalities has been accompanied by a decline in traditional nuclear medicine. The concomitant savings are not apparent when only the new modalities are considered. The main period of rapid growth in CT, MRI, and PET was between 1999 and 2001, on the heels of these advances in technology. 1
Between 2001 and 2005, there was a steep decline in the rate of increase in all 3.1 By 2006, the growth in MRI was only 8%,1 which extrapolates to zero growth by 2009. These numbers are for Medicare reimbursed services and are not influenced by intervention from radiology management corporations. Indeed, it is worth noting that the savings realized by one of the largest of these corporations is as little as $8 per member per year.12 When the financial cost of this oversight is included, the net savings are even less, a sad outcome indeed for all the bureaucratic pain inflicted on physicians. We looked at referring practices of 3 large groups of Stark regulated elf-referring neurologists and found no suggestion of over-utilization. Our very simple straw poll is actually the first direct look at self-referral, and its finding comports with the largest, most comprehensive study to date.8
This is not to say that there has been no abuse of self-referral. Loopholes allowed by the Stark II regulations, such as special leasing arrangements, have been exploited by a few physicians, and we welcome their closing by Stark III.
Finally, we have given examples of how self-referral can lead to meaningful reductions in costs, as well as improved care, by moving patients from the ER and hospital to the outpatient setting. Extrapolating to all diagnoses in all fields, such savings would likely be very substantial. A shift from hospital-based imaging to outpatient imaging is therefore to be expected. Such a positive trend can be delayed by self-interest disguised as ethical behavior. Medicare and other agencies concerned with health-care costs will do well to assess total expenses in a rapidly changing world.
Drs. Chawluk,Hutchinson, Gomez and Preston are self-referring physicians.
References
1. Department of Health and Human Services Report of the Office of Inspector General. October 2007. Growth in advanced imaging paid under the Medicare physician fee schedule. 19.
2. OECDHealth Data. July 2007. Available at (www.oecd.org/health/ healthdata)
3. Levin DC. Me AndMy MRI . New York TimesOp Ed, July 6th 2004.
4. Hillman BJ. Trying to regulate imaging self-referral is like playing whack-a-mole. Am J Roentgenol 2007;189:267-268.
5. Levin DC. High tech imaging utilization soars among nonradiologists. Imaging Biz 2007;2.
6. Hillman BJ, Joseph CA, MabryMR, et al. Frequency and costs of diagnostic imaging in office practices – a comparison of self-referring and radiologist-referring physicians.NewEngl J Med 1990;323:1604- 1608.
7. Hillman BJ, Olson GT, Colbert RW, et al. Responses to a payment policy denying professional charges for diagnostic imaging by nonradiologist physicians. JAMA 1995;247:885-887.
8. Gazelle GS, Halpern EF, Ryan HS, et al. Utilization of diagnostic medical imaging: comparison of radiologist referral versus samespecialty referral. Radiology 2007;245:517-522.
9. Mitchell JM. Utilization trends for advanced imaging procedures: evidence from individuals with private insurance coverage in California. Med Care 2008;46:460-466.
10. Rothwell PM, Giles MF, Chandratheva A, et al. Effect of urgent treatment of transient ischemic attack and minor stroke on early recurrent stroke: a prospective population-based sequential comparison. Lancet 2007;370:1432-1442.
11. Kuhn H. Testimony before The House Subcommittee on Health. July 17 2006. Available at (www.hhs.gov/asl/testify/t060718.html)
12. Brice J. Privileging limits access to imaging, cuts insurers’ costs. Diagnostic Imaging 2006.
