Editor’s Note: In October 2009, Health Affairs published two papers on factors driving imaging utilization. One paper, by Jacqueline Baras and Laurence Baker, analyzes the relationship between MRI supply and care for fee-for-service Medicare patients with low back pain. It finds that increases in MRI supply are related to higher use of both low back MRI and surgery. The other paper, by Joseph Ladapo and coauthors, focuses on the adoption of 64-slice computed tomography, which can be used to image coronary arteries in search of blockages. It finds that early adoption is related to cardiac patient volume but also to hospital operating margins.

We asked two experts on the use of imaging technology to review and comment on these papers. Jonathan Sunshine’s commentary appeared ealier, and a commentary by Jean Mitchell, an economist and Professor in the Georgetown Public Policy Institute, appears below.

Published studies and government reports show that sharp increases in utilization of and expenditures for advanced imaging have occurred among both Medicare beneficiaries and nonelderly persons with private insurance coverage (Government Accountability Office, 2008; Mitchell, 2008; Medicare Payment Advisory Commission (MedPAC) recommendations, 2005 and 2009; Office of the Inspector General-HHS, 2007).

For example, the report by the Office of Inspector General at the Department of Health and Human Services found that between 1995 and 2005, advanced imaging reimbursed under the Medicare Physician Fee Schedule jumped from $1.4 million to $6.2 million, a fourfold increase. The utilization rate per 1,000 beneficiaries was also four times higher: 163 in 2005 compared to 42 in 1995. Even though advanced imaging expenditures represent less than 10 percent of total health care spending, the increased use of all types of services is one of the driving forces behind escalating health insurance premiums and therefore is of concern to patients, insurers, the industry, and policy makers.

Clearly, more widespread use of advanced imaging technologies has — and will continue to have — positive effects on patient care through earlier detection of disease, and possibly reduced morbidity and mortality. It would be counterproductive to reverse these advances. Thus, primary efforts at cost containment should be aimed at weeding out inappropriate utilization that has minimal impact on the diagnosis and treatment of patients.

The use of advanced imaging is discretionary and is frequently performed in situations that do not conform with practice guidelines. Because imaging drives the use of other procedures (i.e., surgery), efforts to ensure that advanced imaging is employed in accordance with clinical guidelines could have ripple effects and control the use of other discretionary procedures. Such efforts may also provide the impetus to develop and follow practice guidelines for other types of discretionary procedures (i.e., back surgery).

In addition, imaging procedures, most notably CT scans and myocardial perfusion imaging, expose patients to ionizing radiation. High cumulative doses of radiation can result from undergoing too many advanced imaging procedures and can have deleterious effects on patient health. And importantly, much of the growth in advanced imaging cited above occurs in situations where the ordering physician also performs and profits from the procedure. This situation is known as “self-referral” and presents a conflict of interest for the referring physician.

With these thoughts in mind, I discuss the implications of the two articles on advanced imaging technologies that were published online by Health Affairs on October 14, 2009.  

Use of MRI for Back Pain

The article by Baras and Baker examines the relationship between MRI availability and the diagnosis and treatment of low back pain among fee-for-service Medicare beneficiaries.  The focus on low back pain is compelling and has important implications for reform of the health care delivery system. Clinical guidelines stipulate that patients with low back pain should wait four weeks after the initial diagnosis before an MRI is performed.  This guideline is based on the recognition that most back pain resolves within this time frame (Chou et al. 2007). The authors extrapolate the results from their logistic regression model and predict that each additional scanner is associated with about forty additional MRI procedures among FFS Medicare beneficiaries with nonspecific low back pain. Furthermore, most of the impact of MRI availability on use happens within the first thirty days after the initial office visit. This finding suggests that a significant share of Medicare patients with low back pain receive care that is inconsistent with practice guidelines.  

The authors’ model assumes that MRI availability is the driver of increased use of MRI—in other words, supply (availability) creates demand. However, they do not address the mechanism for this. Specifically, their model fails to acknowledge the asymmetrical information problem that exists between patients and their physicians. A patient afflicted with low back pain will seek care from his or her physician.  The patient lacks the physician’s clinical expertise and thus relies on the physician to act as his or her agent and determine the best course of treatment.  Physicians therefore control the demand for services.

The authors’ empirical model fails to incorporate any variable to control for the role of the treating or referring physician. This omitted variable is significant because many treating physicians are involved in self-referral arrangements.  Self-referral in this context occurs when physicians who are not radiologists prescribe and bill third-party payers for MRI procedures. Considerable evidence indicates that self-referral is a significant driver of greater use of highly reimbursed advanced imaging procedures (MedPAC, 2009; Mitchell, 2008).  It seems likely that the omitted variable identifying physicians who engage in the practice of self-referral would be highly correlated with the variable that measures the availability of MRI.  Thus, some of the MRI availability effect is probably due to self-referral.

Self-referral arrangements are permissible under the in-office ancillary exception in the federal self-referral prohibition. Recent evidence indicates that despite the federal prohibition, the in-office ancillary exception has enabled self-referral arrangements to persist (Mitchell, 2007; Levin et al., 2008). Much of the growth in MRI availability stems from referring physicians’ incorporating imaging into their practice. 

Manufacturers of high-tech imaging equipment are partly responsible for the increasing prevalence of self-referral arrangements, as they aggressively market their imaging products to referring physicians. Their marketing materials typically include slick brochures that illustrate how physicians can augment their income by referring patients for imaging procedures performed on equipment located in their “office..  The definition of the “office” is vast and includes lease and payment-per-click compensation agreements, both of which many legal experts view as kickbacks for referrals (Armstrong, 2005). In light of these considerations, it would have been more informative if the authors had been able to estimate the number of additional MRI procedures performed on patients with low back pain that could be attributable to self-referral.

Hospital Adoption of 64-Slice CT

The article by Ladapo and colleagues address this question: Why do some hospitals adopt new technologies earlier than others? The authors’ focus on the 64-slice CT merits attention primarily because of its capability to perform coronary CT angiography. Use of coronary CT has been the subject of considerable debate due to mixed evidence regarding its clinical effectiveness and cost-effectiveness (Mitka, 2006; Ladapo et al. 2008). The authors estimate an empirical model which posits that the 64-slice CT is more likely to be adopted by hospitals with high operating margins, that are situated in highly competitive markets. Other factors hypothesized to affect adoption include the prevalence of heart disease in the hospital’s market area, and Medicare and private insurance coverage for coronary CT angiography. The model controls for hospital ownership, system affiliation, teaching status, the presence of certificate of need laws, and demographic variables measured at the census track level.

One key variable—the degree of competition in the marketplace—has a significant impact on the likelihood of early adoption, but in the direction contrary to expectations.  The variables that capture the effects of insurance coverage of coronary CT angiography are either insignificant (Blue Cross indicator) or marginally significant with a contradictory sign (Medicare coverage). Both results can be attributable to model misspecification.  

The theoretical model hypothesizes that hospitals will be more likely to adopt the 64-slice CT if they are located in more competitive markets. The authors’ results indicate the opposite is the case: hospitals located in less competitive markets are more likely to adopt this technology than are facilities in highly competitive markets. This result probably stems from both measurement error and omitted variables bias.

The Herfindahl index (the indicator of market competition) is measured with error because 20 percent of the hospitals in the American Hospital Association (AHA) survey did not report whether they owned a 64-slice CT. Omitted variable bias arises because the model fails to account for acquisition of the 64-slice CT by free-standing centers and physicians’ offices.  This variable is probably highly correlated with hospital adoption of the 64-slice CT. I suspect that hospitals are more fearful of, and will be more responsive to, competition from CT scanners located in nonhospital settings than competition from scanners in other hospitals located in the same market area. 

The lack of significance of the insurance coverage variables can be explained by multicollinearity. Many private insurers follow the coverage decisions adopted by Medicare. This fact implies that coverage of coronary CT angiography by private insurers will be highly correlated with Medicare coverage. While multicollinearity does not bias the parameter estimates, it does produce inflated standard errors; the latter will result in insignificant t-statistics. This specification issue could be resolved by constructing a single variable to capture the effects of coverage by either Medicare or Blue Cross.

The increase of CT scanners in nonhospital settings is of greater concern due to self-referral.  When cardiologists have financial interest in CT scanners, either through direct ownership or some type of lease agreement, they have an incentive to order procedures that may have a negligible impact of the diagnosis of a patient’s heart disease.  This excessive testing drives up health care expenditures and insurance premiums without any evidence that it results in commensurate improvements in patients’ health (Mitchell, 2007; Redberg and Walsh, 2008; Appleby, 2008). Moreover, there is concern that patients who undergo multiple CT scans could be harmed by excessive exposure to radiation.   


Both articles, while informative, fail to address what I believe is the most significant driver of increased use of MRI for nonspecific low back pain and of hospitals’ adoption of the 64-slice CT scanner: self-referral. If Baras and Baker could identify self-referral physicians, I suspect their findings would show that the likelihood of undergoing an MRI is much higher for patients treated by referring physicians who stand to gain financially from each MRI scan performed.  Along the same line, if Lapado and colleagues had the data to capture competition from CT scans performed in nonhospital settings, I expect that this factor would be the driving force behind a hospital’s decision to acquire a 64-slice CT machine.

Our efforts to bend the cost curve and rein in health care spending on procedures and tests that yield minimal improvements in the diagnosis, treatment, and overall health of patients will not succeed unless insurers and policy makers are willing to eliminate the in-office ancillary exception in the existing self-referral law and adopt an evidence-based approach to insurance coverage. (Among its imaging-related provisions, the recently enacted health reform legislation does require self-referring physicians to notify patients in writing that the physician will benefit financially from the procedure and that the patient may receive imaging services outside the physician’s practice.) As Senator Max Baucus noted in his white paper on health reform: “The issue of self-referral must be reviewed in light of how health care is and will be delivered. No serious effort at reform can ignore the potential gaming that financial conflicts may create.”