Can Clinical Decision Support Reduce Inappropriate Utilization of Imaging?

The rapid growth of medical imaging has placed it under scrutiny. ¹ 

 Radiologists responded to the demand in imaging by provider-centric models that emphasized ease of ordering and turn-around of reports. Unsurprisingly, unquestioned emphasis on service contributed to overutilization of imaging: it's often easier repeating a CT than fetching images acquired at another institute.

Efficiency placed radiologists and referring clinicians in different orbits. Radiologists, focusing on timely reporting in a fee-for-service world, paid scant attention to appropriateness of request. Many providers grew complacent with pertinence of information when requesting. It was never demanded of them.

Indications stated for imaging are the folklore of radiology. One purportedly asked for exclusion of "serious pathology." Thankfully, physicians still take pathology seriously.

Payers controlled utilization by interposing radiology benefit managers and requiring pre-authorization for advanced imaging. ²

Concurrently, imaging clinical decision support emerged to optimize diagnosis and reduce errors. Another use of CDS became apparent: management of inappropriate utilization.³

Multiple institutes produced CDS for appropriate imaging including Brigham and Women's Hospital and Massachusetts General Hospital, both in Boston, and Virginia Mason Medical Center in Seattle. The American College of Radiology has also developed a CDS tool, partnering with National Decision Support Company to commercialize a CDS based on its appropriateness criteria, ACR Select.⁴

Does clinical decision support work? 

Remarkably, yes^5.  When reminded of situations where imaging is inappropriate, some physicians' ordering changes.

CDS reduced frequency of CT ordered for suspected pulmonary embolism, according to two studies.^{6, 7} How? By asking if the d-dimer (a biomarker which is almost never negative in PE) was negative. If negative, a pop-up reminded the physician that PE was exceedingly unlikely in the face of a negative d-dimer and low clinical likelihood of disease.

That CDS dissuades physicians from ordering a test like Yelp's low ratings dissuades me from visiting a restaurant makes clinical decision making seem unscientific. Surely, once possessed of a diagnosis the clinician will investigate. A lot of medicine, however, is practiced in the gray zone. This zone is responsible for overutilization of imaging and is amenable to nudges.

How does ACR Select work? 

CDS integrated into computerized physician order entry scores appropriateness for a test for a clinical scenario. Scores have been developed by expert panels from the ACR after review of literature. The nominal scale ascends from one (least appropriate) to nine (most appropriate).

ACR Select abolishes the option of solely using free text, although free text remains for adjunct information. Physicians must choose a clinical condition (e.g., epilepsy) or indication (e.g., headache). Unhelpful information such as "rule out pain" no longer suffices.

 Decision support is either indication-driven or modality-driven. The distinction is important to appreciate. For example, in the investigation of a 22-year-old athlete suspected of arrhythmogenic right ventricular cardiomyopathy, a rare disorder of the heart which causes sudden cardiac death, cardiac MRI receives a score of 9 (see Figure 1).

	Figure 1: The interface of ACR select. In this example the physician wants to investigate ARVC. MRI receives a score of 9 and CT receives a score of 1 .

If clinician wants a cardiac CT and chooses ARVC, because it is alphabetically the first indication (Figure 2), even though not of concern, the score is 1, indicating the test may be inappropriate.

Figure 2. The physician wants a cardiac CT but chooses ARVC as the indication. The appropriateness score is 1. The system warns of the physician of the low appropriateness of the study. In order to proceed with the study physician must speak to the radiologist. This is an example of a “hard stop” in active decision support. This can be turned on or off for a particular physician.

In other words, if a test is chosen to fit the indication the test is likely to score high. If, however, an indication is chosen to fit the test, the test may score low.

Decision support in action: chronic headache

If a clinician investigates a patient with longstanding headache, the CDS will ask if new symptoms are present (Figure 3). If new symptoms, MRI receives a score of 8 and CT a score of 5. If no new symptoms then both receive a score of 4.

This encounter highlights CDS's potential and requirement of interdependent factors for success.

For example, if a co-payment is dependent upon the cost of the test and its appropriateness, with a higher co-payment for an MRI than a CT in the above example, moral hazard can potentially be reduced.

Figure 3 (a). In the investigation of a chronic headache the physician must declare if new symptoms. If new symptoms then MRI scores higher than CT (3 a). If no new symptoms then both MR and CT scores on the intermediate side (3 b).

Stakeholders

Stakeholders include payers, hospitals, referring clinicians and radiologists. The possible response of stakeholders to CDS is considered in the following settings: outpatient, emergency department and inpatient.

Outpatient

If, as some projections suggest, CDS manages utilization at least as well as radiology benefit managers, payers should be supportive, as transaction costs are lower. CDS costs 5 cents per member per month whereas RBMs cost 30 cents PMPM.^8,9 This can be simply understood as the difference between booking travel online versus waiting on hold with a travel agent.

Clinicians who incur time and office costs in obtaining pre-authorization might also prefer CDS. CDS integrated in to EMR would facilitate discussions about imaging with patients at the point of ordering, reducing the uncertainty of obtaining approval for advanced imaging.

Assuming the fee-for-service model, hospitals might view CDS less favorably because reimbursements may decline from imaging studies forfeited. This is offset by greater collection from payers who might be less inclined to deny payment because of insufficient clinical information or low appropriateness.

 Emergency Department

Payers gain from utilization management in the emergency department, where RBMs do not typically intervene.

Hospitals will lose from compensated studies not performed but lose less heavily from uncompensated imaging not performed.

Response of emergency physicians is hard to predict. ED is measured on throughput and timely negative imaging frees a bed for another patient. Some might view CDS as an unhelpful hoop to jump through. Some might find relief in evidence-based reasons explicitly supported by a specialist society for foregoing imaging, which could be exculpatory.

In-patients

As payments are bundled, marginal gain from a diagnostic test is lower than marginal costs. Payers could be indifferent but hospitals would gain from curtailment of inappropriate and uncompensated imaging.

Residents and hospitalists who order many of the diagnostic tests would benefit educationally from the literature used to determine the appropriateness of imaging, which is available via a single mouse click in ACR Select.

Radiologists

ACR Select is a product of radiologist's professional organization. Yet, reaction of radiologists is arguably far from uniform, and depends on attitudes to utilization management. "Gatekeeping" has sinister connotations from era of health maintenance organizations. However, radiologists are the key to the success of ACR Select, which in turn gives them the opportunity to position themselves as the experts who can guide the referrer in ordering the most appropriate test. ¹⁰

It is important to nuance the clinical options according to the specialty, so that physicians face decision support template that contains their most common clinical indications: a fine balance between achieving granularity and avoiding complexity. Radiologists will know this a priori and could work with the medical informatics team so that the clinical indications for cardiologists are different for gastroenterologists.

Once switched on, decision support can be active or passive. In active mode, physicians must discuss with radiologists if they wish to go ahead with imaging deemed of low appropriateness (see Figure 2). In passive mode the physician is warned of the low score but can proceed with ordering.

Not all physicians overutilize equally. Radiologist could identify providers who most benefit from active decision support without punishing other providers, by tailoring the level of decision support.

Analytics

Information stored by CDS can be prescriptive and valuable for research. Providers can be given feedback on the appropriateness and positivity of their imaging requests in relation to local and national norms as well as their peers in their own department. Crucially, CDS can be used to establish these norms.

If providers repeatedly order tests of low appropriateness for a particular indication, this could signal uncertainty in a diagnostic area which might benefit from investment in research in technology that reduces uncertainty.

Institutional variation in ordering patterns can be identified and reasons for variation, such as the risk pool of the institute, explored. CDS can yield information that enables risk adjustment, a crucial variable in assuring fair reimbursement for quality and value.

Moving ahead 

ACR Select is electronic medical record-vendor neutral. It is scalable. Importantly, it must adapt to emerging evidence and feedback. This is a managerial rather than a computing challenge because revision of appropriateness criteria requires convening of experts.

Proof of the pudding lies in implementation. The main impetus for CDS will come from payers, including CMS, who could consider a randomized trial of RBMs versus CDS looking at utilization, costs and effectiveness.

The U.S. House of Representatives is considering a bill to mandate CDS in order to be reimbursed for imaging.¹¹ The bill does not restrict the source of CDS. However, multiplicity of options might confuse stakeholders. Peter Pronovost observed that the absence of a single supplier in healthcare, such as Boeing for the airline industry, means that parts often do not communicate with one another.¹²

The advantages of choice in the market for computers might not be realized in the market for CDS as there is considerable knowledge base to update and synchronize with expanding regulations. The payers and hospitals will be important in determining the sole source of CDS.

Will ACR Select Work? 

The American Board of Internal Medicine and the ACR have campaigned to reduce waste by Choose Wisely and Image Gently initiatives, respectively. Awareness alone is not sufficient to curb overutilization.

The large scale effectiveness of ACR Select requires a resource-conscious medical culture, which still seems distant. Technology cannot change culture but cultural change can be eased by technology. Early adopters will be at an advantage when the economics of imaging change with capitated payment models.

If there is will, then ACR Select could bridge the archipelago of stakeholders who would benefit from quid pro quo understanding. Payers could mandate use of CDS but assure that when imaging dutifully meets appropriateness criteria providers do not incur the wasteful billing merry go-round.

Dr. Jha* is an assistant professor of radiology at the University of Pennsylvania. His scholarly interests include value of imaging, economics of information and appropriate utilization of imaging. He can be reached at saurabh.jha@uphs.upenn.edu. He can be followed on Twitter: @RogueRad

*Dr. Jha is a member of the American College of Radiology. He disclosed no financial relationships with National Decision Support Company.  

1 Emanuel, E.J., & Fuchs, V.R. (2008). The Perfect Storm of Overutilization. JAMA, 299, 23, 2789-2791.

2 Mitchell, J.M., & LaGalia, R.R. (2009). Controlling the escalating use of advanced imaging: the role of radiology benefit management programs. Med Care Res Rev., 66, 339–351.

3 Khorasani, R. (2001). Computerized physician order entry and decision support: improving the quality of care. RadioGraphics, 21, 4, 1015–1018.

4 http://www.acrselect.org/ (accessed February 6, 2014).

5 Blackmore, C.C., Mecklenburg, R.S., & Kaplan, G.S. (2009). Effectiveness of clinical decision support in controlling inappropriate imaging. JACR, 8, 19–25.

6 Raja, A.S., Ip, I.K., Prevedello, L.M., Sodickson, A.D., Farkas, C., Zane, R.D., Hanson, R.,..Khorasani, R. (2012). Effect of computerized clinical decision support on the use and yield of CT pulmonary angiography in the emergency department. Radiology, 262, 68–474.

7 Hoo, G.W., Wu, C.C., Vazirani, S., Barack, B.M. (2011). Does a clinical decision rule using D-dimer level improve the yield of pulmonary CT angiography? AJR, 196, 1059–1064.

8 Vinz, C. A case for decision support of imaging services. http://www.diagnosticimaging.com/radblog/display/article/113619/1932985 (accessed February 6, 2014).

9 Lee, D.W., Rawson, J.V., & Wade, S.W. (2011). Radiology Benefit Managers: Cost Savings or Cost Shifting? JACR, 8, 6, 393 – 401.

10 Clark, C.E. (2010). The role of radiologists in unnecessary imaging. JACR, 7, 746–747.

11 Excellence in Diagnostic Imaging Act, HR 3705.

12http://armstronginstitute.blogs.hopkinsmedicine.org/2013/01/18/connecting-medical-devices-and-their-makers-3/ (accessed February 6, 2014).