Attribution

I had a discussion with an IT expert who was well along in building a software system designed to capture both clinical and financial outcomes from routine patient care in both outpatient and inpatient settings. I raised was the question of attribution with him. Presently, he is accepting the payer’s definition of attribution, but this is not going to be a long-term solution. As CMS moves toward “value-based purchasing,” figuring out who is generating which costs is an important issue.

A recent editorial discussed the government’s first steps in this direction under MACRA—the patient relationship codes.[1] The current scheme has five levels of relationship with descriptive paragraphs outlining which codes might be applicable to given relationships. The editorialists raised several issues that are worthy of further thought.
 
“First, clinician self-reporting creates a moral hazard—or worse, could be subject to gamesmanship—in that clinicians may forgo responsibility in especially complex cases or report exaggerated responsibility in straightforward cases.”
 
Although probably not financially driven, yet, I think there is some evidence for this in hospital care for ill patients, particularly those in the ICU. While the notion of the “attending physician” is given lip service, in many cases neither the patient (or family) nor the clinical staff can identify who is “in charge.” If this already nebulous situation is made even more “high-stakes” by having to accept financial liability, I think the concern raised by the editorialists is understated.
 
“Second, the rollout of an additional billing-code modifier risks adding to the administrative burden of an already overburdened clinician workforce.”
 
Burnout continues to receive a lot of attention, as some have finally recognized our current health care “system” is misusing the skills and training of its most expensive group of people—physicians—with negative impacts for all concerned. The issue of attribution is important but figuring out how to capture it on the fly without further reducing the clinical encounter to a billing code in the computer is, I think, the real challenge. I suspect current “AI” efforts could be helpful here—finding ways to capture in real time the link between ordering physician, diagnosis codes, and costs.
 
“Third, because CMS has tried to make the codes as simple as possible, there is a risk that they will not meaningfully distinguish the roles and responsibilities of certain clinicians in the course of caring for a patient.”
 
The example given is a patient who suffers a complication, perhaps due to inadequate primary care, compared to a patient who suffers a complication of the disease, both requiring specialty help. The specialist would report his/her encounter using the same code, but the costs in the first instance might rightly be attributed to the primary care doctor, whereas the latter would not. Of course, our ability to justify our course of action is such that no one would admit the former possibility in “my” patient, so there would need to be some sort of dispute resolution process.
(Unlikely, given the current climate.)

The editorialists call for more efforts to validate the proposed codes in trial situations with reference to the clinical record but admit the logic of developing attribution codes is inescapable. Is there a realistic way to resolve the dilemma? I think the draft system will work in some cases, but not in others. Let me illustrate another way that might be productive.
Some years ago, the hospital brought in an outside consultant to look at our process for care of patients with pneumonia. As originally presented, the data showed the longer the patient stayed in the hospital, the higher the total charge. (Not cost—that was not possible then, or now, to really determine cost, much less the marginal cost of an additional test, procedure, or medication.) Somehow, this did not strike me as a conclusion worthy of a consultation, so I suggested a different
approach.

What if we supposed the patient went home on the “right” day, and we compared costs for all patients discharged on day 4, for example? This analysis showed the real difference in charges depending on two things: charges for testing, both laboratory and X-ray, and the charges for drugs, particularly antibiotics. By looking at patients sent home on days 3 through 6, and ignoring the others, we were able to identify some practitioners who routinely ordered more tests and/or
more expensive antibiotics. Since we were comparing patients who were discharged on the same
day, the argument “my patients are sicker,” could not be used.

Did this analysis produce changes in ordering behavior and reduce costs? No, because we did not institute a program to monitor these data in real time and never considered using the data for anything other than education. Could it be scaled up and work across institutions? Perhaps.

One obvious issue, though, is the data were based on charges, which vary between institutions and payers. But CMS could establish a “median” price for each item in a local area and report intra-institutional charges as well as inter-institutional charge data. Certainly, these and other issues would have to be addressed by committees of experts, but it would still come closer to identifying practice patterns that are not contributing to improved outcomes. And, of course, this approach is very specific to acute hospitalization and does not cover outpatient care, either episodic or continuous. Perhaps, then, we need multiple approaches to deal with different clinical issues. The sort of analysis that is appropriate for a cholecystectomy or an elective hip replacement, is different from what is need for a hip replacement for a fracture following a syncopal episode, and all are different from the care of a patient with chronic congestive heart failure or ESRD. Likewise, the attribution issues will be different. But, of course, this is yet another layer of complexity at a time when CMS is looking for “easy” systems to try and rein in costs.
 
13 January 2019


[1] Takvorian SU, Bekelman JE, Press MJ. MACRA’s Patient Relationship Codes—Measuring Accountability for Costs. N. Engl J Med 2018;379(24):2288-2230. doi: 10.1056/NEJMp1808427.

Further Reading

Accountability
If asked about the greatest advances I have seen, my outside the box answer would be the insight that quality and safety of medical care is as much about system design as it is about human performance. Current efforts to make providers financially accountable, though, threaten the utility of this insight.

Financing Healthcare

Money in Medicine
Money has always been part of medicine, but it seems both quantitatively and qualitatively different now.

New Payment Methods

Performance Measurement
An expert panel has concluded less than half of current measures used by CMS to assess value for primary care services are valid. What does this tell us about current pay-for-performance efforts?

Productivity & Efficiency in Healthcare
Many are focused on efficiency and productivity in healthcare without a clear understanding that the two are not interchangeable. This article introduces the two concepts as they are commonly used.