Prescription for Change

Using claims to analyze physician performance? First, check your data.

by Eric Fontana

It’s no secret that clinical and process standardization may be one of the most promising levers for cost reduction at many organizations. And with an unprecedented level of data available, many hospital executives are turning their attention to examining claims data to understand physician performance in areas like cost, length of stay, or quality to identify opportunities for organization-wide improvement.

Put simply: better understanding which of their physicians are not performing at the level of their peers and ultimately incurring higher costs for the enterprise.

But claims data has limitations, as we saw last year with the CMS release of physician charge data that ostensibly identified individual physicians performing millions worth of Medicare services. Claims don’t always reflect reality for a variety of reasons. In one instance, a single physician was attributed to every claim that went through a single site of service. Why? Partly because no “gold standard” for physician attribution exists.

So if you’re preparing to sit down with one of your physicians for a hard conversation about performance based on a claims-level analysis, you may want to ensure attribution has been verified with a secondary source, like an EMR. Otherwise you may find yourself in a standoff with a skeptic who (maybe correctly) disputes that your data is a misleading representation of what’s actually happening.

Do you have an attribution problem? Members can get a free, organization-specific Medicare attribution analysis with our new tool. Use the tool now

We conducted some of our own investigation on inpatient claims looking for indicators of erroneous physician attribution with the assistance of our colleagues in Crimson who have also helped dozens of organizations create attribution policies using claims and clinical data. By combining Medicare inpatient claims data with physician detail from CMS, our team was able to understand which physician specialties were attributed to various types of inpatient cases.

Related: Identify your outlier physicians—by name—in one click

We then attempted to find cracks in providers’ attribution practices using a wide variety of indicators. It must be said that these indicators are not watertight markers of problems per se, but they do represent an excellent starting point that can help shine a spotlight on the data to see if any patterns exist.

Four indicators of potential physician attribution errors

Attribution Flag
 Over-responsible physician  A single physician is attributed to a far higher proportion of claims—5 or 10 times more—than physicians treating patients at the same organization
 Emergency medicine physicians outside the ED  Emergency medicine physicians listed as attending on inpatient claims with a length of stay greater than two days
 Non-physician as an attending  Non-physician (example, nurse, occupational therapist etc.) listed as attending physician
 Anesthesiologist as attending for non-ICU cases  Anesthesiologist listed as attending in cases where more than 50% of the inpatient stay occurred outside of the ICU

We've unearthed a few high-level insights:

1. Several facilities have a single physician assigned to far more cases than would be expected.
Nearly 25% of acute care hospitals have at least one attending or operating physician responsible for 10 times the number of expected cases compared to other physicians performing services at the same hospital while a smaller number of physicians had over 20 times the cases that would be expected. A potential cause of this can be when a chief hospitalist is assigned to all cases on the claims, and additional attribution detail that might be found in the EMR is not considered.

2. Significant instances of emergency medicine specialists being attributed to patients with long inpatient lengths of stay (LOS).
While it’s not uncommon that emergency medicine physicians are attributed as the attending on inpatient cases with shorter length of stay (less than or equal to two days), cases greater than this duration are typically attributed to specialty types more focused on inpatient care. Our analysis indicates that roughly one quarter of all inpatient hospitals have attributed claims with a LOS greater than three days to an emergency medicine physician while roughly 260 hospitals have emergency medicine physicians attributed on claims with a LOS of 10 days or more.

Number of hospitals*
>2 days
>3 days
>4 days
>5 days
>10 days
> 259
*Only hospitals with 10 or more instances are counted

3. Non-physicians are being listed as attending physicians at many inpatient hospitals.
Over one quarter of hospitals have at least one claim where a non-physician is listed as the attending physician on the claim. These professions span a wide range, including physician assistants, anesthesiologist assistants, certified nurse midwifes, chiropractors, social workers, nurse practitioners and physical therapists.

4. Anesthesiologists are being attributed to non-ICU cases.
While an anesthesiologist may be listed as an attending when a patient stay is wholly or mostly in the ICU. Generally speaking this will not be true in cases where the majority of the patient stay within a non-ICU location such as a ward or med/surg beds. Granted, exceptions may exist, such as cases where an anesthesiologist has a dual certification in critical care.

What does your attribution data look like?

Our customized physician attribution assessment uses your organization's historical Medicare claims data to help identify performance for each of these attribution flags. We hope this analysis serves as a basis for understanding your attribution performance relative to other acute care hospitals and sparks internal discussion regarding your attribution policies. Furthermore, if you’d like assistance with attribution please reach out to us at for more information.


While you’re at it, check out our full array of customized assessments at