UnitedHealth Group, Inc. vs. the United States of America: The case for CDI

Part 3 (to read part 1, click here. To read part 2, click here.)

By Brian D. Murphy

In part 1 of this series, I introduced the developing story of UnitedHealth Group, Inc. vs. the United States of America. In part 2, I detailed the facts of the case. Part 3 looks at the regulations and commentary regarding claims submission to Medicare Advantage and the ramifications for CDI.

Section III of the complaint United States of American ex rel. Benjamin Poehling, explains the payment methodology which UnitedHealth Group, Inc. allegedly manipulated for financial gain. The Medicare program pays Medicare Advantage (MA) organizations a pre-determined monthly amount for each Medicare beneficiary in the plan. The payment amount for each beneficiary is based on their particular risk adjustment factor (RAF) score, which among other factors including the beneficiary’s demographics is impacted by assigned Hierarchical Condition Categories (HCC).

HCCs are a category of medical conditions that map to a corresponding group of ICD-9/ICD-10 diagnosis codes. An HCC is like a DRG payment in some respects, as it groups like diagnoses together.

From the court document and the Department of Justice (DOJ):

HCCs take into account diagnoses from inpatient hospital stays, outpatient encounters, and physician office visits. The HCC model is prospective, meaning that it relies on diagnoses for certain medical conditions assigned to beneficiaries by their physicians in one year to set the payment for each beneficiary for the following year. The medical conditions included in the model are grouped into HCCs, which are categories of clinically-related medical diagnoses. See 42 C.F.R. § 422.2. The diagnoses grouped into HCCs include major, severe, and/or chronic illnesses.

We can see here the familiar DRG-like bucketing of diagnoses, and the prospective nature of risk adjusted payment, not unlike the IPPS (Inpatient Prospective Payment System), the rules that govern payment for short-term acute care inpatient stays.

However, HCCs are unlike DRGs in some fundamental ways. DRGs are typically assigned from documentation related solely to that, particular, inpatient stay. The diagnoses that comprise HCCs not only derive from inpatient hospital stays, but also outpatient encounters and physician office visits. As a result, CDI specialists are now being employed by some hospitals to review encounters from non-traditional settings such as hospital clinics and physician practices. This is where “outpatient CDI” is starting to make inroads.

The other big difference between HCCs and DRGs is that the latter are assigned from a single encounter; HCCs by their very nature span multiple encounters and significantly widen the reporting window. As seen above HCCs are assigned by the diagnoses documented by physicians over the course of a year. Again, from the court document:

The HCC model is prospective, meaning that it relies on diagnoses for certain medical conditions assigned to beneficiaries by their physicians in one year (referred to by CMS as the “data collection” year but also generally known as the “date of service” or “DOS” year) to set the payment for each beneficiary for the following year (often referred to as the “payment year” or “PY”).

My takeaway: CDI specialists operating in the setting of HCCs need to break some long-established beliefs, and become comfortable reviewing prior medical records and encounter data. The concept of what is a code-able encounter remains the same—encounters must constitute a face-to-face interaction with the provider—but CDI specialists will need to review documentation from previous encounters and work to concurrently query the provider when necessary.

Continuing through the court document, we come to Section IV, Legal Obligation to Submit Valid Claims Data. You can find this on p. 19 of the complaint. To quote from this section (bold emphasis mine):

MA Organizations are entitled to risk adjustment payments based on the diagnosis codes that they submit to CMS only if the codes are from face-to-face medical encounters between the Medicare beneficiary and provider, the encounter occurred during the relevant date of service year, the provider was of a type and specialty acceptable for risk adjustment purposes, and at the time of the encounter, the provider documented the medical conditions identified by the diagnosis codes in the medical record based on acceptable documentation. In addition, codes should be based on documented conditions that require or affect patient care treatment or management.

As seen above codes are derived from face-to-face medical encounters between the patient and the provider, during that relevant date of service year (i.e., the year prior). Code assignment is based on documentation that requires or affects patient care treatment or management.

My takeaway: CDI specialists should perform clinical validation of the medical record to determine if these conditions affect patient care treatment or management. They should educate providers to document their clinical decision making. Since the codes can only derive from face-to-face encounters (with one exception—the professional services of a pathologist), it is not a compliant practice to review a prior encounter (from for example a doctor’s visit the year prior) and ask the physician to simply re-document prior diagnoses—unless of course the physician sees the patient, and documents evidence that he or she has evaluated, monitored, or treated the condition(s).

The next section of the court document is quite confusing, contradictory, and may be an error of interpretation by the DOJ. It reads as follows:

Risk adjustment claims are true and the resulting risk adjustment payments are valid only to the extentthat the diagnosis codes submitted by the MA Organizations are valid. The diagnoses must be coded according to the International Classification of Diseases (ICD) Clinical Modification Guidelines for Coding and Reporting (“ICD-9­ CM” & “ICD-10-CM”) and documented with sufficient clinical specificity.

All good so far, but then we come to this point:

All diagnosis codes submitted by MA Organizations must be supported by medical record documentation. If the medical record is ambiguous, it cannot be relied on for diagnosis information for risk adjustment payments. See 2008 RA Participation Guide at § 7.2.4.1 (stating that risk adjustment claims and payments cannot be based on questionable diagnoses).

…

The ICE RADAR Physician Education Work Group also issued a similar document called “Best Practices for Risk Adjustment,” which advised that “ICD-9-CM coding requires documentation of the diagnosis in the medical record as well as evaluation and management. Documentation should indicate how this diagnosis impacted this episode of care.” In 2012, ICE also issued a Medical Record Documentation Tips sheet, which once again warned MA Organizations and providers not to code diagnoses that are probable, suspected, questionable, or “working diagnoses” and also not to code diagnoses when medical records use “other similar terms indicating uncertainty.”

My Takeaway: This is confusing and appears contradictory to the prior advice above. The DOJ clearly states that the Official Guidelines for Coding and Reporting apply to Medicare Advantage Reporting. Unfortunately, and counter to what the DOJ seems to be stating above, according to the Official Guidelines for Coding and Reporting you can report possible/probable/suspected diagnoses, in the inpatient setting (not the outpatient setting—outpatient diagnoses can only be reported to their highest degree of certainty) and if present and documented at the time of discharge. Remember, HCCs are derived from diagnosis codes derived from inpatient stays as well as outpatient encounters, so it would stand to reason that probable/suspected/questionable diagnoses could be used to support HCC assignment, when a patient is being treated as an inpatient. My personal/very non-legal opinion is that, by “questionable” or “ambiguous” diagnoses, the DOJ means diagnoses that are clinically questionable (i.e., appear to lack clinical support). I welcome commentary and disagreement on this or any other point.

The case continues from here, listing out dozens of examples of diagnoses that the DOJ states were no longer valid and should have been removed prior to claims submission. Some of the at-risk HCCs that were unlikely to validate based on medical record reviews include the following:

• HCC 17 (diabetes with acute complications)
• HCC 96 (specified heart arrhythmias)
• HCC 104 (monoplegia, other paralytic syndromes)
• HCC 111 (chronic obstructive pulmonary disease)

Exhibit 2 (p. 83) of the complaint contains a lengthy list of examples of invalid diagnoses submitted but not deleted by United for the 2011-2014 payment years. These were diagnoses documented by a provider at one point in time—typically the year prior—but the DOJ states that United’s review failed to validate the diagnoses, and no other provider documented them, so they should have been deleted prior to submission. These include diagnoses that mapped to HCCs as wide ranging as HCC 83 (Angina Pectoris/Old Myocardial Infarction) and HCC 44 (Severe Hematological Disorders), among many others.

Again, these and other HCCs and their corresponding diagnoses needed to be clinically validated, but according to the DOJ, United failed to meet that obligation, and in fact willingly and knowingly did so since 2006 using a “look one way” chart review program:

Over the last decade, United has obtained billions of dollars of risk adjustment payments from Medicare from its national Chart Review Program because it only looked one way at the results of its coders’ chart reviews. As part of this program, United’s coders performed “blind” reviews of the beneficiaries’ medical records. That is, the coders did not know which, if any, diagnoses had been reported by the providers who treated the beneficiaries and created the medical records. Thus, rather than confirming diagnoses already reported by the providers or identifying only additional diagnoses supported by the records, the coders were instructed to look for all medical conditions purportedly documented in the records, record all diagnosis codes identifying those conditions, and give all of those codes to United.

Even after Poehling back in 2008 reported concerns that this practice was non-compliant, United failed to take action, the DOJ states. In fact, it even increased the size and scope of its “look one way” program. According to the DOJ after two years of deliberation United finally decided to perform claims validation and begin to “look both ways.” But it took three years for United to fully develop its claims validation program, and according to the DOJ United rolled it out in a half-hearted fashion. In May 2014 United terminated its chart validation program altogether.

Again, from the complaint:

Until late 2012, United did not complete its pilot tests and start to implement its CV (claims validation) Program for charts relating to 2011 medical encounters (i.e., with 2011 dates of service). Even after that, it never fully implemented the program. It also continually changed the program in order to limit its scope and created arbitrary rules to avoid looking at the negative results of many of the blind chart reviews conducted as part of the Chart Review Program.

…

After it terminated CV, United reverted to “looking one way” at the results of its chart reviews, making only ADDS, and knowingly and improperly failing to delete invalid provider-reported diagnoses and repay the Medicare Program for them.

Editor’s note: To read the summary of the case, visit the DOJ website by clicking here. To read the court document in its entirety, click here. Brian Murphy is Director of the Association of Clinical Documentation Improvement Specialists (ACDIS). He can be reached at bmurphy@acdis.org.