The Medicare Improvements for Patients and Providers Act of 2008 (MIPPA) mandated that the prospective payment system for dialysis being implemented in 2011 "shall include a payment adjustment based on case mix that may take into account patient weight, body mass index, comorbidities, length of time on dialysis, age, race, ethnicity, and other appropriate factors."1 In response, the Centers for Medicare & Medicaid Services added 15 new case mix adjusters in the proposed bundle to the existing three, reflecting factors associated with increased dialysis costs. Because all 15 new case mix adjusters increase Medicare payments, CMS "standardized the ESRD PPS payments in order to account for the overall positive effects of the proposed ESRD PPS case-mix patient and facility adjustment factors and wage indexes. ... The standardization factor was calculated to be 0.7827, or a reduction of 21.73%."2 In other words, CMS preemptively took 21.73% (or nearly $1.9 billion) out of dialysis spending to offset the incremental cost of these adjusters.

This presumptive payment reduction assumes that dialysis clinics will reproduce the case mix adjusters that CMS used in its analysis. If clinics are unable to replicate these case mix adjusters, then they will not recover the 21.73% payment reduction and will be significantly underpaid. This de facto payment reduction would be well beyond the 2% authorized under MIPPA.

Determining the case mix adjusters
CMS selected the proposed set of 18 case mix adjusters based on regression analyses run by the University of Michigan Kidney Epidemiology and Cost Center (UM-KECC), in which ESRD reimbursement was the outcome variable and patient-level medical characteristics and comorbid conditions were the predictor variables.2 The predictor variables came from the CMS 2728 Medical Evidence forms, and paid medical claims from the United States Renal Data System.

This introduces a potential flaw in the proposed system: dialysis units do not have access to the extensive claims data used by CMS to generate the case mix adjusters. For example, hospitals forward discharge summaries to dialysis units sporadically at best, and dialysis units have no statutory authority to require delivery of this data. Without access to all the claims data that CMS used to ascertain case mix adjusters, dialysis units will systematically under-report them, resulting in widespread underpayment. The objective of this study was to review all possible sources of data available at DaVita Inc. to ascertain patient-level case mix adjusters at a randomly selected set of clinics, and to compare the clinic mean case mix adjusters to those reported by CMS.3

Methods
We randomly selected one dialysis unit in each of 157 DaVita-defined geographic regions in the United States. We then randomly selected four of these units in each of 27 divisions, resulting in a sample of 108 geographically diverse facilities. We conducted chart reviews at each of these facilities for patients with Medicare as primary payer receiving dialysis between Oct. 1, 2008 and Sept. 30, 2009. Patients were coded for each of the 18 case mix adjusters identified by CMS.

Fixed characteristics (age, gender, initiation of dialysis) were coded once at baseline; all others were coded on a monthly basis. Data sources included electronic medical records, hospital discharge summaries, paper charts, all health care professional notes, and discussions with on-site health care professionals. Comorbidities were coded if the condition was currently documented or if the condition was documentable (e.g., the nephrologist knew that a podiatrist had diagnosed the patient with monoclonal gammopathy, but had not recorded the diagnosis in the chart). Specifically, there were three ways to document each adjuster:

• There was a written diagnosis of that condition (the 2728 form; the medical chart; a hospital discharge summary; physician notes, etc.)
• There was laboratory or other substantial evidence of that condition. For example, a positive surface antigen serology for hepatitis B would substantiate that diagnosis even if the specific diagnosis of hepatitis B were not noted in any other medical record. A patient taking lopinavir, ritonavir, zidovudine and lamivudine, for example, was presumed to have HIV.
• A medical professional with credentials and knowledge (nurse or physician, or in the case of substance dependence, a psychologist or social worker) verbally reported the presence of the condition and when it was diagnosed.
  

Data was collected by a specially trained and clinically experienced team at DaVita. Completed abstractions were reviewed by a facility nurse familiar with each patient to verify and check for completeness. Case mix adjusters were computed using an Excel spreadsheet program. Case mix adjusters were assigned to each treatment for each patient at the clinic, and a clinic mean was computed as a 12-month average of these patient-level adjusters. CMS clinic mean case mix adjusters were abstracted from the CMS "CY2011 Proposed ESRD PPS Facility Level Impact File."3

Results

Sample and Patient Characteristics
Eight centers were excluded from analysis due to incomplete data, e.g., the clinic changed ownership. Of the 7,349 charts reviewed, nine were missing essential data (e.g., number of dialysis sessions) and were excluded from analysis. The final sample included 100 clinics and 7,340 patients.

The 2009 CMS proposed rules did not include a full description of the analytic cohort used to conduct the case mix adjuster analysis, but such descriptors were included in the 2008 UM-KECC Report.4 Given the size of the population being analyzed (>250,000 patients per year over multiple years), we assumed a relatively stable prevalence of the conditions defining the case mix adjusters. Thus, we included a comparison of the DaVita study sample to the sample in the UM-KECC report (see Table 1).

Though race/ethnicity is not currently included as a case mix adjuster, it is interesting to note that there were fewer Caucasians and more Hispanics in the DaVita sample compared to UM-KECC. Among the demographic case mix adjusters, DaVita patients tended to be younger, were more likely to be incident patients, and were more likely to have low BMI, all associated with higher payments. DaVita patients were also less likely to be female, had lower BSAs, and none of the DaVita units surveyed met the criteria for low volume, all of which are associated with lower payments. Taken together, there was no systematic bias among these objective case mix adjusters-half were favorable for DaVita patients, half were not.

Among the comorbidities, the differences were more striking. The DaVita units were lower on eight out of the 12 comorbidities: hepatitis B, septicemia, cancer, HIV/AIDS, hemolytic or sickle cell anemia, monoclonal gammopathy, myelodysplastic syndrome, and pericarditis. In addition, the magnitude of the absolute differences were large in many cases: -7.0% for hepatitis B; -6.7% for septicemia; -4.4% for HIV/AIDS. DaVita reported higher prevalence of four co-morbidites (cardiac arrest, pneumonia/other, alcohol-drug dependence, GI bleeds), but in each case the difference was <2.0%.

Clinic-level case mix adjusters
Of the 100 clinics surveyed, 11 were not included in the CMS flat file (likely due to clinics that were newly opened or changed ownership). Therefore, all comparisons between CMS and DaVita case mix adjusters are based on 89 clinics. The mean CMS case mix adjuster for the 89 units was 1.283. The mean case mix adjuster for these same units in the DaVita study was 1.213. In other words, on average, the DaVita clinic case mix adjuster was 25% lower than the CMS case mix adjuster for the same clinics. Figure 1 displays the difference between the CMS and DaVita case mix adjusters for all 89 units surveyed (each bar represents a clinic). The case mix adjuster was lower in 75 of the 89 clinics (84%) and higher in 14 of the clinics (16%). The mean increment for clinics with a higher case mix adjuster was 0.045; the mean decrement was 0.092, or twice as large. Because the monies for the case mix adjusters were preemptively taken out of the base payment, this trend would generalize to a 7% reduction in dialysis payments in the United States, on top of the 2% reduction authorized under MIPPA.

Discussion
Despite extensive effort, DaVita was unable to ascertain the full extent of case mix adjusters used by CMS to model the impact of the proposed bundled payment system on dialysis clinic reimbursement. In the proposed rules, CMS has indicated that it will preemptively reduce the base payment to offset the costs of the 18 case mix adjusters. Without corrections of those adjusters to reflect more realistic patient characteristics, as demonstrated in this study, the implementation of this payment system will result in a significant reduction in dialysis reimbursement across the country, beyond the 2% authorized by Congress under MIPPA.1

The prevalence of case mix adjusters for objective variables (e.g., age, sex, initiation of dialysis, BMI, BSA) did not vary systematically between the UM-KECC/CMS and the DaVita estimates, i.e., some were higher, some were lower, and all were within a couple of percentage points. The case mix adjusters for comorbid conditions were systematically lower in the DaVita study, and several came short by 4% to 7%. This finding is not surprising: dialysis units do not have access to the complete Medicare claims data used to generate these adjusters. Units without extensive electronic data capture may be at further disadvantage. In addition, all 18 case mix adjusters explain only 8.7% of the variance in separately billable costs.2 In the UM-KECC published analysis, the five objective case mix adjusters (plus race) explained 5.52% of the variance in separately billable costs.5 The 12 comorbid conditions explain little incremental variance in cost, add significant administrative burden, and ultimately clinics may not be able to replicate them. It would be both parsimonious and prudent to eliminate the comorbid conditions all together.

The faulty case mix adjusters have other implications. CMS created a "Transition" variable for each facility that "Identifies facilities that are estimated to opt out of transition." This variable was based on whether a clinic will be reimbursed more under the old payment system, or the new payment system based on the 18 case mix adjusters. CMS assumed that facilities paid more under phase-in would choose that option, and those paid more under opt-in would choose that option, which would increase Medicare spending. In order to avoid this increase, CMS applied a 3% "transition budget-neutrality adjustment factor to all payments." In other words, all dialysis sessions from 2011 through 2013 will be taxed 3% to pay for the incremental costs of selective phase-in, regardless of whether a unit opts in or phases in. However, as units discover that they cannot recoup the 21.73% "standardization factor" payment reduction (which anticipates that all case mix adjusters will be found and billed), we expect very few will choose to opt in 100%. If very few units opt in, Medicare will save money, negating the need for this 3% "transition budget neutrality adjustment factor."

The results reported here are from a large dialysis organization with more than 1,500 units and more than 110,000 patients. Smaller and independent organizations should replicate these results. Small units without electronic records may show even greater discrepancies in detectable case mix adjusters, while hospital-based units with access to inpatient records may fare better.

Conclusion
The currently proposed system may inadvertently reduce dialysis payments as a result of the inability of clinics to replicate the case mix adjusters, coupled with the preemptive reduction of the base payment to offset the costs of the adjusters. The objective and currently reported case mix adjusters account for the majority of the 8.7% of variance in separately billable costs. The comorbid conditions add little to the proposed payment system outside of administrative burden and uncertainty.

Acknowledgements
We would like to thank the DaVita teams that completed the case mix ad-juster chart review survey and the leadership team of Amy Young, Ginger Hanson, Heather Ashbaugh, Irina Goykhman, Maribeth Sommer, Mary Mann, Nancy Culkin, and Stephanie Nagel.

References
1. H.R. 6331, The Medicare Improvements for Patients and Providers Act of 2008 (MIPPA) enacted on July 15, 2008

2. CMS Medicare Programs; End-Stage Renal Disease Prospective Payment System; Proposed Rule, September 2009. Accessed at http://www.cms.hhs.gov/ESRDPayment/PAY/itemdetail.asp?filterType=none&filterByDID=99&sortByDID=4&sortOrder=descending&itemID=CMS1228517&intNumPerPage=10

3. CMS CY2001_Proposed_ESRD_PPS_Facility_Level_Impact_File, September 2009. Accessed at http://www.cms.hhs.gov/ESRDPayment/PAY/itemdetail.asp?filterType=none&filterByDID=99&sortByDID=4&sortOrder=descending&itemID=CMS1228517&intNumPerPage=10

4. University of Michigan Kidney Epidemiology and Cost Center, ESRD Payment System : Results of Research on Case-Mix Adjustment For Expanded Bundled, Februrary 2008, accessed at http://www.sph.umich.edu/kecc/assets/documents/UMKECC_Expanded_ESRD_Bundle.pdf

5. Hirth RA, Turenne MN, Wheeler JRC, Pozniak AS, Tedeschi P, Chuang CC, Pan Q, Slish K, Messana JM. Case-Mix adjustment for an expanded renal prospective payment system. J Am Soc Nephrol 18: 2565-2574, 2007


Dr. Mayne is Senior Director of Health Economics and Medical Informatics at DaVita Inc. Ms. Burgess is Manager of Special Projects at DaVita Clinical Research. Mr. Weldon is Manager of Analytics at DaVita.