«DEPARTMENT OF HEALTH AND HUMAN SERVICES Centers for Medicare & Medicaid Services 42 CFR Parts 409, 424, and 484 [CMS-1560-F] RIN 0938-AP55 Medicare ...»
Such a behavioral response on the part of home health agencies would therefore have contributed to the 1997-2000 13.4 percent change in the average case-mix. As we indicated in our discussion, it is very possible that a certain amount of nominal
the period October 28, 1999, through September 30, 2000, which is the period after the proposed rule was issued.
Comment: Some commenters had specific criticisms of the real case-mix change model. Some wrote that the methodology for assessing changes in patient characteristics relies on DRG changes, but only half of HHA patients are discharged directly from a hospital to an agency. In commenting on the case-mix change model, some commenters stated that data on ownership structure were not related to patient characteristics. They went on to write that the methodology gave no consideration to changes in care delivery in other health sectors (for example, the growth in Medicare Advantage), or in reimbursement methodologies that drive patients into home health care.
Response: Far greater than half of the observation units— that is, episodes -- in the samples had hospital discharge data.
The model uses data from the last hospital stay the patient had before the home health episode. Approximately 90 percent of the random sample of episodes in the case-mix change model, regardless of the time period (FY2000 or CY2007), had a hospital stay record. Not all of these hospital stay records were classifiable to a specific DRG because of sample size
into a medical or a procedure group, based on information in the hospital stay record. For patients with multiple episodes, the last discharge did not necessarily lead directly to home health admission, but it would still reflect fairly recent health characteristics. For a small proportion of episodes, the hospital stay may have occurred distantly in time (but no more than four years earlier). In alternative models described in the Abt Associates Final Report (April, 2008), hospital stays for some conditions were not used if they did not occur relatively close in time to the home health episode, but the results did not change the essential conclusions we drew from the analysis.
The predictions of the case-mix weight from the model were adjusted for the ownership/affiliation category of the agency that delivered the care under the episode. We made this adjustment to account for the historically different coding practices and apparent case-mix levels associated with different kinds of ownership. We did this out of an abundance of caution, because of a paucity of literature explaining these differences.
It is plausible that the large decline in hospital-based agencies that occurred after the last year of the IPS could have
had we not made the ownership/affiliation adjustment, we would have found less real case-mix change from our analysis.
We disagree with the commenter’s conclusion that we have ignored the effects of reimbursement methodologies that drive patients into home health care. Variables in the model account for prior utilization in acute care hospitals, long-term-care hospitals, inpatient rehabilitation facilities and skilled nursing facilities. The model relates these various kinds of utilization to the case-mix weight in the ensuing home health episode. We used the model and the levels of prior utilization that occurred by CY2007 to make predictions of the real case-mix weight for that year. In fact, the net effect of all the Medicare cost and utilization variables in the model was to raise the predicted average case-mix weight, consistent with what appears are the commenter’s assumptions. However, the increase was small. To the extent that the nature of the relationship between the specific kind of prior utilization and the ensuing episode’s case-mix weight has changed, the case-mix prediction methodology may not capture the entire impact of reimbursement changes in other parts of Medicare. However, in its Final Report (April, 2008), Abt Associates conducted a test
variables and case-mix, and this test did not support the idea that changes in the model variables’ relationship to case-mix had occurred. Moreover, we believe we have captured some of the other settings’ reimbursement effects by measuring change in utilization of prior settings. In addition, the model includes an array of other demographic and health-related variables that are expected to detect change in the health status of the user population, which is the real underlying issue raised by reimbursement changes.
As we indicated in the proposed rule, we intend to test changes to the model that may represent the growth in Medicare Advantage.
Comment: Several commenters wrote that CMS’ methodology for estimating nominal case-mix change is imprecise and relies on limited sources of data. One commenter noted that the methodology was not based on clinical analysis but on statistical inferences in a complex model that is so abstract and complex that significant data errors were undetected. The commenter noted that it is plausible that the average case-mix continues to grow, since the ratio of for-profit to nonprofit agencies increases each year, and for-profit agencies have
change estimate is a guesstimate, and is not sufficient or accurate. Some commenters suggested CMS engage additional consultants to use alternative methods of evaluation, and crosscompare outcomes, before the proposed 2011 adjustment is finalized. Another commenter asked for an independent audit of Abt’s work.
Response: We believe that our methodology for quantifying the contribution of real case-mix change to total case-mix change between FY2000 and CY2007 is a reasonable approach, but it is only part of the evidence base for our conclusion that nominal case-mix change has been pervasive. As we noted in the proposed rule, the full evidence base was presented in a series of regulations, beginning with the May 4, 2007, proposed rule (72 FR 25393). We discussed a variety of statistical data, including but not limited to resource use measures in comparison to case-mix weight changes, shifts among severity levels of the clinical, functional, and service dimensions of the case-mix system, shifts in the share of high-therapy episodes, differential changes in responses among various OASIS items (payment-related items and non-payment-related items), and a detailed analysis of the evolution of OASIS guidance and manual
item responses. We presented admission rates over time for five specific conditions suggested by commenters, and examined the time to admission for those conditions. These results were updated in the proposed rule, and suggested that changes were insufficient to explain the substantial upward trend in casemix. We also noted the steep learning curve faced by agencies in adapting to the new environment presented by OASIS, resulting in improved coding. We also pointed out that coding changes are not foreign to any payer system when payment methodology becomes more dependent on provider ascertainment of health status information. The evidence base is the best available, given the infeasibility of auditing large chart samples from both time periods, which may be assumed to be the type of clinical analysis that a commenter suggests. As we noted in the proposed rule, we are investigating enhancements to the model to capture more elements of real case-mix change that may be unmeasured.
However, whether these enhancements will reveal any additional real case-mix change than we have already measured is unclear at this time.
From the point of view of statistical methodology, the model is a basic linear model and not complex; although it
on large, representative samples. The preparation of the data has been subject to some technical corrections, but the basic approach has remained the same and is not subject to significant error. Furthermore, insofar as there have been data errors, they have not been so significant as to alter by large amounts the size of the payment reductions we made based on the model findings. As we have noted elsewhere in our responses, the model does allow for the contribution of for-profit agencies to real case-mix change.
We have no plans for undertaking alternative methods of evaluation. An independent audit is not necessary because the model and results of the application of the model have been presented in detail in the Abt Associates reports. However, we do intend to test enhancements to the model (described in the proposed rule) and welcome suggestions from the public for modifications to the statistical approach and additions to the data that are cost-efficient to make.
Finally, as a point of clarification, the 2.71 percent reduction for CY 2011 is not a proposed adjustment. In the CY 2008 final rule (at 72 FR 49843) we promulgated our policy of a
2.75 percent reduction for 3 years (CY 2008, CY 2009, and CY
final rule changes what was finalized in the above rule, with regards to payment reductions to address the increase in nominal case-mix.
Comment: Some commenters believed that the increased therapy needs or increased involvement of physical therapists in assessing patients have contributed to appropriate growth in HHRGs. They wrote that the change in focus from disease management to restorative therapy has increased HHRGs and benefited patients. A few suggested that the process for evaluating case-mix change related to therapy utilization must include in-depth review of the merits of individual claims, as the limited use of proxies is unreliable. Several commenters believed that the analysis failed to adequately evaluate whether changes in case-mix are due to abusive over-utilization of therapy, fraudulent or abusive coding, erroneous coding, revised coding instructions, or improved quality coding. Where changes are due to abusive or fraudulent practices, several commenters suggested that CMS address those abuses with the specific providers, rather than applying a punitive adjustment to all agencies. Alternatively, commenters suggested CMS use enforcement to conduct targeted claims review and deny payment
Response: We agree that there has been a shift toward rehabilitative services, but we believe commenters are confusing a change in the home health “product” with actual change in the health status of the treated population. As MedPAC has noted for years, with the implementation of the HH PPS, the service payment unit underwent changes: the unit of payment changed from visits to 60-day episodes, and the content of the home health product changed from that of the 1997-2000 period--consisting of fewer visits, shorter stays, and more therapy with less aide care (MedPAC, March 2004, “Report to Congress: Medicare Payment Policy”, Section 3D, “Home Health Services”). In any future enhancement of the real case-mix change model, we may investigate allowing for the possible increased use of physical therapists as the assessing clinician. We would do this on the assumption that increased use of therapists to make assessments is a change that is not a consequence of the agencies’ learning curve in the HH PPS environment or of new financial incentives that began in October 2000. We would do this despite the fact that it could be stated that differing assessment results arising from the use of nurses vs. therapists as assessing clinicians do not signify differences in the health status of
to the model would have a very small impact on our conclusions.
To the extent that abusive over-utilization of therapy and fraudulent or abusive coding are responsible for case-mix growth between FY2000 and CY2007, it would be preferable to remove agencies engaging in these activities from the data analysis.
However, it is difficult for us to identify these agencies on a large scale, so we find the commenter’s suggestion impractical.
Furthermore, we believe that the overwhelming majority of providers are not committing fraud, which would mean that eliminating the fraudulent providers would not have a large impact on our results. If commenters know of fraud being committed in their areas, we urge them to inform the Office of the Inspector General and the CMS Regional Office. As stated earlier, CMS is committed to addressing suspect fraudulent activities, especially those in areas where we see suspicious outlier payments, and will monitor and aggressively pursue actions towards agencies where inappropriate billing of outlier payments is identified.
Comment: Several commenters suggested we conduct an impact analysis of the proposed rule relative to case-mix, include an evaluation of access in each year of any adjustment, and
that the impacts in the proposed rule were factually and legally inadequate and therefore violated the Regulatory Flexibility Act.
Response: We appreciate the commenter’s suggestion;
however, our current approach to impact analysis does include the effect of the rate reduction related to nominal case-mix change. Our impact analysis is subject to OMB review and meets legal requirements. We will consider how to increase our monitoring of access going forward. We would appreciate any specific suggestions from commenters on ways to do this.
Comment: A few commenters questioned the assumptions surrounding LUPA episodes which were used in the case-mix change analysis. One wrote that nearly all “creep” may have been offset if CMS had modified its actuarial assumption of 5 percent LUPA incidence to actual occurrence once PPS was in place. The commenter asked that we disclose the LUPA incidence for 2001 through 2006. The commenter felt that using a 5 percent LUPA incidence, rather than the higher, actual LUPA incidence, has led to agencies being underpaid. This commenter added that instead of lowering rates using a “creep” theory of justification, CMS should have raised the base rate calculation