Evaluating BPCI Advanced Target Factors in Episode Selection

Submitted by jonpearce on Sun, 2018-07-08 19:49

Along with the BPCI Advanced claims data, CMS provided to applicants with an Excel workbook named 2018_05_31_bpid-0000_Baseline_MY_1_2_Prelim_TP.xlsx containing the preliminary targets for all episode types in which the hospital had sufficient baseline episode volume. Those workbooks also contained the three factors that are used to adjust each hospital's baseline cost to compute the final target amount. Analyzing these factors individually and comparing them to the ranges of these factors in other hospitals can provide insights into the difficulties of reducing cost to levels achieve these targets. Our comparisons originate from the targets provided to about 100 client hospitals who have engaged us for BPCI analytics services. This group is primarily composed of large metropolitan hospitals and may not be representative of all hospitals throughout the country. This article describes how such analysis can be performed.

Clinical vs Analytical Opportunity

This article focuses on "analytical opportunity", which deals solely with target amounts and episode costs, and not with the revisions to clinical care that will be necessary to achieve success. While analytical opportunity can indicate areas in which success may be more likely than others, participating in episodes based solely on analytical opportunity has never been shown to be a consistently effective success strategy in bundled payment programs. Success in these programs invariably originates from physician champions and other clinicians becoming heavily involved in restructuring the care of patients to improve care and reduce cost. Analysis of targets as described here should accompany clinical analysis, and may be limited to accepting or eliminating "borderline" episodes in which moderate clinical opportunity exists. In assisting our clients to select episodes, we rely heavily on the clinical analyses described in this Singletrack Analytics blog article. The factors described in that article should be considered along with those described below in selecting episodes with an opportunity for success.

Overview of factors

From the regression equations that incorporate a myriad of inputs described in the target documentation on the CMS website, three primary factors are critical for understanding the development of targets. These factors are:

  • The acute-care hospital (ACH) Efficiency Factor, which compares the hospital’s actual episode cost to a regression-derived theoretical episode cost amount. The factor is computed by dividing the theoretical cost into the actual cost, resulting in a percentage in which more efficient hospitals have a lower factor. Since these factors are multiplied together to compute the target, a lower efficiency factor (meaning a more efficient hospital) will have a lower target, all other factors being equal. This factor will remain constant into the first performance period.
  • The Patient Case Mix Adjustment (PCMA) factor, which measures the complexity of the case mix for each hospital in each episode type. Hospitals having a higher PCMA have a more complex case mix and thus a higher target, all other factors being equal. Notably, this factor will be recomputed in each performance period.
  • The Peer Adjusted Trend (PAT) factor, which incorporates two primary components; persistent differences in clinical episode spending levels across ACH peer groups and trends each peer group’s clinical episode spending to the performance period. This factor remains constant into the first performance period.

Evaluating the Efficiency Factor

As noted above, the efficiency factor measures the theoretical cost efficiency of each hospital in each episode type. Since more efficient hospitals have a lower factor, those hospitals will have lower targets. This appears to be CMS’ way of avoiding rewarding hospitals for efficiencies created in the past, and only rewarding them for efficiencies created during the performance period.

Efficiency factors vary widely among hospitals and episode types. The graph below shows the range (computed as plus and minus two standard deviations of the mean) of efficiency factors among our client hospitals for several episode types. The range of efficiency factors appears to vary significantly among episode types.

 

The graph below shows the efficiency factors for each hospital in the CHF episode type. Of the two hospitals highlighted, the hospital having the green bar would have an easier task at reducing cost to meet the target than the hospital having the red bar. This occurs for two reasons; first because the green bar hospital would have a higher target than the red bar hospital, and also because the red bar hospital has apparently already created cost-efficient care pathways and therefore the "low hanging” cost has already been pruned from these episodes. Therefore, evaluating an applicant hospital’s efficiency factor can provide some insights into the difficulty of meeting the target. Note that this factor doesn't change in the performance period.

Patient Case Mix Adjustment (PCMA)

The PCMA adjusts targets for differences in patient complexity that the CMS regression models have identified as affecting episode cost. It allegedly puts to rest the age-old argument in healthcare as to whose patients are “sicker”. Importantly, the PCMA distributed from the baseline data will be revised in the performance period, and therefore this factor will cause baseline targets to differ from those used in the performance period. Like the efficiency factor, the PCMA varies among hospitals as shown in the graph below.

 

Notable in the above graph is a wide variation in PCMA for major joint replacement episodes. This is caused by a small number of relatively high PCMA factors for some hospitals, many ranging from 1.05 to 1.2. These higher factors are not found in the other episode types and may be driven by regression inputs specific to major joint replacement episodes. In partial recognition of the emergence of outpatient total knee arthroscopy cases and their effect on the cost of the remaining inpatient episodes, CMS has added several additional PCMA regression inputs that apply only to the major joint replacement episodes family, and those inputs may be responsible for the higher range of values in this episode.

What's not currently known, and would be important in evaluating this factor, is how much the PCMA might change between the baseline and performance periods. Unfortunately we have no data to assess this change, since only data for the baseline period is available. However, applicants can compare their PCMA values with the ranges shown above to estimate the possibility that their performance period PCMA will increase or decrease, since it's unlikely that the performance period PCMA will exceed the limits shown above.

The PCMA is influenced by a variety of patient related factors, including the presence and number of various hierarchal condition categories (HCCs) of patients in these episodes. Since HCCs are determined from the diagnosis codes on patient bills, it's important that all diagnoses be coded for all patients during the episode in the performance period.

Peer-Adjusted Trend Factor (PAT)

The PAT factor adjusts for variations in hospital type as well as trends in episode cost throughout the baseline period. This factor appears to be highly influential in target determination for some types of episodes. As shown in the graph below, almost all of the PAT factors for these episode types center around a value of 1, indicating that some hospitals will have factors that exceed or are lower than 1. A value of 1 would indicate that this factor did not influence the target from the baseline amount, other factors being equal.

The major joint replacement episode is again notable in this graph because all of the PAT factors for this episode type are lower than 1. This appears to be the result of the trend factor, which is applied to accommodate changes in episode cost during the baseline period. Because of the high level of participation in these episodes in the BPCI program along with the large number of hospitals participating in the Comprehensive Care for Joint Replacement (CJR) initiative, reductions in costs in these episodes appears to be widespread, and consequently this factor adjusts targets downward to compensate for cost trends that will already have taken place by the performance period. This means that hospitals evaluating participating in major joint replacement episodes should evaluate their own cost trends throughout the baseline period and continuing into the present to determine if their own episode costs have already been reduced to those near the target amount. If their own costs have not dropped, achieving financial success in these episodes may be difficult.

Case Study

Now that we have some context for evaluating different factors, let's take a look at some examples that might occur at a BPCI Advanced applicant. Those factors appear in the table below.

Episode Type

Efficiency

PCMA

PAT

Chronic obstructive pulmonary disease, bronchitis, asthma

1.03

0.98

0.99

Congestive heart failure

0.97

0.95

1.02

Major joint replacement of the lower extremity

0.84

0.97

0.88

Stroke

1.11

1.02

1.00

 

COPD

The efficiency factor of 1.03 indicates that this hospital is slightly less efficient in treating these patients than is predicted by the regression model. This may mean two things; first, that there is some amount of cost reduction opportunity in these episodes that has not yet been achieved by this hospital, and also that COPD target will be slightly higher as a result. The PCMA factor of .98 indicates that patients are slightly less clinically intense that of other hospitals, and the PAT factor indicates that the target will be slightly reduced due to hospital characteristics and episode trends. Overall, these target factors do not provide compelling reasons for or against participation in this episode.

CHF

The efficiency factor of .97 shows that this hospital has been moderately efficient at reducing costs in this episode, and therefore a slightly reduced target will result. The PCMA of .95 indicates a slightly lower clinical intensity, and therefore a potentially lower target. However, the PAT of 1.02 indicates the cost trends in these episodes may be increasing, which will slightly elevate the target. Again, these factors do not provide compelling reasons for or against participation in this episode.

MJR

The efficiency factor of .84 indicates that this hospital’s costs in this episode type are already significantly low, which will result in a lower target. The patient case mix is also slightly lower, which will moderately reduce the target. But the major factor PAT adjustment, which itself will reduce the target below the baseline cost by approximately 12%. In this case, the target factors create a relatively strong case not to participate in this episode.

Stroke

The efficiency factor of 1.11 indicates that this hospitals cost per stroke patients slightly higher than predicted by the regression model, which may indicate a clinical opportunity and will also create a slightly elevated target. The PCMA and PAT factors are inconsequential because of their small difference from 1. If detailed examination of the clinical data indicates opportunities for cost reduction, this episode may be a good candidate for participation.

Conclusion

While identifying clinical opportunity remains the primary means of creating success in bundled payment episodes, an analysis of the target factors may provide additional input into episode types that should or should not be selected for participation.