CMS CMMI Bundled Payments for Care Initiative Pilot Resources

CMS CMMI Bundled Payments for Care Initiative Pilot Resources

A Suggested Framework for Contracting for Bundled Payments Under the CMMI BPCI

The CMMI Bundled Payment for Care Improvement is an opportunity for providers and other organizations to contract for a “user-defined” episode of care. During the past several months, CMMI has provided guidance on desired features of pilot proposals. In particular, the need to include all patients in a selected MSDRG, and all MSDRGs in a MSDRG class. These two features are important because, on their face, they might avoid the potentially perverse incentive of shifting patients from one MSDRG to another, or from selecting certain patients and not others. There are other important principles and features to consider when contracting for bundled payments. And in prior work, the Health Care Incentives Improvement Institute has outlined these principles of contracting and encapsulated them in a template contract. While this template is geared towards implementations of the PROMETHEUS Payment model, many of the clauses included in it are germane to any contract around bundled payments. As BPCI Applicants put finishes touches on their proposals to CMS, they should consider these important elements:

  1. Management of patients and not probabilities – the goal of bundled payment contracts is to focus providers on the good management of patients and to reward them for that management. We therefore recommend the following:
    1. Pricing each bundle based on the principle procedure code (especially for those focusing on MSDRGs 469 and 470), or the principle diagnosis code. Here’s why. Pricing bundles by MSDRG, blending all patients, irrespective of their reason for admission (as evidenced by the principle diagnosis code), or the specific procedure done (as evidenced by the principle procedure code), imputes variability in the price that is simply due to patient mix. While severity adjustment might mitigate some of that mix, it leaves the provider vulnerable to the odd few patients with an uncommon diagnosis code, for which a severity or case mix adjustment model will have limited effect. And given the thin margins available in some bundles, a few bad cases can tip the scales to a loss.
      Creating a price by principal diagnosis code reduces the patient mix variability and creates more clinical homogeneity around the pricing, e.g. patients with an AMI undergoing a PCI versus patients with arrhythmias undergoing a PCI. The tables in Appendix illustrate the difference in pricing by procedure or by diagnosis code for specific MSDRG classes.
    2. Eliminating outliers, both high and low. In particular, eliminating patients who die during the stay, and eliminating cases that are linked to trauma or other uncontrolled event. These two types of outliers might, over the course of three years, balance themselves out. However, there’s no assurance that will be the case. As a result, including the 2009 mix of such patients might represent either a high or low anomaly which will result in potential gains or losses simply based on the luck of the drawn (or the selection of patients) during the pilot year. While all patients should be included during the implementation, with the provider having an opportunity to request adjustments based on adverse selection, the negotiated episode price should exclude these patients. The tables in Appendix illustrate the effect of leaving these patients in and excluding them on two MSDRG classes.
    3. Observing but not pricing episodes with inadequate sample sizes. It’s common for providers to have small sample sizes of patients either in a MSDRG, or for specific procedures or with specific principal diagnoses. Creating an episode price based on a small sample size will leave both payer and provider at total risk of a random draw. As such, episodes with less than 25 or 30 patients should not be priced, but rather included for observation. To the extent the number of patients in these episodes goes above the minimum agreed sample, then they would become subject to a bundled price based upon the agreed-upon formula for that episode.
  2. Managing the financial risk – bundled payment contracts in the private sector almost always include a provision for stop loss, the ceiling per episode (or across episodes) above which the provider is no longer at financial risk. There are two types of stop loss, one that would be episode-specific and is oftentimes expressed as a number of standard deviations above the mean historical price for the bundle, and one that would be an aggregate stop loss for all contracted episodes. For example, from Table 2, a stop loss for Total Hip Replacement surgery episodes would be $38,804, which is the historical average plus three standard deviations. The aggregate stop loss is simply a number above which the providers feel that they would be at serious financial harm. While there exist some nascent commercial arrangements to purchase stop loss insurance, BPCI applicants should consider proposing an additional discount in exchange for a stop loss. For example, instead of a 3% discount on a 30-day episode, an applicant might propose a 4% for a 30-day episode with an individual episode stop loss of three standard deviations and an aggregate stop loss of $5 million.
  3. Ensuring continuous feedback and appeal on reconciliations – the template contract, which has been used in PROMETHEUS Payment implementations, includes a section on timeliness of reports and appeals on financial reconciliations. These are important elements to ensure that providers have timely access to claims information in order to understand their gains and losses on specific episodes. BPCI applicants should request quarterly claims files and reconciliations of closed cases, as well as the right to appeal for the exclusion of certain outliers.

The CMMI BPCI presents a great opportunity for providers to engage in value-based payments, and we encourage all to submit proposals.

Example of HRC XX for patients in Reattachment of Lower Extremities MSDRG class

(1) The CMMI Guidance







Average Episode $






CMMI requires applicants to take all patients across all related DRGs in a severity family. Provider applicants’ ability to win or lose under this model is completely dependent on their case mix, which is out of their control. Risk/Reward shoud hinge on the provider organization’s ability to manage clinical risk, not probability risk.

(2) The HCI3 Suggested Approach


Total Hip Replacement – 81.51


Average Episode $




Partial Hip Replacement – 81.52


Average Episode $




Total Knee Replacement – 81.54


Average Episode $




Total Ankle Replacement – 81.56


Average Episode $




HCI3 recommends that rather than calculating an average across all patients within each related DRG, applicants should examine the mix of underlying Principal Procedure or Diagnosis codes within each DRG and calculate a target price by Principal Procedure or Diagnosis across all related DRGs in a severity family. In order to arrive at the target price, applicants should also exclude outliers (patients who die before discharge and other low/high cost outliers, which could be defined as 2 standard deviations from the mean). By looking within Principal Procedure or Diagnosis across DRGs, the incentive for provider applicants shifts to managing patients into the lower cost DRG by avoiding complications rather than trying to game based on case mix which is out of their control. HCI3 also recommends proposing a stop loss at 3 standard deviations above the mean in exchange for an additional % discount in the target price.


Bene ID DRG Principal Px Avg Episode $ Exclusion Reason (either patient expired or the case was a high cost outlier)
Patient 1 469 81.51 $66,547 Expired; Prin Dx: 733.42, Aseptic necrosis of head and neck of femur; Unspecified septicemia, Acute respiratory failure, Acute Kidney Failure, Pressure ulcer, lower back
Patient 2 469 81.52 $54,714 Admitted via ED; Prin Dx: 820.8, Closed fracture of unspecified part of neck of femur; Acute posthemorrhagic anemia, Acute respiratory failure
Patient 3 470 81.52 $44,994 Admitted via ED; Prin Dx: 820.21, Closed fracture of intertrochanteric section of femur; Acute posthemorrhagic anemia; 2 months at a SNF
Patient 4 469 81.52 $25,487 Admitted via ED; Expired; Prin Dx = 820.09, Other closed transcervical fracture of femur; Anoxic brain damage. AMI, Acute Respiratory Failure, Acute Kidney Failure, Shock, Cardiac Complications


Example of HRC XY for patients within the PCI MSDRG Class

CMMI Guidance            
HRC XY 246 247 248 249 250 251


N 298 1636 140 552 64 282 2972
Average Episode $  $22,054  $13,477  $22,119  $12,721  $23,107  $12,629
STDEV  $9,126  $5,023  $12,077  $6,548  $10,465  $5,766
HCI3 Suggested Approach (includes outliers and expired patients) HCI3 Suggested Approach (with outliers and expired patients removed)
HRC XY       HRC XY      
Dx N Average STD Dx N Average STD
410 814 $17,529  $8,947 410 768 $16,194  $5,723
 414 1763 $13,332  $5,748  414 1693 $12,517  $3,289
996 66 $13,398  $5,784 996 64 $12,564  $3,050
427 227 $14,845  $8,473 427 214 $13,252  $4,794
All other Dx (21 Dx Codes) 102
HCI3Suggested Approach (with outliers removed)  
HRC XY      
Dx N Average STD
410 781  $16,252  $5,717
 414 1697 $12,528  $3,294
996 64 $12,564  $3,050
427 215 $13,246  $4,783