When we heard that CMS has selected a new data contractor for the BPCI Advanced program, we remain optimistic about the quality of the data that we would receive for this program. Data for the BPCI and CJR programs was relatively complete, containing demographic information about patients (name, medical record number, zip code, etc.), payment components (DRG outlier payments, IMD, DSH, etc.), and other fields describing utilization levels. This data was arranged similarly to its source in the CMS claims data warehouse, and contained header and detail records for each claim. We were hopeful that this format would continue in BPCI Advanced, since it provided a robust basis for strong analytics.

Missing Claims Data Fields

Unfortunately our hopes were dashed when we saw the recently-released data specifications from the new contractor. This data format contains significantly less data than was provided in the other two programs, and the omissions may be problematic for several reasons. For example, the BPCI and CJR inpatient data files contained 78 data elements (although some data elements overlapped in different files), while the BPCI Advanced inpatient data file contains only 19 data elements. Notably missing from this data are the values of payment reductions for value-based purchasing, excess readmissions, and uncompensated care, ACO overlap and others. Similar omissions occur in the other data files. No “beneficiary” file is provided that would identify the names of the beneficiaries; the only patient identifier is the Medicare identifier.

Missing Monthly Episodes Files

Of significant note is the apparent plan to omit an “episode” file from the monthly data, and provide it only with the baseline data. This is especially problematic for two reasons; first, that the data layout for this file contains fields indicating potential episodes that would be eliminated because of patient ineligibility or for other reasons. This appears to indicate that claims for ineligible patients will be included in the monthly data, with no ability to eliminate those episodes. This can result in significant over-counting of episodes on an ongoing basis. The BPCI files contained data that would allow identification and elimination of these excluded episodes, but those data elements do not exist in the BPCI Advanced data.

Inability to Exclude Ineligible Episodes

The other problem with the absence of episodes file is the resulting need to identify episodes in the monthly data from the inpatient and outpatient services files. In the BPCI and CJR programs each episode was identified by beneficiary ID, DRG, admission date, and other related information in the episodes file. The CJR data provided additional information summarizing the costs of the episodes into claim categories such as SNF, HHA, and others, which could be used to provide some general guidance on cost and utilization. None of this information will be provided in the monthly BPCI Advanced data. Instead, users of the data, or their respective data vendors, will need to synthesize the episodes from inpatient admissions and qualifying outpatient procedures. This process is fraught with potential error; an admission that might appear to initiate in episode (i.e., an "index admission") might actually be a readmission from a previous episode in another hospital. This would cause the episode initiated by that admission to be incorrect, and could potentially affect other later episodes for that beneficiary.

There are other data-related issues – the omission of a field containing SNF days and other utilization counts, the inclusion of almost 300 columns of HHA revenue data for no apparent reason, and others. CMS has noted that this data set contains the minimum necessary data for participating in the program, the ‘minimum necessary” being a HIPAA requirement, which would be more persuasive if they weren’t already releasing significantly more data for the BPCI and CJR programs. Because of the extremely tight timeframe between the release of the data (expected in late May) and the deadline for applications (August 1) we hope that CMS doesn’t delay the data release to make changes to the formats. However, we hope that these issues will be addressed before the monthly data in released late in 2018.