At least week’s Pharmaceutical Regulatory and Compliance Congress, conference attendees received some insight into the Open Payments program from its source. Doug Brown, the acting director of the data sharing and partnership group within the Center for Program Integrity at CMS, spoke on a number of topics that compliance professionals have spent the last year working through. He also touched on what to expect in 2015. As we note below, the Open Payments database will contain a LOT of records.
As a reminder, applicable manufacturers and GPOs can now download a "Removed Records Report" from the Open Payments system showing the records that were removed. A guide for correcting records is also available. CMS will host a webinar on November 13, 2014 at 1:00pm EST to address the re-submission of data.
Challenge in the First Reporting Cycle
Brown first detailed the challenges in the first reporting cycle. We chronicled the struggles with the Open Payments system from a data integrity standpoint and also the covered recipient perspective during the review and dispute period. Brown provided CMS’s point of view during the Open Payments roll-out.
“We thought of these challenges as speed bumps, or bumps in the road,” he said. “A lot of things came to our attention during the review and dispute period that we needed to take corrective action on.”
The primary issue that brought CMS to action was that some physicians were seeing other physicians’ identifiers. “We took that very seriously, and as a result, shut the system down for a period of time,” stated Brown. Open Payments was shut down between August 3 through August 15, during which CMS began an “exhaustive review” of “every single record” they received—a total of 4.7 million records. This analysis revealed a significant number of intermingled physician identifiers which allowed some physicians to view other physician payment records.
Brown notes that CMS’s responsibility was “to make sure on a record-by-record level that [industry] was talking about someone that we could recognize.” Thus, CMS implemented more stringent matching criteria to ensure that payment records were attributed to a single, consistent physician or teaching hospital. Based on this matching logic, CMS flagged inconsistent records in the Open Payments system. They flagged 1.7 million records as potentially inconsistent.
“It was still important for us to publish the data, but based on the fact that we were uncertain about who exactly the payment went to, we felt the best approach was to de-identify payments within that category,” states Brown. This included removing the physician’s name, state license number, address, and any data that could identify a physician. This method also afforded all physicians the chance to dispute records that may have been removed from view during the review and dispute process.
De-Identified Data, and December Publication
CMS ended up with the following numbers for their September 30th release:
- 2.7 million identified records, totaling $1.3 billion in payment
- 1.7 million de-identified records, totaling $2.2 billion
- 199,000 unpublished records, totaling $1.1 billion – these included records still in dispute status at the end of the review period, and research records that are marked for delay in publication
In terms of the 199,000 unpublished payments, Brown stated that 9,000 are disputed records. “Our December publication will include the final resolution applied to those 9,000 records,” he notes. Of the $1.1 billion total in unpublished records approximately half (~$550 million) is tied up in the 9,000 disputed records.
Thus, come December, a fair amount of new records will be published. The other 190,000 are associated with delayed publication. Brown noted that manufacturers and GPOs will need to re-request for those payments to be delayed.
Old Matching Logic
Brown went into detail about CMS’s old matching logic and their transition to a new system.
CMS first used a matching process that worked in a very mechanical way. During the reporting period, companies could submit a record with, for example, a physician named Jane Smith and a state license number and NPI number. “What we would do is take that record and in a very serial fashion, look at the NPPES system in order to validate that the NPI, state license, and first and last name were exactly correct.” If that didn’t pass, they would run the data through the PECOS system. If neither was an exact match, Brown noted that they validated the information through their third party vendor, Truven, which also has a broad data set of physician information.
“In some cases we were seeing where Jane Smith was reported with her NPI, but her state license may have been attributable to some other J Smith.” CMS’s system would then use a “probabilistic matching algorithm,” to solve close cases. For example, where the NPI’s matched, the names were very close, and only the state license number was a bit off, the system would consider that to be a match. “That was where everything started falling apart in our system,” notes Brown.
New Matching Logic - Combination of Systems
Going into the data return and data correction phase, CMS is implementing a new matching logic, with two primary objectives: (1) minimize the effect that inconsistent submitted data have on the program and (2) efficiently and effectively absorb all of the records and reduce the rate of data rejection in future submission periods, while preserving data integrity. In other words, according to Brown: “How do we take in as much as we can, but not have records that contain this intermingled information?”
Now, instead of a sequential process, CMS will combine the NPPES and PECOS data fields. “When you look at each system in isolation, you are failing more records,” Brown states. This new list pulls all the identifiers from NPPES into PECOS. Thus, validation of records against NPPES and PECOS can occur in parallel.
CMS has provided applicable manufacturers and GPOs with a "profile validation list" using combined information from NPPES and PECOS. This pre-validated list of physician identifiers will "equip manufacturers with the answer key" to have their submissions accepted. Such a process allows for variation on valid first and last names as seen in the two systems and "pre-established profiles within the System, linked on Name, NPI, and State License."
Brown believes that combining PECOS information will help rectify much of the inaccuracies inherent in the NPPES database. One audience member noted that their company received many payment records back as inaccurate, despite matching 100% with PECOS. Brown, without offering his guarantee, noted that most likely these payments would be accepted under the new data matching system.
Finally, if no NPI exists for a particular physician, CMS will push the record over to Truven.
Brown did note that even with their newly robust matching system, some potentially correct records will come in that CMS cannot validate because their system will be unable to find a necessary match.
September 30 Roll-Out
“On September 30th, the data explorer was not met with all the pomp and circumstance we would have liked,” notes Brown. “Most of the feedback wasn’t saying that this was the best thing since sliced bread. There were a lot of things that we could have done better.” Brown notes that CMS has taken steps to continue to improve the usability of the data by users and industry.
"September 30 was our first shot at getting the data out there," he states. “We knew there was a long way to go to make the data usable for the public.” On October 17, CMS rolled out their search tool. “This was met with slightly better public response,” notes Brown. Now users could look for their particular doctor, and then dig deeper using the original data sets. Furthermore, companies can add context comments on a payment-by-payment level about the circumstances that led to that payment being made.
October 30 Data Return
On October 30th, AMs and GPOs were able to log-in to the Open Payments system in order to access their data and start going through all of the records CMS returned. “The explanations on a payment by payment level need to be improved,” notes Brown, but in essence the returns are based on all of the edits CMS put in place during the two week period where the system was offline and the generic reasons CMS applied to each. CMS is anticipating a lot of individual question from industry members and has scheduled a webinar for November 13 to address specific issues.
Data Refresh in December
All corrective actions taken by industry on disputed records prior to October 31 will be reflected in the refreshed data publication in December (scheduled sometime before 12/31).
What will be included:
- Disputes initiated by physicians and teaching hospitals as of September 11, 2014
- Display as “Disputed” if unresolved as of October 31
- Display with corrected data if resolved as of October 31
What is not included this time, and will be published later
- Any change to disputed records involving the covered recipient
- Any records submitted to the Open Payments system for the first time
- Any modification to records currently on display--these changes will be put into next year’s review and dispute process.
Brown states that the December refresh will contain “the final resolution on disputes,” but “with a small exception: If that dispute said you didn’t pay me, you actually paid someone else, then that someone else never got the opportunity to review and dispute that record.” These will be included in the next review and dispute cycle. All other changes will be part of the refresh. In response to an audience question, Brown stated that once a physician has had the opportunity to review and dispute a payment, they will not get the opportunity to re-dispute old payments in this next reporting cycle.
2015- A Big Year for Open Payments
Next year, the Open Payments database may contain three times as much identified information as it does now.
In June of 2015, CMS will refresh the 2013 data publication, removing all of the de-identified files, and replacing them with “brand-new corrected identified files representing 2013" as well as "only identified files representing the payments or transfer of value files from 2014.”
Brown notes that CMS’s new matching logic will be implemented in late January 2015. Right now, CMS is using the current 4.7 million entries it received to test their logic. Data submission for the next reporting year will likely be submitted in February/March of next year. “This will also be the opportunity to report and have the corrections processed for data [companies] reported in 2013.” March 31, 2015 is the deadline for data collection, followed by the review and dispute period.
Thus, by June of next year, Open Payments will show 1.7 million identifiable records from the de-identified 2013 data, corrected 2013 dispute records, and, conservatively, another 10 million 2014 records (5 months of 2013 data brought in approximately 4.7 million records).
On the issue of whether the de minimis amount of $10 will be changed, Brown noted that in the first year of reporting, CMS didn't have much of an idea as to how many small payments they would receive. He believes the threshold question will likely be a continuous point of discussion. Last month we noted that two-thirds of the general payments were under $20.