Veterans Choice Program - Program Integrity Tool (PIT)
Disclosure: This page is a work in progress and information may become out of date quickly. Please check last updated date at bottom of page.
Reconciliation and Auditing: Program Integrity Tool (PIT)
It is our understanding that the Program Integrity Tool (PIT) is an IBM product purchased by VA that aggregates many sources of data and allows for checking for fraud, waste, and abuse. This data source has an elevated privilege and operations users can request access via ePAS. It appears that starting in FY2016, Choice data began bypassing FBCS and residing in the PIT. By June 2017, no Choice stays are found in FBCS.
PIT went online in 2013 and FBCS claims in PIT can have a claim create date back to 2013. Providers and TPAs can submit older claims so you can see a claim submitted in 2018 with a date of service as far back as the start of PIT. We believe the data flow into PIT begins with non-VA providers rendering services and submitting claims to TPAs. They submit EDI837 files to EMDEON and from there it goes into ERepos in VA, then to FPPS to FBCS and finally to PIT. Within each step of data processing, the potential for data loss is great. In one instance, a TPA was not passing any of the Service Facility NPIs to VA for a few years so for those claims, it looked like the provider was the TPA; which is incorrect. There are known issues with the PIT mappings for Rendering Provider NPI and Service Facility NPI. Efforts have been made to correct some such errors in the past and more checks are in place to ensure such errors are not allowed moving forward.
VA payments to non-VA providers are limited to Medicare payments, by law. In general, providers can bill VA whatever amount they feel like and it is up to the TPA, VA, or FSC to ensure the billed amount is correct, prior to releasing payments. Choice and PC3 payments should be based on the Medicare rates times the CLIN (which can be 85%, 100%, 150%, 250%, and a few other multipliers, based on the provider contracts) but the price validation was not set up during the initial processing of Choice and PC3 claims, so VA paid the billed amounts on submitted claims. The result is that for a single CPT code, you may find different payments which vary from a couple hundred dollars to tens of thousands of dollars. Since the onset of Choice there have been efforts to go back to reprice and collect money from overpayment to providers but it is a time-consuming effort and is being remedied moving forward.
The PIT Domain was released into the CDWWork environment on September 25, 2018. The domain has 66 views, 27 dimensions, and 39 fact views. The PIT domain contains VA claims that have been processed through the PIT, including but not limited to Choice claims. CDW documentation includes:
- PIT Release documents (https://vaww.cdw.va.gov/metadata/Metadata%20Documents/Forms/AllItems.aspx?RootFolder=/metadata/Metadata%20Documents/PIT%201.0&FolderCTID=0x0120007BD83FE7EC890F42B79E1DA11A744B1E&View=%7bA9968955-5886-4DEC-A6BF-0CB219ADD175%7d)
- CDW Metadata Report (https://vaww.cdw.va.gov/metadata/_layouts/15/ReportServer/RSViewerPage.aspx?rv%3aRelativeReportUrl=/metadata/Reports/Metadata%20Report.rdl&Source=http%3a//vaww.cdw.r02.med.va.gov/metadata/Reports/Forms/AllItems.aspx&DefaultItemOpen=1)
- How to Get Access (https://vaww.cdw.va.gov/Pages/CDWHome.aspx) (select "Data Access Information")
How to identify Choice utilization in PIT tables
We have summarized 11 different methods for identifying Choice claims in the PIT tables. Below is a file that shows you the T-SQL search strings for each method. Please note that the summary table figures were calculated on 2/13/2019. Since PIT data are active and updated nightly, the numbers may have changed slightly.
We extracted FY15 professional and institutional PIT claims data for cataract surgeries using the CPT codes 66982 and 66984. To illustrate how we joined the PIT tables together, we have created two Excel files. One is for the professional component, and the other is for the institutional component.
The PIT data, along with the metadata for these tables, can be found on the VHACDWA01.VHA.MED.VA.GOV server in VINCI. Specifically, these Excel files are adapted from the metadata from [CDWWork].[Meta].[DWViewField]. Our Excel files contain information as of October 2018. For the most up-to-date variables and variable descriptions, please check the original metadata.
Description: Each tab in the Excel files represents a particular PIT table. The yellow rows denote the linking keys between each subsequent table. For the professional component, the orange rows are the variables we chose to keep for our analysis. For the institutional component, this is denoted by the blue rows.
Assumptions: In an earlier analysis, we extracted data using three filters:
- 1. WHERE CurrentFlag = “Y” AND
- 2. ClaimStatus = “Accepted” AND
- 3. PayFlag = “Y”
This will return current, accepted, and paid claims. If you are interested in looking at acted upon utilization and not only paid claims, you might want to consider leaving out the third filter, PayFlag = “Y”. However, this may leave you with duplicate records. Keep in mind that PIT only has records on which there has been some action. If there was no action on the claim, then it will not show up in PIT.
Caveats: Please note that the PITProcedureCode variable can be messy. In our analysis that focuses on using cataract surgeries, we only found three variations, “66982”, “66984”, and “66.984” when we intended to search for CPTs 66982 and 66984. We assume that “66.984” is a data entry error for “66984”. Thus, it was relatively easy to re-code “66.984” to lump with the “66984” records. However, users should be aware that the data could be more complex when one expands to other procedure codes that might contain typos and/or other variations.
Additionally, we found that only ~ 40-50% of institutional and professional claims from FY15-18 have non-missing SSN and PatientICN in the [CDWWork].[SVeteran].[PITPatient] table. This figure is for all claims and not only cataracts. However, we discovered that the MemberID variable in SVeteran.PITPatient actually contains the patient’s SSN as well. This variable is rarely missing. Thus, MemberID is recommended to be used as a patient identifier.
Methods:There appears to be “intermixing” in the professional and institutional claims data. For example, we are seeing many instances of professional claims that have a rendering provider who appears to be a facility/institution. Similarly, we also observe a large percentage of institutional claims where the rendering provider appears to be an individual/person. Thus, we chose to separate out the data by what we deem to be “true” professional and “true” institutional claims before merging the data.
One reason why this is occurring is because data are placed into one or the other schema based on what type of claim was submitted by the provider via the 837 files. Thus, the providers themselves inform PIT which claim they are submitting by providing the claims data either in the 837P or 837I forms. PIT receives these data from FSC or FBCS in either institutional format or professional.
In order to determine the difference between a professional record versus an institutional record, we use the NPPES file from CMS to merge in the entity type code variable. This variable can take on the values of 1 and 2. 1 represents an individual NPI, whereas 2 represents an organizational NPI. This file is publicly available from CMS and can be found here. If the record corresponds to an individual NPI, we deem that to be a “true” professional record. Conversely, if the record corresponds to an organizational NPI, we deem it to be a “true” institutional record.
There is currently no specific flag or value that allows one to connect the claims. However, we hypothesized that using an authorization number could assist in identifying a unique episode of care. After we separated out the data, we merged the “true” professional and “true” institutional records by MemberID, ServiceFromDate, and AuthID. Note that AuthID (authorization ID) is a variable that can be found in [CDWWork].[PIT].[PITVAAuthorization]. Initially, we tested merging on MemberID, ServiceFromDate, and PITProcedureCode (e.g. CPTs), and that yielded comparable results too.
For FY15, we were able to merge about a third of the “true” professional and “true” institutional records (at the time of the run). This left us with about another third of “true” professional records without a match, and about 5% that were “true” institutional records without a match. Ultimately, we stacked these resulting three datasets into one final dataset. We calculated the total cost for a cataract surgery to be the sum of the PaidAmount variables from the professional and institutional datasets.
We have found this process to be the gold standard method for our analytical purposes. However, it should be noted that the AuthID variable may contain a considerable amount of missingness. For Choice claims that were processed at FSC, AuthID will be in the PITVAAuthorization table. However, data travel from many different places in order to arrive at PIT. Depending on the origin of the data and where they get processed, records that are not from FSC will be missing AuthID.
Last updated: 02/19/2019