Where the largest signal appeared
The largest pool of potentially duplicate transportation payments identified in a federal review of veterans’ travel benefits appeared between two payment systems, rather than within either system alone. That distinction puts the ability to match records across systems at the center of the accountability problem.
The Government Accountability Office identified $5.8 million in potentially duplicate invoices and claims shared between VA’s Invoice Payment Processing System and Electronic Claims Administration and Management System. Its separate tests identified $1.7 million within the first system and $1.3 million within the second.
Adding the three rounded dollar figures gives $8.8 million; the cross-system category represents about 66% of that combined testing amount. This is our comparison of the reported categories, not a new estimate of proven losses. The records can contain false positives, some payments had already been identified by VA, and the report warns that testing categories can overlap.
A payment check can exist and still miss a match
The systems served different payment workflows. The invoice system primarily handled contracted special transportation, while the claims system handled noncontract emergency transportation. Vendors could submit through both channels, creating a risk that one transport could be paid twice.
VA did have matching controls. According to the report, a comparison of new and historical transactions was intended to flag possible duplicates across the two systems. But VA officials confirmed one potential duplicate identified by GAO and said variations in how information was stored may have prevented the comparison from finding it.
Within the claims system, officials also confirmed previously unrecognized duplicates. Two likely passed because provider information differed between the original and repeat claims. These examples show why the existence of an automated match is not sufficient evidence that its matching rules work across variations in names or other fields.
The units in the testing also matter. GAO listed 365 invoices within the invoice system, 2,616 invoices and claims across systems, and 675 claims within the claims system. An invoice could cover many rides. The combined count therefore cannot be presented as the number of veterans, appointments or duplicate trips.
The agency’s response contains real progress
VA reported that more than $26.3 million associated with over 15,000 claims duplicating invoice-system submissions had been prevented or recovered as of December 2025. The matching report became daily in 2025, after having run weekly. Those reported results demonstrate that controls did catch problems; they do not negate the weaknesses found in GAO’s separate testing.
VA also began moving transportation invoices to VetRide in November 2024. Officials told auditors that most, but not all, medical facilities had transitioned by February 2026. Unlike the older consolidated invoices, VetRide itemizes transportation and was intended to connect individual trips to appointments.
The report described a September 2026 goal for moving VA-initiated noncontract special transportation and a December 2026 goal for rideshare services. The reviewed records do not independently establish that the September goal was met. A scheduled conversion should not be reported as a completed one.
The underlying claims and invoice data covered fiscal years 2018–23, with different start dates for newer systems. The June 29, 2026 report thus combines historical testing with later descriptions of control changes; it is not a measurement of the duplicate-payment rate in October 2026.
The test is whether the replacement works
GAO recommended monitoring and evaluating duplicate-payment controls across and within the systems, along with clearer invoice-review instructions and other fraud-risk safeguards. VA agreed with all nine recommendations. The recommendation tracker reviewed for this article listed them as open; that status means GAO had not confirmed implementation, not that VA had taken no action.
The larger program’s estimated improper payments also declined substantially, from about 24% of outlays in 2018 to about 8% in 2024. Those estimates include underpayments and payments lacking sufficient documentation. They are not fraud totals, and they should not be added to the targeted duplicate-payment findings.
The original comparison identifies a practical oversight priority: testing whether the same ride can still be paid through different channels after the transition. Publishing verified exception rates, confirmed duplicate totals and documented recoveries would show whether a software change has closed the gap that mattered most in the auditors’ three tests.