VA Doesn’t Really Know What It Costs To Run VistA

As the Veterans Affairs Department makes its 10-year, $10 billion transition to the Cerner Millennium electronic health records system, the agency will have to maintain its proprietary VistA system at a cost of $4.89 billion. However, as with VA’s past EHR programs, that estimate is likely well below what the actual cost will be.

In order to arrive at the estimate, VA officials looked at the costs of maintaining and updating VistA over the last three years, a figure they pegged at $2.3 billion. But a Government Accountability Office analysis of VistA costs casts that figure in doubt.

By GAO’s accounting, only $1 billion of that figure is reliable.

“The source data for the remaining $1.3 billion—which largely accounted for VistA’s infrastructure, related software and personnel costs—were not well documented,” Carol Harris, director of GAO’s IT and cybersecurity division, told the House Veterans Affairs Subcommittee on Technology Modernization. “As a result, VA’s subject matter experts were unclear on how to account for VistA versus non-VistA costs. Furthermore, the department omitted costs related to additional hosting and data standardization in testing from the total spend.”

Based on GAO’s analysis, spending on VistA over those three years was likely much higher, Harris said.

For example, in June, VA officials told GAO the agency spent $238 million on additional hosting services each year. Shortly thereafter, VA officials came back to GAO with a new number: $950 million. Ultimately, officials ended up leaving the line item blank.

“When we talked with VA subject matter experts, they agreed that the $950 million was off base, but the fact that that additional hosting line item was not included in the $2.3 billion estimate suggests that the number is higher,” Harris said.

The central issue is VistA’s decentralized nature, she explained. As the system has developed over the decades, each facility and office has customized it to their own needs, resulting in at least 130 distinct versions across 1,500 locations. That reality coupled with poor inventory management has led to an unreliable accounting of costs.

“As a result of that decentralization—which began in the ‘80s—VA is not in a position to be able to effectively draw that circle—that perimeter—around what is and isn’t VistA,” Harris said. “The inability to draw that perimeter is why they don’t have accurate costs and they don’t have an accurate basis for an ROI for moving to the Cerner system.”

Furthermore, the record of poor accounting casts doubt that VA’s $10 billion estimate for the Cerner rollout, lawmakers said.

“We don’t have any confidence a) in what VistA actually entails, so I don’t think we have any confidence in that $4.8 billion; but then, more importantly, that makes me have less confidence in the $10 billion estimate for Cerner, as well,” Subcommittee Chair Rep. Susie Lee, D-Nev., said. “Money does not grow on trees. At what point do we lay out exactly what the costs are?”

Early into the planning phases of the Cerner deployment, the agency increased the cost estimate by $350 million to cover government salaries that previously weren’t included.

One way VA plans to address gaps in the VistA cost estimate is through use of the Technology Business Management framework—a set of accounting standards designed to link IT investments to specific outcomes. Agencies are under mandate to use TBM in their IT reporting, which Paul Tibbits, executive director of the Office of Technical Integration, said VA is implementing in this year’s upcoming budget requests.

Using that framework, Tibbits said VA will begin to properly report personnel and infrastructure costs that were previously wrapped in with other areas.

But TBM itself won’t be enough, according to Harris.

“Until VA can fully define VistA, they will not be in a position to accurately report the costs. The two go hand-in-hand and the definition of VistA is foundational,” she said. “Whether they use TBM or another methodology, the core issue remains that the definition of VistA is not fully defined, and that’s the problem.”

“TBM is only part of the answer,” Tibbits agreed. “The definitional boundary of [the collocated hosting services] is clearly an important part of the answer, as well. The two of those combined together is what’s going to end up being our methodology.”

Tibbits added that VA expects to have that methodology finalized in the transition plan update coming this fall.

“There are many unknowns in this transition,” Lee said during her opening remarks. “The fact that this plan is still being formulated is concerning.”