> ## Content Index
> Fetch the complete content index at: https://www.healthdatacon.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# VA innovation chief says rural hospitals need robots their own staff can maintain
- URL: https://www.healthdatacon.io/articles/va-rural-hospitals-robot-maintenance-pilot-deployment/
- Published: 2026-11-18T19:30:00.000Z
- Updated: 2026-11-19T11:02:50.000Z
- Description: Beth Ripley described repair capacity and repeatable performance as barriers to expansion, while ARPA-H’s Kate Arnold explained how its fetal-monitoring program is testing what hospitals will actually buy.
- Author: HDM Staff
- Tags: Operations, Americas, HLTH

Robots can end up unused in rural hospitals when no one is available to repair them, a Veterans Health Administration innovation leader said Wednesday. Beth Ripley urged developers to design physical AI systems around the skills of the staff already working in the hospital, making maintainability part of the path from a successful pilot to wider use.

Ripley said the technology needs to be “simple, trustworthy, and reliable.” Describing a barrier the VA has encountered, she said equipment can sit idle because “there’s no one to come and fix them.” She pointed to biomedical engineering staff who already maintain complex imaging equipment as people whose capabilities developers should consider when designing support and repair processes.

The comments came during [From Breakthrough to Every Hospital](https://thefuturist.global/PhysicalAI?ref=healthdatacon.io), part of The Futurist Physical AI Summit at HLTH USA 2026 in Las Vegas. Host Ian Khan moderated the discussion with Ripley, acting chief officer of VHA’s Office of Healthcare Innovation and Learning, and Kate Arnold, a program manager at the Advanced Research Projects Agency for Health. The exchange linked research funding to the operational work needed to keep a device useful in care.

## A working pilot still has to earn expansion

Ripley described two places where promising technology can stall. The first is the transition from an early prototype to a device ready for clinical use, when testing, regulatory work and integration can require funding that a team has not secured. The second comes after a pilot, when the conditions that helped a system work at one location do not carry over to another.

She said health systems should make a pilot an iterative test of readiness, leaving room to change the product or its use before expanding. A team can learn that the initial workflow is too demanding, that the equipment needs a different design or that the next site has different staffing. Treating wider deployment as an automatic next step risks spending more money before those problems are understood.

Ripley’s preferred near-term use for physical AI was dependable transportation within hospitals. Some facilities have used that technology for years, she said, while other attempts have been less successful. Her test was whether it could function reliably beyond flagship sites, including hospitals with fewer people available to manage another system.

That shifts part of an evaluation toward the service around the machine. Someone has to recognize a fault, determine whether the equipment can continue operating and restore it when it cannot. A maintenance process requiring a specialist to travel from a distant city creates a different operating risk from one the local team can carry out with training and support.

## What manufacturing taught the VA

Ripley’s [VA biography](https://www.innovation.va.gov/ecosystem/views/programs/fellowship/fellow/beth-ripley/?ref=healthdatacon.io) identifies her leadership in the department’s 3D-printing network and its work on personalized healthcare technologies. Khan asked what that experience could teach developers of robots and other physical AI systems. Ripley said the answer depends partly on where the work should happen.

Manufacturing can sometimes benefit from concentration at a smaller number of locations, she said. Physical AI intended to assist in routine care may need to work at each hospital where the task occurs. That would require testing across the differences in those hospitals, including their layouts, staff, equipment and support capacity.

The FDA’s [human-factors materials](https://www.fda.gov/medical-devices/human-factors-and-medical-devices/human-factors-considerations?ref=healthdatacon.io) likewise treat users, use environments and interfaces as connected parts of device design. They include the people who install, maintain and repair equipment among possible users. That provides a framework for considering local support during development, alongside what happens when a clinician or patient operates the product.

## ARPA-H is asking whether a hospital wants the result

Arnold described a related test in Making Obstetrics Care Smart, or MOCS, the ARPA-H program developing new ways to assess fetal and placental health. She said teams have required quarterly engagement with rural, urban and academic hospitals. That gives people working in different care settings a way to question the technology before the design and workflow become harder to change.

“We have required end user engagement,” Arnold said. She described a later program assessment that would ask nurses, clinicians, patients and hospital leadership about the output, ease of use and cost. The question would include, “Is this something that you would buy for your hospital?” She said those findings would help determine which work receives continued support.

ARPA-H’s [program description](https://arpa-h.gov/explore-funding/programs/mocs?ref=healthdatacon.io) distinguishes a proposed rapid assessment of risk before labor from a proposed noninvasive fetal-monitoring system during labor. Arnold’s discussion concerned how teams would learn whether the resulting information is useful in care. She offered no completed trial result showing that the new technologies improve outcomes or are ready for routine use.

The purchasing question adds evidence about adoption to the program’s technical questions. A sensor can produce a signal a developer intended to measure while leaving a clinician uncertain about what to do with it. A useful output may also require a workflow and price a hospital can support, making staff participation relevant before the product reaches a purchasing committee.

## Resilience includes what happens when software stops

Arnold also raised the risk of an outage. Clinicians are accustomed to interpreting existing fetal heart-rate patterns, she said, while new sensing methods could produce information they cannot interpret without software. A product based on those signals would need a plan for the information and functions that remain available during a disruption.

She described ARPA-H bringing different kinds of expertise into program design to examine risks, including possible malicious uses and cybersecurity threats. Those questions continue during a multiyear program because the technology and threat environment change. The discussion supplied a development concern and a process for addressing it, with no claim that a particular outage safeguard has already been validated for the proposed monitors.

Ripley said conversations with veterans and caregivers had also challenged assumptions about where people would welcome robotic assistance. She described interest in support for daily activities and giving caregivers respite, including situations where more independence could preserve dignity. Those preferences, she said, belong in the design conversation along with the technology’s behavior and reliability.

The speakers described several kinds of readiness that have to meet at the same point. The equipment needs to perform its task, people need to understand and trust its behavior, and the hospital needs a way to operate and support it. Ripley’s rural maintenance example gave that broader question a practical test, whether the machine can stay in service where the patient actually receives care.