AI Imagines CRT
- By Laurie Watanabe
- Aug 29, 2023
As an English major, I am on the front lines of “Do we even need editors anymore? Can’t artificial intelligence do your job faster, more accurately… and with less deadline drama?”
While looking over my shoulder to see how quickly artificial intelligence (AI) might be coming for my job, I wondered how well AI would do at understanding and recommending seating and wheeled mobility for clients with complex and unique positioning needs.
I did background reading on how to communicate with AI, as every source I consulted said AI’s answers are only as good as the questions it’s asked. After I learned to optimally phrase my “prompts” — i.e., my questions — I signed onto ChatGPT, perhaps the best known of the publicly accessible chatbot platforms. ChatGPT, which launched in November, is a product of OpenAI, which describes itself as “an AI research and development company. Our mission is to ensure that artificial general intelligence benefits all of humanity.”
Here’s the disclaimer you knew was coming: This is an exploratory exercise to demonstrate ChatGPT’s responses to questions about Complex Rehab Technology (CRT). Do not use this information to build your own seating or mobility system. Consult human seating experts instead.
Getting Started with ChatGPT
To get going, I said to ChatGPT, “Tell me about Complex Rehab Technology.”
Within moments, ChatGPT replied: “Complex Rehab Technology (CRT) refers to specialized medical equipment and assistive devices that are designed to meet the unique needs of individuals with significant disabilities or medical conditions. It encompasses a wide range of products that are specifically tailored to enhance mobility, independence, and quality of life for people with complex medical and functional needs.
“CRT devices are prescribed by healthcare professionals such as occupational therapists, physical therapists, and rehabilitation specialists. These professionals work closely with individuals to assess their specific requirements and determine the most appropriate equipment to meet their needs.
“The provision of Complex Rehab Technology involves a collaborative effort between healthcare professionals, assistive technology suppliers, and funding sources such as health insurance providers and government programs. The goal is to ensure that individuals with complex medical needs have access to the appropriate equipment and support to maximize their independence, functional abilities, and overall well-being.”
Analysis: ChatGPT’s definition was largely correct, though it sounds as if the chatbot believes OTs and PTs routinely write prescriptions...? ChatGPT then listed “examples” of CRT, as directly quoted:
• “Custom Wheelchairs: These are individually configured wheelchairs that are designed to accommodate the unique postural and mobility needs of the user. Customizations may include adjustable seating, specialized backrests, and controls tailored to the user’s abilities.”
• “Power Wheelchairs: These electrically powered wheelchairs provide mobility for individuals who have limited or no independent movement. Power wheelchairs can be customized with various seating options, control systems, and accessories to enhance comfort and functionality.”
• “Adaptive Seating Systems: These systems provide additional postural support and pressure relief for individuals who have difficulty maintaining a stable sitting position. They can include specialized cushions, backrests, and positioning accessories.”
• “Standing Devices: Standing devices allow individuals with limited mobility to achieve a standing position, promoting various health benefits such as improved circulation, bone health, and range of motion. These devices can be manual or powered, depending on the individual’s abilities.”
Much of the first part of ChatGPT’s CRT list was accurate or close. Like ChatGPT, I am not a clinician, ATP, RTS, or technician. But I noticed flecks of inaccuracies or oddities, such as ChatGPT’s reference to custom wheelchairs. Since ChatGPT separately listed power wheelchairs, I assume custom wheelchairs are something different, i.e., not power chairs. By process of elimination, I’m guessing that ChatGPT believes custom wheelchairs are manual wheelchairs, with custom referring to the ability to customize ultralightweight and tilt-in-space models.
But then ChatGPT said custom wheelchairs have “controls tailored to the user’s abilities.” Handrims aren’t typically referred to as controls, and neither are they typically described as tailored, although ergonomic handrims do have grippier textures. Possibly I’m nitpicking.
Further along in its CRT definition, ChatGPT misstepped in listing “specialized computer input devices, environmental control systems, and augmentative and alternative communication devices” as CRT. Nope. Still, all of those technologies are CRT adjacent and certainly used by many wheelchair riders.
Recommend Seating & Mobility for C5-6 SCI
Using better prompts, I asked ChatGPT for suggestions for a seating clinic client: Recommend CRT for a man, age 24, with a spinal cord injury (SCI) at the C5-6 level.
ChatGPT first gave an “I’m not a medical professional, but I can provide some general information on Complex Rehab Technology” disclaimer, followed by a recommendation for a power chair: “A power wheelchair can provide mobility and independence for individuals with limited or no use of their lower extremities. Depending on the user’s specific requirements, a power wheelchair can be customized with features such as tilt-in-space, recline, and elevating legrests to enhance comfort and positioning.”
ChatGPT didn’t mention an optimally configured ultralightweight chair, though an Internet search brought up articles discussing that as an option for some wheelchair riders with C5-6 SCI.
ChatGPT did recommend pressure relief equipment, noting, “Individuals with limited mobility are at increased risk of developing pressure ulcers. Specialized cushions, mattresses, and pressure-relief systems can help distribute pressure, reduce the risk of skin breakdown, and promote healing.”
Less helpfully, the chatbot suggested augmentative and alternative communications devices (AAC), noting that they “enable individuals with speech impairments or difficulties in verbal communication to express themselves. These devices can include speech-generating devices, tablets, or specialized software that convert text or symbols into spoken words.”
AAC is indeed wonderful, but is it typically needed solely because of a C5-6 SCI?
Recommend Seating & Mobility for T7 SCI
Next, I asked ChatGPT to recommend CRT for a client with a T7 SCI.
Recognizing the lower injury level, ChatGPT’s first recommended item was a manual wheelchair, which “can provide mobility and independence. Depending on the user’s needs and abilities, a manual wheelchair can be customized with features such as lightweight construction, adjustable seating, and specialized controls for self-propulsion.”
Once again, ChatGPT mentioned controls, presumably for an ultralightweight manual wheelchair.
The next suggestion was a “power-assist wheelchair,” which ChatGPT said, “combines the benefits of a manual wheelchair with the assistance of a power system. This technology helps individuals conserve energy and overcome physical limitations by providing motorized assistance during propulsion. It can be particularly beneficial for individuals with limited upper-body strength.”
The phrasing was unusual — power-assist wheelchairs? — but the power-assist suggestion sounds good.
ChatGPT also recommended a standing frame; environmental control systems; aids for activities to daily living, such as “adaptive utensils, specialized handles, and dressing aids”; and pressure relief systems.
In this case, ChatGPT recognized that a self-propelled manual wheelchair would likely be a good choice for a client with a lower-level SCI, and did not mention power chairs. But the chatbot also suggested adaptive eating utensils, much less likely to be needed by a client because of a T7 SCI.
ChatGPT’s Report Card
So in this limited sample, did ChatGPT pass the test? It offered broad seating and mobility suggestions in line with what clinicians have suggested in interviews I’ve done over the years, though the chatbot didn’t always seem to know why it was recommending particular types of equipment. While ChatGPT incorrectly referred to a wide range of products — from AAC devices to ramps and environmental controls — as CRT, it did capture the complex, custom nature of CRT seating and wheelchairs.
ChatGPT’s answers were broad and often repetitive: It suggested adaptive forks and spoons for every mobility-related disability I asked about, including hemipelvectomies. Some recommendations seemed random: ChatGPT suggested smart-home controls for clients with lumbar-level spinal cord injuries, because those controls could be voice activated. But the chatbot did not recommend smart-home controls for clients with cervical or thoracic spinal cord injuries who, because of greater impairment, would probably find those home controls even more useful.
And a few recommendations didn’t make sense, such as when I asked for a power wheelchair suitable to a smaller, older home that couldn’t be renovated to improve accessibility. ChatGPT recommended choosing a power chair with “lightweight construction,” because “lightweight models are easier to maneuver and transport, which can be beneficial in navigating within a small home.”
I’m not sure how light weight impacts maneuverability in a power wheelchair.
ChatGPT can identify core types of CRT and can name typical considerations, such as pressure relief. But in these examples, ChatGPT didn’t demonstrate the ability to incorporate context or to “connect the dots” — possibly why it kept recommending adaptive spoons for everyone. It didn’t acknowledge differences in clients’ functional goals, environments, or potential progressions.
Therefore, I give ChatGPT a final grade of… incomplete. AI appears far from ready to make seating clinic decisions optimal for each client’s unique needs.
But maybe ChatGPT should get extra credit for its answer to my final question of whether or not CRT produces good outcomes. “Yes, CRT has been shown to produce positive outcomes for individuals with complex medical needs,” it said. “It is important to note that the outcomes of CRT can vary depending on individual needs, appropriate assessment and prescription, access to timely and customized equipment, and ongoing support and training. Collaborative efforts among healthcare professionals, suppliers, and individuals with complex medical needs are crucial in maximizing the positive outcomes and benefits of CRT.”
This story originally appeared in Mobility Management’s eBook digital edition. To subscribe, visit the newsletter subscription page. Subscribing to the eMobility newsletter also subscribes you to future eBook digital editions. To see the Innovation edition of the eBook, click here.