Kaya Kesim is a little shy when meeting a stranger, retreating behind the legs of his mother, Katie Kesim. But no amount of gentle greeting, coaxing, or even clowning will elicit a reply from this brown-eyed 4-year-old, because Kaya doesn't talk.

"After he turned one, he stopped talking altogether," says Kesim. Soon after, he was diagnosed with autism.

But Kaya can communicate. He uses a speech-generating device, or "SGD," which at first glance looks like an Etch-a-Sketch, its gray touch screen framed by red plastic. He carries his SGD around and uses it to request snacks, sing songs, and play.

"It has saved us a lot of temper tantrums," says Kesim. "He used to beat his head on the wall repeatedly out of a lot of frustration."

Life without communication is unimaginable for most of us. SGDs offer a means of communication to those with neurological conditions such as cerebral palsy, amyotrophic lateral sclerosis (think Stephen Hawking), stroke, and autism, which is a developmental disorder marked by social and communication difficulties. In giving Kaya a voice, his mother says, the device has not only decreased his meltdowns, but it has increased his social inclusion at school and his language comprehension.

Yet even as autism awareness in particular grows, some Washington health insurance companies have refused to cover SGD costs, calling the devices "investigational," which implies that they haven't been shown to work for autism. This riles parents and professionals alike.

Marci Revelli, a speech language pathologist who works with many different patients at Children's Hospital in Seattle, says that these policies unfairly discriminate against those with autism. "Why are we just picking on this population?" she asks.

Costly devices, reluctant insurers

Kaya has one of the more flexible types of SGDs. It has a dynamic display and synthesizes speech sounds, rather than relying on pre-recorded sounds. The display shows pictures and their associated words: an apple for "to eat," a pointing finger for "you," two red strings criss-crossed in an "x" for "don't."

Kaya presses these types of pictures to assemble a sentence, and then finishes with another button when he is ready to broadcast it to his communication partner. "I want bubbles," was one of the first things he "said." These types of devices have the greatest potential for growth, says Revelli. But they are also costly, ranging from $4,500 to $9,000.

The Kesims received their device through Medicaid, but others insured through Regence Blue Shield, and until recently Premera Blue Cross, are not so fortunate.

Regence's current policy says SGDs are investigational in cases of autism and mental retardation, arguing that the devices have not yet been scientifically demonstrated as effective for people with these conditions. In support of this, it cites a handful of studies.

But the authors of those studies say that Regence got it wrong. "If they're quoting the [study] to say that it's investigational ... they're incorrect," says Diane Millar, of Radford University in Radford, Va. Her review article summarizing 23 studies is used by Regence to justify its policy.

Millar says she was never contacted by Regence about her article, which she points out was "collateral" to the issue of whether SGDs were useful for people with autism. She was interested in whether using SGDs and other alternative communication methods would inhibit speech development, and her analysis indicated that they did not.

As for their value to people with autism, Millar says the data are clear. "The benefits for communication, and for helping with challenging behaviors, [are] really obvious in the research at this point," she says. "It's very clear cut research with great outcomes."

Even after receiving reproving letters from some of the scientists who did the studies cited in the policy, Regence is unmoved. The letters formed part of an appeal by a client in Snohomish County, whose claim for an SGD was rejected by Regence. Lew Golinker, a lawyer in Ithaca, NY, filed the appeal, arguing that Regence misread and missed the relevant research, and that it is illegal to discriminate based on a person's diagnosis in Washington.

But Regence turned down the appeal in April.

"Regence ... ignor[ed] everything that I cited, referencing none of what was cited in the appeal, stating that the policy is the policy. Information I had asked Regence to explain they didn't address," says Golinker, who has been representing people with communication impairments for 26 years.

Ask the bosses

"It's not even their money," Golinker says, referring to Regence, which is the plan administrator for his client's employer, Snohomish County. In employer-based insurance plans, the employer foots the bill for what is recommended by its plan administrator.

Golinker asked the employer to reconsider Regence's decision, which is a strategy that has worked before for his clients with employer-based plans. In 2006, after rejections of three different appeals to Premera, he turned to the employers.

In two cases, the employer was Microsoft. "I said, 'Are you aware that your own plan is being used to deny something that your company helps create?'" Golinker recounts, referring to Microsoft's long-standing involvement in the alternative communication field. Microsoft rejected Premera's recommendation and covered the devices.

In the third case, the employer was Skagit County. "They sent the appeal out for a truly independent expert review," says Golinker. In the end, the county also decided to cover the device.

A few months later, Premera withdrew its conclusion about autism, but only to reinstate it in late 2007.

That is when Golinker found himself going back to Microsoft when another one of its employees had an SGD claim denied by Premera. Microsoft promptly reversed the denial and approved the device, without even waiting for the appeal. Soon after, Premera removed the autism exclusion from its policy again.

The story with Regence may unfold similarly. Earlier this month, Snohomish County decided to cover the device that had been denied by Regence. "This decision by the county sends a message to Regence that one of its customers is not persuaded the autism conclusion is correct," says Golinker.

According to Regence spokeswoman Lee Therriault, "This policy is actually under revision," and the revised policy will be issued by mid-summer. In the meantime, kids with autism wait for their devices.

Blinded with science

According to the scientists, Regence's policy goes against the professional consensus, and its literature review amounts to shoddy scholarship. For an ironic example, Millar's review article, used by Regence to justify excluding SGD coverage, was actually cited by Premera as a reason to cover SGDs in cases of autism.

"So you have the same article being cited by two sister Blue Cross organizations within four months of each other reaching completely opposite conclusions," says Golinker. "Neither knows what this article says."

This begs the question of who exactly is doing the reading and analyzing of these studies at the health insurance companies. At Regence a group of nurse practitioners initially drafts policy revisions, says Richard Rainey, a Regence medical director based in Boise, Idaho. The draft is then reviewed by an unpaid external physician.

Although the most relevant specialist for reviewing SGD use in autism is not clear, Rainey suggests it might be a physiatrist, who rehabilitates people with muscular or neural conditions such as stroke. The physician's recommendation is then considered by Regence's medical work group, which Rainey heads, when deciding whether to approve the revision.