IM this article to a friend!

March 7, 2006

Dial-up deaf test goes international

From: Radio Netherlands - Netherlands - Mar 7, 2006

by Laura Durnford

Loss of hearing is one of those things that increases with age, but the majority of people never seek medical help because they don't realise they have a problem. A do-it-yourself test by phone or internet, which was devised by Dutch researchers, has not only proved successful in the Netherlands, but has recently been adapted for use in other countries, too.

Dr Cas Smits taps out a number on a telephone keypad and, after a moment, a voice welcomes him to the Dutch national hearing test. He gives a quick translation as the voice requests some basic information - the age and sex of the caller - and then Dr Smits has to enter a number between one and nine, representing how good he thinks his hearing is. This allows the computerised system to set the difficulty level for the first test.

"The test is an adaptive test, so the level is adapted to the response," explains Dr Smits, who's an audiologist at the Free University Medical Centre (VUMC) in Amsterdam. What you hear is a series of three numbers between zero and nine, but these spoken 'digit triplets' are masked by added background sound. You have to key in the three numbers you think you heard; if you were correct, the next triplet is harder to hear, if you get it wrong, the next set of numbers is easier to distinguish from the noise.

The background noise itself "has the same spectral shape as average speech", according to Dr Smits, a bit like "thousands of people all saying the same digits". Not that you're aware of this as a caller; the resulting static-like din - which the audiologist describes at "the optimal masking noise", succeeds in making the more difficult triplets so hard to decipher that you really do concentrate on the task itself.

This is an adaptation of a classic 'speech in noise test' in which sentences are spoken aloud, rather than numbers. "We used digits that you can enter on the phone and then a computer can easily judge whether the response is correct or incorrect."

Because the difficulty level is correspondingly adjusted by the computer, all callers will, on average, understand about 50 of the triplets correctly. "That means that the technical details of the test are a bit complex. It took a lot of work to implement at the telephone company, but we succeeded."

Speech-noise ratio
"What we want to measure is the ratio between the speech level and the noise level," says Dr Smits, explaining that trying to simply measure the volume level at which people hear is not feasible by phone, as many callers can adjust how loud their machine plays the message. With this test though, changing the volume doesn't alter the ratio between the spoken digits and the background noise, "so that makes the test reliable by telephone and also by internet".

On average the test lasts about three minutes, involving 23 separate sets of numbers being presented to the caller. At the end, the user's performance is assessed as 'good', 'insufficient', or 'poor', and those at the lower end of the scale are advised to consult a professional for further help. Based on one research project, Dr Smits knows that, in the Netherlands at least, only 22 percent of hearing-impaired adults actually use a hearing aid - "that's the reason I started this research".

After running for several years in Dutch, via phone and internet, the system is now being developed in other languages for use in other countries, as part of a Europe-wide project called 'HearCom'. An English-language phone version was launched at Christmas by the Royal National Institute for the Deaf (RNID) in Britain, and has already attracted many thousands of callers. Dr Smits, whose idea the test was, is naturally pleased that it has taken off in this way - especially since problems of hearing-loss and lack of attention to it are comparable in all countries - "but it was already nice to see that it was so successful in the Netherlands," he smiles.

And Dr Smits knows exactly how successful the test has been, because in addition to devising and implementing it, he's also spent several years investigating its reliability and effectiveness. His results were presented as part of his PhD, completed in mid-January.

The first task was to compare the new number-based test against the standard, sentence-based version. "The results were very comparable, so for us that was an important result." Dr Smits and colleagues also checked whether different types of telephone would impair the test. "Fortunately that effect is so small that we could not detect a significant difference," he says, although he adds that the ensuing growth in telephoning via internet could affect sound quality, "so maybe we have to do a new project to check whether that's a problem."

High percentage
But for the audiologist, the ultimate proof of his test lies in the behaviour of the many thousands of people who have used the Dutch language systems over recent years. The in-built questions about age, sex and - on the internet version - whether participants were using headphones or speakers, have allowed him to gauge something about who is using the system and how. More detailed questionnaires that were posted to 2000 phone users revealed that of those who had received the message that their hearing was insufficient or poor, 50 percent had followed the recommendation to consult a specialist within the following three months.

"Maybe it's a bit biased in the data," muses Dr Smits, since the people who are willing to return a questionnaire may also be more open to following advice, "but 50 percent is really high for such a self-test, and finally we hope that they go to a shop and buy a hearing aid!"

Copyright Radio Netherlands 2006