United Health extends its direct-to-consumer hearing aid offer to a larger market

What the various hearing aid professional groups fail to understand about the hi HealthInnovations online hearing test is that it is .NOT. designed to diagnose hearing loss: Instead, just like the Master Hearing Aid function on audiometers, their web-based app is a tool designed to aid in setting the gain and response of the hearing aids.

In an e-mail responding to this thread from program development leader Dianne Van Tasell PhD, she asked me to post the following:

You mentioned in your first e-mail that there was a technical discussion going on. As I read through the comments, though, it seemed to me that the bulk of the discussion had to do with professional and business issues, very important ones that are stimulating a lot of discussion around delivery models. I led the development of the two versions of the hi Health Innovations hearing tests, so I can respond to the technical issues. I am not an employee of hi Health Innovations, however, so I don’t have much information about how they have set up their order fulfillment or customer service system, so I can’t respond in those areas.

Here are a few comments I have in regard to the questions of a technical nature:

1. I just want to clear up a few misconceptions evident from Dr. [Jason] Galster’s comments and those of some others. The challenge of the home test is to get enough information in order to program a hearing aid so that the user will find it beneficial. [Emphasis added: DLS] I personally think that there are probably many different paths to this that would lead to success (say, filtered speech tests or speech in noise tests or just having the user adjust the hearing aid him/herself), but the method that is the most extensively researched requires an audiogram. Since we wanted to stick with the method that had the largest body of supporting research, the problem became how get a sufficiently accurate estimate of the audiogram over the internet. The interesting issue that arose immediately was: how accurate to you need to be? Remember, the purpose of the test is not to get the audiogram, but to put appropriate gain into the hearing aid. [Emphasis added: DLS] If you take any gain prescription method for nonlinear hearing aids (NAL-NL2 is the one we used) you will quickly see that there are many audiograms — themselves quite different — for which the gain prescription is essentially identical. This is because those methods use thresholds at all frequencies (not just one as in the early linear methods) to determine a frequency response that will maximize speech intelligibility and still provide acceptable loudness. Turns out that there are a lot of audiograms for which this frequency response is very similar. A different way of saying this is that a large audiogram space collapses into a much smaller prescribed gain space.

So, it became clear that one would not have to get an absolutely accurate audiogram in order to get very close to the NAL-NL2 gain prescribed for the user had one actually known what the accurate audiogram was. For purpose of hearing aid programming the method does not, as Dr. Galster describes, sort users into groups. On the contrary, it works like this: 1) we begin by assuming that the actual audiogram is one of the modal audiograms reported by Ciletti & Flamme (Journal of the American Academy of Audiology, 2008. 19(9): p. 672-85). There are different possible audiograms for males and females; since we know the gender of the user we know which set to use. 2) we then estimate the PTA from the user’s score on the Hearing Screening Inventory in the manner described in the Hearing Review article; this allows us to eliminate all the candidate audiograms that have PTAs outside the expected error range of the estimate. 3) we use the estimate of audiogram slope between 2 and 4 kHz to eliminate further all audiograms that have slopes outside the expected error range of the estimate. This process usually results in only one or two candidate audiograms, although there are some cases in which the set of candidate audiograms reduces to 3. We cannot know which is the best estimate, so we program each of the 3 hearing aid memories with gain corresponding to each of the candidate audiograms and instruct the user to listen to all until he/she decides which is the most beneficial. As reported in the article, when we analyzed our subject population we discovered that in at least one of the 3 memories we were within a dB or so of NAL-NL2 prescribed gain.

Since we can’t know the exact audiogram, we never report one to the user. [Emphasis added: DLS] We can, however, estimate what general degree of hearing loss the user has once we know what the set of candidate audiograms is. This is no different, in principle, from the estimate of hearing loss severity that one gets when one takes the BHI Quick Check and then receives the results. So we do give the user some general information about how much they are hearing based on the results of our estimation procedure. But we do not program the hearing aids that way from the home test. In the office test, we know what the audiogram is so we don’t need to use the estimation procedure.

2. In response to Dr. Gosalia: what can I say? If he had read the article he would have known that the threshold measurement procedure produces thresholds very close to standard audiometric techniques. This is because the method of adjustment used in both tests is modeled after the Hughson-Westlake search procedure: the user first hears tones audibly, then they are reduced to inaudible levels, and then there are at least two ascending threshold measurements. The user has to respond consistently on two ascending runs in a row before the threshold can be estimated. So it is not surprising that this procedure produces thresholds similar to those obtained with manual techniques.

3. The other comments in the thread all have to do with the business model and the fact that there are many best clinical practices excluded when the user does not interact with an audiologist. This latter cannot be disputed, of course. But as [BHI Director] Sergei [Kochkin] has pointed out, the important issue is whether the hi HealthInnovations method (or any direct-to-consumer method, for that matter) produces results that provide equal or more benefit to the user than a visit to an audiologist or dispenser. You might argue that even if you hit NAL-NL2 targets every time, individual users require adjustment and I do not dispute that. But how many audiologists/dispensers are there out there who are hitting “first fit” and then calling it a day? Given the significant effect of price on users’ perception of benefit (there’s that now-famous Killion figure), I would submit that in those cases the differences in benefit would be very small, and might even significantly favor direct to consumer.

Dan Schwartz,
Editor, The Hearing Blog