Healthcare and fast networks

I've written about this before in a different forum, but that was ages ago and it still holds true.

People thinking that fibre isn't necessary for telehealth don't understand just how many kinds of applications there are, and the truly vast scale of the data that can be produced.

If you live in any reasonable suburb, there's a fair chance that it has a medical imaging joint servicing the local GP. There are more than 200 in greater Sydney, Google tells me – and most of these aren't in the lousy-with-fibre streets near major hospitals. They're in suburban high streets or the upper levels of ordinary shopping centres.

Go and get a CAT scan, look at those images, think how many there are if you've got lots of pictures in the scan, and take a guess at the amount of data there is. Individual images – depending on resolution – might range from 60 MB to multi-GB per image.

Which is why you go back to the joint to collect the images, go home, and then drive to your GP to deliver them.

Which is why, because the private imaging joint I have in mind didn't have a fibre link to Royal Prince Alfred hospital, I once found myself asked by a specialist treating my wife to walk to the place, pick up the DVD, and return it to him as quickly as possible, since they needed the images for a procedure due to start in 20 minutes.

It's sneakernet, sometimes with cars, all because there isn't fibre.

Now, consider this in economic terms – by which I mean stuff like productivity and unnecessary travel.

This report tells us that per 1,000 head of population in 2009-2010, there were 816 imaging procedures. For 22 million of us, that means there are nearly 18 million imaging procedures taking place.

Let's assume that just 25 percent of these happen in the suburban imaging service – 4.5 million. And let's assume a standard model, with all trips being 1km each way at an easy-to-calculate 50c per km:

1. Trip to GP to get referral for imaging
2. Trip to imaging centre for imaging
3. Trip to imaging centre to collect images
4. Trip to GP to deliver results

If the images can be delivered directly from the centre to the GP, the travel is reduced by 25 percent. Each year, in other words, connections suitable for lots of gigabytes would (in this very simple model) save 9 million kilometres of unnecessary travel.

That's nearly two billion tons of carbon, not to mention the cost of the travel, and there's the productivity cost to be considered. If you include time hanging around in the waiting room, those 4.5 million trips could easily represent a couple of million hours of unproductive time.

Even that's only $50 million a year, if we assess a productive hour as being worth $25. Plus (say) $9 million in travel. But get this: it's $59 million a year, probably rising – and it goes on forever. There won't be a year in the future in which there will suddenly be no medical images created.

And that's just one fibre-medicine opportunity. There are others, and every single one of them delivers real-money benefits.

If you try to imagine “telemedicine” as a single monolithic entity, you're missing the real point. Medicine – in any form – involves a huge number of different interactions, and many of those interactions are some kind of “sneakernet”. It's just that it's not easy to write a single article, or post a single video, that explains all the interactions.

“But you don't need to transmit that much data!” isn't an argument, it's a reflex built from a simplistic understanding of the billions of interactions involved in healthcare that get better with better connectivity.

Wireless spectrum scare-story: $400 per year per user?

-->

If Ord Minnet’s data are to be believed, the federal government’s spectrum reserve price is going to add $400 per year to “broadband consumers”.

If, of course, the Daily Telegraph got its understanding of the Ord Minnet numbers right.

Let’s take the number as read, and since I can’t see why a mobile investment would be recovered from fixed broadband users, let’s confine the calculation to the country’s 5.8 million wireless subscribers. Finally, let’s take one other number Ord Minnet’s Brad Dunn gave to the Tele:

“The bill to purchase enough spectrum to cover the 22.6 million population will be approximately $1.5 billion.”

With 5.8 million subscribers, $400 per subscriber would yield $2.3 billion in the first year alone. Or to put it another way: to recover $1.5 billion in a year, the additional cost to users would only need to be $258, not $400.

So that didn’t work.

Not only that, but spectrum is a multi-year investment. Telstra doesn’t have to recover $1.5 billion in the first year; $150 million over ten years would pay it back (yes, I am ignoring the cost of capital).

If you ignore any kind of ramp-up, then Telstra’s current 4G customers – around 375,000 last August according to ComputerWorld – would have to repay $150 million in the first year.

Would you be surprised to find out that $150 million divided by 375,000 is $400?

But that’s only the first year (and from the laziest calculation possible). If you presume that Telstra wants to have 10 million 4G customers after ten years, and use a simple linear ramp-up, the per-year spectrum cost recovery by the tenth year falls to $15 per user.

A more complex calculation is to assume that Telstra would spread the $1.5 billion over ten years so that each user pays the same amount towards spectrum costs. Then, ignoring the cost of capital, the extra burden on subscribers is a paltry $29 per year (or less than $2.50 per month).

Even with a 10% cost of capital – if my spreadsheet is correct, which I don’t guarantee, but it's good enough for this case – the annual per-user recovery with a ten-year ramp up is $132 – or a little over $10 per user per month.

There’s plenty of room in a 4G premium alone to cover that.

If any of my numbers are close to reality - closer, say, than $400 per user per year - then perhaps the government's spectrum floor price isn't as stupid as people are saying.

Will the coalition’s FTTN get people off dial?

This is odd: There’s a considerable gap between two different ABS measures of the number of dial-up Internet customers still in Australia.

According to the bureau’s Internet Activity, Australia, there were 379,000 household dial-up subscribers at the end of 2011. However, according to the Census data (collected in August 2011), there were only 235,000 households reporting themselves as dial-up Internet users.

I suspect that there are a lot of people who kept a backup dial-up account at some point, and who haven’t gotten around to cancelling it.

This is an aside, however. 

This post is in the context of Malcolm Turnbull's initiative to survey users on broadband speeds.

I was interested that Turnbull's concerns include the dial-up customer (which led me to notice the data discrepancy). To quote Tony Eastley’s introduction* in AM, “the opposition says close to half a million households still use dial-up technology, and they’re being forced to wait too long for the NBN.”

Given the speed of dial-up’s decline in Australia, I doubt it. The dial-up sector lost nearly 100,000 customers in the six months to December 2011 anyhow – if that rate is maintained, it will die of its own accord in a couple of years.

(I would guess that most of the remaining 90,000-ish businesses with dial-up accounts use them only for backup, by the way. That application is being replaced rapidly with mobile connections.)

My other point is best illustrated by a rough map:

This shows the Census SA1 areas in which more than ten percent of the households still have dial-up connections. I haven’t calculated exchange distances, but the illustration should tell you enough.

Dial-up is at its most stubborn in remote places – areas that would be the slowest, most expensive and difficult to upgrade to FTTN.

*Note: it's perfectly possible that AM misunderstood something Mr Turnbull said, but that the program didn't put to air. If so, my apologies to Mr Turnbull.