This is a well-written and very clear paper. It covers several aspects that went unmentioned in the Gizmodo article, and highlights that although building such a device may be trivial, evaluating and studying it is not.
Some highlights: the paper covers how the device was built in detail, including information on the three safety mechanisms used to ensure the operator doesn't get exposed to infected blood (shatterproof plastic capillaries, epoxied sample holders made from drinking straws, and sealing of the capillaries inside two paper discs). It is exclusively made from low-cost materials, but it's more than just a piece of paper, handles and some string (fishing wire).
The paper also covers the physical dynamics of the "paperfuge" in great detail, analyzing its rotational dynamics and building a theoretical model of motion that agrees well with the physical observations (captured with a 6000fps high-speed camera). It also shows that the max RPM varies with disc size, with 125000 RPM for a small disc (5mm diameter). The paper even mentions that this was submitted as a Guiness World Record.
Finally it shows that the paperfuge produces sample separation results on par with electromechanical centrifuges using similar spinning time (1.5min for paperfuge, 2min for centrifuge for plasma separation), and does an analysis of the resulting blood samples.
An aside: If you are not reading Nature you may be missing out on a lot. I would guess the same applies to Science.
Only the back half of Nature is actual research papers, which often are indeed very slow going. The other half is science news written clearly - it's fascinating, not a chore at all - and with a level of knowledge and sophistication unmatched elsewhere (AFAIK), and most importantly it will completely change your perspective: There is a world of research and knowledge that you won't even know exists if you read the more 'popular' science press.
A little time reading Nature will save you much more time reading less informative publications.
Wouldn't want to waste words on silly shit like methods, references, discussions, etc in a CNS paper, would we?
LIGO doesn't even bother sending papers to Nature or science any more. The editors at both are unbearably pompous.
-  http://www.nature.com/news/
Not including JSTOR and other subscriptions for these Journals. The bill 10 years ago was about $70,000 for journal access on computers where I had the print copies in boxes behind the journals or in the storage area collecting dust.
Almost no organization can afford to pay for all the stores, so even universities have limited access in the sense that they can only offer their students and staff a limited selection of what is available. Subscription fees are too high, single journals are often only available in bundles with other journals/topics, increasing the price further. Publishers act as gatekeepers to (often publicly funded!) knowledge.
I really like that there's a comparison to commercial centrifuges, since that shows the quality of medical care isn't going down simply to be cheap. Overall it looks like a well thought out design.
On top of that, the g-force would be
a = v^2 / r
v = 2 * pi * 5mm * 125000/60 s^-1`
r = 5mm
I start to wonder if general relativity effects (frame-dragging?) start to become noticeable at that acceleration.
9,000 MPH = 0.00001342 C
Threshold for relativistic effects is .01 C according to this paper:
EDIT: Rereading that paper, they're focused at the particle level, which may or may not make this irrelevant. I know anecdotally that GPS satellites have to take relativity into account. Geostationary satellites move at 1.9 miles per seconds, which is 6840 mph - quite a bit lower than our centrifuge. That being said, the precision required for GPS means that very small changes due to relative effects have a rather large impact. In short, where there is motion, there is relativity. Is it enough to measure here? Possibly - sticking a small microcontroller and having it report the time would be interesting. Is it enough to matter? The answer is relative.
So although you are technically correct (45 > 7) they are both significant.
EDIT: I might be wrong on the “11km” part. See reply
It's also less relevant when solving using the more common method which is time-tagging at the receiver, relativity might be a bigger pain if you are doing time-tagging at the transmitter but since the clocks of different satellites are not synchronized and can differ by as much as 1ms that delta would be more of an issue than relativity.
If you are doing normal GPS accuracy which is 15m you don't need to use relativistic correction in the receiver, is you are solving for almanac data you do not need to do relativistic correction in the receiver, you only really do relativistic corrections when you are using centimeter level accuracy GPS and even then you have considerably bigger corrections to do.
Yes relativity was accounted for when GPS was designed, but it's by no means would send you to China, not even to the wrong town.
What GPS needs to work is for the receivers to be able to calculate a proper geometric range delay or at least be constantly and consistently wrong when calculating the delay for different satellites in the cluster.
Discounting relativity would reduce accuracy, but it would be consistent and constant and so would not present that much of an issue, it's also by far not the biggest offender when it comes to corrections.
The other "relativistic" correction is one done in the receiver and is needed for geometric range delay calculation, basically taking into account that light doesn't travel in a straight path, we've kinda known that since the early days of classical optics - Fermat's principle or the principle of least time.
Now while it's true that we are using the "relativistic" version of this in our calculation solving the classical or even ignoring it all correctly as long as we account for other effects would still allow you to find a starbucks.
Since GPS was initially intended for military use it had other uses such as time keeping that can be used to synchronize communication, encryption and other things these errors can be cumulative whilst navigation errors are often not since the drift would be more or less uniform with all satellites you see.
Fill in the blanks, m/s:
v (in m/s) = pi * D * rpm / 60 (D in m)
3.141593 * 0.01 * 125,000 / 60 = 65.45 (m/s)
v (in km/h) = pi * D * rpm * 60 / 1000 (D in m)
3.141593 * 0.01 * 125,000 * 60 / 1000 = 235.62 (km/h)
2 * Pi * 0.005m * 125000/60 s^-1 = 65.4m/s = 235km/h.