Thanks to everyone who has publicly or privately suggested improvements to the location of the facilities in the placemark collections. I have tried as described above to improve the quality of the coordinates, but the fact is that a small proportion of the NPI coordinate data is incorrect. I suggest you go straight to the source and send them suggested corrections at their comments page. I have included the address of each facility in the bubble so you can try to find it yourself if you like.

On another note, I found that using RPD resulted in rather flat and uniform columns rising from each site. After a little statistical research I decided to assume the data is normal (it isn't) and calculated Z-scores which I then summed and amplified for each site to provide a height for the columns.

As a result of the above I have updated all of the relevant posts and attachments.