[Imports] US - HIDFL (DHS) Hospital Import

[Mike Thompson]

On Wed, May 3, 2023 at 6:27 PM William Edmisten <wcedmisten at gmail.com>
wrote:

> Hello,
>
> My name is William Edmisten (wcedmisten). I've been mapping in OSM for
> almost a year, but have no relevant affiliation with any other
> organizations.
>
> Hi William, thanks for engaging with the community!


> I am writing to propose an import of the HIDFL Hospital dataset
> <https://hifld-geoplatform.opendata.arcgis.com/datasets/hospitals/explore>
> (8,013 records), and to discuss possible ways in which this dataset could
> be used.
>

> I've documented some details about the Hospital dataset and how it might
> be applied on this wiki page
> <https://wiki.openstreetmap.org/wiki/HIFLD/Public_Health>.
>
It appears, as I think Marc has pointed out, that you have already started
this import.  You need to get community buy-in *before* starting to
actually load data into OSM via an import.  If this is the case, please
stop the import until the community has had a chance to comment on your
proposal.

Also, I don't see this import listed in the Import Catalog:
https://wiki.openstreetmap.org/wiki/Import/Catalogue

You, and others who participate in the import, must use an account other
than their regular OSM account, e.g. for you that might be
wcedmisten_imports

>
> Looking at the archives, the entire HIDFL dataset was previously discussed
> here in October 2022, but I would like to specifically focus on the
> Hospitals dataset. One of the previous concerns was that the scope of HIDFL
> overall was too broad, so I would like to narrow the discussion to this
> single dataset for now.
>
> I believe this data is appropriate for OSM, and can be verified
> independently of the dataset, making it a good choice for import or for
> reference while mapping. In particular, hospitals' websites can be used to
> verify most or all of the data aggregated here.
>
I have some serious concerns about the likely quality of this dataset.  The
metadata says that the dataset was "geocoded", while
that can have different meanings, what it means here is likely that a list
of hospitals with addresses was fed to a computer program that matched them
up with an address (reference) database to obtain coordinates.  While this
seems simple, a lot can go wrong.  For example, a transposition of a digit
in a hospital's zip code can cause it to be placed 100's of miles away
from its true location.

Duplicates are another problem, I have checked a few hospitals in my local
area, and I have already found two sets of duplicates (cases where one
entity purchased another, and now they both show up in the dataset at
essentially the same location).

While the metadata says that the dataset was updated in January of 2023, I
don't think that means that a complete update was done, just that those
maintaining the data changed something at that time.  Thus, large parts of
this dataset could be out of date. Some of the validation dates (VAL_DATE)
are from 2013, 10 years ago! Also, the VAL_METHOD for many records is
listed simply as "IMAGERY", and while a large general hospital is probably
obvious from imagery, this is not necessarily true for a smaller facility.

I would like to see at least a sample of the data converted to the OSM
schema.

Mike

>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/imports/attachments/20230504/71dd9abe/attachment-0001.htm>