Stream: Covid-19 Response
Topic: IDs for US Hospitals
Michael Donnelly (Apr 06 2020 at 18:58):
What sort of IDs are out there for hospitals in the US? Is there an exhaustive list anywhere?
Michael Donnelly (Apr 06 2020 at 18:59):
When we're referring to targets in our MeasureReports, if we just use Location IDs from the submitting orgs, the receivers won't know what they are.
Gino Canessa (Apr 06 2020 at 19:14):
For basic info, we can use additional fields in the Reference (MeasureReport.reporter or MeasureReport.subject), e.g., filling in the display with the facility name or address.
The other thing (which I think Keith has shown before) is adding extensions to the same for lat/long (though I know you want to avoid extensions on MeasureReport).
Otherwise:
- we could use contained to add in the needed info (if we use an Address, we wouldn't even be misusing it - edit: I think)
- we could make a primary bundle that includes the org, the facility location, and the report
- other ideas?
Thoughts?
Carl Anderson (Apr 06 2020 at 19:20):
There's a Medicare Facility ID in this data set:
https://data.medicare.gov/Hospital-Compare/Hospital-General-Information/xubh-q36u
Carl Anderson (Apr 06 2020 at 20:31):
Also, could we use a ZIP+4 in the location? This could be useful to point to makeshift locations (like the Javits Center).
Abbie Watson (Apr 06 2020 at 22:12):
Not all hospitals take Medicare payments, but it's a pretty good proxy. It covers something like 95% of the hospitals in the US, I think?
Abbie Watson (Apr 06 2020 at 22:19):
I'm inclined to use the IDs from the ArcGIS Homeland Infrastructure Foundation-Level Data (HIFLD) hospital data file. Pros: it presumably includes all hospital buildings, regardless of whether they're operational or not, and whether they accept Medicare or not. Cons: might include shuttered hospitals.
https://hifld-geoplatform.opendata.arcgis.com/datasets/hospitals
Abbie Watson (Apr 06 2020 at 22:21):
Also, one of the referenced data sets from the SANER Wiki.
Michael Donnelly (Apr 08 2020 at 20:59):
Those are great. Thanks @Carl Anderson and @Abigail Watson
Eric Haas (Apr 09 2020 at 15:56):
The CDC/ NHSN has ids I think - check with Austin
Austin Kreisler (Apr 09 2020 at 16:11):
Yes, NHSN assigns an ID for each facility that submits to NHSN. That ID is treated as PHI and isn't made public. Individual facilities will know their own NHSN Facility ID.
Eric Haas (Apr 09 2020 at 16:21):
@Austin Kreisler will the receiver of this information know it too... ?
Austin Kreisler (Apr 09 2020 at 16:26):
Only if the facility chooses to share that information. It won't be shared by NHSN.
Eric Haas (Apr 09 2020 at 16:32):
confused why is a big secret if is a good way to identify the facility? can you explain in 10 words or less?
Austin Kreisler (Apr 09 2020 at 16:42):
Virtually every hospital in the US reports to NHSN and have for years. Some of those facilities are very small, and much of the information reported to NHSN is clearly PHI, identifying the facility might inadvertently identify the patient. That's why the facility id is PHI.
Josh Mandel (Apr 11 2020 at 21:08):
Looks like the current federal guidance is asking for
"2. Hospital name Name of hospital and CMS Certification Number
(CCN)"
(from HHS_Ltr_to_Hospital_Admins___FAQs.pdf ).
Josh Mandel (Apr 11 2020 at 21:09):
They now seem focused on 17 data elements (3 of which are identifying the facility, so 14, if I'm reading correctly) -- actually just the original CDC list plus N95 mask counts, I think.
Keith Boone (Apr 12 2020 at 02:47):
Yeah, I think we need to talk about how to either the reportingOrganization or location, and what to do about things like stratified classifications (N95 mask responses and the next two measures in order to get them to aggregate up).
Robert McClure (Apr 12 2020 at 15:29):
@Austin Kreisler I'm not following the logic on facility ID being PHI. Are you saying that since the facility ID is in the report that contains PHI, that makes the ID PHI? Are the reports publicly accessible, so the concern is using facility ID could expose other PHI? PHI is something that comes in aggregate. Street addresses are certainly PHI in the context of a medical record but the post office would be a bit confused if you made them PHI everywhere. Help me understand this.
Josh Mandel (Apr 13 2020 at 16:15):
https://www.cdc.gov/nhsn/pdfs/covid19/57.132-covid19-sup-blank-p.pdf has details on N95
Josh Mandel (Apr 13 2020 at 16:34):
@Andrea Pitkus, PhD, MLS(ASCP)CM, CSM I'd be interested to get your take on whether HHS_Ltr_to_Hospital_Admins___FAQs.pdf sufficiently addresses the responsibilities of hospitals vs labs
Josh Mandel (Apr 13 2020 at 16:34):
(Under "Testing Data: Hospitals that Send COVID-19 Tests Data to State Public Health")
Josh Mandel (Apr 13 2020 at 16:36):
https://data.cms.gov/inpatient-provider-lookup has CMS Certification Number data openly...
Josh Mandel (Apr 13 2020 at 16:37):
For example this query shows me the CMS Cert Number for facilities in Madison, WI, or this one for Boston.
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Apr 13 2020 at 16:50):
Thanks for sharing this doc. Will take a look at later (this eve). FYI, the CLIA# is typically used to identify the performing lab as required by CLIA law on lab reports and part of ELR reporting criteria. It would be especially useful for those labs that are independent of hospitals. For those that are a centralized offsite lab for a health system, they would have their own CLIA number and based upon the patient location (inpatient hospital A, inpt Hospital B, outpt at clinic C, employee health), it could be used to "filter" reporting to what a particular hospital's numbers.
Not sure if folks have seen my Scenarios for patients from last week? It delineates most of these scenarios. (Sorry, lab is hard ;) )
Josh Mandel (Apr 13 2020 at 16:54):
Thanks @Andrea Pitkus, PhD, MLS(ASCP)CM, CSM ! Is there a public list of all existing CLIA #s somewhere? Or is this "secret" information?
Josh Mandel (Apr 13 2020 at 16:54):
Looks like https://www.cms.gov/Regulations-and-Guidance/Legislation/CLIA/CLIA_Laboratory_Demographic_Information has it
Josh Mandel (Apr 13 2020 at 16:55):
So... https://www.cms.gov/Regulations-and-Guidance/Legislation/CLIA/Downloads/regis2018.pdf has a full registry (but I'm confused about how this could be only 100 pages long; I might not understand what this is about.
Josh Mandel (Apr 13 2020 at 16:56):
Maybe this is just facilities "in trouble" in some sense.
Riki Merrick (Apr 14 2020 at 01:19):
CDC has an OID directory for facilities that have CLIA - I am not sure they give out all of them at a given time - direct a question to PHINTECH@cdc.gov
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Apr 14 2020 at 02:40):
Good question @Josh Mandel
CMS has a lookup of their CLIA database of labs here: https://www.cms.gov/Regulations-and-Guidance/Legislation/CLIA/CLIA_Laboratory_Demographic_Information Didn't see downloadable version, but they have summary stats here: https://www.cms.gov/Regulations-and-Guidance/Legislation/CLIA/CLIA_Statistical_Tables_Graphs showing ~9000 hospital labs. I'll mention that NY and WA are exempt from CLIA regulatory meaning those states have more stringent requirements, but all labs performing testing in US are required to have a CLIA #.
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Apr 14 2020 at 02:42):
Yes, that is correct for the registry, @Josh Mandel Thanks @Riki Merrick Wonder if you emailed CMS if they'd give you the full database or perhaps, just the hospital slice?
Also see here: https://www.cdc.gov/clia/LabSearch.html There's also email here.
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Apr 14 2020 at 03:34):
Regarding updated HHS Letter, @Josh Mandel, HHS wants hospital reporting of info daily. It's not asking for the independent labs that might be performing testing for a hospital to report, but the hospital to report testing it does in house and sent out to labs other than public health labs and the 6 commercial labs listed.
It does not specifically address how to handle the scenario where there is centralized testing for a health system, shared by multiple hospitals. It seems like one hospital could claim the lab as their own for in house testing, and the others would consider it a reference lab for their reports. One lab administrator of a large integrated delivery network indicatd they built their own custom report pulling data from the LIS and EHR for FEMA/WH/HHS.
Another resource that might be helpful in testing queries, is AACC's directory of laboratory's performing COVID-19 testing in the US. The HHS report wants hospital lab test info, as well as test sent to independent labs that are not one of the 6 commercial labs listed, nor a public health lab. https://www.aacc.org/global-health-outreach/how-labs-can-prepare-for-coronavirus-and-other-outbreaks/covid-19-laboratory-directory
Abbie Watson (May 06 2020 at 19:11):
Hi. So, I’m back working on the SANER IG with all sorts of new progress. First up, can I get a sanity check on the Organization IDs? I’ve been importing the IPPS file that Josh Mandel recommended, and am only coming up with 3184 inpatient facilities. Can anybody confirm a checksum of 3184 for Medicare inpatient facilities?
SANER-IPPS-Medicare-Hospital-File-Import.png
Abbie Watson (May 06 2020 at 19:11):
Meanwhile, the HIFLD file is listing 7582 hospitals locations. And Andrea is referencing ~9000 hospital laboratories. Can anybody confirm that less than 50% of the hospital locations in the US participate in Medicare reimbursement for inpatients? A 3K to 9K range of how many hospitals operate in the US seems a bit… imprecise.
Gino Canessa (May 06 2020 at 19:54):
I'll preface this with the fact that I haven't used any of the resources prior to COVID work, so if people with more experience/insight want to jump in, please do.
- The IPPS appears to only be current through FY 2017, and their criteria descriptions for what is in the list is lengthy. So, I'd say it is a high-quality source for a specific band of data.
- HIFLD lists 7581 Hospitals (has a header row in the CSV). There is a description of the criteria used to list hospitals in that set here. This is also very specific (e.g., doesn't include the HIFLD Urgent Care Facilities - 4810 records). So another input across a different band.
- The most thorough (widest band) of data I found would be the CMS Provider of Services dataset. The categories are listed there, but it includes labs, hospitals, vendors, suppliers, etc. Descriptions of the fields are in the zip files. This set also includes facilities outside the US that have provided services (e.g., COVANCE CENTRAL LABORATORY SERVICES SARI in Geneva Switzerland). I haven't filtered it down to just providers see where it lands, but the raw data has ~670K records.
Unfortunately, they all seem about right for their context to me :shrug:
Note: also not listed is that set of fake organizations (~55) for the connecathon.
Abbie Watson (May 06 2020 at 20:05):
Oy vey. One would think that understanding how many hospitals are in the US wouldn't be so difficult.
FWIW, the Urgent Care Facility file roughly matches the Dartmouth Health Service Area (HSA) file, which has something like 5.1K records. So that makes sense. Inpatient < Urgent Care / Outpatient < Service Providers.
Okay, I'm going to continue parsing these. Thank you much!
Gino Canessa (May 06 2020 at 20:54):
Good luck!
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (May 07 2020 at 00:10):
Can't speak to the number of hospitals, but that is likely the correct number of hospital labs. If labs are decentralized across a hospital (i.e. blood bank near surgical suite or ED), main lab on 1st floor or basement, cancer center lab near the cancer clinic, etc., they may have the same or different CLIA numbers. In other words, may not be a 1:1 match to the hospital. Also in some health systems with multiple hospital sites, they may have a stat lab at each hospital, but used a central shared lab offsite for routine and esoteric testing. There are many different combinations possible.
Last updated: Apr 12 2022 at 19:14 UTC
