NCHS Mortality DB

[justin]

Hi,

Just looked around the NCHS mortality stats site, and unfortunately
the data might not be what we're looking for. The biggest road block
is that over time the data has gotten less and less precise:

Prior to 1989, precise geographical data were associated with each
death record, along with exact date-of-death.

From 1989-2004, precise geographical data were only associated with a
death record if the location was in a city or county with population
greater than 100K, and the date is partially obscured, giving year,
month, and weekday but not date.

After 2004, the records contain no geo data at all. The dates are
unchanged from the 2004 format.

So, we won't get recent data at the district level using the publicly
available data. OTOH, it is possible to apply for geographically-
detailed data, provided that you give a reason and a full list of
people that will see the data, as well as a time range that the data
will be used. For the purposes of watchdog, though, that sounds like
an unsuitable course of action.

Do you all think that this dataset is suitable despite its
limitations?

Justin

[Alex Gourley]

What if you put in the request for the data just for yourself for a
weekend for the purpose of research. IANAL but do they have a
copyright on this data?

-Alex

[Aaron Swartz]

> What if you put in the request for the data just for yourself for a
> weekend for the purpose of research. IANAL but do they have a
> copyright on this data?

Since we only want the aggregate data, perhaps it would be reasonable
to request the data for research and then publish the aggregate
results you drew from it and then we could include those in watchdog.

Re copyright: US Government works don't get copyrighted and you can't
copyright data anyway.

[Aaron Swartz]

Also, NICAR seems to have gotten a copy:

http://data.nicar.org/node/24

We can purchase it from them if that will help.

[garryj]

Hi All!

Did you know that the top three causes of death in San Francisco
county from 1999-2005 are:
1. Pituitary gland
2. Meninges, unspecified
3. Brain, unspecified :-)

I agree with everything Justin said about the specific limitations of
the public-use data and it's also nice to know we can get a dump from
nicar.org. However, if we need only aggregate data, we can scrape it
from http://wonder.cdc.gov. For our purposes, I think we can just use
a query that groups by County, Cause of Death and break down the
queries by state (otherwise it takes too long and sometimes times
out).

I have a call and email into Charles Rothwell @ the NCHS who's the
contact for detailed data requests. Asked for geographically-detailed
data for 2005+ and explained we'd only use it in aggregate. Letcha
know whta happens.

Garry

[garryj]

Perhaps someone more familiar with districting can help..? Can you
map county data -> districts?

[Aaron Swartz]

> Perhaps someone more familiar with districting can help..? Can you
> map county data -> districts?

Not cleanly.

[christopher bdnk]

On Thu, Sep 4, 2008 at 12:21 PM, garryj <gar...@gmail.com> wrote:

Perhaps someone more familiar with districting can help..?  Can you
map county data -> districts?

I haven't tried this yet, but it should be somewhat possible with a rough hack. You can use district and county population statistics to combine the data from a few different counties (or a partial county) into a single district, but it won't be accurate.