HICDEP Development (#1) - Coding of causes of death (#61) - Message List

Coding of causes of death
 unsolved

In COHERE we have asked the cohorts to indicate what has been the source of information for coding the causes of death.

We identified 5 possible options:

1) Did not apply any method, just had the codes mapped into HICDEP/CoDe

2) Applied full Cause of Death (CoDe) protocol within own cohort

3) Used data received back from the ART-CC

4) Used data received back from DAD

5) A mixture of the above.

If the answer was “5) A mixture of the above”, we asked the cohorts to verify for each patient, submitted to COHERE in 2013, whether the source of information for the coding of causes of death was:

A=Code method within own cohort

B=Data from ART-CC

C=Data from DAD

X=Other - please specify

Therefore we introduced a variable == source == in the table LTFU. Could we introduce this variable also in HICDEP? See also the WP4 minutes of 11 June 2013.

  • Message #118

    Hi

    Shouldn't we use a numeric coding? I think this fits the Hicdep encoding style better.

    Is it possible that a cohort uses a mixture of 1 and 2? It may have used a own method until at some date it rolled out the full CoDe protocol. If so, then we should have a code for "own method" and one for "CoDe protocol".

    If the answer is other, where is the description provided? I think we should add a SOURCE_OTH field to do so.

    Lukas

    • Message #119

      I would prefer this to be free text, not even letter codes, simply because we haven't explored all the possible sources yet and the information is more of a background type than something that will be used for analysis.

      /Jesper

      • Message #120

        Fine for me. Any other comments?

        • Message #121

          Hi,

          Is it possible that cohorts could submit causes of death from more than one method? i.e. if the cohort has a code from both ART-CC and DAD, would this have been submitted?

          Many thanks, Sue

          • Message #122

            Hi Sue

            As datamanager I would tend to say yes (submit all you have).

            As statistician I don't like to decide myself between two potentially conflicting results.

            My suggestion would be to report the result from the best coding procedure. In your example this clearly is the full CoDe procedure applied by D:A:D.

            The WP4 members will discuss this at tomorrows TC.

            Best,

            Bruno

Attachments

No attachments created.