The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Malaria » Identifying Fevered Children Who Received an Antimalarial
Re: Identifying Fevered Children Who Received an Antimalarial [message #9507 is a reply to message #9447] Wed, 06 April 2016 20:49 Go to previous message
Trevor-DHS is currently offline  Trevor-DHS
Messages: 766
Registered: January 2013
Senior Member
The general rule currently for the h37 series and the ml13 series is that ml variables are used whenever the malaria module is used in a questionnaire and the h37 variables are not used in that case. For both sets of variables the questions are applicable for children with fever or cough in the last 2 weeks (h22 == 1 | h31 == 1 | h31 == 2) [Note h31 code 1 (cough in last 24 hours) is rarely used but is still permitted where the data are collected for 24 hours and for 2 weeks.]

In some surveys though, particularly older ones, this general rule has not been followed for various reasons, and we find non-standard denominators for some variables. In earlier surveys there were some differences with ml13 sometimes including treatment for fever and convulsions, sometimes fever or convulsions, and sometimes fever alone, while the h37 series ignored the questions related to convulsions. Additionally, in many earlier surveys, questions about treatment were asked separately about treatment for fever and for cough, and the variables were typically presented separately in that case.

Consequently it is necessary to check the questionnaires and to review the data and see which cases are applicable for each variable.

To check data for a particular survey I would suggest doing some simple crosstabs, such as:
gen fevercough = (h22 == 1 | h31 == 1 | h31 == 2)
tab h37a fevercough,m
tab h37a h22,m
tab h37a h31,m
* and similarly with ml13a instead of h37a.
tab ml13a fevercough,m
tab ml13a h22,m
tab ml13a h31,m
tab ml13a ml12,m

Look at the cases which are not missing for h37a or ml13a compared with the cases that are code 1 on fevercough, code 1 on h22, code 1 or 2 on h31, or code 1 on ml12. For example, for Malawi 2004 I can deduce that the h37 series are applicable only for children with fever in the last 2 weeks from the following table:
. tab h37a h22,m

  fansidar |
 taken for |              had fever in last two weeks
     fever |        no        yes         dk          9          . |     Total
-----------+-------------------------------------------------------+----------
        no |         0      2,823          0          0          0 |     2,823 
       yes |         0        879          0          0          0 |       879 
         . |     5,995          0        157          4      1,056 |     7,212 
-----------+-------------------------------------------------------+----------
     Total |     5,995      3,702        157          4      1,056 |    10,914

as the only cases that are not missing for h37a are those where h22 is 1, and when h22 is not 1 then h37 is missing (not applicable).

In this survey ml12 is labeled as convulsions in the last 2 weeks, but crosstabbing ml12 with h22 shows that they are identical and ml12 should just say fever in the last 2 weeks. If you look at s472 you will find information about convulsions in the last 2 weeks. However, looking at the questions following question 472 in the questionnaire, no information was asked about treatment with different drugs so there is no data on treatment with, say, Fansidar for convulsions.

On checking the programs used for the Malawi 2004 survey it appears that the h37 series uses data from questions 466C, 466H, 466M, 466R, 466W. In contrast, the ml13 series only use data from question 466C which is only asked if the child was given medicine from home or from a pharmacy without prescription. In this case I would say that the ml12 and ml13 series are mis-labeled.

Where the questionnaire differ from standard questionnaires (and the Malawi 2004 questionnaires differ quite a bit) you may find differences like this. Unfortunately I cannot give you a hard and fast rule and I suggest that you use some simple crosstabs, such as above, and you review the country specific variables (e.g. s472) to work out who certain questions apply to. I hope this helps.
 
Read Message
Read Message
Read Message
Read Message
Previous Topic: Computing the number of U5 children by age classes in Burundi
Next Topic: Citing Malaria Indicator Survey
Goto Forum:
  


Current Time: Mon Jan 17 01:42:09 Coordinated Universal Time 2022