care seeking behaviours [message #8303] |
Mon, 05 October 2015 13:35 |
lucianabrondi
Messages: 18 Registered: October 2014 Location: Scotland
|
Member |
|
|
Dear DHS Forum colleagues,
I doing an analysis of care seeking behaviours for children with diarrhoea, ARI and fever using the India DHS2005_06 dataset in Stata. I saw another post in this Forum (message #2774) asking about ARI and care seeking behaviours and I am bit confused. I am looking at care seeking behaviours by sex and other risk factors like rural setting, mother's education, birth order, etc.
1. If I wanted to use initially the "summary" variables for diarrhoea, for instance, I am talking about h12y , h12z, h21a and h21 for DHS recode V in this example. Can I rely on these "summary" variables to assess if the child carer looked anywhere or gave the child any treatment? Or do I have to create a new variable using the preceding variables?
2.If I want care seeking for ARI specifically, I suppose I will have to create a new summary treatment variable because h32y and h32z refer to fever and/or cough.
3. What would be the disadvantage of dropping all the dead children from my child recode dataset (since these questions are only relevant to live children) instead of creating new variables using the live children as a denominator?
Thanks,
Luciana Brondi
CPHS, University of Edinburgh
|
|
|
Re: care seeking behaviours [message #8348 is a reply to message #8303] |
Wed, 14 October 2015 10:53 |
Liz-DHS
Messages: 1516 Registered: February 2013
|
Senior Member |
|
|
Dear Luciana,
Here is a response from one of our expert analysts, Dr. Rebecca Winter:
Quote:In response to your first question: Yes, the summary variables for care seeking are reliable, the only question is whether these summary indicators summarize the information in exactly the way you wish to summarize it. You could also create your own summary variables, specifying exactly which sources of care you wish to include. For example, you could use something like the code below, to identify children for whom care was sought from any source (this is just an example that will match the summary variable h32y):
**Generate a variable to identify children for who treatment was sought from ANY source for fever
g trtany=0
foreach x in a b c d e f g h i j k l m n o p q r s t u v x {
replace trtany=1 if h32`x'==1
}
replace trtany=0 if h22==1 & trtany !=1
*Prepare a weight variable
gen wgt=v005/1000000
* Tabulate the weighted percentage of children with fever who received any treatment
tab trtany if h22==1 [iw=wgt]
In response to your second question: If you wish to look at care seeking for symptoms of ARI, you will want to restrict your analysis of care seeking to the children with reported symptoms of ARI. I recommend that you first generate a variable to identify children with symptoms of ARI (see sample code below), and then, when you tabulate the variable h32y, restrict the tabulation to those children. [Note that we recommend defining "symptoms of ARI" as children with a reported cough with rapid/difficult breathing for whom the difficult breathing is related to a problem in the chest, or to a problem in the chest and to a blocked/runny nose, as in the sample code below.]
Example code:
*Generate a variable to identify children with symptoms of ARI
gen ari=1 if h31c==1 | h31c==3
* Tabulate the weighted percentage of children with symptoms of ARI who received any treatment
ta h32y if ari==1 [iw=wgt]
In response to your third question: Since female respondents are only asked questions about their children's recent illness for living children under age five, your analysis is already restricted to living children. It should not make any difference whether you drop the dead children from the recode file, because your tabulations should be restricted to living children with recent illness anyways.
I hope this is helpful!
[Updated on: Fri, 16 October 2015 18:50] Report message to a moderator
|
|
|