The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Dataset use in Stata » Infant data(less then 1 yr) from Children data
Infant data(less then 1 yr) from Children data [message #1421] Mon, 24 February 2014 15:28 Go to next message
Malachi Arunda is currently offline  Malachi Arunda
Messages: 30
Registered: February 2014
Member
Dear experts
Could you guide me know how I could only use infants data from the children recode. Data n the the age at death B6, of children and am looking at association of antenatal care and infant mortality. please advice accordingly. Kenya 2008-9
Thanks
Malachi
Re: Infant data(less then 1 yr) from Children data [message #1436 is a reply to message #1421] Tue, 25 February 2014 14:34 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3016
Registered: February 2013
Senior Member
Following is a response from one of our DHS experts Shea Rutstein.

Antenatal care in the Kenya 2008-09 DHS was only asked for the last live birth. For a simple tabulation of infant mortality by antenatal care, select last-born children who were born 12 to 59 months prior to the survey [v008-b3(1)>11 and v008-b3(1)<60]. The reason for selecting children born 12 or more months prior to the survey is to avoid a bias created by censoring of the time exposed to mortality.

Depending on the data file selected, you may have to link the maternity record to the birth history record using the index to the birth history [midx(1)=1]. Some children's files will not need linking but it depends on which file you are using.

Then recode age at death (imputed) into a new dichotomous variable, infmort, such that infmort is 0 if child is still alive or died at 12 or more months of age and 1 if child died at less than 12 months of age [infmort=0 if b7(1)=notappl or b7(1)>11; infmort=1 if b7(1)<12].

Cross tabulate by the antenatal variable of interest, for example, with m2a(1) if had antenatal care by a doctor. Remember that the difference in infant mortality by antenatal care may be confounded by other characteristics, such as wealth, education, residence, sex and birth order of child, mother's age at birth, etc. Multivariate analysis is usually need to sort out the confounding characteristics. For example, it was found that in Ghana children delivered by doctors died more often that those delivered by nurses. However, once wealth of the household was controlled, doctors had a better record of survival. The confounding was due to poorer mothers going late to get assistance from doctors when deliveries were complicated but richer women would go directly to doctors even for uncomplicated deliveries.
Re: Infant data(less then 1 yr) from Children data [message #1471 is a reply to message #1436] Sun, 02 March 2014 16:49 Go to previous messageGo to next message
Malachi Arunda is currently offline  Malachi Arunda
Messages: 30
Registered: February 2014
Member

Dear DHS Expert Shea Rutstein,
Thanks a lot for your elaborate reply, however, it was my fault I did not mention that I am using SPSS and NOT Stata, I think because I've just started learning how to use this helpful forum. Of course the guidance is still very useful in my work. I am going to use try use the information on SPSS, however there could be something more you have while using SPSS, Once more my question was as below, kindly let me be sure what to do while using SPSS.

My question was

Could you guide me know how I could only use infants data from the children recode. Data on the the age at death B6, of children and am looking at association of antenatal care and infant mortality. Kenya 2008-9
THANK YOU
Malachi
Re: Infant data(less then 1 yr) from Children data [message #1670 is a reply to message #1471] Wed, 26 March 2014 09:32 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3016
Registered: February 2013
Senior Member
Following is another response from Shea Rutstein.

Use the SPSS .sav file, which is in KEBR52SV.zip. Then use SELECT IF (MIDX=1) to select only the last born children who have maternal data (M vars). To create the dichotomous variable for infant mortality use RECODE B7 (0 THRU 11=1) (ELSE=0) INTO INFMORT. Remember that children born less than a year prior to the survey have not been fully exposed to the risk of infant mortality (i.e. censored) and so will have a lower . To avoid this bias, create a variable to filter out children born less than 12 months prior to the survey date. Use COMPUTE INFEXP=0. And IF (V008-B3>11 AND V008<60) INFEXP=0. Then use FILTER BY INFEXP.

You can then tabulate by using CROSSTABS TABLES=INFMORT BY M2A or other variable of interest. You could also use MEANS TABLES=INFMORT BY M2A.

One hint: To get the infant mortality rate per 1000, just use COMPUTE INFMORT=1000*INFMORT before doing the MEANS procedure.
Re: Infant data(less then 1 yr) from Children data [message #1895 is a reply to message #1670] Sun, 06 April 2014 18:16 Go to previous messageGo to next message
Malachi Arunda is currently offline  Malachi Arunda
Messages: 30
Registered: February 2014
Member
Thank you for your reply, Considering the SPSS analysis.Children or birth recode 2008-9. The BIDX/MIDX=1 are 6102 who were the ones whose antenatal care data were collect. BUT the ANC data has only TOTAL of 4073 children.1) How do I go about this. 2) I have done the procedures you adviced me accordingly, however, when I'm doing cross tabs &chi square test (infant mortality vs ANCvisits)or INF_MORT vs Tetanus injection or Iron Folic Acid, do I use this information to make conclusion OR Instead I use the ANC vs B5 (childalive or Dead IF lastborn-BIDX=1)

Thank you,
Malachi
Re: Infant data(less then 1 yr) from Children data [message #2024 is a reply to message #1895] Tue, 15 April 2014 11:29 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3016
Registered: February 2013
Senior Member
Following is Shea Rutstein's response to your post:

MIDX is the index to the maternal health section. As the standard recode is structured, the child of the section with MIDX(1) is the most recent child born in the five years preceding the survey. BIDX is the index to the birth history section. BIDX(1) refers to the most recent child of the mother. However, this child could have been born prior to the five years preceding the survey. Therefore the number of children is not the same since many mothers will have their youngest child whose age is five years or higher (or a dead child born five or more years ago). Thus the count of children with BIDX(1)=1 can be greater than that of children with MIDX(1)=1. Matching the sections needs to be done so that there is a one-to-one correspondence.
Re: Infant data(less then 1 yr) from Children data [message #2025 is a reply to message #2024] Tue, 15 April 2014 14:51 Go to previous messageGo to next message
user-rhs is currently offline  user-rhs
Messages: 132
Registered: December 2013
Senior Member
^^^ Thank you for clarifying the difference between BIDX and MIDX. I have always wondered why there were two separate sets of variables.
Re: Infant data(less then 1 yr) from Children data [message #2076 is a reply to message #2024] Fri, 25 April 2014 06:38 Go to previous messageGo to next message
Malachi Arunda is currently offline  Malachi Arunda
Messages: 30
Registered: February 2014
Member
A thank you is really not enough, but let me just say it..THANK YOU, really grateful. I'm progressing with analysis
Re: Infant data(less then 1 yr) from Children data [message #2078 is a reply to message #2076] Fri, 25 April 2014 07:20 Go to previous messageGo to next message
Malachi Arunda is currently offline  Malachi Arunda
Messages: 30
Registered: February 2014
Member
Dear Shea/Bridgette
On a closer look at the data informatio I found out this information below on Individual Recode Documentation that shows that 6079 were children who were born less than 60 months before the survey. Please clarify. Then could the MIDX(1) mean the 1st position of the child among the children and MIDX(6) also mean 6th position in the family (Although they are all Lastborns). Thank you
 
Kenya (DHS-V) Version 1 Doc 01 KE5
 
Name of Survey Kenya Demographic and Health Survey 2008-2009
 
Executing Agency Kenya National Bureau of Statistics
Year of fieldwork 2008/2009
 
Universe All women 15-49
Coverage National
 
Size Households: 9057
Women: 8444
Children: 6079 ( <60 months)
Births: 22534 ( all children)
Household Members: 38515
Men: 3465
Couples: 1431
Re: Infant data(less then 1 yr) from Children data [message #2079 is a reply to message #2078] Fri, 25 April 2014 07:21 Go to previous messageGo to next message
Malachi Arunda is currently offline  Malachi Arunda
Messages: 30
Registered: February 2014
Member
MIDX(1) Gives 4082
Re: Infant data(less then 1 yr) from Children data [message #2104 is a reply to message #2079] Mon, 28 April 2014 18:12 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3016
Registered: February 2013
Senior Member
Please see comments below from Shea Rutstein:

MIDX is the index to the maternal health record (REC41), of which there is one for each birth that a woman had in the 5 years preceding the survey. If a woman had 3 births in the 5 year period, then there would be three records for her MIDX(1), MIDX(2) and MIDX(3), in reverse birth order (last birth, next to last, next to next to last). However, to match with the last birth information from the birth history (REC21), one should use the association between the two sets of records. For example the matched birth date (b3) would be b3(midx(1)). Another way to match would be to use records where BIDX = MIDX. For only last births, select the data file do that there are only cases where MIDX$01=1.

Re: Infant data(less then 1 yr) from Children data [message #2117 is a reply to message #2104] Sat, 03 May 2014 12:08 Go to previous messageGo to next message
Malachi Arunda is currently offline  Malachi Arunda
Messages: 30
Registered: February 2014
Member
Thanks alot, let me keep forward and see
Re: Infant data(less then 1 yr) from Children data [message #2118 is a reply to message #2104] Sat, 03 May 2014 12:32 Go to previous message
Malachi Arunda is currently offline  Malachi Arunda
Messages: 30
Registered: February 2014
Member
Please clarify this too, does it mean that in case of MIDX(5) & MIDX(6) some children were twins or triplets. Thank you
Previous Topic: Multilevel modelling (ML)
Next Topic: Combined men/women/hiv dataset
Goto Forum:
  


Current Time: Thu Mar 28 04:10:53 Coordinated Universal Time 2024