| Still births-infant mortality [message #22145] |
Sat, 06 February 2021 16:04  |
A.Z_Epi
Messages: 2
Registered: February 2021
|
Member |
|
|
I am doing an analysis on DHS for the following question.
"maternal education associated with child mortality" (stillbirths, infant). I have restricted the analysis to the births in the last 5 years preceding the survey.
1)Initially, I am just trying to calculate the numbers for infant and still births (to calculate power).
For example for the number of children who died in the neonatal period, I am using the command
tab b7 exposure v if b19<60 where I have recoded b7 into 0/1=neonatal 3/12 post-neonatal number of deaths. (What can I do for the older DHS where b7 variable is not present).
2)I wish to know the code I need to calculate the number of stillbirths.
3)Later if plan to calculate odds for this association in logistic regression. Do I need to create an event file (tutorial 5 of contraceptive calendar) ?
I am not an expert Stata user. I have already tried the tread message #15076] in this forum but it did not work.
I hope you can guide me in the right direction
Thanks in advance.
[Updated on: Mon, 08 February 2021 13:17] Report message to a moderator |
|
|
|
| Re: Still births-infant mortality [message #22170 is a reply to message #22145] |
Tue, 09 February 2021 07:57  |
Bridgette-DHS
Messages: 3218
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
Neonatal deaths are those with b7=0. Post-neonatal have 1<=b7<=11. Months of age at death are completed months, so in the first month a child's age is 0 months. Deaths under 5 years, i.e. before the 5th birthday, have b7<=59.
Stata code for the child mortality chapter of the main report, including stillbirths and perinatal mortality, is in our new GitHub site. Go to www.dhsprogram.com, the Data tab, and click on Data Users Forum. The link to the GitHub site is in the lower left corner of that page. You should be able to do most of what you want by using the BR (or KR) file, in which births are cases. To get at stillbirths you will probably need to get into the calendar data. I suggest that you start by doing as much as you can without the calendar.
|
|
|
|
| Re: Still births-infant mortality [message #22202 is a reply to message #22145] |
Thu, 11 February 2021 22:10 |
A.Z_Epi
Messages: 2
Registered: February 2021
|
Member |
|
|
Thank you for your reply it has clarified the problems. I have questions for a stata code I have used to calculate stillbirths and it has worked but I am second guessing if it exactly applies.
I have calculated total stillbirths using the IR file.
use caseid v001 v002 v003 v005 v008 v011 v013 v017 v018 v021 v022 v023 v024 v025 v106 v190 v231 v242 b3*
> vcal_1 using"C:PKIR71FL.DTA", clear
. gen stillbirths = regexm(vcal_1,"TPPPPPP")
.
.
.
. gen early_neonatal_deaths = 0
.
. gen infant_deaths = 0
.
. gen child_deaths = 0
.
.
.
. * rename birth variables for use in forvalues loop
.
. rename b*_0* b*_*
.
.
.
. * Any early neonatal deaths, infant deaths, child deaths
.
. forvalues b = 1/20 {
2.
. * early neonatal - days 0-6
.
. replace early_neonatal_deaths = 1 if inrange(b6_`b',100,106)
3.
. * infant - from birth trough age 11 months
.
. replace infant_deaths = 1 if inrange(b6_`b',100,211)
4.
. * child from 12 months up through 4 completed years
.
. replace child_deaths = 1 if inrange(b6_`b',212,304)
5.
tab stillbirths = 261 (also after using the iweight command)
the table Pakistan Demographic and Health Survey 2017-18 shows a total of 249 stillbirths.
Can you please guide where I might be making an error?
|
|
|
|
| Re: Still births-infant mortality [message #22239 is a reply to message #22202] |
Wed, 17 February 2021 08:02 |
Bridgette-DHS
Messages: 3218
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
You are identifying stillbirths with a string TPPPPPP, that is, a termination preceded by 6 (or more) months of pregnancy. T identifies terminations other than a live birth, and includes miscarriages, abortions, and stillbirths. However, if T is preceded by 6+ months of pregnancy, it is almost sure to be a stillbirth. That would be the usual assumption.
If a T occurs near month 80 of the calendar, there may be fewer than 6+ months of pregnancy. In fact, you will find some T's right in month 80, with NO preceding P's. (This point is probably not serious because you are probably not using the full calendar for your reference period.)
You are probably looking for stillbirths within the past 3 years. This means that the T occurred in that interval, not that TPPPPPP occurred in that interval. You need to identify these strings and THEN restrict to those for which T is in the past 3 years.
Note that the number of events given in the report will be weighted by v005. If your number is unweighted, and the number in the report is weighted (and rounded to the nearest integer) there will be a difference.
Please let us know if you still cannot get a match, taking these things into account.
Did you check how the GitHub code identifies stillbirths as a component of perinatal deaths?
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Re: Still births-infant mortality [message #22376 is a reply to message #22367] |
Thu, 04 March 2021 07:54 |
Bridgette-DHS
Messages: 3218
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
In survival analysis, the time variable is the elapsed time between entering a state and either leaving that state or being censored by the date of observation. For completed pregnancies, we have the duration of the pregnancy in months, but the range of months is narrow. You can use the calendar to look at duration of pregnancy and outcome of the pregnancy--for example, a stillbirth, or a live birth followed by neonatal death, or a live birth without a death. By definition, stillbirths are non-live birth terminations of pregnancies of duration 7+ months. That is, the duration of the pregnancy is built in to the definition of the outcome. You have the duration of a current (censored) pregnancy, but we don't know the date of conception and early pregnancies are under-reported. For these reasons, I don't see a useful application of survival analysis to pregnancy terminations. But perhaps another user has suggestions. We do apply survival analysis to the calendars to analyze contraceptive discontinuation, failure, and switching.
|
|
|
|
|
|
| Re: Still births-infant mortality [message #22379 is a reply to message #22376] |
Thu, 04 March 2021 13:28 |
SamH79
Messages: 11
Registered: October 2019
|
Member |
|
|
Dear Tom,
Thank you for your response, this is helping me so much, I will consider changing my research question of survival analysis in perinatal mortality.
My further queries,
I will also investigate factors correlated with perinatal mortality in a simple way using logistic regression (Socio-demographic, maternal health, etc). You know this.
My questions:
Is there any possibility to do further analysis after the logistic regression, such as PATH ANALYSIS, to model the correlations between variables with the perinatal mortality?
Thank you for your help
Best Regards,
Sam
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Re: Still births-infant mortality [message #22646 is a reply to message #22638] |
Thu, 15 April 2021 16:48 |
Bridgette-DHS
Messages: 3218
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
I don't understand your question. The IR file is a file of interviewed women. The BR file has one record for each child ever born to these women. The perinatal mortality rate is calculated for children born recently, e.g. in the past 3 years. The numbers you list are supposed to be different from one another.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
