The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Mortality » U5M and IMR
U5M and IMR Mon, 02 May 2016 09:44
 shu2013 Messages: 5Registered: June 2013 Member
I am using the following code to examine U5M and IMR and specifically to compare urban/rural mortality and mortality within urban settings (by wealth); The trouble is, that the results are somewhat contrary to expectations... rural mortality is lower than overall urban mortality, and within the city, there is a non-linea relationship, whereby the thrid quintile has the highest mortality of the 5 wealth groups. This has led me to wonder if perhaps I am not estimating U5M and IMR correctly. For now, I have just used the Dominican Republic 2013 file, but I plan to expand the analysis to other LatAm countries. Can anyone comment? I have checked out other posts on this subject and the DHS statistics guide and I am still unable to see if/what I am doing wrong. Many thanks.

****Calculating the IMR and CMR
*V008: date of interview (CMC)
*B3: date of birth (CMC)
*B7: age at death (month imputed)
*B5: whether the child is still alive

gen hypage=(v008-b3)/12
gen timeyears=.
replace timeyears=hypage
replace timeyears=b7/12 if b5==0

sort urbano
Re: U5M and IMR [message #9708 is a reply to message #9668] Mon, 09 May 2016 12:01
 Trevor-DHS Messages: 744Registered: January 2013 Senior Member
Please see the Guide to DHS Statistics, pages 90-94, and pay particular attention to the Calculation section. We use a synthetic cohort life table approach as described in point 3 on page 91, and we estimate mortality for a five year period preceding the survey (see the diagram on page 93) where tu would be the month before interview and tl would be 5 years before interview.

In your code you are not limiting the time period at all. You need to restrict the time period to the children exposed to the risk of mortality in the last 5 or 10 years. That isn't just limiting to children born in the last 5 years, but includes child born more than 5 or 10 years ago who contributed some exposure in the last 5 or 10 years (late entries in the life table).

Also note that we don't use a classic life table approach, and your results, even when limited to exposure in the last 5 or 10 years, will differe slightly from our approach.
Re: U5M and IMR [message #9711 is a reply to message #9708] Mon, 09 May 2016 13:40
 shu2013 Messages: 5Registered: June 2013 Member
Thank you. This would address the problem?

gen hypage=(v008-b3)/12
gen timeyears=.
replace timeyears=hypage
replace timeyears=b7/12 if b5==0
ltable timeyears dead if hypageyrs <=10, int(.5) gr
Re: U5M and IMR [message #9712 is a reply to message #9711] Mon, 09 May 2016 13:43
 Trevor-DHS Messages: 744Registered: January 2013 Senior Member
No - this just limits the analysis to births that took place in the last 10 years. It does not address the late entries. Please read the Guide to DHS Statistics.
Re: U5M and IMR [message #10911 is a reply to message #9712] Tue, 04 October 2016 23:40
 Olutosin Messages: 5Registered: October 2016 Location: Nigeria Member
Expaciate more sir because it relates to the statistical model I want to use for my project survival analysis wilbul proportional hazard cox regression

Olutosin

[Updated on: Tue, 04 October 2016 23:41]

Report message to a moderator

Re: U5M and IMR [message #10916 is a reply to message #10911] Wed, 05 October 2016 13:53
 Trevor-DHS Messages: 744Registered: January 2013 Senior Member
Sorry, I do not understand your question.
 Previous Topic: Adult Mortality based on Orphanhood Next Topic: Under 5 Mortality
Goto Forum:

Current Time: Tue Apr 13 22:07:54 Coordinated Universal Time 2021