The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Mortality » Birth selection
Birth selection [message #29154] Wed, 01 May 2024 10:31 Go to next message
Eve is currently offline  Eve
Messages: 2
Registered: April 2024
Member
Hello,

I am new to DHS and SPSS and I am seeking guidance for my analysis of Malawi's child mortality, focusing on infant and neonatal mortality, utilizing the MDHS 2015-2016 (BR dataset).

Currently, I am trying to isolate births five years before the survey period, from 2016-2012. However, I have encountered discrepancies between the sample I am obtaining and the figures presented in the final report table D.4. The filter I applied yields 13,584 births, whereas the report indicates 13,659 (both live and deceased)

Here is the filter condition I utilized in SPSS: Select if (B2 >= 2012 & B2 <= 2015).

Any insights or recommendations on how to refine my approach would be appreciated.

Thank you,

Eve
Re: Birth selection [message #29174 is a reply to message #29154] Fri, 03 May 2024 09:08 Go to previous messageGo to next message
Eve is currently offline  Eve
Messages: 2
Registered: April 2024
Member
Anyone, please help
Re: Birth selection [message #29177 is a reply to message #29154] Fri, 03 May 2024 15:01 Go to previous message
Janet-DHS is currently offline  Janet-DHS
Messages: 880
Registered: April 2022
Senior Member
Following is a response from DHS staff member, Tom Pullum:
You are not using the sampling weights. Below I will paste the lines (I use Stata, not SPSS) to get the weighted numbers of births from 2012 to 2016, inclusive.

The DHS calculation of under-5 mortality rates is complex. It is described in the Guide to DHS Statistics ( https://www.dhsprogram.com/Data/Guide-to-DHS-Statistics/inde x.cfm). It is based on a different definition of "the past 5 years", the window for the deaths, and it includes any exposure to risk from births that were before the window for the deaths. For example, a child who died as an infant in 2012 could have born in 2011.

Windows of time for DHS rates, such as the past 3 years (for fertility), the past 5 years (for under-5 mortality), the past 7 years (for adult mortality) are based on the calendar month of interview. The month of interview is not included, because it is incomplete. For example, if a woman is interviewed in May of 2024 (where we are now), a five-year window would end with April of 2024, and would include exposure and outcomes in that month. It would begin with May of 2019. There are 60 full months between May 2019 and April 2024, inclusive. The window is not fixed, but is different for women who were interviewed in different calendar months.

For your analysis of neonatal and infant mortality, there is no law saying that you have to calculate the rates exactly the way that DHS does, but you will only get an exact match if you use the programs available on our GitHub site (https://github.com/DHSProgram). The under-5 rates are in generic Chapter 8, and we provide SPSS versions as well as Stata and R. (There are equivalent non-DHS programs; google "syncmrates".) Rates are calculated for aggregates, and if you want an individual-level version of neonatal and infant mortality you just have to define a cohort of births that has had full exposure to the risk of dying within the age interval, or use some other method to adjust for censoring (incomplete exposure). I encourage you to proceed with your analysis.
. tab b2 if b2>=2012 & b2<=2016 [iweight=v005/1000000]

 

    year of |

      birth |      Freq.     Percent        Cum.

------------+-----------------------------------

       2012 |3,602.90957       26.38       26.38

       2013 | 3,435.8864       25.15       51.53

       2014 | 3,428.7698       25.10       76.63

       2015 | 3,157.2184       23.11       99.75

       2016 |  34.642183        0.25      100.00

------------+-----------------------------------

      Total | 13,659.426      100.00

[Updated on: Fri, 03 May 2024 15:02]

Report message to a moderator

Previous Topic: missing results
Next Topic: Working through Odds Ratios
Goto Forum:
  


Current Time: Wed Nov 13 05:19:31 Coordinated Universal Time 2024