The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Mortality » survival analysis
survival analysis [message #14526] Tue, 17 April 2018 07:39 Go to next message
fngendahimana55@gmail.com is currently offline  fngendahimana55@gmail.com
Messages: 1
Registered: April 2018
Location: Kigali
Member
Hello members
is there someone who can help me out
I want to conduct a survival analysis of under-five children using a retrospective study
but I don't know how to select an appropriate birth cohort because I have 2 different DHS data file (BR, and KR)
I'm not aware also of how I will graph survival graphs (kaplan or hazard)

Thanks


Frederic
Re: survival analysis [message #14773 is a reply to message #14526] Fri, 04 May 2018 12:38 Go to previous message
Liz-DHS
Messages: 1516
Registered: February 2013
Senior Member
A response from Dr. Shea Rutstein,
Quote:

The BR file of births is the one to use and includes living and dead children. The KR file only has living children.
Kaplan-Meier survival analysis gives a survival curve taking into account right-censoring (i.e. loss to follow-up). We have that if, for example, we want under-five mortality rates for the most recent five-year period since not all children will have been 5 years of age or older at the time of the survey: their experience is truncated by the interview and we don't know what will happen to them at ages less than 5 years in the future. Note that the experience of the older children under age 5 is being used to estimate mortality at the older ages. They will have been born earlier but died recently. There may be a problem with sample size for these estimates. For the mortality at younger ages, say under age 1 year, the experience of most of the children born in the last five years is used.

Kaplan-Meier is usually used to compare two or three groups (e.g. control versus treatment, urban vs. rural, male vs. female).

The Cox proportional hazard regression procedure estimates the hazard rate of a group with a selected characteristic compared with a baseline and is generally used with multiple correlated variables to find the effect of each.


Previous Topic: Mortality history before birth of index child
Next Topic: stata codes for infant mortality
Goto Forum:
  


Current Time: Thu Mar 28 04:42:30 Coordinated Universal Time 2024