| South Africa 2016 - HIV prevalence [message #16871] |
Mon, 11 March 2019 18:11  |
Suzanne Ryan-Ibarra
Messages: 2
Registered: December 2018
|
Member |
|
|
Hello,
I'm calculating HIV prevalence among women ages 15-49 years, using the variable HIV03 in the South Africa 2016 HIV dataset called ZAAR17FL.sas7bdat, which I merged with the individual recode file called ZAIR71FL.sas7bdat.
My code is below:
proc surveyfreq data=ZA_hiv;
table hiv03/row cl;
cluster v021;
stratum v022;
format hiv03 hiv.;
title 'HIV Prevalence - South Africa DHS 2016';
weight hiv_wgt;
run;
In the SAS output (attached), the sum of weights is 2485, which matches the number in Table 13.3 in the report FR337 available here: https://dhsprogram.com/pubs/pdf/FR337/FR337.pdf. However, I am concerned that the prevalence estimates calculated using my code above for women 15-49 who are HIV positive are 27.7%, and these do not match exactly the prevalence reported in the FR337 report, which is 27.3%.
Can anyone please advise on next steps to troubleshoot?
Thanks!
|
|
|
|
| Re: South Africa 2016 - HIV prevalence [message #17083 is a reply to message #16871] |
Fri, 15 March 2019 11:33  |
Bridgette-DHS
Messages: 3203
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS Specialist, Joy Fishel:
You are treating the inconclusive code "9" as missing and excluding it from the denominator. The HIV prevalence should include inconclusive cases in the denominator.
This denominator represents every case that completed the HIV testing algorithm with a final result. Indeterminate and inconclusive, according to the lab folks, are valid final results that are not HIV positive, and they should be treated the same as a negative result for the purpose of prevalence calculation.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
