Menu ENN Search
Language: English Fran├žais

How to manage don't know and missed information during analysis

This question was posted the Assessment and Surveillance forum area and has 3 replies.

» Post a reply

Anonymous 22505

Public Health Nutritionist

Normal user

14 Nov 2022, 10:06

Is there any agreed approach to handling "don't know" and "missed/blank"? This is an issue when we want to dichotomous the variable into two groups. So, do we need to consider them as "No"? Or do we consider "don't know" as missed and then exclude from denominator? I can understand the issue when these are too many. The big issue is when analysis of indicators by combining multiple responses like for MDD, MMF or MAD in IYCF. For example, when there is "don't know" or "missed" information in one of the eight food groups, can we analyse MDD, MMF or MAD by considering them as "NO"? if not treated as "NO", what will happen let's say BF, cereal intake, eggs intake is missed or "don't know?" we consider them as "Missed", how do we exclude them in the denominator given mixture response?  

Of course in the current WHO/UNICEF indicator for assessing IYCF, the following statement is there "Missing information includes: 1) “don’t know” responses; 2) questions accidentally left blank; and 3) responses with inconsistent or illogical codes owing to recording or data entry errors". 

Bradley A. Woodruff


Technical expert

25 Nov 2022, 01:04

In box 3 on page 32, the new IYCF recommendations published by WHO and UNICEF in 2021 specifically address how to treat unknown or missing values. Although the language is somewhat complicated, the basic recommendation is to assume the worst. That is, if current breastfeeding information is missing, assume the child is not breast-fed. When calculating any of the indicators which use consumption of specific food groups, any missing value for a specific food group is assumed to be “No” (code 0).

Of course, prevention is the key. Survey managers should do everything possible to ensure the absolute minimum number of missing or “Don’t know” responses. Interviewers should be thoroughly trained in interview technique and completion of the questionnaire so that questions are not skipped and responses are accurately recorded. Interviewers should carefully probe if respondents who initially report “Don’t know”. One of the best ways to minimize the number of missing responses is to use electronic data collection and a program which forces interviewers to record a response to all questions which are not skipped for legitimate reasons, such as a built-in skip pattern.

Milka Khayendi

Nutritionist/Ministry Of Health

Normal user

25 Nov 2022, 06:34

Managing i dont know and missed information can be very challenging but not an impossible situation.

However this can be managed through the following:

1.Taking the family history with keen interest on health seeking behaviour ,how often the child is sick,how often do they attend the clinic and generally when do they visit the hospital.

2.Look into the CWC booklet or Child welfare clinic booklet,it can give a rough idea on the health seeking behaviour and health status ,it will show what the child has been give in terms of immunization,supllimentation, deworming, treatment and  all matters health.

3.Ecomomic Assesment,can easily tell you what they are able to access in terms of health seeking , economic status can tell us where they go ,what they do and where you can get the records of the health status.

4.Follow up records at the facility that they visit frequently will help answer the I don't knows.

Jay Berkley


Frequent user

25 Nov 2022, 10:19

I agree with Bradley's comments and would add that it is important to think about what 'don't know' means for each variable. In some cases, we find that 'don't know' is an important category to include in analyses.

For example for birth weight, 'don't know' means they may be a home birth or less engagement or the child is older with health providers which means this group has  some different charactersitics and not including them introduces bias. Interestingly, for HIV , we find that 'refused' or 'not tested' children have higher mortality because this group includes children whose mother knows she is or may be positive but doesn't want this exposed.

If you have any problem posting a response, please contact the moderator at

Back to top

» Post a reply