1

Emergency Infant Feeding Emergency Infant Feeding SurveysSurveys

Assessing infant feeding as a Assessing infant feeding as a component of emergency nutrition component of emergency nutrition surveys: Feasibility studies from surveys: Feasibility studies from

Algeria, Bangladesh and EthiopiaAlgeria, Bangladesh and Ethiopia

Marko KeracMarko Kerac11, Marie McGrath, Marie McGrath33, Fathia Abdalla, Fathia Abdalla22, , Andrew SealAndrew Seal11

1 UCL Centre for International Health and 1 UCL Centre for International Health and Development London;Development London;

2 UNHCR Geneva; 3 ENN, Oxford2 UNHCR Geneva; 3 ENN, Oxford

Supported by ENN & Supported by ENN & Funded by UNICEF-led IASC Nutrition ClusterFunded by UNICEF-led IASC Nutrition Cluster

2

AimAim

Investigate feasibility & utility of Investigate feasibility & utility of including including

standard indicators of standard indicators of

infant feeding practice infant feeding practice

in routine nutrition surveys. in routine nutrition surveys.

3

ObjectivesObjectives

1)1) Describe the sample size Describe the sample size assumptions and calculations assumptions and calculations

required required

2)2) Assess the precision achieved Assess the precision achieved when measuring the indicators in when measuring the indicators in

4 emergency nutrition surveys4 emergency nutrition surveys

4

BackgroundBackground

Why good quality data is important:Why good quality data is important:- Correct response to vulnerable situationCorrect response to vulnerable situation

- Start programme when indicated (‘threshold’)Start programme when indicated (‘threshold’)- No programme when there is no need for one No programme when there is no need for one

(efficient use of resources)(efficient use of resources)

- Assessing programme impactAssessing programme impact- Correct baselinesCorrect baselines- True impact vs artefact (poor validity; poor True impact vs artefact (poor validity; poor

precision)precision)

- Assessing trendsAssessing trends- True differences vs artefactTrue differences vs artefact

5

MethodsMethods

Study designStudy design~ Descriptive~ Descriptive

~ Summary of key methodological features & ~ Summary of key methodological features & results of: results of: 4 recent emergency nutrition 4 recent emergency nutrition surveys.surveys.

~Selected purposively~Selected purposively data on infant feeding (0 to 5.9m & 6 to 24 data on infant feeding (0 to 5.9m & 6 to 24

months)months)

A.Seal, CIHD/ICH ~ lead investigator on all A.Seal, CIHD/ICH ~ lead investigator on all surveyssurveys

6

MethodsMethods

SettingsSettings Refugee populations in Refugee populations in

Algeria Algeria BangladeshBangladesh

Resident populations in Resident populations in Ethiopia (highland)Ethiopia (highland) Ethiopia (lowland) Ethiopia (lowland)

Sampling (within each survey)Sampling (within each survey) ‘‘Traditional’ 2 stage, 30x30 cluster Traditional’ 2 stage, 30x30 cluster

design. design.

7

MethodsMethods

ParticipantsParticipants Children aged 6 to 59.9 months Children aged 6 to 59.9 months

== main population of interest in main population of interest in most nutrition surveys, including the most nutrition surveys, including the four described. four described.

Young infants aged 0 to 5.9 monthsYoung infants aged 0 to 5.9 months

= = additional toadditional to the above the above

8

MethodsMethods

MeasurementsMeasurements 3-4 day team training ( = standard)3-4 day team training ( = standard) Included anthropometry, morbidity questions Included anthropometry, morbidity questions

and 24 hour recall food frequency and 24 hour recall food frequency questionnairequestionnaire

ESTABLISHED / CONSISTENT / VALIDATED ESTABLISHED / CONSISTENT / VALIDATED

(Mary Lung’aho et al – previous presentation)(Mary Lung’aho et al – previous presentation) currentcurrent feeding practices (all infants, ages 0 to feeding practices (all infants, ages 0 to

23.9m) 23.9m) Focus groups / key informants for inclusion of Focus groups / key informants for inclusion of

specific local food itemsspecific local food items Questionnaires were translated into local languages Questionnaires were translated into local languages

and piloted prior to the start of each survey.and piloted prior to the start of each survey.

9

MethodsMethods

Sample size (1)Sample size (1) Emergency nutrition cluster surveys, where Emergency nutrition cluster surveys, where

prevalence data limited, prevalence data limited, 900 children aged 900 children aged 6-59 m6-59 m

To calculate the number of infants required:To calculate the number of infants required:1) likely prevalence, 1) likely prevalence, 2) required precision, 2) required precision, 3) anticipated ‘design effect’ (=loss of power in a 3) anticipated ‘design effect’ (=loss of power in a cluster sampling method vs simple random sample)cluster sampling method vs simple random sample)

~ routine to assume 2 for standard ~ routine to assume 2 for standard anthro indicators (cases localized, not random) anthro indicators (cases localized, not random) x2 sample sizex2 sample size

~ we assumed infant feeding practices ~ we assumed infant feeding practices not localised not localised design effect=1 design effect=1 no sample no sample size increasesize increase

10

MethodsMethods

Sample size (2)Sample size (2) To determine prevalence of EBF (0-5.9m):To determine prevalence of EBF (0-5.9m):

30% prevalence 30% prevalence assumedassumed

based on based on globalglobal statistics, [ref: statistics, [ref: UNICEF Statistics UNICEF Statistics http://www.childinfo.org/eddb/brfeed/index.htm] http://www.childinfo.org/eddb/brfeed/index.htm]

Design effect = 1Design effect = 1 desired precision of +/- 15 %, desired precision of +/- 15 %,

adequate for a baseline needs adequate for a baseline needs assessment assessment

*** sample size = 36 infants ****** sample size = 36 infants ***

11

MethodsMethods

Sample size (3)Sample size (3) To determine prevalence of continued BF at 12 To determine prevalence of continued BF at 12

and 24 months:and 24 months: 60 % prevalence assumed, 60 % prevalence assumed,

also based on available global estimates, and a also based on available global estimates, and a precision of +/- 20 %. precision of +/- 20 %.

sample size:sample size:

*** 24 children aged 12 to 15.9 months ****** 24 children aged 12 to 15.9 months ***

*** 24 children age 20 to 22.9 months ****** 24 children age 20 to 22.9 months ***

Population pyramid Population pyramid ? recruit from the 900 ? recruit from the 900 ‘core’ survey‘core’ survey

12

MethodsMethods

Statistical methods ~ for individual Statistical methods ~ for individual surveyssurveys

Data entry, validation, cleaning Data entry, validation, cleaning EpiInfo EpiInfo v.6.04dv.6.04d

Separate files for:Separate files for:

0-5.9 month 0-5.9 month

& &

6-59.9 months6-59.9 months

Analysis Analysis EpiInfo v.6.04d and SPSSv11 EpiInfo v.6.04d and SPSSv11

13

MethodsMethods

Statistical methods ~ key to this Statistical methods ~ key to this paper….paper….

For each indicator…For each indicator…

in each survey… in each survey…

we retrospectively we retrospectively calculated:calculated:

Design effectDesign effect Standard errorStandard error Actual precision achievedActual precision achieved

14

ResultsResultsSurvey site Algeria

Saharawi Refugee Camps, Tindouf

Bangladesh Myanmar Refugees Camps,

Cox’s Bazar

Ethiopia Aroressa Woreda,

Sidama Zone *

Ethiopia Aroressa Woreda,

Sidama Zone **

Date of survey 12th - 22nd Sept. 2002

18th - 24th Aug. 2003

12th - 25th Mar. 2004

12th - 25th Mar. 2004

Total Population 154,670 19,804 84,655 40,675 Survey sample***

Infants (0-5 m) 92 98 52 46 Children (6-59 m) 907 923 918 921

Sample ratio 0-5 m:6-59 m 1:10 1:9 1:18 1:20

Survey measures Infants (0-5 m) FFQ, WH, HA FFQ FFQ FFQ Children (6-23 m) FFQ, WH, HA FFQ, WH, HA FFQ, WH, HA FFQ, WH, HA Children (24-59 m) WH, HA WH, HA WH, HA WH, HA

15

Results (t.b.c…)Results (t.b.c…)

etc… for 10 indicatorsetc… for 10 indicators

Indicator*

Age group analysed

Algeria Bangladesh Ethiopia

(highland areas) Ethiopia

(lowland areas)

Ever breastfed

0-23 m

n=371 93.3%

(95.3, 99.4) DE=tbc SE=tbc RP=tbc

n=360 97.8%

n=261 95.0%

(92.1-97.9)

n=238 99.6

(98.8-100.0)

Timely initiation of breastfeeding

0-23 m n=369 12.7%

(8.8-16.7)

n=357 42.3%

n=260 93.1%

(88.5-97.6)

n=235 91.1%

(85.1-97.1)

Exclusive breastfeeding

<6 m n=87 2.3%

(0.0-6.8)

n=92 53.3%

n=52 71.5%

(58.2-84.1)

n=46 47.8%

(32.6-63.0) Continued BF at 12 m

12-15 m n=82 84.1%

(75.3-93.0)

n=67 95.5%

n=64 95.3%

(90.7-99.9)

n=57 96.5%

(92.0-100.0)

16

DiscussionDiscussion

Key result and interpretationKey result and interpretation Successful inclusion of infant feeding Successful inclusion of infant feeding

indicators into a standard nutrition indicators into a standard nutrition survey is feasible and achievable. survey is feasible and achievable.

Diverse physical and social settings:Diverse physical and social settings:

refugee camps ~~> resident populationsrefugee camps ~~> resident populations

Sahara desert ~~> Ethiopian highlands.Sahara desert ~~> Ethiopian highlands.

17

DiscussionDiscussion

Mortality & morbidity consequencesMortality & morbidity consequences n=4 surveys too small to reliably n=4 surveys too small to reliably

interpret the mortality and morbidity interpret the mortality and morbidity implicationsimplications

BUT notable thatBUT notable that All 4 sites far short of ideal infant All 4 sites far short of ideal infant

feeding practice feeding practice e.g.e.g. EBF as low as 2% in AlgeriaEBF as low as 2% in Algeria Best EBF, in the Ethiopian highlands only 71.5% Best EBF, in the Ethiopian highlands only 71.5%

potential for harm (6-59.9m MAM/SAM high)potential for harm (6-59.9m MAM/SAM high)

need for interventionsneed for interventions

18

DiscussionDiscussion

Including IF indicators important Including IF indicators important because:because:

Better planning Better planning Identify & address potential negative effects Identify & address potential negative effects

of emergency interventionsof emergency interventions e.g. effects of code violationse.g. effects of code violations

Increased awareness of infant feeding issues Increased awareness of infant feeding issues in communities surveyedin communities surveyed

( In principle ), problems can be addressed ( In principle ), problems can be addressed proximally, proximally, beforebefore MAM/SAM evolves MAM/SAM evolves

19

DiscussionDiscussion

Other issues Other issues (work in progress)(work in progress)

Anthropometry in 0 – 5.9mAnthropometry in 0 – 5.9m Difficult in this age! (e.g Difficult in this age! (e.g

scales)scales) Only 1 of 4 surveys Only 1 of 4 surveys

measured young infant measured young infant anthropometryanthropometry

InterpretationInterpretation NCHS vs WHO standardsNCHS vs WHO standards

Binns C, Lee M. Will the new WHO growthreferences do more harm than good? Lancet

2006; 368: 1868–69 (figure)

20

DiscussionDiscussion

Other issues Other issues (future work)(future work)

Survey methodologySurvey methodology

LQAS vs 30x30LQAS vs 30x30

21

ConclusionsConclusions Our preliminary results suggest that Our preliminary results suggest that

inclusion of already available, validated inclusion of already available, validated questions about infant feeding practice questions about infant feeding practice is feasible and achievable is feasible and achievable

These may be integrated within current These may be integrated within current emergency nutrition survey designsemergency nutrition survey designs

We suggest that there are strong We suggest that there are strong arguments for routine inclusionarguments for routine inclusion

However, we acknowledge that all data However, we acknowledge that all data collection and analysis has a cost collection and analysis has a cost

Any data collection should only take Any data collection should only take place in an emergency context when it place in an emergency context when it will be used to inform decision making.  will be used to inform decision making.  

22Thank YouThank You