Transcript Theoretical ratio of global, moderate and severe malnutrition

Analysis of surveys
• Effect of measurement error
• Shape of distribution as population gets
more malnourished.
• Possibility of development of new methods
• Age/ height profile during famine
• The problem of oedema
• Implications for relief programs
Effect of measurement errors on survey results
Suppose an imprecise error moves
a value from one segment to
another (up or down)
If the errors are random then the
same number of values will
move from one half of the
distribution to the other (orange
to green and green to orange).
There will be no change in the
mean of the distribution
provided that there are as many
positive as negative
measurement errors
Effect of measurement errors on survey results
•
If there is an error that moves
a value from one segment to the
other in the tail then there will
be more points moving from the
orange to the green than from
the green to the orange in
relation to the respective areas
•
There will be an increase in
the Standard Deviation.
•
There will be an increased
prevalence of values below –2Z
and also below –3Z
Effect of measurement errors on survey results
• A change in the standard
deviation from 1.0 to 1.2 will
have a major effect upon the
prevalence of moderate and
severe malnutrition.
• Imprecise measurements are
potentially a major cause of
error in surveys.
• The prevalence will be
exaggerated even if the positive
and negative errors balance
each other out.
Effect of SD of survey on
prevalence of wasting
Norm al distributions, Mean at -1.0 Zscore
Global
20
Severe
% malnutrition reported by survey
• Wide SD, from
measurement error can
increase prevalence
dramatically
• Narrow SD from
“over cleaning”,
selection or bias can
reduce the prevalence
of malnutrition
15
10
5
0
0.8
0.9
1
Population SD
1.1
1.2
Effect of measurement errors on survey results
MonteCarlo Simulation
1 Take a normally distributed population with mean Z-score of minus 1 Z-score
WFH and Standard deviation of 1 Z-score unit.
2 Introduce an error height with a mean of 0.0cm and an SD of 1.0cm.
3 Introduce an error in weight with a mean of 0g and an SD of 100g
4 Introduce a 5 cm height error and 500g weight error to 0.25% of population
5 Introduce a 10cm height error and 1kg weight error to 0.15 of population
•
•
Introduce a random error in height of 0.5cm to account for diurnal variation
Introduce a random error in weight of up to 200g (mean error 0, SD error
200g) to allow for meal, fluid/excreta, and diurnal changes.
5 Remove all flags (plus or minus more than 4SD from population mean)
Effect of measurement errors on survey results
Treatment
none
height
weight
Height + weight
all errors
%global %severe
15.8%
2.2%
17.1%
2.6%
16.6%
2.8%
18.8%
3.0%
21.9%
6.1%
SD of surveys does not change as the
population becomes more malnourished
1.3
100.0
90.0
80.0
Cumulative distribution (% )
Survey standard deviation (z Score)
1.2
1.1
1.0
0.9
70.0
60.0
50.0
40.0
30.0
20.0
0.8
10.0
0.7
-2.5
0.0
-2.0
-1.5
-1.0
-0.5
0.0
0.5
Survey Mean Weight for Height (Z score)
1.0
0.7
0.8
0.9
1.0
1.1
Survey standard deviation
1.2
1.3
SD does not change as the population
becomes more malnourished
• All individuals in the population are affected by
the situation to a similar extent
• The whole distribution moves downwards
• As this happens more and more are “recruited to
the ranks of the malnourished”
• Movement in the mean WFH (and prevalence of
malnutrition) is not necessarily a “trailing
indicator” in population terms.
The Kurtosis of the distribution
becomes more, not less, normal
• p<0.001.
2 .0
Ku r t o s is = 0 .3 3 w h z + 0 .5
r = 0 .4 1
1 .5
1 .0
Sur ve y Kur tos is
• Kurtosis of weight-forheight distributions in
surveys of children
aged 6-59 months as
the population becomes
progressively more
wasted.
0 .5
0 .0
- 2 .5
- 2 .0
- 1 .5
- 1 .0
- 0 .5
0 .0
- 0 .5
- 1 .0
S u r v e y M e a n W HZ
0 .5
1 .0
There is slight Skewness at with
a very thin of fat population
S k e w = 0 .1 7 w h z 2 + 0 .1 5 w h z + 0 .0 8 , r = 0 .3 3
0 .5
Sur ve y Sk e w ne s s
• Skewness in weightfor-height
distributions in 228
surveys of children
aged 6-59 months as
the population
becomes
progressively more
wasted.
1 .0
0 .0
- 3 .0
- 2 .0
- 1 .0
0 .0
• r=0.33 p < 0.001
- 0 .5
S u r v e y M e a n W HZ
1 .0
The population remains normally
distributed as wasting increases
3.0
2.0
1.0
Expected value (SND)
• Normal probability
plots of weight-forheight z-scores of
individual subjects
from illustrative
surveys.
• The skewness of the
reference population
may be because of
inclusion of obese
individuals in the
NCHS population.
0.0
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
-1.0
-2.0
-3.0
Observed weight-for-height (Z)
Mozambique
kenya
Ethiopia
Rwanda
Uganda
Nicaragua
Bosnia
Croatia
4.0
Deviation from normality of the
distribution curves
100
80
Cumulative surveys (%)
• The maximum deviation of
individual data points from a
normal (Gaussian)
distribution, determined by the
Kolmorgorov-Smirnov
procedure. SND = Standard
Normal Deviate
• The mathematical calculation
of the prevalence of
malnutrition gives the same
values as with counting the
number observed.
• Smaller numbers are then
needed to give an estimate
with the same confidence
interval
60
40
20
0
0.000
0.025
0.050
0.075
KS- Maximum Difference (SND)
0.100
The relative importance of moderate (SFC) and
severe (TFC) malnutrition varies with the level of
malnutrition in the population
50
45
% m o d e r a te
% s e ve r e
40
% M o d e r a te + s e ve r e
m o d e r a te :s e ve r e r a ti o
35
30
25
20
15
10
5
0
- 2 .5
-2
- 1 .5
-1
- 0 .5
P o p u l a ti o n m e a n Z-S c o r e
0
0 .5
1
The actual Observed and theoretical prevalence of
malnutrition are related within the confidence intervals of
the survey – maybe calculation would be more precise?
50
45
40
% m o d e r a te
35
% s e ve r e
30
25
20
15
10
5
0
-2 . 5
-2
-1 . 5
-1
-0 . 5
P o p u la tio n m e a n Z-S c o re
0
0.5
1
Wasting by height group as the population nutritional state deteriorates:
All groups are affected – as the situation becomes desperate older children have a
high prevalence
The problem of oedema
• 82% of cases in surveys are not wasted
• This is so even when an adjustment is made
for the weight of the oedema
• A population without excess global
malnutrition can develop odematous
malnutrition rapidly
• SFP will not prevent oedematous
malnutrition in the population because
moderate wasting is not a forerunner.
The problem of oedema 2
• The time course of oedema is short -- days
whereas that of marasmus is weeks or
months. Even a low prevalence of oedema
indicates a reasonably high incidence.
• Surveys can not be used to estimate the
number of patients that need treatment or
the “coverage” of programs.
• The implications of oedema and severe
wasting are different and should be reported
seperately.
Anthropometric surveys give prevalence:
incidence is needed to plan curative services
50 red cases of Kwashiorkor and 20 cases of wasting
0
1
2
3
4
5
6
The problem of oedema 3
• The counts of children with wasting in each
cluster follows a Poisson distribution in a
well taken survey.
• The counts of children with oedema in each
cluster follows a negative-binomial
distribution. That is there are small
“pockets” of oedema with many cases in a
few clusters and no cases in most clusters
• The estimate of oedema prevalence is much
less certain than that of wasting.
Oedematous malnutrition does not have any
relationship to the prevalence of wasting
50
45
" % s e v e r e w a s t in g "
40
" % m o d e a r t e w a s t in g "
% oedema
35
30
25
20
15
10
5
0
- 2 .5
-2
- 1 .5
-1
- 0 .5
P o p u l a ti o n m e a n Z-S c o r e
0
0 .5
1
Conclusions
• Surveys give a measure of the severity of a crisis
• They require careful planning, experienced
supervisors, good training and quality assurance
• The internal structure of the data can be used to
examine how reliable an anthropometric survey is.
There are sufficient “good” surveys to set limits
on the deviation that is acceptable
• New windows-based, user-friendly software
should be developed specifically to analyse
anthropometric surveys. It should indicate if one
team is giving aberrant results in time to correct
the defects.
Conclusions 2
• Who becomes malnourished may relate
more closely to location than other
vulnerabilty indicators
• unreliable to judge the numbers of
beneficiaries who will require relief
(incidence-vs-prevalence)
• should be combined with surveillance data