Advanced Nutrition Assessment and Monitoring_Final Paper

10

Click here to load reader

Transcript of Advanced Nutrition Assessment and Monitoring_Final Paper

Page 1: Advanced Nutrition Assessment and Monitoring_Final Paper

1

Final Report: Analysis of Underweight and Intervention Recommendations

Western Kenya and Nyanza

GCHB 7090

Megan Hall

Abstract

The western region of Kenya experiences high rates of underweight and stunted children; Table

1 shows the prevalence of underweight and stunting by district. Focusing on an aimed 5% rate of

underweight, this report examines the factors associated with underweight, specifically analyzing

the questions: are source of drinking water, toilet facility and roof type associated with

underweight. The dataset used in this analysis is extracted from DHS 1994 results. After cleaning

the dataset, independent variables—including source of drinking water, toilet facility, roof type

(as proxy for SES), place of delivery (as proxy for access to health services), education and

occupation were dichotomized and severity of underweight was compared between categories of

each variable. Linear regression of the independent variables revealed that source of drinking

water shows the most reliable significant effect on underweight, especially among high-educated

households. The effect of toilet facility on underweight is unreliable due to the small sample size

of households with improved (flush or ventilated) toilet facilities. The effect of SES on

underweight is confounded by source of drinking water, education and access to health services.

Agriculture and unskilled occupations are associated with underweight, however this effect is

accounted for by education, source of drinking water and SES. To reduce underweight among

children under five years old in western Kenya and Nyanza, nutrition interventions must target

households in South Nyanza—which show the greatest prevalence of underweight—and focus

on improving source of drinking water, taking into account the behavioral differences between

high- and low-educated households that lead to the differential effect of improved source of

drinking water on underweight severity among these households.

Introduction

In western Kenya and Nyanza, 22.1% of people reside in Kisii or Nyamira (n = 488); see Table 1

for distribution of households among other districts, as well as distribution of households among

dichotomized independent variables. Respondents in the DHS dataset range from 15 to 49 years

old, with the average age 28 years (n = 2,209) and 72.25% of households include a children

under five years old. The severity of underweight ranges from -5.29 to 3.72 z-score, with a mean

of -0.9524 (n = 1,023). Stunting ranges from -6.00 to 5.41 z-score, with a mean of -1.281(n =

1,023). Using a cut-off of -2.00 z-score for underweight and stunting, the prevalence of

underweight and stunting in western Kenya and Nyanza is 20.6% (n = 211) and 30.4% (n = 311),

respectively. Table 1 shows the prevalence of underweight and stunting in each district, as well

as the independent variables under analysis in this report.

Targeting Priorities

As see in Table 1, South Nyanza has the highest prevalence of underweight (35%, n = 130) and

stunting (42%, n = 130) and is the only district with these prevalences substantially greater than

the national averages of 21% and 30%, respectively. With 130 households of the total 1,023 in

the sample dataset, South Nyanza comprises 12.7% of the population of western Kenya. For

these reasons, households experiencing underweight in South Nyanza are the target priority for

this analysis.

Page 2: Advanced Nutrition Assessment and Monitoring_Final Paper

2

Results

Source of Drinking Water

Surface water as source of drinking water is significantly associated with a greater prevalence of

underweight (24%, n = 564), compared to well or piped water (17%, n = 454); this difference has

p = 0.009, as seen in Table 1, and shows an effect size of -0.244 (n = 1,017, p = 0.004), as seen

in Table 2. This effect size predicts a 7.8 percentage point increase in underweight prevalence.

Linear regression of source of drinking water revealed a significant interaction with toilet facility

(p = 0.041, n = 1,015) as well as with education (p = 0.034, n = 1,017), but no interaction with

access to health services, as seen in Table 2. Table 3 shows the two-by-two analysis of the

interaction between source of drinking water and toilet facility; because there are so few

households that report having a flush or ventilated toilet, this interaction is not reliable for

planning a nutrition intervention.

The two-by-two analysis between source of drinking water and education, found in Table 4,

revealed a significant difference in the severity of underweight between high-educated (z = -

0.4940, n = 124) and low-educated (z = -0.9417, n = 330) households using well or piped

drinking water, with p = 0.001. Additionally, there was a significant difference in the severity of

underweight between well or piped drinking water (z = -0.4940, n = 124) and surface drinking

water (z = -1.0370, n = 116) among high-educated households, with p = 0.001. Chart 1 illustrates

that underweight improves among households with well or piped drinking water compared to

those utilizing surface water, but this improvement is only significant when those households are

high-educated.

Table 5 shows the linear regression of source of drinking water, access to health services and

SES. The effect of source of drinking water on the severity of underweight remains significant

after controlling for these factors, diminishing from -0.244 (n = 1,017, p = 0.004) to -0.207 (n =

1,003, p = 0.013), indicating that access to health services and SES do not confound the effect of

source of drinking water on underweight.

Toilet Facility

Traditional pit latrine as toilet facility is significantly associated with a greater prevalence of

underweight (22%, n = 926), compared to flush or ventilated toilet facility (11%, n = 91); this

difference has p = 0.017, as seen in Table 1, and shows an effect size of -0.493 (n = 1,017, p =

0.001), as seen in Table 2. This effect size predicts a 15.8 percentage point increase in

underweight prevalence. Linear regression of toilet facility showed no significant interactions

between toilet facility and SES, education or access to health services, as seen in Table 6. Table

7 shows that, after controlling for SES, education and access to health services, the effect of

toilet facility on underweight is no longer significant, diminishing from -0.493 (p = 0.001, n =

1,017) to -0.227 (p = 0.158, n = 1,003), indicating that these factors confound the effect of toilet

facility on underweight.

Roof Type as Proxy for SES

Grass or thatch roof is significantly associated with a greater prevalence of underweight (23%, n

= 563), compared to corrugated iron roof (18%, n = 443); this difference has p = 0.035, as seen

in Table 1, and shows an effect size of -0.257 (n = 1,005, p = 0.004), as seen in Table 8. This

Page 3: Advanced Nutrition Assessment and Monitoring_Final Paper

3

effect size predicts an 8.2 percentage point increase in underweight prevalence. Linear regression

of roof type, as seen in Table 8, showed no significant interactions between roof type and source

of drinking water, education or place of delivery. After controlling for source of drinking water,

education and place of delivery, the effect of roof type on the severity of underweight decreased

slightly to -0.204, but remained significant (p = 0.018, n = 1,004), indicating that source of

drinking water, education and place of delivery do not confound the effect of roof type on

underweight. The -0.204 effect size of roof type on underweight predicts a 6.5 percentage point

increase in underweight prevalence.

Agriculture and Skilled Occupations

As seen in Table 1, Agriculture and skilled occupations are associated with a greater prevalence

of underweight (23%, n = 487) compared to professional, sales, domestic and unskilled

occupations (20%, n = 333) and clerical and services occupations (18%, n = 109), however these

differences are not significant. As seen in Table 10, linear regression of agriculture and skilled

occupations revealed that the effect of agriculture and skilled occupations on the severity of

underweight is confounded by education, water source and roof type, indicating that these

factors—education, source of drinking water and SES—account for the effect of occupation on

underweight status.

Discussion and Conclusions

This analysis found that a nutrition intervention in the Western Kenya and Nyanza region must

target South Nyanza district, as this district has the highest prevalence of underweight children.

An intervention in this region must focus on source of drinking water as the primary intervention

variable, with SES as a secondary intervention variable. Toilet facility is not a reliable

intervention variable as the sample of households with improved toilet facilities was not large

enough in this dataset to conclude that the observed effect of toilet facility on underweight was a

real effect. Agriculture and skilled occupations are associated with underweight, but this

association is explained by the tendency of these households to have low education, low SES and

surface drinking water.

Source of drinking water had the most reliable effect on the severity of underweight in Western

Kenya and Nyanza, but this effect was modified by the education level of the household so that

improved source of drinking water (well or piped water) was significantly associated with an

improvement in underweight only among high-educated households. Interventions aimed at

reducing the severity and prevalence of underweight in Western Kenya and Nyanza must

prioritize improving source of drinking water, but take into consideration the behavioral

differences between high- and low-educated households. Formative research is needed to

develop an understanding of these behavioral differences in order to contextualize nutrition

interventions, as well as to develop an appropriate educational piece of the intervention in order

to ensure that improving households’ source of drinking water leads to the anticipated

improvement in underweight. Additionally, source of drinking water was associated with

improvement in underweight regardless of access to health services and SES, indicating that

improvement in source of drinking water will predict an improvement in underweight across all

income levels and throughout rural regions that may have limited health facilities. SES must be a

secondary intervention variable because improvement in SES is associated with an improvement

in underweight regardless of source of drinking water, education or access to health services.

Page 4: Advanced Nutrition Assessment and Monitoring_Final Paper

4

Tables and Charts

Table 1. Distribution of population among independent variables and prevalence of underweight

and stunting by district, water source, toilet facility, SES, access to health services,

education and occupation.

Table 2. Linear regression of underweight by source of drinking water (badwater) and

interactions with toilet facility (badtoilet), education (lowed) and access to health services

(ddelpl): B Coefficient (t, p).

Table 3. Mean underweight z-score by source of drinking water and toilet facility (n).

Table 4. Mean underweight z-score by source of drinking water and education (n).

Chart 1. Interaction between source of drinking water and level of education.

Table 5. Linear regression of underweight by source of drinking water (badwater), access to

health services (ddelpl) and SES (badroof): B Coefficient (t, p).

Table 6. Linear regression of underweight by toilet facility (badtoilet) and interactions with SES

(badroof), education (lowed) and access to health services (ddelpl): B Coefficient (t, p).

Table 7. Linear regression of underweight by toilet facility (badtoilet), SES (badroof), education

(lowed) and access to health services (ddelpl): B Coefficient (t, p).

Table 8. Linear regression of underweight by SES (badroof) and interactions with source of

drinking water (badwater), education (lowed) and access to health services (ddelpl): B

Coefficient (t, p).

Table 9. Linear regression of underweight by SES (badroof), source of drinking water

(badwater), education (lowed) and access to health services (ddelp): B Coefficient (t, p).

Table 10. Linear regression of underweight by agriculture occupation (dag), education (lowed),

source of drinking water (badwater) and SES (badroof): B Coefficient (t, p).

Page 5: Advanced Nutrition Assessment and Monitoring_Final Paper

5

Table 1. Distribution of population among independent variables and prevalence of

underweight and stunting by district, source of drinking water, toilet facility, SES, access to

health services, education and occupation.

Underweight

(<= -2SD)

Stunting

(<= -2SD)

n % n % n %

District 2,209 1023 21 1023 30

S. Nyanza – rural 257 11.6 130 35 130 42

Kisii/Nyamira – rural 488 22.1 195 21 195 28

Bungoma – rural 396 17.9 215 21 215 36

Siaya – rural 408 18.51 182 20 182 34

Kakamega - rural 381 17.2 182 16 182 25

Other rural 161 7.3 69 9 69 12

Other urban 118 5.3 50 14 50 20

Source of drinking water (badwater) 2,195 1018 1018

Well, piped 955 43.5 454 17* 545 26*

Surface water 1,240 46.5 564 24 564 34

Toilet facility (badtoilet) 2,192 1017 1017

Flush, ventilated 209 9.5 91 11* 91 15*

Traditional, bush 1,983 90.5 926 22 926 32

SES (badroof) 2,152 1006 1006

Corrugated iron 1,082 50.3 443 18* 443 26*

Grass, thatch 1,070 49.7 563 23 563 34

Access to health services (ddelpl) 1,214 1022 1022

Government, private 426 35.1 370 16* 370 28

Home 788 64.9 652 23 652 31

Education (lowed) 2,209 1023 1023

Secondary+ 472 21.4 241 17 241 23*

None, primary 1,737 78.6 782 22 782 30

Occupation (partocc) 1,565 929 929

Agriculture, skilled 833 53.2 487 23 487 34

Professional, sales, domestic, unskilled 539 34.4 333 20 333 31

Clerical, services 193 12.3 109 18 109 21

* Denotes p-value < 0.05

Page 6: Advanced Nutrition Assessment and Monitoring_Final Paper

6

Table 2. Linear regression of underweight by source of drinking water (badwater) and

interactions with toilet facility (badtoilet), education (lowed) and access to health services

(ddelpl): B Coefficient (t, p).

Variable Model 1 Model 2 Model 3 Model 4

badwater -0.244

(-2.920, 0.004)

0.443

(1.373, 0.170)

-0.543

(-3.189, 0.001)

-0.301

(-2.183, 0.029)

badtoilet -- -0.247

(-1.430, 0.153)

-- --

badwater*badtoilet -- -0.685

(-2.048, 0.041)

-- --

lowed -- -- -0.448

(-3.224, 0.001)

--

badwater*lowed -- -- 0.415

(02.124, 0.034)

--

ddelpl -- -- -- -0.366

(-2.859, 0.004)

badwater*ddelpl -- -- 0.101

(0.583, 0.560)

Constant -0.819 -0.201 -0.494 -0.589

Adj R sq 0.008 0.024 0.018 0.018

N 1017 1015 1017 1016

Table 3. Mean underweight z-score by source of drinking water and toilet facility (n).

Source of drinking water

Toilet

facility

Well, piped Surface TOTAL p

Flush -0.6109 (69) -0.1677 (22) -0.5037 (91) 0.215

Bush -0.8574 (384) -1.099 (541) -0.9990 (925) 0.006

TOTAL -0.8199 (453) -1.063 (563) -0.9456 (1016) .004

p 0.143 0.002 0.001

Table 4. Mean underweight z-score by source of drinking water and education (n).

Source of drinking water

Education

Well, piped Surface TOTAL p

High -0.4940 (124) -1.0370 (116) -0.7564 (240) 0.001

Low -0.9417 (330) -1.0699 (448) -1.0135 (782) 0.182

TOTAL -0.8194 (454) -1.0631 (564) -0.9524 (1023) 0.004

p 0.001 0.816 0.008

Page 7: Advanced Nutrition Assessment and Monitoring_Final Paper

7

Table 5. Linear regression of underweight by source of drinking water (badwater), access

to health services (ddelpl) and SES (badroof): B Coefficient (t, p).

Variable Model 1 Model 2 Model 3

badwater -0.244

(-2.920, 0.004)

-0.237

(-2.853, 0.004)

-0.207

(-2.483, 0.013)

ddelpl -- -0.311

(-3.614, 0.000)

-0.235

(-2.675, 0.008)

badroof -- -- -0.214

(-2.524, 0.012)

Constant -0.819 -0.624 -0.579

Adj R sq 0.008 0.019 0.020

N 1017 1016 1003

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0

Well Surface

Chart 1. Interaction between source of drinking water and level of education.

High Low

Page 8: Advanced Nutrition Assessment and Monitoring_Final Paper

8

Table 6. Linear regression of underweight by toilet facility (badtoilet) and interactions with

SES (badroof), education (lowed) and access to health services (ddelpl): B Coefficient (t, p).

Variable Model 1 Model 3 Model 2 Model 4

badtoilet -0.493

(-3.391, 0.001)

-0.423

(-2.423, 0.016)

-0.619

(-2.988, 0.003)

-0.533

(-2.865, 0.004)

badroof -- -0.932

(-2.260, 0.024)

-- --

badtoilet*badroof -- 0.745

(1.766, 0.078)

-- --

lowed -- -- -0.541

(-1.934, 0.053)

--

badtoilet*lowed -- -- 0.408

(1.359, 0.174)

--

ddelpl -- -- -- -0.583

(-1.998, 0.046)

badtoilet*ddelpl -- -- -- 0.348

(1.137, 0.256)

Constant -0.504 -0.463 -0.272 -0.305

Adj R sq 0.010 0.016 0.016 0.018

N 1017 1005 1017 1015

Table 7. Linear regression of underweight by toilet facility (badtoilet), SES (badroof),

education (lowed) and access to health services (ddelpl): B Coefficient (t, p).

Variable Model 1 Model 2 Model 3 Model 4

badtoilet -0.493

(-3.630, 0.001)

-0.295

(-1.857, 0.064)

-0.265

(-1.645, 0.100)

-0.228

(-1.412, 0.158)

badroof -- -0.220

(-2.549, 0.011)

-0.202

(-2.311, 0.021)

-0.182

(-2.077, 0.038)

lowed -- -- -0.127

(-1.239, 0.216)

-0.072

(-0.684, 0.494)

ddelpl -- -- -- -0.201

(-2.199, 0.028)

Constant -0.504 -0.571 -0.511 -0.458

Adj R sq 0.010 0.011 0.014 0.019

N 1017 1005 1005 1003

Page 9: Advanced Nutrition Assessment and Monitoring_Final Paper

9

Table 8. Linear regression of underweight by SES (badroof) and interactions with source of

drinking water (badwater), education (lowed) and access to health services (ddelpl): B

Coefficient (t, p).

Variable Model 1 Model 2 Model 3 Model 4

badroof -0.257

(-3.072, 0.002)

-0.381

(-3.036, 0.002)

-0.278

(-1.543, 0.123)

-0.273

(-1.941, 0.053)

badwater -- -0.335

(-2.677, 0.008)

-- --

badroof*badwater -- 0.228

(1.356, 0.175)

-- --

lowed -- -- -0.179

(-1.334, 0.183)

--

badroof*lowed -- -- 0.061

(0.229, 0.765)

--

ddelpl -- -- -- -0.276

(-2.200, 0.028)

badroof*ddelpl -- -- -- 0.086

(0.489, 0.625)

Constant -0.822 -0.701 -0.701 -0.669

Adj R sq 0.009 0.012 0.012 0.017

N 1005 1005 1005 1004

Table 9. Linear regression of underweight by SES (badroof), source of drinking water

(badwater), education (lowed) and access to health services (ddelpl): B Coefficient (t, p).

Variable Model 1 Model 2 Model 3 Model 4

badroof -0.257

(-3.072, 0.002)

-0.254

(-3.038, 0.002)

-0.230

(-2.694, 0.007)

-0.204

(-2.370, 0.018)

badwater -- -0.209

(-2.498, 0.013)

-0.203

(-2.425, 0.015)

-0.204

(-2.441, 0.015)

lowed -- -- -0.142

(-1.402, 0.161)

-0.078

(-0.753, 0.452)

ddelpl -- -- -- -0.218

(-2.407, 0.016)

Constant -0.822 -0.709 -0.616 -0.537

Adj R sq 0.008 0.014 0.015 0.019

N 1006 1005 1005 1004

Page 10: Advanced Nutrition Assessment and Monitoring_Final Paper

10

Table 10. Linear regression of underweight by agriculture occupation (dag), education

(lowed), source of drinking water (badwater) and SES (badroof): B Coefficient (t, p).

Variable Model 1 Model 2 Model 3 Model 4

dag -0.208

(-2.408, 0.016)

-0.186

(-2.138, 0.033)

-0.164

(-1.876, 0.061)

-0.135

(-1.535, 0.500)

lowed -- -0.199

(-1.920, 0.055)

-0.163

(-1.755, 0.800)

-0.072

(-0.675, 0.500)

badwater -- -- -0.201

(-2.303, 0.022)

-0.174

(-1.996, 0.046)

badroof -- -- -- -0.270

(-3.028, 0.003)

Constant -0.866 -0.724 -0.640 -0.613

Adj R sq 0.005 0.008 0.013 0.017

N 929 929 924 911