Ctown 5 Interpreting data emergencies long term descriptive trends for causality and intervention decisions
Access to Food Source:
Area level survey results, Kenya: GAM % by season
Area level survey results, pooled & smoothed: Kenya, GAM%, by season
Area level survey results, pooled & smoothed: Ethiopia, GAM%, by season
Figure 2. Trends in Underweight and HIV prevalence by region in Ethiopia
Figure 6. Trends in Underweight and HIV prevalences by region in Uganda
Child underweight prevalences are higher in lower HIV prevalence areas
Figure 8. Kenya, Ethiopia, Uganda: Scatterplot of Underweight and HIV prevalences by country
Under 5 mortality with HIV prevalence by area
+ Increases + SES HIV Malnutrition Decreases –
Table 4. Associations of underweight with HIV and SES variables Ethiopia, Kenya, and Uganda, pooled. Variable Model 1Model 2Model 3 Prevalence of HIV (%) (-3.114, 0.003) (-2.290, 0.027) (-1.721, 0.092) Improved flooring in household (%) (-6.088, 0.000) (-5.141, 0.000) Safe water (%) (2.391, 0.021) Safe excreta (%) (-2.145, 0.037) Constant Adj R sq N6050 Dependent variable =underweight prevalence (%).
Interaction between drought and HIV on changes in child underweight.
Figures and Tables Figure 1. Kenya, Ethiopia, and Uganda: Drought (negative y-values) plotted over time
Figure 9. Scatterplot of Underweight and Drought with HIV results for all countries pooled (hivcat07=1 refers to high prevalence HIV).
Differences in stunting and wasting in two regions of Kenya
Differential growth patterns in Uganda and Somalia
Different relations between GAM% and child mortality in different populations Hence interpret GAM within populations, not across...
Conclude Horn: effect of drought > HIV Srn: drought and HIV interact – both together give rapid deterioration Both: HIV still associated with lower malnutrition, because of assocn with SES.
Size of effects: Season: about 4 ppts wasting (GAM) Drought: >= 10 ppts wasting (GAM)
Wasting in different populations. Similar mortality risk (e.g. U5MR of 2.0) at 5% GAM in (e.g. Uganda), 25% N Kenya Hence interpret for specific population, and use trends more About 10 ppts change in GAM suggests 0.5—1.0 increase in U5MR
Policies and programmes to improve nutrition Long-term: high priority for community-based (CBHNPs) with CHNWs Reduce vulnerability, esp to drought Mitigate emergencies
Levels of food aid
Estimating food aid levels: per need Denominator changes can give large fluctuations
Estimating food aid levels: per population For descriptive, easier to see what’s happening
Food aid levels per population: Lesotho and Mozambique
Food aid levels per population: Malawi and Swaziland
Food aid levels per population: Zambia and Zimbabwe
Effects of HIV by area on underweight, controlling for SES
Underweight with SES
HIV with SES
CountryCoefficient (B)P-valueR sqN All Lesotho Malawi Mozambique Swaziland Zambia Zimbabwe Underweight with HIV
+ Increases + SES HIV Malnutrition Decreases –
Variable \ Models 1234 Prevalence of HIV (%) (hivprev4) % head of hh with more than primary education (eduprim2) % urban population (urban) % hhs with electricity (electric) % children >= 12 mo immunized for measles (measles) % hhs with safe water (safewatr) % hhs with safe excreta disposal (safexcrt) Constant N Adj R squ Dep = underweight HIV is less associated with underweight controlling for SES In cells B T P
CountryCoefficientP-valueR sqN All * Lesotho Malawi Mozambique * Swaziland Zambia Zimbabwe Removing SES from underweight, association with HIV becomes insignificant Coefficient smaller and less significant
Significant overall To recap …
Effects of HIV by area on change in underweight, controlling for SES
+ Increases + SES HIV Malnutrition Decreases – To recap …
HIV with change underweight: no clear relation
Variable \ Models 1234 Dhivcat Eduprim Uwt Safeexctr urban Constant N50 Adj R squ F = 4.463, n = 50, p = Dependent = change in underweight, ppts/yr Means adjusted for SES (education) Change in underweight deteriorated more in higher HIV areas when SES is controlled. In cells B T P
Effects of food aid
Lower need Higher coverage Higher need Higher coverage Lower need Lower coverage Higher need Lower coverage Targeting: actual food aid coverage vs VAC need Some areas had high coverage with low need: these did worst in terms of child nutrition Most targeting worked, to high need areas A few high need areas had low coverage
Change in underweight (ppts/yr) by need and food aid coverage More coverage improves nutrition in high need areas Low need areas do OK anyway Need to look into …
Change in underweight by RDA met by food aid Findings as before when stratified by VAC need
Report CHUWY Change in underweight per year (percentage points) (+=deter; -=improv) ( )(1) FA1CATUS Food aid 1 category (unstandardized) (low & high) 1 Low (<=11.50) 2 High (>11.50) Total 1 Low (<=11.50) 2 High (>11.50) Total 1 Low (<=11.50) 2 High (>11.50) Total HIVCAT3 HIV category (relative to the country) (low & high) 1 Low 2 High Total MeanNStd. Deviation Effect of food aid in high need group, by HIV (areas) High need, low food aid, high HIV, deteriorates fast; high food aid helps
100% Coverage 50% Coverage 80% Coverage Coverage > 50% need Coverage < 50% need Need <50%>50% How coverage of food aid met assessed need in drought Change in Underweight ppts/yr
BVACAT (beneficiaries / need category) Need Lo (Vaccat2=0) Need Hi (Vaccat2= 1) Low (<0.5) 1.30 (9)4.49 (9) High (>=0.5) 3.44 (10)-0.47 (13) Change in underweight prevalence by beneficiary/need coverage
Effects of food assistance on child nutrition (24-59 mo) in Zimbabwe, : regressions (HLM) of wt/age, relating changes between May 2002 to Feb 2003 to levels of supplementary feeding (SF) and food distribution; in cells – coefficient (B, unstandardized), t value, p, n. May 2002 Feb 2003 WAZ Z-score Test if slopes are different High Supplementary Feeding Low Supplementary Feeding
Intervention, outcome Supplementary Feeding WAZ Food distrn WAZ Area by HIVLow HIVHigh HIVLow HIVHigh HIV Model (HLM), number 1256 Year (DYEAR) Food aid category (SFEDCAT or FDXCAT) Interaction (INDYRSFE or INTDYFD) Constant Total N Chi-sq, P < < Interaction by OLS NS In cells B T P N
Changes in mean WAZ by supplementary feeding coverage group, low and high HIV areas together, May 2002 – Feb 2003; results calculated from regression. Higher supplementary feeding improved No supplementary feeding deteriorated, but remained best