Result of Stata software showed that of the total 8768 respondents 3441 (3.2%) stunted while 5327 (60.8%) were normal and 2508 (28.6%) were underweight while 6260 (71.4%) were normal children.
Similarly 1425 (16.3 percent) were wasted while 7343 (83.7 percent) were normal children at the time of the survey.
From 39.5 percent of stunted children, more than half percent were aged less than 20 months years and also highly wasted and underweight children observed in this age group. This is viewing that as the age of children increasing the malnutrition is falling.
Mothers’ education exposed the effects on nutritional status. The results mothers with no formal education 40.9 percent had had stunted children out of the given. Mothers’ with secondary and higher education whose children were stunted 36.9 percent and 35.2 percent respectively. Thus, the effect of education might diminish when analysis is done based on the status of malnutrition.
The results revealed that the prevalence of stunting, wasting and underweight among the regions were variable with Tigray (48.1%), Amhara and Somali regions 46.6% having highest proportion, while the lowest proportion of stunted was observed in Addis Ababa and Dire Dawa (24.9% and 30.8%) respectively. Somali and Afar regions were where highest proportion of wasted of children (23.9% and 21.3%) respectively while Addis Ababa and Harari were lowest proportion of wasted observed (6.8% and 9.4%) respectively. Likewise Somali and Afar confirmed that highest proportion of underweight (34.5% and 34.1%) respectively. Households who lived in rural area had highest proportion of stunted under-five children (39.4%) while who lived in urban had lowest proportion of stunted under-five children (35.7%).
Also those lived in rural area with highest proportion of wasted and underweight of children from the total were (23.9 percent and 37.1%) respectively while those lived in urban were 13.6 percent and 28.1 percent respectively. It is thought that exposure of media like radio, television and newspapers helps to improve health care and practiced of feeding. Accordingly, households who were not exposed to any kind of media (40.0%) were found with stunted children than those who were exposed (35.5%). Households who were not exposed to any kind of media (16.8% and 29.5%) were found with more wasted and underweight children than those exposed to media (13.1% and 23.4%) respectively. Some 42.5 percent of the households who had not been visited by health workers during last 12 months before the survey, while 37.1 percent of these who had been visited by health workers were have stunted under five-children.
Wealth index also showed effects on status of malnutrition of under-five children. From 39.5 percent of stunted children, 42.6 were poor, 38.8 percent were middle level and 32.0 percent were rich of under-five children. From total of wasted children, 20.0 percent were poor, 12.3 percent were middle level and 13.2 percent were rich of under-five children. Also highest prevalence of underweight children observed in those failed under poor, middle and rich status of the societies (34.1%, 24.6% and 20.8%) respectively.
Table 1
Descriptive analysis on some selected variables of malnutrition among under-five children, n = 8768
Variables | Status of Stunting | Status of Wasting | Status of Underweight |
Stunt | Normal | (P value) | Wasting | Normal | (P value) | Underweight | Normal | (P value) |
Count (%) | Count(%) | Count (%) | Count (%) | Count (%) | Count(%) |
Age of Children < 20 months 21–40 months 41–59 months | 1709(52.0%) 907(33.3%) 825(29.9%) | 1579(48.0%) 1816(66.7%) 1932(70.1%) | 364.245 (.000) | 771(23.4%) 295(10.8%) 359(13.0%) | 2517(76.6%) 2428(89.2%) 2398(87.0%) | 204.241(.000) | 1039(31.6%) 827(30.0%) 751(27.6%) | 2249(68.4%) 1930 (70.0%) 1972(72.4%) | 10.703(.005) |
Mothers education No education Primary Secondary Higher | 2230(40.9%) 955(36.6%) 117(36.9%) 139(35.2%) | 3219(59.1%) 1652(63.4%) 200(63.1%) 25664.8%) | 17.71(.001) | 925(17.0%) 343(13.2%) 44(13.9%) 44(11.1%) | 4524(83.0%) 2264(86.8%) 273(86.1%) 351(88.9%) | 26.397(.000) | 1754(32.2%) 671(25.7%) 107(27.1%) 76(24.0%) | 3695(67.8%) 1936(74.3%) 288(72.9%) 241(76.0%) | 41.991(.000) |
Fathers education No education Primary Secondary Higher | 1721(38.5%) 1149(40.7%) 256(40.1%) 313(37.4%) | 2745(61.5%) 1674(59.3%) 385(59.9%) 523(62.6%) | 48.205(.000) | 565(12.7% 457(16.2%) 165(19.7%) 169(26.3%) | 3901(87.3%) 2366(83.8%) 671(80.3%) 474(73.7%) | 97.406(.000) | 1189(26.6%) 916(32.4%) 279(33.4%) 224(34.8%) | 3277(51.6%) 1907(67.6%) 557(66.6%) 419(65.2%) | 43.936(.000) |
Region Tigray Afar Amhara Oromia Somali Benishangul SNNP Gambela Harar Addis Ababa Dire Dawa | 448(48.1%) 314(38.2%) 410(46.6%) 502(36.8%) 531(46.6%) 240(33.2%) 420(38.6%) 191(34.6%) 157(34.4%) 99(24.9%) 129(30.8%) | 483(51.9%) 507(61.8%) 470(53.4%) 863(63.2%) 609(53.4%) 482(66.8%) 667(61.4%) 361(65.4%) 299(65.6%) 298(75.1%) 288(69.1%) | 146.895(0.00) | 144(15.5%) 175(21.3%) 130(14.8%) 222(16.3%) 172(23.9%) 90(12.5%) 175(16.1%) 91(16.1%) 43(9.4%) 27(6.8%) 56(13.4%) | 787(84.5%) 646(78.7%) 750(85.2%) 1143(83.7%) 868(76.1%) 632(87.5%) 912(83.9%) 461(83.5%) 413(90.6%) 370(93.2%) 361(86.6%) | 117.510(0.000) | 273(29.3%) 280(34.1%) 255(29.0%) 397(29.1%) 393(34.5%) 211(29.2%) 313(28.8%) 157(28.4%) 123(27.0%) 43(10.8%) 63(15.1%) | 658(70.7%) 541(65.9%) 625(71.0%) 968(70.9%) 747(65.5%) 511(70.8%) 774(71.2%) 395(71.6%) 333(73.0%) 354(89.2%) 354 (84.9%) | 131.119(0.000) |
Place of Residence Urban Rural | 367(35.7%) 2817(39.4%) | 1037(64.3%) 4338 (60.6%) | 85.838(.000) | 220(13.6%) 1710(23.9%) | 1394(86.4%) 5445(76.1%) | 108.345(.000) | 453(28.1%) 2655(37.1%) | 1160(14.0%) 4500(62.9%) | 51.674(.000) |
Exposure to Media Not at all At least once a week | 2985(40.0%) 456(35.3%) | 4490(60.0%) 837(64.7%) | 10.067(.001) | 1255(16.8%) 170(13.1%) | 6220(83.2%) 1123(86.9%) | 10.740(0.004) | 2206(29.5%) 302(23.4%) | 5269(70.5%) 991(76.6%) | 20.450(0.220) |
Visited by HEW within12 months No Yes | 2601(42.5%) 982(37.1%) | 3519(57.5%) 1666(62.9%) | 4.313(0.020) | 947(15.5%) 409(15.4%) | 5173(84.5%) 2239(84.6%) | 5.111(0.013) | 1778(29.1%) 830(31.3%) | 4342(70.9%) 1818(68.7%) | 4.648(.017) |
Wealth index Poor Middle Rich | 1819(42.6%) 1043(38.8%) 579(32.0%) | 2453(57.4%) 1644(61.2%) 1230(68.0%) | 59.880(0.000) | 856(20.0%) 331(12.3%) 238(13.2%) | 3416(80.0%) 2356(87.7%) 1571(86.8%) | 88.256(0.000) | 1471(34.4%) 660(24.6%) 377(20.8%) | 2801(65.6%) 2027(75.4%) 1432(79.2%) | 40.675(0 000) |
Spatial Data Analysis
Global spatial autocorrelation was assessed using the Global Moran’s I statistic (Moran’s I) to evaluate whether the pattern was clustered, dispersed or random across the study area using ArcGIS version 10.3.
A positive value for Moran’s index indicates a cluster pattern of childhood malnutrition. The result of Global Moran index value of malnutrition of children in Ethiopia (for stunting I = 0.204, P-value < 0.0001, for wasting I = 0.152, P-value = < 0.0001 and for underweight I = 0.195, P-value = < 0.001). These results indicate that there was a positive spatial autocorrelation or cluster adjacent having similarities in the prevalence of child malnutrition. In other word, we can say that the neighboring clusters influenced the prevalence of child malnutrition in Ethiopia.
A further peak is spatial mapping which is used to reveals variation across regions of Ethiopia. The geographical distribution of the indexes of the stunting (a), underweight (b) and wasting (c) displayed in Fig. 2 shows regionally spatial pattern. In view of that highest rate of stunting observed in Tigray, Amahar and Somali. Afar and Somali had the highest rates of underweight and wasting in the under-five children. Generally the result reveals that some Northern and Eastern part of the country had higher malnutrition proportion than central south-west regions.
Multivariate Multilevel Regression Model
To identify determinant factors of child malnutrition, a two level mixed effects logistic regression analysis was used. The intercept only model (empty model) was constructed without determinant factors to check the application of multilevel analysis to the data set. This model allows us to evaluate the extent of cluster variation influencing child nutritional status. In model two, only individual-level determinants factors were included and in model three, only cluster level determinants were included. The level of heterogeneity was evaluated at each model.
The data was tested to determine whether there is heterogeneity between clusters. The result of chi square were 147.28, 211.55 and 201.43 respectively for stunting, wasting and underweight with p < 0.0001 providing evidence of heterogeneity among clusters with respect to the status of nutrition of under-five children (Table 2).
Table 2
Estimates for variance components model for Stunting, Wasting and Underweight.
Response variable | Indicators | OR | S.E. | Z-value | P-value | [95% CI] |
Stunting | = intercept | 1.580 | 0.0512 | 14.03 | 0.000 | 1.4.789 1.680 |
| 0.295 | 0.0405 | | | 0.226 0.386 |
Deviance-based chi-square | 147.25 | | | < 0.0001 | |
ICC | 0.082 | 0.011 | | | 0.064 0.105 |
Wasting | = intercept | 6.240 | 0.308 | 37.14 | 0.000 | 5.670 6.880 |
| 0.624 | 0.0798 | | | 0.485 0.801 |
Deviance-based chi-square | 211.55 | | | < 0.0001 | |
ICC | 0.117 | 0.016 | | | 0.089 0.152 |
Underweight | = intercept | 2.717 | 0.1030 | 26.49 | 0.000 | 2.523 2.925 |
| 0.411 | 0.0498 | | | 0.295 0.492 |
Deviance-based chi-square | 201.43 | | | < 0.0001 | |
ICC | 0.104 | 0.012 | | | 0.082 0.130 |
From Table 2 we observe that the result estimated intra class correlation (ICC) used to assess the variation by levels and the values which is different from zero indicates that appropriateness of multilevel modeling analysis [14]. Thus the computed ICC = 0.082 shows that 8.20 percent of the variation in the stunting under-five children can be explained by cluster (level two). The remaining 91.8 percent of the variation of stunting under-five children is explained within the same cluster. Likewise the ICC for wasting and underweight were 11.7 and 10.4 percent of variation between clusters (Table 2).