Bivariate Analysis
A Correlation Test Between ME in Single Years and IM was done to determine how infant mortality correlated with ME single years for the period 1992 to 2018 using point-biserial correlation at 0.01 or 0.05 levels of significance.
Table 2: Pearson Point-biserial Correlations between ME in Single Years and IM: 1992 to 2018
ZDHS Phase
|
Education in single years
|
Infant Death
|
ZDHS1992
|
ME in Single Years
|
Pearson Correlation
|
1
|
-.037**
|
Sig. (2-tailed)
|
|
.003
|
IM
|
Pearson Correlation
|
-.037**
|
1
|
Sig. (2-tailed)
|
.003
|
|
ZDHS1996
|
ME in Single Years
|
Pearson Correlation
|
1
|
-.029*
|
Sig. (2-tailed)
|
|
.013
|
IM
|
Pearson Correlation
|
-.029*
|
1
|
Sig. (2-tailed)
|
.013
|
|
ZDHS2001/02
|
ME in Single Years
|
Pearson Correlation
|
1
|
-.027*
|
Sig. (2-tailed)
|
|
.028
|
IM
|
Pearson Correlation
|
-.027*
|
1
|
Sig. (2-tailed)
|
.028
|
|
ZDHS2007
|
ME in Single Years
|
Pearson Correlation
|
1
|
-.006
|
Sig. (2-tailed)
|
|
.642
|
IM
|
Pearson Correlation
|
-.006
|
1
|
Sig. (2-tailed)
|
.642
|
|
ZDHS2013/14
|
ME in Single Years
|
Pearson Correlation
|
1
|
-.014
|
Sig. (2-tailed)
|
|
.111
|
IM
|
Pearson Correlation
|
-.014
|
1
|
Sig. (2-tailed)
|
.111
|
|
ZDHS2018
|
ME in Single Years
|
Pearson Correlation
|
1
|
-.005
|
Sig. (2-tailed)
|
|
.594
|
IM
|
Pearson Correlation
|
-.005
|
1
|
Sig. (2-tailed)
|
.594
|
|
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
|
Table 2 presents the correlations between ME in single years and IM from 1992 to 2001-2 and as indicated in the table, increase in ME years was associated with a decrease in IM with Point-Biserial correlation coefficients of -.037**, -.029* and -.027* which were statistically significant at the P = 0.003, P = 0.013 and P = 0.028 respectively.
Multivariate Analysis
By use of Binary Logistic Regression, three models were fitted using the ENTER method to determine the influence of ME and its interaction with the selected explanatory variables on IM. In the first model all the predictor variables were entered to determine their single general influence on IM. However, the first model is not presented in this paper because the main aim of the paper was to determine how ME related with IM and how this key independent variable with its interactions influenced IM for the period under study. In the second model, ME as a single variable and its interactions with other predictor variables were entered to determine their general interaction influence on IM. In this model however, only variables which were found to be associated with ME were included. In the third-final model, ME as a single (key independent) variable and the interactions which influenced IM throughout the period under study were entered. The goodness-of-fit of the final model was done for prediction accuracy at -2 log likelihood, Cox & Snell R squared and Hosemer and Lemeshow tests. Results are presented with respect to the macro-variables: social-economic, demographic, mother and child health and water and sanitation factors.
Second Model
Table 3: Interaction Influence of ME and other Predictors on IM
Variables in ZDHS
|
ZDHS1992
|
ZDHS1996
|
ZDHS2001/2
|
ZDHS2007
|
ZDHS2013/14 ZDHS2018
|
B
|
Sig.
|
B
|
Sig.
|
B
|
Sig.
|
B
|
Sig.
|
B
|
Sig. B Sig.
|
M.E / Me * Social-EC- Factors
|
ME in single years
|
-.21
|
.011*
|
-.35
|
.031*
|
-.41
|
.023*
|
-.27
|
.346
|
-7.10
|
.921 -.571 .100
|
ME * Province
|
|
.000*
|
|
.001*
|
|
.000*
|
|
.031*
|
|
.019* .004*
|
Copperbelt (1)
|
-.17
|
.099
|
-.04
|
.703
|
-.01
|
.892
|
-.00
|
.982
|
-.23
|
.085 -.033 .324
|
Eastern (2)
|
-.24
|
.014*
|
-.13
|
.153
|
-.03
|
.808
|
-.05
|
.708
|
-.15
|
.236 -.008 .796
|
Luapula (3)
|
-.07
|
.459
|
-.14
|
.125
|
-.11
|
.303
|
.27
|
.032*
|
.08
|
.533 .041 .156
|
Lusaka (4)
|
.12
|
.216
|
.12
|
.207
|
.30
|
.001*
|
.24
|
.075
|
-.12
|
.360 -.074 .025*
|
Muchinga (5)
|
|
|
|
|
|
|
|
|
-.04
|
.33 .032 .302
|
Northern (6)
|
-.03
|
.740
|
-.04
|
.631
|
.07
|
.463
|
.12
|
.323
|
-.05
|
.763 -.018 .611
|
North Western (7)
|
-.10
|
.459
|
-.07
|
.564
|
-.16
|
.239
|
-.02
|
.882
|
.14
|
.342 -.093 .021*
|
Southern (8)
|
-.28
|
.003*
|
-.26
|
.010*
|
-.24
|
.027*
|
-.01
|
.047*
|
-.10
|
.032* -.051 .115
|
Western (9)
|
-.24
|
.062
|
.01
|
.879
|
-.04
|
.743
|
-.03
|
.855
|
-.12
|
.018* -.012 .988
|
ME by Lived in Urban (1)
|
.05
|
.469
|
.00
|
.974
|
-.01
|
.912
|
-.11
|
.223
|
-.06
|
.469 .018 .385
|
ME by Used Contraceptive (1)
|
-.19
|
.001*
|
-.22
|
.000*
|
-.26
|
.000*
|
-.12
|
.036*
|
-.11
|
.025* -.069 .000*
|
M.E *Demographic Factors
|
M. E * M. Age In 5-Year
|
|
.035*
|
|
.004*
|
|
.378
|
|
.022*
|
|
.001* .000*
|
(20-24)(1)
|
-.23
|
.318
|
.12
|
.618
|
-.19
|
.286
|
-.24
|
.274
|
-.09
|
.709 -.019 .002*
|
( 25-29)(2)
|
-.07
|
.744
|
.14
|
.527
|
-.26
|
.122
|
-.34
|
.082
|
-.41
|
.055 -.095 .001*
|
(30-34)(3)
|
-.07
|
.723
|
.10
|
.660
|
-.13
|
.444
|
-.28
|
.152
|
-.42
|
.046* -.088 .005*
|
(35-39)(4)
|
-.07
|
.738
|
.06
|
.794
|
-.18
|
.292
|
-.47
|
.018*
|
-.53
|
.013* -.109 .002*
|
( 40-44)(5)
|
-.18
|
.390
|
-.22
|
.350
|
-.16
|
.345
|
-.33
|
.099
|
-.53
|
.015* -.064 .118
|
(45 - 49)(6)
|
-.09
|
.694
|
-.08
|
.753
|
-.22
|
.243
|
-.26
|
.229
|
-.32
|
.166 -.122 .173
|
ME by Pre-Birth Interval24+(1)
|
-.20
|
.000*
|
-.15
|
.001*
|
-.15
|
.002*
|
-.09
|
.016*
|
-.15
|
.007* -.107 .000*
|
M.E * Environmental Factors
|
ME By Sourced from Pipe (1)
|
-.01
|
.824
|
-.03
|
.676
|
-.02
|
.810
|
-.15
|
.098
|
-.04
|
.612 .052 .170
|
ME By Flushable Toilet (1)
|
.07
|
.205
|
.06
|
.294
|
.04
|
.662
|
-.02
|
.862
|
.16
|
.037* -.05 .070
|
M.E * M.C Health Factors
|
ME By Breastfed (1)
|
-1.49
|
.000*
|
-1.82
|
.000*
|
-1.81
|
.000*
|
-1.66
|
.000*
|
-7.26
|
.019* -.460 .000*
|
ME By Delivered from Hospital (1)
|
-.09
|
.077
|
-.04
|
.439
|
-.04
|
.441
|
-.08
|
.222
|
-.00
|
.954 .34 .067
|
ME By Received Antenatal Care (1)
|
-.23
|
.000*
|
-.00
|
.02*
|
-.22
|
.000*
|
-.24
|
.000*
|
-.05
|
.046* -.159 .000*
|
ME By Received Tetanus Injection (1)
|
-.01
|
.822
|
-.05
|
.195
|
.03
|
.600
|
.03
|
.615
|
.09
|
.188 -.003 .900
|
Intercept
|
2.03
|
.000*
|
1.95
|
.000*
|
2.12
|
.000*
|
2.23
|
.000*
|
.09
|
.001* 3.107 .000
|
In table 3, the intercepts (constants) show that from 1992 to 2018 infant mortality increased by [2.0, 2.0, 2.1, 2.2 .1 and 3.1 log odds, P < 0.05: CL: 95%] respectively, holding all predictor variables constant.
When ME (single years) was added to the second model some reductions in the coefficients were observed compared to the first model (table 3). Results show that, overall, the interactional effect between ME and Province contributed significantly to the reduction in IM throughout the period under study. However, holding everything else constant, disaggregated effects from 1992 to 2018 shows that the interaction between ME and being in Eastern Province reduced infant mortality (the least) by [.2 log odds, P < 0.05: CL: 95%] in 1992 only. The interaction between ME and Luapula and Lusaka Provinces increased infant mortality by [.3 log odds, P <0.05: CL: 95%] in 2007and 2001-02 respectively. The interaction between ME and Southern Province reduced infant mortality (the most) from 1992 to 2013/14 by [.3, .3, .2, .0, and .1 log odds, P < 0.05: CL: 95%] respectively. The interaction between ME and place of residence was found to have no significant effect on IM throughout the period 1992 to 2018. Throughout the period under study, the interaction between ME and contraceptive use was found to reduce infant mortality by [.19, .2, .2, .1, .1 and .07 log odds, P < 0.05: CL: 95%] respectively.
Overall, the interaction between ME and maternal age significantly contributed to the model in 1992, 1996, 2007 and 2013/14. However, holding all other variables constant, from 1992 to 2018 the interaction between ME and maternal age group 35-39 reduced IM the most by [.5, .5 and .02 log odds, P < 0.05: CL: 95%] in 2007, 2013-14 and 2018 respectively. The interaction between ME and maternal age groups 20-24 and 40-44 reduced IM the least by [.02 and.5 log odds, P <0.05: CL: 95%] in 2013-14 (for age group 40-44) and 2018 (for age group 20-24) respectively. Additionally, from 1992 to 2018, the interaction between ME and birth interval reduced IM by [.2, .2, .2, .1, .2 and .12 log odds, P < 0.05: CL: 95%] respectively.
Holding all other variables constant, for the periods from 1992 to 2018, the interaction between ME and source of drinking water and, type of toilet facility did not significantly affect infant mortality.
From 1992 to 2018, holding other variables constant, the interaction between ME and breastfeeding, significantly reduced infant mortality by [1.6, 1.8, 1.8, 1.7, 7.3 and .46 log odds, P < 0.05: CL: 95%] respectively. The interaction between ME and place of delivery and tetanus injection did not affect infant mortality. The interaction between ME and antenatal care, significantly reduced infant mortality by [ .2, .00, .2 .2, .1 and .16 log odds, P < 0.05: CL: 95%] from 1992 to 2018 respectively.
Third Model (Final Model)
Table 4: Binary Logistic Regression Final Model
Variables in ZDHS
|
ZDHS1992
|
ZDHS1996
|
ZDHS2001/2
|
ZDHS2007
|
ZDHS2013/14
|
ZDHS2018
|
B
|
Sig.
|
B
|
Sig.
|
B
|
Sig.
|
B
|
Sig.
|
B
|
Sig.
|
B Sig.
|
ME/ M.E * Social-Economic And
Demographic Factors
|
|
ME Single Years
|
-.01
|
.041*
|
-.05
|
.051*
|
-.18
|
.025*
|
-.32
|
.346
|
-7.00
|
.911
|
-.186 .120
|
ME By Used Contraceptive (1)
|
-.65
|
.001*
|
-3.2
|
.000*
|
-1.7
|
.000*
|
-.62
|
.016*
|
-.91
|
.025*
|
-.071 .000
|
ME By Pre-birth_Interval_24+(1)
|
-.80
|
.000*
|
-.75
|
.001*
|
-.56
|
.002*
|
-.98
|
.016*
|
-.57
|
.007*
|
-.101 .000
|
M.E * Mother and Child Health And
Water and Sanitation Factors
|
|
ME By Breastfed (1)
|
-2.49
|
.000*
|
-.91
|
.000*
|
-3.81
|
.000*
|
-1.88
|
.000*
|
-8.62
|
.019*
|
-.434 .000
|
ME By Received Antenatal Care (1)
|
-1.23
|
.000*
|
-.07
|
.002*
|
-.72
|
.000*
|
-.94
|
.000*
|
-.65
|
.046*
|
-.153 .000
|
Intercept
|
6.03
|
.000*
|
2.65
|
.000*
|
.61
|
.000*
|
1.23
|
.000*
|
.19
|
.001*
|
-3.111 .000
|
When ME single years was added to the third-final model further reductions in the coefficients were observed compared to the coefficients in the first and second models (table 4). Variables which significantly contributed to the reduction in infant mortality throughout the period under study, (in the second model), including maternal education as the key independent variable, were advanced to the final (third) model. Table 4 shows the model with ME, ME by use of contraceptives (1), ME by preceding birth interval 24+ months (1), ME by breastfeeding (1) and ME by received antenatal care (1). Apart from ME which only influenced to the reduction in IM from 1992 to 2001-02, all the interactions in the model significantly reduced IM throughout the period under study.
Goodness-of-fit of the Model
From 1992 to 2018 model 3 illustrated 91.4%, 92.4%, 93.7%, 95.1%, 96.5% and 96.2% accuracy of data prediction, respectively. The model indicates that variations of 8.4% (Cox & Snell R²: 0.084) 12.9% (Cox & Snell R²: 0.129) 14.7% (Cox & Snell R²: 0.147), 13.8% (Cox & Snell R²: 0.138), 21.2% (Cox & Snell R²: 0.212) and 6.9% (Cox & Snell R²: 0.212) correspondingly, (in IM) were explained by the model. Hosmer and Lemeshow goodness of fit model of [.058, .098, .276, .091, .070 and .197 (P > 0.05)], respectively, showed that the model estimates were acceptable and the -2 Log Likelihood values of 1512.26a, 1355.74b,1098.33c, 1172.71d, 1083.017e and 2656.024f, correspondingly, gave an indication of a good model fit.