The contribution of spatial analysis to understanding HIV/TB mortality in children: a structural equation modelling approach

Figures & data

Fig. 1.  Agincourt sub-district in 2004 showing villages (grey shading) schools (flags), and health facilities (circled grey crosses).

Fig. 2.  Conceptual framework for modelling childhood HIV/TB-related mortality [adapted from Meade 24]. The red arrows reflect a Fig. 2. Conceptual framework for modelling childhood HIV/TB-related mortality [adapted from Meade (24)]. The red arrows reflect a direct effect on mortality and the purple (explanatory variables) and orange (spatial random effects) arrows depict an indirect effect.

Fig. 3.  Child HIV/TB mortality: posterior adjusted odds ratio map, Agincourt sub-district, 2004.

Table 1. Child HIV/TB mortality: descriptive statistics and results of two multiple logit regression analyses, Agincourt sub-district 2004

Variable	Summary number (%)	Univariate logit coefficient (standard error)	Non-spatial multiple logit coefficient (standard error)	Spatial Multiple logit coefficient (standard error)
Age of child in years^1	2.77±1.20	−0.617 (0.131)**	−0.601 (0.146)**	−0.630 (0.137)**
Gender
Female	3,379 (50.52)	0	0	0
Male	3,310 (49.48)	0.556 (0.283)*	0.698 (0.303)*	0.641 (0.291)*
Child's parents former refugee
Yes	2,388 (35.71)	0
No	4,299 (64.29)	−0.137 (0.280)
Mother's age in years^1	28.70±7.59	−0.003 (0.018)
Cumulative household deaths^2	0.68±0.96 (0–6)	0.684 (0.086)**
Mother deceased by 2004
No	6,634 (99.13)	0	0	0
Yes	58 (0.87)	1.508 (0.732)*	1.504 (0.867)*	1.386 (0.707)*
Minimum distance to health facility^1	2.33±1.61	0.078 (0.083)
Household size^1	4.88±2.92	−0.011 (0.048)
Gender of household-head
Female	2,434 (36.37)	0	0	0
Male	4,258 (63.63)	−0.865 (0.418)**	−0.880 (0.302)**	−0.895 (0.30)**
Parity^2	2.12±1.43 (1–10)	0.085 (0.090)
Socio-economic status quintiles
Most poor	1,823 (27.24)	0
Very Poor	1,073 (16.03)	−0.112 (0.393)
Moderately poor	1,120 (16.74)	−0.516 (0.444)
Poor	1,316 (19.67)	−0.425 (0.406)
Least poor	1,360 (20.32)	−0.458 (0.406)
Non-spatial variance			0.10; 95% BCI (0.001; 0.68)	0.04; 95% BCI (0.001–0.40)
Spatial variance estimate				0.04; 95% CI (0.001–0.28)
Deviance information criteria			457.26±15.56	555.87±11.65

*p<0.05; **p<0.01.

¹mean±standard deviation, ²mean±standard deviation (range). 95% BCI=95% Bayesian credible interval.

Fig. 4.  Conceptual framework pathways results for Agincourt sub-district child mortality determinants in 2004. Key for variable names: very poor versus most poor (vpoor_mostpoor), moderately poor versus most poor (modpoor_mostpoor), least poor versus most poor (leastpoor_mostpoor), poor versus most poor (poor_mostpoor), age of child in years (age_ch), household size (hh_size), cumulative household deaths (cum_hh_deaths), gender of household head (hh_gender), mother age in years (age_mum), mother deceased by 2004 (mother_deceased), child born to former refugee parents (refuge), gender (gender), minimum distance to health facility (min_dist) parity (parity), child deceased in 2004 (dead). Other: the arrows pointing from the exogenous (explanatory) to endogenous (dependent) variables and the error terms ( ɛ ) placed on all five endogenous variables.

Table 2. Direct, indirect and total effects of child HIV/TB mortality in 2004, Agincourt sub-district

	Direct effects on the deceased child as shown in Fig. 2 conceptual framework
	Gender of household-head	Child a former refugee	Child deceased	Mother deceased by 2004	Age of child in years	Indirect effects on child HIV/TB death	Total effects on child HIV/TB death
Very poor vs most poor	−0.029 (0.019)	0.140 (0.015)***				2.6×10^−4 (1.6×10^−4)^ns!	2.6×10^−4 (1.6×10^−4)^ns
Moderately poor vs most poor	−0.059 (0.020)***					4.5×10^−4 (2.1×10^−4)**	4.5×10^−4 (2.1×10^−4)**
Poor vs most poor	−0.052 (0.020)***					4.0×10^−4 (2.0×10^−4)**	4.0×10^−4 (2.0×10^−4)**
Least poor vs most poor	−0.111 (0.018)***					8.4×10^−4 (3.0 ×10^−4)***	8.4×10^−4 (3.0×10^−4)***
Household size vs most poor	0.008 (0.002)***					−5.9×10^−5 (2.5×10^−5)**	−5.9×10^−5 (2.5×10^−5)**
Cumulative household deaths	−0.083 (0.006)***					6.4×10^−4 (2.1×10^−4)***	6.4×10^−4 (2.1×10^−4)***
Minimum distance to health facility		0.031 (0.004)***				7.72×10^−6 (4.88×10^−6)^ns	7.72×10^−6 (4.88×10^−6)^ns
Male household-head			−0.0076 (0.002)***				−0.0076 (0.002)***
Mother deceased by 2004			0.027 (0.012)*				0.027 (0.012)**
Age of child in years			−0.0046 (0.001)***				−0.0046 (0.001)***
Gender			0.0052 (0.0023)**				0.0052 (0.0023)**
Mother's age (years)				0.001 (0.000)***		1.9×10^−5 (9.99×10^−6)*	1.9×10–05 (9.99×10^−6)*
Parity				−0.003 (0.001)**		−7.0×10^−5 (4.0×10^−5)^ns	−7.0×10–05 (4.0×10^−5)^ns
Parents former refugee					−0.054 (0.032)*	−2.5×10^−4 (1.5×10^−4)*	−2.5×10^−4 (1.5×10^−4)*
Intercept	1.719 (0.017)***	1.537 (0.011)***	−0.002 (0.014)	0.995 (0.005)***	2.814 (0.054)***

*p<0.10; **p<0.05; ***p<0.01; ns: not significant; !: indirect effects computed as the product along the related pathways, i.e −0.029×−0.0076+0.14×−0.054×−0.0046=2.55×10⁻⁴≈2.6×10⁻⁴ .

Meade MS. Medical geography as human ecology: the dimension of population movement. Geogr Rev. 1977; 67: 379–93.10.3402/gha.v6i0.19266.

 Google Scholar