Housing materials as predictors of under-five mortality in Nigeria: evidence from 2013 demographic and health survey

The study was carried out in Nigeria, with a human population figure of above 180 million. Nigeria is characterised with high infant and childhood mortality. The maternal mortality ratio is also one of the highest among developing nations worldwide [36]. High fertility is a key factor for high under-five mortality across countries [37]. In Nigeria, the total fertility rate remains high (TFR = 5.5) [36]. Unfortunately, the contraceptive prevalence rate of 10% is still considered as low [36]. Nigeria is seen on the global page as poverty stricken country with a striking gap between the rich and the poor. Administratively, Nigeria is made up of 36 states and the Federal Capital territory. Each state has local government areas (third level of government) which are further divided into localities. The country is also stratified into 6 geopolitical zones; North-Central, North-East, North-West, South-East, South-South and South-West. The reason being that Nigeria comprised approximately 400 ethnic groups and 450 dialects. There was the need for the government to unify similar groups into zones for effective allocation of resources. The inhabitants in each of the geopolitical zones are homogeneous and share similar socio-cultural characteristics and unique in other health-related characteristics like access to health care, environment, housing system etc.

The Demographic and Health Surveys (DHS) and alike surveys that collect information on birth outcomes from women are one of the main bases for data collection on under-five mortality estimation in developing countries like Nigeria, where reliable and adequate vital registration system is lacking [17]. The 2013 Nigeria Demographic and Health Survey data were used for this study. The original data was collected to provide information on the population, health and fertility levels in Nigeria. A 3-stage stratified cluster design consisting of 904 clusters (372 in urban areas and 532 in rural areas) was used for sample selection. A representative sample of 40,680 households was selected for the survey and a fixed sample of 45 households was selected in each cluster. In this study, the sample used was 18,516 women of reproductive age who had given birth in the past 5 years prior the survey. However, the attention was focused on the survival status of the index child, the most recent delivery the women had in the past 5 years prior the study.

The dependent variable was childhood mortality and this was captured with the question on the survival status of the most recent birth (dead = 1 or alive = 0) in the last 5 years preceding the survey. Under-five mortality was defined as the death of a live-born child before its fifth birthday [27]. The main predictor was the type of housing materials. This was obtained as aggregate score based on information from roof materials (Improved – cement, roofing sheets, ceramic tiles; Unimproved – natural, no roof, palm leaf, sod, rudimentary, rustic mat, bamboo, cardboard), wall materials (Improved – cement, stone with cement, cement blocks, bricks; Unimproved – natural, no wall, palm/trunks, dirt, rudimentary, bamboo with mud) and floor materials (Improved – cement, ceramic tiles, vinyl asphalt strips, parquet, polished wood, finished; Unimproved – natural, earth, sand, dung, rudimentary, wood planks, palm, bamboo, others). The improved categories assumed a score of 1 while unimproved scored [36, 38]. The overall score (13-point maximum and 0-point minimum) for a woman was disaggregated into three categories: inadequate (<50% of the overall score), moderate (50% ≤ x < 75% of the overall score) and adequate (75% ≤ x ≤ 100%) of the overall score). They are based on quartile classification: 3rd quartile is 75% and 2nd quartile is 50%.

Other independent variables include mother’s age, highest educational level, religion, ethnicity, marital status, place of residence, region, wealth index and media exposure. others are; number of antenatal visits, tetanus injection, gender of a child, size at birth, birth order, preceding birth interval, prenatal care provider, delivery assistance, place of delivery, cooking fuel (Clean – electricity, liquefied petroleum gas, natural gas, biofuel; Unclean/biomass – coal, charcoal, wood, straw/shrubs/grass, agricultural crop, animal dung, kerosene), source of water (Improved - piped into dwelling, public tap, borehole, protected well and spring, rain water and bottle water; Unimproved – other sources not listed as improved sources) and toilet facility (Improved – flush/pour flush to piped sewer system, septic tank or pit latrine, ventilated improved pit latrine, pit latrine with slab, composting toilet; Unimproved – other toilet types not listed as improved). We adapted the groupings of environmental factors documented in the 2013 Nigeria National Demographic Health Survey (NDHS) and the 2010 WHO and UNICEF document on progress on sanitation and drinking water [36, 39]. These variables were used as covariates during multivariate analysis to determine the association between housing materials and childhood mortality.

The dataset was weighted before data analyses. The weight is an inflation factor which extrapolates the sample to the target population [40]. Analysis of data was done at bivariate and multivariate levels using Chi-square, Cox proportional hazard model and Brass 2-parameter model. The Chi-square test was used to examine the relationship between child’s survival status and the independent variables at 5.0% level of significance. At the multivariate level of analysis, Cox proportional hazard model was used to detect the predictors of childhood mortality. The proportional hazards model assumes that the time to event and the covariates are related as; \( { \log}_{\mathrm{e}}\left\{\frac{\upgamma_{\mathrm{i}}\left(\mathrm{t}\right)}{\upgamma_0\left(\mathrm{t}\right)}\right\}={\upbeta}_0+{\upbeta}_1{\mathrm{x}}_{\mathrm{i}1}+{\upbeta}_2{\mathrm{x}}_{\mathrm{i}2}+\dots +{\upbeta}_{\mathrm{p}}{\mathrm{x}}_{\mathrm{i}\mathrm{p}} \). Where; γ_i(t) is the hazard rate for the i^th case of a woman having lost her child before age five; γ₀(t) is the baseline hazard at time t when the death of the child occurs; β_j is the value of the j^th regression coefficient; x_ij is the value of the i^th case of the j^th covariate.

Further analysis using an indirect method to ascertain the influence of housing materials on childhood mortality was done. Brass [41] reported that the probability of dying between birth and exact age (a) can be estimated as q(x) = k(x) × D(x) where D(x) is the number of dead children in each age group and k(x) is a multiplier and is estimated as; k(x) = a(i) + b(i)(P₁/P₂) + C(i)(P₂/P₃). The number of non-surviving children for women in age groups 15–20, 20–25, 25–30,…, 45–50 were used to calculate childhood mortality at exact ages (q(x)); 1, 2, 3, 5, 10, 15 and 20. Regression equations which relate the multipliers k(x) to indices of fertility schedule were formulated from mathematical simulations [42]. The time reference to which the q(x) values refer were also formulated. Due to limitations in the estimates produced by this method, we further adjusted the childhood mortality using Brass 1-parameter model (Y = ∝ + βY_s) where β = 1. This procedure made use of logit equation: \( \mathrm{logit}\left\{\mathrm{q}\left(\mathrm{x}\right)\right\}=\frac{1}{2} \log \left\{\frac{1-{\mathrm{l}}_{\mathrm{x}}}{{\mathrm{l}}_{\mathrm{x}}}\right\} \). This was transformed to Brass relational system of life tables; logit{l(x)} = ∝ + βlogit{l_s(x)} usually transcribed as Y = ∝ + βY_s. The logit relational system smoothens the estimated values of l(x) (survival probability) in contrast to the values from the model life-table. Therefore, if β = 1, \( \widehat{\propto}=\mathrm{Y}\left(\mathrm{x}\right)-{\mathrm{Y}}_{\mathrm{s}}\left(\mathrm{x}\right) \), thus generating; \( \mathrm{Y}(1)={\widehat{\propto}}_1+{\mathrm{Y}}_{\mathrm{s}}(1);\mathrm{Y}(2)={\widehat{\propto}}_2+{\mathrm{Y}}_{\mathrm{s}}(2);\dots; \mathrm{Y}(20)={\widehat{\propto}}_{20}+{\mathrm{Y}}_{\mathrm{s}}(20) \). Estimate of \( \widehat{\propto} \) was obtained from the average values of x = 2, x = 3 and x = 5 which produce reliable values of l(x). Therefore, if \( \overline{\mathrm{Y}} \) is the average of Y(2), Y(3) and Y(5) and \( {\overline{\mathrm{Y}}}_{\mathrm{s}} \) is the average of Y_s(2), Y_s(3) and Y_s(5), then \( \widehat{\widehat{\propto}}=\overline{\mathrm{Y}}-{\overline{\mathrm{Y}}}_{\mathrm{s}} \). This produced the adjusted survival probabilities at childhood. The infant and under-five mortality were estimated using the models; \( \mathrm{infant}\ \mathrm{mortality}\ \mathrm{rate}=\frac{{\mathrm{q}}_0}{1-0.7\times {\mathrm{q}}_0} \) and \( \mathrm{under}-5\ \mathrm{mortality}\ \mathrm{rate}=\frac{2{\mathrm{q}}_5}{2\times \left(1-{\mathrm{q}}_5\right)} \) [43].

All analysis were done using SPSS version 20.0 and Excel software.

At the multivariate level of analysis, as shown in Table 4, four models were generated. The first model included only the key independent variable (Housing materials) while the housing materials and other environmental factors were included in the second model. The third model involved the housing materials, other environmental factors and health/child-related factors while the socio-demographic and all other factors were included in the last model. The models were built in this manner so as to explore the possible interaction of the housing material and childhood mortality while other factors were used as covariates.

The data revealed that the type of housing materials was significantly associated with childhood mortality across the first three models. For instance, in model 1, the hazard ratio of childhood mortality was 1.23 (C.I = 1.24–1.71, p < 0.001) and 1.46 (C.I = 1.02–1.47, p < 0.001) higher among children who lived in houses built with moderate and inadequate housing materials respectively than those living in houses built with adequate housing materials. This pattern was found across the models but in the last model, housing materials were found to be insignificantly associated with childhood mortality showing that socio-demographic factors are strongly important in the relationship between childhood mortality and housing materials. The identified predictors of childhood mortality are; the size of the child at birth, preceding birth interval, prenatal care provider, place of residence and education. The likelihood of childhood mortality was 0.82 (C.I = 0.72–0.92; p < 0.001) lower among females than males, among children who were average (HR = 0.76; C.I = 0.64–0.90) or larger (HR = 0.71; C.I = 0.60–0.84, p < 0.001) than average in size at birth than those who were small. The data further show that childhood mortality hazard was 1.35 (C.I = 1.13–1.62, p < 0.001) more in the rural areas than the urban.

Globally, scientific evidence on the associations between housing and health has grown considerably in recent years. This information has been used as preventive measures that are linked to better housing condition and human health. From a policy point of view, there is a need for formulation of guidelines on healthy housing by the government and other relevant stakeholders to assist in averting a wide range of preventable illnesses that are often associated with poor housing conditions. This position has earlier been echoed at the international consultation of 40 specialists drafted from 18 countries by the WHO on 13–15 October 2010 in Geneva [78]. In this regard, a national standard on “healthy housing would enable action that is scientifically based, and protects and advances the health of under-five children in households” [78]. The existing national housing policy that sees housing as a determinant of health and social stability should be strengthened in Nigeria. However, the most healthy and safest house will not support the health of its inhabitants if not maintained and used in a manner that promotes sound health. Therefore, individuals should be advised to use good housing materials while building their houses. Even the poor can get good housing materials from the little resources they have if properly managed.

The major shortcoming of this study is the cross-sectional nature of the data where information on housing materials was based on verbal reporting instead of the use of check listing approach which could have revealed the true state of housing condition in the study area. Although this limitation was minimized since data were collected at the individual homes, the data collectors were able to assess the structures and match their observations with the report from the respondents. Also, the reporting of under-five children’s death by women may include omission of few dead children. Culturally, in sub-Saharan Africa, people dislike talking about their dead children as they see this as a bad event they don’t want to remember. However, in this study, we used an indirect method that accounts for such limitation for the estimation of childhood mortality across the classes of housing materials.