• Users Online: 197
  • Print this page
  • Email this page

 Table of Contents  
Year : 2020  |  Volume : 1  |  Issue : 1  |  Page : 15-20

Spatial clustering and impact of household characteristics on under-five mortality in India: A secondary data analysis

1 Department of Community and Family Medicine, AIIMS, Rishikesh, Uttarakhand, India
2 Department of Architecture and Planning, IIT, Roorkee, Uttarakhand, India
3 Department of Nephrology, AIIMS, Rishikesh, Uttarakhand, India

Date of Submission06-Jun-2020
Date of Decision17-Jun-2020
Date of Acceptance18-Jun-2020
Date of Web Publication20-Jul-2020

Correspondence Address:
Dr. Jatin Chaudary
Department of Community and Family Medicine, AIIMS, Rishikesh, Uttarakhand
Login to access the Email id

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/JME.JME_40_20

Rights and Permissions

Background: Under-five mortality rate (U5MR) is one among the best social indicators that identifies social development and well-being. Various risk factors are seen contributing to its increase. Aim: The present correlational study aims to identify the spatial clustering of U5MR in different states and union territories (UTs) of India and determines its association with household characteristics. Patients and Methods: The present study incorporates primary data for 29 states and 2 UTs (Delhi and Chandigarh) of India from the National Family and Health Survey-4 (2015–2016) for secondary data analysis. The data include the outcome variable which was the U5MR and predictor variables such as households with electricity; improved source of drinking water; toilet facility; solid fuels being used for cooking; anyone smoking and living in a pucca house. Primary data were analysed using GeoDa software, employing Univariate Local Indicators of Spatial Association, Spearman's correlation coefficient, ordinary least square (OLS) and spatial error model (SEM). Results: The study showed significant spatial clustering of high U5MR in six states and one UT, namely Rajasthan, Madhya Pradesh, Chhattisgarh, Uttar Pradesh, Bihar, Jharkhand and Delhi and clustering of lower U5MR in southern states of Tamil Nadu and Karnataka. The Spearman's correlation showed a significant positive association of U5MR with households using solid fuel for cooking and negative association with households using electricity, with toilet facility and Living in a pucca house. The OLS and SEM spatial regression models model showed an association between households with toilet facility and in which anyone smokes at home with under-five mortality. Conclusions: U5MR shows a significant clustering geographically. This mortality indicator is influenced by the external environment such as household characteristics.

Keywords: India, spatial analysis, under-five

How to cite this article:
Chaudary J, Akshay S, Reddy D S, Sharma A, Jha N, Kaushal P, Malik A. Spatial clustering and impact of household characteristics on under-five mortality in India: A secondary data analysis. J Med Evid 2020;1:15-20

How to cite this URL:
Chaudary J, Akshay S, Reddy D S, Sharma A, Jha N, Kaushal P, Malik A. Spatial clustering and impact of household characteristics on under-five mortality in India: A secondary data analysis. J Med Evid [serial online] 2020 [cited 2022 Nov 30];1:15-20. Available from: http://www.journaljme.org/text.asp?2020/1/1/15/290139

  Introduction Top

The single best indicator of child health in low- and middle-income countries is the mortality rates. They are often also used as indicators of general social and economic development. In recent years, the most widely used measure of child mortality has been the under-five mortality rate (U5MR). It is defined as the probability of dying between live birth and age 5 years.[1]

Globally, there was a reduction of 50% in 1990–2016 in U5MR with significant declining rates since 2000. There has been encouraging advancement in the reduction of child mortality rates, but progress has been uneven across and within the countries. The fourth Millennium Development Goal (MDG) aimed at reducing the U5MR by two-thirds between 1990 and 2015.[2] This goal was not achieved in most countries, including India, which according to the National Family and Health Survey 4 (NFHS-4), had a U5MR of 49.7, which is higher than the target set in MDG of 42/1000 live births by 2015. There are still a lot of children facing very low odds of surviving their first 5 years of life. Since MDGs were not achieved by their destined time, to continue past efforts of reducing child mortality and complete the unfinished MDG agenda to improve child survival, the Sustainable Development Goals were called upon. These goals call for an end to preventable deaths of newborns and children by 2030. The major agenda to be followed by all countries is to reduce U5MR to at least 25 deaths per 1000 live births.[3]

Healthy housing can be defined as dwellings and premises that are designed to provide shelter and protection from hazards resulting from the physical and social environments and are maintained in ways that support the health of its occupants.[4] An adverse household environment impacts the physical and mental health of its occupants and also deteriorates the social and economic well-being and quality of life of individuals. Among the occupants of an unhealthy household, children are predominantly susceptible to housing-related hazards than adults since they spend most of their time inside the house. They require a higher amount of air inhalation than adults, and their organs are not fully developed, which puts them at a higher risk of having any morbidity or mortality due to unhealthy housing.[5]

Public health research has started focusing on understanding the impact of different geographical regions or space as possible contextual factors on the health of the population. There are evident systemic and potentially remediable differences in one or more aspects of health across populations disaggregated by social, economic, demographic, and geographical characteristics in India.[6] With the help of geospatial analysis, it is now possible to account for the effect of spatial diffusion in particular health parameters. It has also helped in the past in understanding several contiguous and contagious biophysical and geographical variables in the causation of the disease.[7] Thus, the present study performs secondary data analysis for the state-wise information gathered in NFHS-4 (2015–2016) data and aims to identify the geospatial pattern of U5MR in different states and union territories (UTs) of India. Furthermore, the study determines its association with household characteristics.

  Patients and Methods Top

Ethical consideration

The study was conducted using anonymous public use data set with no identifiable information on the survey participants; therefore, no ethical approval is required for this work.

Study design

The present correlational study carried out secondary data analysis for the information gathered in NFHS-4 (2015–2016). It executed the concept of Geographic Information System (GIS), in ascertaining the spatial distribution of U5MR in India.

The GIS as defined by Burrough in 1986 as 'Set of tools for collecting, storing, retrieving at will, transforming and displaying spatial data from the real world for a particular set of purposes'. It describes the first law of Geography also known as Tobler's Law, which states, 'Everything is related to everything else, but near things are more related than distant things'. It means that virtually all human interactions, natural and man-made features, resources, and populations have a geographic component. The development of choropleth maps, which permits the dynamic visual examination of a dependent variable and potential predictor variables, highlights the intersection between GIS, statistics, and visualisation in an application to generate well-informed, relevant hypotheses.[8] Thus, in this study, the dependent variable, i.e., U5MR was visualised in 29 states and 2 UTs of India, and the states/UTs showing higher and lower U5MR were stated along with the significant map showing spatial dependence. The association was found out between the dependent variable and the predictor variables, such as household characteristics.

Source of data/outcome/predictor variables

The primary data for the present study was taken from the NFHS-4 (2015–2016). The NFHS is a set of multi-dimensional large-scale surveys conducted periodically on a representative sample of households in India; these surveys deliver essential data on health and family welfare and emerging issues in this area for both the national and the state level. NFHS-4 collected information from 601,509 households.[9] For the present study, the outcome variable selected was UF5MR and predictor variables were mainly household characteristics that include households with (i) electricity; (ii) improved source of drinking water; (iii) toilet facility; (iv) solid fuels being used for cooking; (v) anyone smoking and living in a pucca house. Since these were the household characteristics for which data was available in the NFHS-4 survey.[10]

Operational definitions as given in National Family and Health Survey-4 to define the predictor variables[10]

  • Improved source of drinking water - Includes piped water, public taps, standpipes, tube wells, boreholes, protected dug wells and springs, rainwater, and community reverse osmosis (RO) plants
  • Solid fuels used for cooking - Includes coal/lignite, charcoal, wood, straw/shrubs/grass, agricultural crop waste, and dung cakes
  • Pucca house - These are made with high-quality materials throughout, including the floor, roof and exterior walls.

Statistical analysis

GeoDa software was used to implement various exploratory spatial data analysis, including data manipulation, data mapping, and spatial regression analysis.[11] In this study, spatial autocorrelation statistics were applied. To implement the statistics, spatial weight had to be generated. GeoDa software was used to compute spatial weights. It could either be based on contiguity from polygon boundary files or based on the distance between points. In this study, contiguity-based spatial weights were used. There are two types of spatially contiguous weights, rook's weight, and queen's weight. In this study, rook's weight was used for estimating all the geospatial statistics and geo-spatial regressions. It uses common boundaries to define spatial neighbours.[12]

Spatial autocorrelation measures the degree to which data points are similar or dissimilar to their spatial neighbours. It is measured using Moran's-I, which is the Pearson coefficient measure of spatial autocorrelation. A positive spatial autocorrelation indicates that points that have similar attribute values are closely distributed in space, whereas negative spatial autocorrelation indicates that closely associated points are more dissimilar. Its values range from −1, which indicates perfect dispersion to +1 indicating perfect correlation. A zero value indicates a random spatial pattern.

The Univariate Local Indicators of Spatial Association (LISA) tells about the clustering present in the data. The LISA functionality in GeoDa offers two important options, cluster maps and significance maps. The cluster map shows those locations with a significant local Moran statistic as classified by the type of spatial correlation: Bright red color shows a high-high association, bright blue is for low-low, light blue shows low-high and light red for high-low. The high-high and low-low suggest clustering of similar values, whereas high-low and low-high locations indicate spatial outliers. On the other hand, the significance map is a special map showing those locations with a significant local Moran statistic in different shades of green depending on the significance level. The significance levels are shown as P < 0.05, P < 0.01, P < 0.001, and P < 0.0001.[12]

SPSS (Statistical Package for the Social Sciences) version 23 was used to find Spearman's correlation coefficient. The coefficient was used to find the strength and direction of the relationship between the outcome and the predictor variables.

Spatial Regression models were applied to generate residual maps. There are three types of regression models in GeoDa to examine the relationship between the outcome variable and a set of predictors: ordinary least square (OLS), spatial lag model (SLM), and spatial error model (SEM).[13],[14] In this study, OLS and SEM models were run, and residual maps were generated to model the spatial clustering of the outcome variable that is not explained by the set of predictor variables.

  Results Top

The Univariate LISA results for U5MR are presented in Figure 1. [Figure 1]a represents the geographic clustering of high U5MRs in northern, central, and eastern India that includes six states – Rajasthan, Madhya Pradesh, Chhattisgarh, Uttar Pradesh, Bihar, Jharkhand, and one UT – Delhi. On the other hand, two southern coastal states (Karnataka and Tamil Nadu) show comparatively lower clustering of U5MR. [Figure 1]b shows the significance map for the cluster map in [Figure 1]a with a significant local Moran statistic in different shades of green depending on the significance level.
Figure 1: *These maps do not portray the political/administrative boundaries of India. Univariate Local Indicators of Spatial Association (cluster and significance) maps depicting spatial clustering and spatial outliers of under-five mortality across 29 states and 2 Union Territories of India, 2015–2016. (a) Univariate Local Indicators of Spatial Association cluster map of under-five mortality rate across 31 regions in India, 2015–2016, (b) Univariate Local Indicators of Spatial Association Significance map of under-five mortality rate across 31 regions in India, 2015–2016. Jatin Chaudary, Department of community and family medicine. 05 May, 2020

Click here to view

[Table 1] shows a positive significant correlation between U5MR and households using solid fuel for cooking. Negative significant correlations were seen for those with electricity, improved sources of drinking water, toilet facility, and pucca houses.
Table 1: Correlation between Under-five mortality and household characteristics in India (2015-2016)

Click here to view

[Table 2] shows that there was geospatial clustering in U5MR and household characteristics such as households with electricity, using solid fuel for cooking, in which anyone smokes at home and pucca house. The highest geospatial clustering was observed for those living in a pucca house (Moran's I = 0.581). Least geospatial clustering was, however, observed for those with improved sources of drinking water.
Table 2: Moran's I statistics for under-5 mortality and household characteristics in India, 2015-2016

Click here to view

[Table 3] shows the output of OLS and SEM models. Both models show a significant association between U5MR and households with toilet facilities. Other parameters, such as AIC, show minimum value in the SEM model. The value of R square is also seen increasing in the SEM model from 0.599 to 0.653; therefore, the SEM model was preferred.
Table 3: Ordinary least square and spatial error model to assess the association between under-five mortality and household variables, India, 2015-2016

Click here to view

[Figure 2]a and [Figure 2]b shows residual maps for OLS and SEM spatial regression models, respectively. It can be observed that residuals show clustering in the OLS model in [Figure 2]a, whereas they are randomly distributed in the SEM model, as shown in [Figure 2]b. The value of Moran's I also reduced from 0.147 in OLS model to-0.032 in SEM model.
Figure 2: *These maps do not portray the political/administrative boundaries of India. Residual maps of OLS and spatial error models for under-five mortality, India, 2015–2016. (a) Univariate Local Indicators of Spatial Association cluster map (Moran's I = 0.147) plotting residuals of OLS regression model. (b) Univariate Local Indicators of Spatial Association cluster map (Moran's I = −0.032) plotting residuals of spatial error regression model. Jatin Chaudary, Department of community and family medicine. 05 May, 2020

Click here to view

  Discussion Top

In the present study, six northern states of India showed significant spatial distribution of U5MR in India. These include Rajasthan, Madhya Pradesh, Chhattisgarh, Uttar Pradesh, Bihar and Jharkhand. These six states are also the part of eight socioeconomically backward states known as the empowered action group (EAG) states that lag behind in the demographic transition and have high mortality rates in the country.[15] These states in India have been recognised as poor-performing states since the 1980s, based on their low health indicators. EAG is the group of most backward and deprived states that need greater health-care facilities at an affordable cost.[16] EAG states are already receiving special funds from NHM in a 90:10 ratio since 2000 compared to other non-EAG states in 60:10 ratio; still, after the completion of two decades, EAGs continue to lag. This could be due to barriers in extensive coverage of integrated packages and inadequate operational management, especially at the district level.[7]

The southern states of India, specifically Karnataka and Tamil Nadu, showed a spatially significant geographical distribution of low U5MR. This could be attributed to the better health facilities in these states. Among the high-performing states in India, Tamil Nadu is often ranked the best after Kerala in terms of various health indicators. The state has low mortality rates in addition to the effective health-care infrastructure; health workforce and provides good-quality health services at an affordable cost.[17],[18] The southern states have better health facilities as well as infrastructure and have higher densities of health workers to cater to the need of the population.[19]

The study showed that there was a positive association between households using solid fuel for cooking with under-five mortality. Most of the deaths occur in low- and middle-income countries are attributed to Indoor air pollution. The use of solid fuels for household indoor cooking could result in high levels of chemical components (e.g., carbon monoxide, sulphur oxides, nitrogen oxides, particulates, benzene, formaldehyde, polyaromatic compounds, arsenic, lead) whose effect on health is not well understood.[20] Children under 5 years of age are one of the vulnerable groups most likely to experience ill health caused by solid fuel use, as they accompany their mothers while they are cooking. Studies done in the past show that children <5 years of age living in homes using solid fuels for cooking are at a greater risk of dying from acute respiratory illnesses.[21],[22]

The present study revealed that households with electricity show low under-5 mortality. Literature shows that expanding access to electricity accompanied by reliability, measured using hours of supply and voltage stability, can have much larger welfare effects, including impacts on health. Various studies have shown that the availability of electricity is one of the determinants of receiving health information and the utilisation of health services.[23],[24]

In the present study, the household living in a pucca house showed low U5MR. The building materials from which the house is built is a reflection of the well-being of household occupants. The quality of the material used to build the house is mostly linked with its durability and the health of its occupants.[25] Findings from a study in Nigeria show that the risk of dying before the age of 5 years was higher among children who lived in houses built with inadequate housing materials than those in moderate and adequate housing materials.[5] In another study, households with floors made of mud or sand were more likely to experience under-five mortality than households with cement residence floors.[26]

The study showed an association between a household with toilet facility and under-five mortality using OLS and SEM model. The unimproved sources of water and poor sanitation practices have been implicated in the death of a child every 15 seconds from the diarrhoeal disease.[27] According to the WHO, there are annually 1.7 million morbidities and 760,000 under-five children's deaths globally due to diarrhoea. It remains the second-leading cause of death among children under-five globally.[5]

Limitation of the study

The present correlational study design cannot control for the possible effects of potential confounders, other than household characteristics, thus representing average U5MR rates for an overall population. SEM model was used to reduce residuals effects but unable to omit spatial effects in data completely.

  Conclusions Top

This study finds spatial distribution and association of household characteristics and U5MR in India. Unfavourable and adverse housing conditions can trigger a range of diseases, including lung diseases, neurological disorders, mental and behavioural dysfunction, mostly affecting children. Housing characteristics can be an important predictor of U5MR and its impact in the increase/decrease of this important mortality indicator cannot be neglected.

Financial support and sponsorship


Conflicts of interest

There are no conflicts of interest.

  References Top

Black RE, Levin C, Walker N, Chou D, Liu L, Temmerman M, et al. Reproductive, maternal, newborn, and child health: Key messages from Disease Control Priorities 3rd Edition. Lancet 2016;388:2811-24.  Back to cited text no. 1
Hill K, You D, Inoue M, Oestergaard MZ, Technical Advisory Group of United Nations Inter-agency Group for Child Mortality Estimation. Child mortality estimation: Accelerated progress in reducing global child mortality, 1990-2010. PLoS Med 2012;9:e1001303.  Back to cited text no. 2
Mejía-Guevara I, Zuo W, Bendavid E, Li N, Tuljapurkar S. Age distribution, trends, and forecasts of under-5 mortality in 31 sub-Saharan African countries: A modeling study. PLoS Med 2019;16:e1002757.  Back to cited text no. 3
Lawrence RJ. Housing and health: From interdisciplinary principles to transdisciplinary research and practice. Futures 2004;36:487-502.  Back to cited text no. 4
Adebowale SA, Morakinyo OM, Ana GR. Housing materials as predictors of under-five mortality in Nigeria: Evidence from 2013 demographic and health survey. BMC Pediatr 2017;17:30.  Back to cited text no. 5
Acharya S. Health Equity in India: An Examination Through the Lens of Social Exclusion. J Soc Inclus Stud 2018;4:104-30.  Back to cited text no. 6
Kumar C, Singh PK, Rai RK. Under-five mortality in high focus states in India: A district level geospatial analysis. PLoS One 2012;7:e37515.  Back to cited text no. 7
Wieczorek WF, Delmerico AM. Geographic information systems. Comput Stat 2009;1:167-86.  Back to cited text no. 8
Dhirar N, Dudeja S, Khandekar J, Bachani D. Childhood morbidity and mortality in India Analysis of National Family Health Survey 4 (NFHS-4) findings. Indian Pediatr 2018;55:335-8.  Back to cited text no. 9
NFHS-4 (2015-2016) Available From: http://rchiips.org/NFHS/NFHS-4Reports/India.pdf (). [Last accessed on 2020 May 01].  Back to cited text no. 10
Anselin L, Syabri I, Kho Y. GeoDa: An introduction to spatial data analysis. Geographical Analysis 2006;38:5-22.  Back to cited text no. 11
Singh A, Pathak PK, Chauhan RK, Pan W. Infant and child mortality in India in the last two decades: A geospatial analysis. PLoS One 2011;6:e26856.  Back to cited text no. 12
Griffith D, Anselin L. Spatial econometrics: Methods and models. Econom Geography 1989;65:160.  Back to cited text no. 13
Akaike H. A new look at the statistical model identification. IEEE Trans Autom Control 1974;19:716-23.  Back to cited text no. 14
Arokiasamy P, Gautam A. Neonatal mortality in the empowered action group states of India: Trends and determinants. J Biosoc Sci 2008;40:183-201.  Back to cited text no. 15
Ministry of Health and Family Welfare. Family Welfare Statistics in India 2011. Statistics Division. New Delhi: MoHFW, Government of India; 2011. Available from: https://nrhm-mis.nic.in/SitePages/HMIS-Publications.aspx. [Last accessed on 2020 May 01].  Back to cited text no. 16
Kumar V, Mishra AJ. Quality of health care in primary health care system: A reflection from Indian state. Int J Health Syst Disaster Manage 2015;3:136-40.  Back to cited text no. 17
  [Full text]  
Parthasarathi R, Sinha SP. Towards a better health care delivery system: The tamil nadu model. Indian J Community Med 2016;41:302-4.  Back to cited text no. 18
[PUBMED]  [Full text]  
Kasthuri A. Challenges to healthcare in India The five A's. Indian J Community Med 2018;43:141-3.  Back to cited text no. 19
[PUBMED]  [Full text]  
Salvi SS, Barnes PJ. Chronic obstructive pulmonary disease in non-smokers. Lancet 2009;374:733-43.  Back to cited text no. 20
Smith KR, Samet JM, Romieu I, Bruce N. Indoor air pollution in developing countries and acute lower respiratory infections in children. Thorax 2000;55:518-32.  Back to cited text no. 21
Gajate-Garrido G. The impact of indoor air pollution on the incidence of life threatening respiratory illnesses: Evidence from young children in Peru. J Dev Stud 2012;49:500-15.  Back to cited text no. 22
Aklin M, Cheng CY, Urpelainen J, Ganesan K, Jain A. Factors affecting household satisfaction with electricity supply in rural India. Nat Energy 2016;1:1-6.  Back to cited text no. 23
Chen YJ, Chindarkar N, Xiao Y. Effect of reliable electricity on health facilities, health information, and child and maternal health services utilization: Evidence from rural Gujarat, India. J Health Popul Nutr 2019;38:7.  Back to cited text no. 24
Alnsour J. Illegal housing in Jordan. Jordan J Soc Sci 2011;4:339-54.  Back to cited text no. 25
Izugbara C. Whose child is dying? Household characteristics and under-5 mortality in Nigeria. South Afri J Child Health 2014;8:16-22.  Back to cited text no. 26
UNICEF/WHO. Diarrhoea: Why Children are Still Dying and what Can be Done. Geneva: UNICEF and WHO; 2009.  Back to cited text no. 27


  [Figure 1], [Figure 2]

  [Table 1], [Table 2], [Table 3]


Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

  In this article
Patients and Methods
Article Figures
Article Tables

 Article Access Statistics
    PDF Downloaded217    
    Comments [Add]    

Recommend this journal