|Year : 2018 | Volume
| Issue : 1 | Page : 32-38
Spatio-temporal assessment of infant mortality rate in India
Amitha Puranik1, VS Binu2, Seena Biju3, Sonu H Subba4
1 Research Scholar, Department of Statistics, Prasanna School of Public Health, Manipal Academy of Higher Education, Manipal, Karnataka, India
2 Associate Professor, Department of Statistics, Prasanna School of Public Health, Manipal Academy of Higher Education, Manipal, Karnataka, India
3 Associate Professor, Department of Operations Management, T. A. Pai Management Institute, Manipal, Karnataka, India
4 Professor, Department of Community and Family Medicine, All Institute of Medical Sciences, Bhubaneswar, Odisha, India
|Date of Web Publication||6-Mar-2018|
Dr. V S Binu
Department of Statistics, Prasanna School of Public Health, Manipal Academy of Higher Education, Manipal, Karnataka
Source of Support: None, Conflict of Interest: None
| Abstract|| |
Background: Infant mortality rate (IMR) is globally identified by the policymakers as the marker of health of a population. Objectives: This study aimed to detect the change in hotspots of IMR in Indian states from the year 2000 to 2012, identify hotspots of IMR at district level in selected states from each of the six regions of India and determine the potential predictors of IMR after accounting for spatial autocorrelation. Methods: Ecological study design was used to analyze state and district level data on IMR of India. For the first objective, the data were obtained from Sample Registration System. For the second objective, we classified India into six regions and selected a state in each region that had the highest IMR. The district level data on IMR and potential predictors were obtained from surveys, namely, Annual Health Survey, District Level Household and Facility Survey and Census. Spatio-temporal hotspots of IMR were examined using local indicators of spatial association statistic. Spatial regression was used to identify the potential predictors of IMR after accounting for spatial autocorrelation. Results: Temporal hotspots of IMR were found in the central part of India. Spatial hotspots were identified in districts of Uttar Pradesh. A negative association of IMR existed with female literacy rate, mothers receiving antenatal checkup (%), and people living in urban areas (%). Conclusion: IMR continues to be a problem in the states that have previously shown to be poor performing. Certain districts within these states need emphasis for focused activities.
Keywords: India, infant mortality, local indicators of spatial association, Moran’s I, spatial regression model, Spatio-Temporal
|How to cite this article:|
Puranik A, Binu V S, Biju S, Subba SH. Spatio-temporal assessment of infant mortality rate in India. Indian J Public Health 2018;62:32-8
|How to cite this URL:|
Puranik A, Binu V S, Biju S, Subba SH. Spatio-temporal assessment of infant mortality rate in India. Indian J Public Health [serial online] 2018 [cited 2019 Dec 6];62:32-8. Available from: http://www.ijph.in/text.asp?2018/62/1/32/226620
| Introduction|| |
Infant mortality reflects the effect of economic and social conditions on health of mothers and newborn babies, as well as the effectiveness of health systems in a country/region. Globally, the infant mortality rate (IMR) has decreased from an estimated rate of 63 deaths per 1000 live births in 1990 to 32 deaths per 1000 live births in 2015. Among the developing countries, India alone contributes to one-third of the five million children who die before reaching their first birthday. Currently, India ranks third in IMR among the South Asian countries superseded by Afghanistan and Pakistan. IMR in India has declined from 88 deaths per 1000 live births in 1990 to 38 deaths per 1000 live births in 2015. Even then, the goal to bring down IMR by two-thirds between 1990 and 2015 has not been achieved.
Several epidemiological studies have looked at the distribution and predictors of IMR quality of medical care, poverty, female literacy, etc., in the states of India.,, However, one can also look for the role of space/location on the occurrence and distribution of IMR since a predictor of IMR in one region can influence or be influenced by the same predictor in the neighboring region. This can result in spatial autocorrelation among the values of that predictor which in turn can contribute to spatial autocorrelation in IMR. Hence, when assessing the distribution of IMR or identifying the potential predictors for IMR, accounting for spatial autocorrelation can be more informative. Spatial analytical methods are used for this purpose. Identification of hotspots of IMR using spatial analytic methods help in identifying the regions with high IMR surrounded by regions with high IMR.,,
In modern epidemiology, spatiotemporal analysis has gained currency to map the events and associated risk factors. In addition, factors related to operational as well as implementation challenges of different health programs can be mapped and analyzed. The mapping and analysis can be both spatially and temporally to formulate tailor-made prevention and control strategies.
In India, studies are scarce to determine spatiotemporal distribution of IMR. Only a few studies were conducted in the past on spatio-temporal distribution and hotspots at state and district levels.,, However, with scaled-up implementation of national programs related to maternal and child health, IMR has declined progressively after 2000. Therefore, it is expected that the hotspots will change as well. In this context, we conducted this study to detect the change in hotspots of IMR in Indian states from the year 2000 to 2012, identify hotspots of IMR at district level in selected states from each of the six regions of India, namely, North, South, East, West, Central, and North East and determine the potential predictors of IMR after accounting for spatial autocorrelation.
| Materials and Methods|| |
This is an ecological study that utilized aggregate level data from Sample Registration System, Annual Health Survey (AHS), District Level Household and Facility Survey– 4 (DLHS-4), and Census. We used ArcGIS 10.3 (ESRI, Redlands, CA, USA) and Open GeoDa 1.6.6 (Arizona State University, Tempe, AZ, USA) software packages for the analysis.
The outcome variable of this study is IMR. For the first objective, the data on IMR at state level were obtained from Sample Registration System (https://data.gov.in/). Spatial analysis method was used to explore the temporal hotspots of IMR of India after accounting for spatial autocorrelation. Hotspots are regions with high IMR and surrounded by regions with high IMR. Local indicators of spatial association (LISA) statistic (which is in the form of map) were used to identify the temporal hotspots of IMR at state level. A LISA map is categorized based on the pattern of spatial autocorrelation: high-high (hotspot) and low-low (coldspot) districts suggest the clustering of similar values of IMR, while the high-low and low-high districts indicate spatial outliers. A significance map shows the areas with a statistically significant LISA statistic value.,
For the second objective, we classified India into six regions, namely, North, South, East, West, Central, and North East. A state with highest IMR from each of the six regions was identified based on 2001 and 2011 census report. The states selected in this study are Uttar Pradesh, Orissa, Rajasthan, Madhya Pradesh, Assam, and Andhra Pradesh. IMR data for each district were obtained from the AHS (2012–2013), and Census 2011. LISA statistic was used to assess the spatial hotspots of IMR at district level.
For the third objective, we considered eight predictor variables of IMR, namely, children with birth weight <2.5 kg (%), female literacy rate, mothers receiving antenatal checkup (%), mean age at marriage for girls, women with unmet need for spacing (%), i.e., currently married women who are not using contraception but want to wait for 2 years or more before having another child, safe delivery (%), i.e., institutional deliveries and home deliveries conducted by doctor/nurse/auxiliary nurse midwife/female health visitor, per capita gross district domestic product (GDDP), and people living in urban areas (%). The data on per capita GDDP and people living in urban areas (%) were retrieved from Census 2011. We used DLHS – 4 (2012–2013) and Census 2011 as the source for all data on districts of Andhra Pradesh since the state was not covered under AHS. We performed bivariate Moran’s I to assess the spatial autocorrelation of IMR with predictor variables and bivariate LISA to visualize the relationship between IMR and predictor variables. Spatial regression was used to identify the potential predictors of IMR after accounting for spatial autocorrelation.
We explored the relationship between IMR and set of predictor variables using the spatial regression approach. There are two distinctions of spatial regression modeling, i.e., spatial lag model and spatial error model. Spatial lag model is appropriate when the value of an event in one region is directly influenced by the values of the same event in its neighboring regions. Spatial error model is appropriate when the concern is to adjust for the bias occurring from spatial autocorrelation, due to the use of spatial data.
Before calculating the spatial autocorrelation measure, it is mandatory to specify the weight matrix. A spatial weight matrix specifies the neighborhood set for each observation of the event variable. In this study, we performed all the spatial analysis based on queen’s contiguity weight matrix. Under queen, areas are neighbors if they share either a border or point.
| Results|| |
The IMR in 35 states in the year 2000 varied from 14/1000 live births in Kerala to 95/1000 live births in Orissa. Kerala being consistent with the least IMR, the highest IMR was observed in Madhya Pradesh in the year 2004 and 2008. In the year 2012, Madhya Pradesh had the highest IMR of 56/1000 live births and the lowest IMR of 10/1000 live births was observed in both Goa and Manipur. The LISA map for the years 2000, 2004, 2008, and 2012 revealed the presence of hotspots of IMR [Figure 1]. The hotspot of IMR was formed by Rajasthan, Madhya Pradesh, Uttar Pradesh, Chattisgarh, Orissa, and Jharkhand in the year 2000. Since 2004, the hotspot was consistently formed by Rajasthan, Madhya Pradesh, Uttar Pradesh, and Chhattisgarh until the year 2012.
|Figure 1: Local indicators of spatial association map of infant mortality rate in different years. (a) 2000. (b) 2004. (c) 2008. (d) 2012.|
Click here to view
The district level analysis was performed in the selected six states of India based on AHS 2012–2013 and Census 2011. The IMR in the 223 districts of six selected states ranged from 22/1000 live births in Hyderabad district of Andhra Pradesh to 97/1000 live births in Balangir district of Orissa. Among the 223 districts, 210 districts had IMR more than the national average. The Moran’s I statistic for IMR was 0.46, which indicated that there exists a moderate positive spatial autocorrelation in IMR among the districts of the selected six states in India. This means that higher values of IMR are clustered together and so are all the lower values of IMR.
LISA and significance map confirmed the presence of clustering (hotspots and coldspots) of IMR [Figure 2]. Three hotspots were found in Uttar Pradesh formed by 26 districts and one in Madhya Pradesh formed by two districts. Coldspot was formed by 16 districts out of which 15 districts belonged to Andhra Pradesh and one to Orissa. The significance map showed that the hotspots and coldspot identified were statistically significant in the districts of Uttar Pradesh, Madhya Pradesh, Andhra Pradesh, and Orissa. It could be pointed out that Sultanpur in Uttar Pradesh was an outlier since it had a lower IMR of 45/1000 live births and was sandwiched between two hotspots with an average IMR of 83.5 and 78.3/1000 live births, respectively.
|Figure 2: (a) Univariate local indicators of spatial association map of infant mortality rate. (b) Univariate local indicators of spatial association significance map of infant mortality rate.|
Click here to view
Ordinary least squares (OLS) regression and spatial regression models were constructed to find the relation between IMR and its potential risk factors. The regression coefficients and corresponding P values for each predictor variables obtained from OLS regression and spatial regression models are shown in [Table 1].
|Table 1: Estimated regression coefficients for ordinary least squares, spatial lag and spatial error models|
Click here to view
The model diagnostic measures, namely, akaike information criterion (AIC) and mean-squared error (MSE) were compared between spatial error model and spatial lag model. It was observed that spatial error model had minimum AIC (1661.28 vs. 1663.25) and MSE (86.79 vs. 89.56) values. Hence, spatial error model was considered to have the best fit.
Based on the spatial error model, it was found that female literacy rate, mothers receiving antenatal checkup (%), people living in urban areas (%) had statistically significant effect on the spatial distribution of IMR. Using these significant variables, a bivariate LISA map was constructed with respect to IMR. The bivariate LISA map for IMR and female literacy rate showed the presence of high – low outliers which imply that districts that have high IMR have low rate of female literacy and the low-high outliers imply that districts with low IMR have high rate of female literacy. It can be inferred that the hotspots and coldspots were statistically significant. Similar interpretation can be made for the bivariate maps of IMR with mothers receiving antenatal checkup (%) and people living in urban areas (%) [Figure 3].
|Figure 3: (a) Bivariate local indicators of spatial association map of infant mortality rate with female literacy rate. (b) Bivariate local indicators of spatial association significance map of infant mortality rate with female literacy rate. (c) Bivariate local indicators of spatial association map of infant mortality rate with mothers receiving antenatal checkup (%). (d) Bivariate local indicators of spatial association significance map of infant mortality rate with mothers receiving antenatal checkup (%). (e) Bivariate local indicators of spatial association map of infant mortality rate with people living in urban areas (%). (f) Bivariate local indicators of spatial association significance map of infant mortality rate with people living in urban areas (%).|
Click here to view
| Discussion|| |
Spatial analytic method such as hotspot detection of IMR over time and space gives valuable information about the vulnerable regions that require immediate attention with respect to IMR. It will also reveal changing hotspots over time, and further analysis for risk factors can be instrumental in helping to reduce IMR in those areas. Such studies have been conducted in some parts of the world, and few studies have been done in India too. Sartorius and Sartorius identified the spatial clustering of IMR at national level based on data from 192 countries. A study conducted in a state of Northeastern Brazil by Rodrigues et al. identified clusters of IMR at municipal level. In India, Singh et al. examined the infant and child mortality for two decades at district level in the 76 natural regions of India. The temporal analysis in their study revealed the decline in the infant and child mortality from 1993 to 2004. Both Kumar et al. and Gupta et al. studied the high-focus states for under-five mortality rate and IMR, respectively and identified hotspots in various districts.
The present study is the first one which has included all the six regions of India while assessing the situation of IMR by accounting for spatial autocorrelation and utilized the latest report of AHS data. An important finding of this study is hotspot detection of IMR at both state and district levels. The result obtained from this study has revealed the presence of significant hotspot of IMR. There were consistent hotspots of IMR that were concentrated in a central part of India during the 12 years. The hotspots were formed by Rajasthan, Madhya Pradesh, Uttar Pradesh, and Chattisgarh. A similar pattern of hotspots was observed in other studies that dealt with infant and child mortality in India.,, Gupta et al. analyzed the spatial clustering and risk factors of infant mortality across high-focus states of India at district level and found hotspots of IMR in Uttar Pradesh and Madhya Pradesh. The present study showed that 35% of the districts in the Northern and Eastern part of Uttar Pradesh formed the hotspot. This emphasizes that continued attention is required in the central part of India since the problem of IMR persists in those regions even after introducing several control measures by the Government of India. Detecting these hotspots should help the policymakers in identifying the areas/regions that need immediate attention. Knowing the exact geographic location of vulnerable regions is an added advantage since it would aid in effective management of resources and time.
Spatial regression analysis performed to identify the potential predictors of IMR after accounting for spatial autocorrelation revealed that female literacy rate, mothers receiving antenatal checkup (%), and people living in urban areas (%) had a significant association with IMR. Similar findings have been reported by Sartorius and Sartorius, who found that maternal mortality, poor female education, unavailability of drinking water, and toilet facility were the determinants for infant mortality. In India, Singh et al. reported that the regions which/that were poor in child nutrition, wealth, or female literacy had high infant and child mortality rates. Studies in India have especially elucidated that maternal education has a significant negative association with IMR.,,, Female literacy is essentially the channel for good practice in child care and health-care utilization. On the other hand, place of residence has a strong impact on the IMR. Usually, the health-care services in urban areas receive a larger share of public resources, resulting in lower investments in rural health services., In addition to the socioeconomic factors, health indicator such as the antenatal checkup for mothers has a vital impact on IMR.
In the present study, it was observed that children born with low birth weight (%), mean age at marriage for women, unmet need for spacing, safe delivery, and per capita GDDP of the districts were not associated with IMR. Several studies highlighted the problem of low birth weight as the leading cause of infant mortality in both developed and developing countries., A recent study conducted in India found an association between age at marriage of the mother and mortality of infants. The economic characteristics of administrative units have also been identified to have a significant effect on infant and child survival. However, in our study, the results were not in line with the findings of other studies. This may indicate the presence of other factors that operate for infant mortality in these areas and they need to be delved into, through a combination of qualitative and quantitative methods.
The bivariate LISA graphically showed the negative association of IMR with female literacy, mothers receiving antenatal checkup (%), and people living in urban areas (%). Although the negative association between maternal education and IMR is established, six districts in Assam formed high–high clusters which implied that these districts had a high rate of female literacy but that has not advantaged in reducing the IMR in the neighboring districts. This indicates that providing education to females alone may not help in controlling the rate of infant deaths. There are many factors that were observed to be associated with IMR like accessibility to hospitals, exclusive breastfeeding, etc., which were not considered in this study., It is of importance to note that several districts in Uttar Pradesh were found to have high IMR and low percentage of mothers receiving antenatal checkup. The 28 districts forming the outliers have relatively low percentage of urban population as compared to the rest of the districts of Uttar Pradesh. A recent study conducted in Uttar Pradesh confirmed that there exists a large gap between urban and rural areas in receiving antenatal care. The mean number of antenatal care visits is higher among women living in urban areas than the rural women.
Although the present study tried to address the situation of IMR using the available data and advanced analytical techniques, there are certain limitations in this study. This is an ecological study with state and district level data as units of analysis. The use of aggregate level data on IMR can lead to ecological fallacy, and hence, the findings cannot be generalized at individual or even household levels. In addition, the study is based on secondary data from multiple sources of different period, and hence, the correlations among different variables could not be measured at real time. Due to unavailability of data, the analysis was not performed at smaller levels of administrative units such as village or at ward level. Another limitation of the study is that environmental factors such as hygiene, climate, rainfalls, temperature, etc., are not considered in the analysis. Although it is important to consider these factors, we could not address them due to lack of data. Further studies must make use of information on such factors while assessing the situation of IMR in India using spatial analytic methods.
| Conclusion|| |
A couple of studies in India have explained the spatial distribution of IMR in India and have identified some risk factors. The present study not only corroborates their findings but also adds by elucidating the temporal hotspots of IMR and identifying the existence of hotspots in the central part of India for more than a decade. The study also analyzed the hotspots of IMR at district level by including a state from each of the six regions of India. This can help in looking for differentials in IMR and its associated factors caused due to the heterogeneity that exists at regional level in India. It has also for the first time taken all six regions of India and found that common notions may not always hold ground; like female literacy alone may not necessarily lead to reduced IMR. Such findings will help to look deeper into issues to find solutions beyond what is apparent till now.
Financial support and sponsorship
Authors (Binu V.S, Seena Biju and Amitha Puranik) acknowledge the support by Department of Science and Technology, India (NRDMS/01/122/015).
Conflicts of interest
There are no conflicts of interest.
| References|| |
Reidpath DD, Allotey P. Infant mortality rate as an indicator of population health. J Epidemiol Community Health 2003;57:344-6.
Kumar A, Singh A. Is economic inequality in infant mortality higher in urban than in rural India? Matern Child Health J 2014;18:2061-70.
You D, Hug L, Ejdemyr S, Beise J. Levels and Trends in Child Mortality: Report 2015. Estimates Developed by the UN Inter-Agency Group for Child Mortality Estimation. New York: United Nations Children’s Fund; 2015.
Kamath SS. Reducing the infant mortality rate of India to 20 by 2020: Together we can do. Indian Pediatr 2015;52:193-4.
Pradhan J, Arokiasamy P. High infant and child mortality rates in Orissa: An assessment of major reasons. Popul Space Place 2006;12:187-200.
Jain AK. Determinants of regional variations in infant mortality in rural India. Popul Stud 1985;39:407-24.
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.
Sartorius BK, Sartorius K. Global infant mortality trends and attributable determinants – An ecological study using data from 192 countries for the period 1990–2011. Popul Health Metr 2014;12:29.
Rodrigues M, Bonfim C, Portugal JL, Gurgel IG, Medeiros Z. Using spatial analysis to identify areas vulnerable to infant mortality. Rev Panam Salud Publica 2013;34:36-40.
Sartorius BK, Kahn K, Vounatsou P, Collinson MA, Tollman SM. Young and vulnerable: Spatial-temporal trends and risk factors for infant mortality in rural South Africa (Agincourt), 1992-2007. BMC Public Health 2010;10:645.
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.
Gupta AK, Ladusingh L, Borkotoky K. Spatial clustering and risk factors of infant mortality: District-level assessment of high-focus states in India. Genus 2016;72:2.
Jong PD, Sprenger C, Veen FV. On extreme values of Moran’s I and Geary’s C. Geogr Anal 1984;16:17-24.
Anselin L. Local indicators of spatial association – LISA. Geogr Anal 1995;27:93-115.
Anselin L. Spatial econometrics. In: Baltagi BH, editor. A Companion to Theoretical Econometrics. Oxford: Basil Blackwell; 2001. p. 310-30.
Chen Y. On the four types of weight functions for spatial contiguity matrix. Lett Spat Resour Sci 2012;5:65-72.
Kateja A. Role of female literacy in maternal and infant mortality decline. Soc Change 2007;37:29-39.
Saurabh S, Sarkar S, Pandey DK. Female literacy rate is a better predictor of birth rate and infant mortality rate in India. J Family Med Prim Care 2013;2:349-53. [Full text]
Subramanian SV, Ackerson LK, Subramanyam MA, Sivaramakrishnan K. Health inequalities in India: The axes of stratification. Brown J World Aff 2008;14:127-39.
Balarajan Y, Selvaraj S, Subramanian SV. Health care and equity in India. Lancet 2011;377:505-15.
Hollowell J, Oakley L, Kurinczuk JJ, Brocklehurst P, Gray R. The effectiveness of antenatal care programmes to reduce infant mortality and preterm birth in socially disadvantaged and vulnerable women in high-income countries: A systematic review. BMC Pregnancy Childbirth 2011;11:13.
Lau C, Ambalavanan N, Chakraborty H, Wingate MS, Carlo WA. Extremely low birth weight and infant mortality rates in the United States. Pediatrics 2013;131:855-60.
Centers for Disease Control and Prevention (CDC). Increasing infant mortality among very low birthweight infants – Delaware, 1994-2000. MMWR Morb Mortal Wkly Rep 2003;52:862-6.
Raj A, Saggurti N, Winter M, Labonte A, Decker MR, Balaiah D, et al.
The effect of maternal child marriage on morbidity and mortality of children under 5 in India: Cross sectional study of a nationally representative sample. BMJ 2010;340:b4258.
Chuang KY, Sung PW, Chang CJ, Chuang YC. Political and economic characteristics as moderators of the relationship between health services and infant mortality in less-developed countries. J Epidemiol Community Health 2013;67:1006-12.
Dummer TJ, Parker L. Hospital accessibility and infant death risk. Arch Dis Child 2004;89:232-4.
Vennemann MM, Bajanowski T, Brinkmann B, Jorch G, Yücesan K, Sauerland C, et al.
Does breastfeeding reduce the risk of sudden infant death syndrome? Pediatrics 2009;123:e406-10.
Singh NK, Jha KK, Sharma RK, Verma N. Differential in antenatal care visit: A case study of Uttar Pradesh. Indian J Matern Child Health 2015;17:1-7.
[Figure 1], [Figure 2], [Figure 3]