This study uses the data from the National Family Health Survey round-2. Using hierarchical linear models the role of community level variables in predicting neonatal deaths at individual level is analyzed. SAS sub-routine PROC GLIMMIX is used for hierarchical linear modeling. Findings re-enforce the importance of mother’s literacy and tetanus toxid vaccination during pregnancy in predicting neonatal deaths at individual level. At community level, though the magnitude of the presence of either a private doctor, or a visiting doctor, or a village guide, or a traditional health attendant, or the presence of mobile health van unit within community is in an expected direction, the effect is statistically not significant. In order to capture true impact of community level intervention of program success it is imperative that the large demographic surveys should incorporate more and more community level indicators while designing these surveys.
High levels of neonatal mortality in Empowered Action Group (EAG) States in India has been well documented. The individual level determinants explaining the variations in these neonatal deaths have also been well researched. The recent stagnation in the declining trend in-spite best of governmental interventions at systemic and individual level to tackle neonatal deaths has not yielded much result. Community based large scale interventions in India have been extremely limited until the launch of National Rural Health Mission Programs (NRHM) in 2005. The NRHM saw tangible shift in the health policy planning not only by increased program funding and central assistance through grant-in-aid to States but also by institutionalizing the community based health workers named Accredited Social Health Activist (ASHA) per 1000 population selected from the community. Prior to the Accredited Social Health Activist (ASHA) the Auxiliary Nurse Midwives (ANM) of the sub-centre health facility were the only first level direct contacts with the community. One ANM as envisaged had a catchment of 5000 population. However currently on an average ANMs cover anything between 7000 - 12,000 population and even more depending on the catchment population growth. The importance of these community level frontline workers though always existed, it was scientifically not looked into at-least in EAG States in an Indian context. Until the launch of NRHM programs in 2005, all the new government health schemes that were launched for the betterment of the community health they got padded up on the shoulders of the ANMs. Post recruitment of ASHAs, lot of community outreach burden got shared up and the ANMs got extra time for other health sector programs including record keeping. Under the NRHM and during the XIth five year plan realizing the importance of the community based health worker i.e. ASHA in saving new born lives in the year 2011 it was envisaged to have home based neonatal care (HBNC) to reduce the neonatal mortality. The XIth plan [
This chapter thus focuses statistically on studying the linkages between availability of community private doctor, visiting doctor, village health guide, availability of trained attendants and mobile health units on neonatal mortality. The only available data through which one can study the impact of community level variables on individual outcome is the NFHS-2 data where village file is merged with the individual file to study the community context of individual outcome.
Home based neonatal care in resource poor settings is designed considering the fact that not all deliveries occur at an institution and even if they did under the guidance of trained medical professional at an institution, majority do not prefer staying at an institution post delivery. In India and more so in demographically backward EAG States lot many women leave hospital in less than 24 hours of delivery. Evidences emanating from the Health Management Information System (HMIS) data of the EAG States point towards the fact that even today lot of women leave hospital within 24 hours of normal delivery. In-spite of mandatory 48 hours stay at health facility as mandated under the Janani Sishu Suraksha Karyakram (JSSK) program under the NRHM programs launched in 2005. The 2010 study by Indian Council of Medical Research (ICMR) [
The main objectives of the study are;
・ To study the clustering effect by estimation of the Intra-class correlation in order to analyze the community effect
・ To study the effect of Level-1 variables (individual level) on the neonatal deaths
・ To study the community presence of either the village level private doctors or the visiting doctor or the village health guide or the trained attendant or the availability of mobile health van i.e. (level-2 variables) on neonatal survival chances.
Data collected from the National Family Health Survey (NFHS-2) [
The variables used in the analysis are as under
a) Outcome Variable: Neonatal Deaths 0 = Failure (Non occurrence of neonatal deaths 1 = Event (occurrence of neonatal deaths)
b) Independent Variables included in the model based on relative influence on outcome variable: Level-1 predictors
i) Sex of the child at the time of birth (Biological) 0 = Male 1 = Female
ii) Mothers literacy (Socio-economic-demographic) 0 = Illiterate 1 = Literate
iii) Tetanus Toxid injection during pregnancy (Government intervention) 0 = Received no TT injection 1 = Received any number of TT injection
c) Community (Village) level variables: Level-2 predictors
i) Availability of private doctor in the village 0 = No 1 = Yes
ii) Availability of visiting doctor in the village 0 = No 1 = Yes
iii) Availability of village health guide in the village 0 = No 1 = Yes
iv) Availability of traditional attendant in the village 0 = No 1 = Yes
v) Availability of mobile health unit 0 = No 1 = Yes
vi) Availability of either private doctor or a visiting doctor or a traditional attendant or a village health guide or a mobile health van (hwmu) 0 = No 1 = Yes
Findings
Observation | Proportion | |
---|---|---|
Neonatal Deaths (Dependent Variable) | ||
Yes | 6065 | 6.78 |
No | 83,390 | 93.22 |
Level 1 (Individual) (Independent Variable) | ||
Sex of the child | ||
Male | 46,636 | 52.13 |
Female | 42,819 | 47.87 |
Mothers Literacy | ||
Illiterate | 71,938 | 80.44 |
Literate | 17,494 | 19.56 |
Ante Natal checkups during pregnancy | ||
Yes | 5068 | 43.05 |
No | 6704 | 56.95 |
TT injection before birth | ||
Yes | 7245 | 61.29 |
No | 4576 | 38.71 |
Type of toilet facility at home | ||
Open | 80,673 | 90.19 |
Flush or pit | 8776 | 9.81 |
Caste | ||
SC/ST/OBC | 62,193 | 70.63 |
Others | 25,862 | 29.37 |
Marriage to First Birth Interval (BI) | ||
≤33 months | 15,549 | 66.32 |
>33 months | 7895 | 33.68 |
Level 2 (Village = 1120) (Independent variable) | ||
Availability of private doctor in the village | ||
Yes | 343 | 30.7 |
No | 775 | 69.3 |
Availability of visiting doctor in the village | ||
Yes | 252 | 22.6 |
No | 865 | 77.4 |
Availability of village health guide in the village | ||
Yes | 351 | 31.4 |
No | 766 | 68.6 |
Availability of traditional attendant in the village | ||
---|---|---|
Yes | 545 | 48.8 |
No | 572 | 51.2 |
Availability of mobile health unit in the village | ||
Yes | 92 | 8.24 |
No | 1024 | 91.76 |
Combined Variable at Level 2 (Villages = 1111) | ||
Availability of doctor or attendant or mobile health unit in the village (hwmu) | ||
Yes | 812 | 73.09 |
No | 299 | 26.91 |
visiting doctors in the village. Thirty one percent of the villages had the facility of village health guide while the remaining sixty nine percent does not have village health guide. Forty nine percent of the villages had the facility of traditional attendant while the remaining 51 percent of the village does not have any traditional attendant. Eight percent of the villages had mobile health unit in the village while the remaining ninety two percent does not have the facility of mobile health unit in the village. Based on these five village level variable a combined variable was created. Villages that have private doctor, visiting doctor, traditional attendant, village health guide or mobile health unit was coded as 1 and rest as zero. Seventy three percent of the villages had either private doctor, visiting doctor, traditional attendant, village health guide or mobile health unit while twenty seven percent had none. This constructed variable was used to study the village level effect on neonatal survival.
the value is ( π 2 3 ) .
GLIMIX Estimates | Estimate | SE | Pr > |t| | Confidence Interval | |
---|---|---|---|---|---|
Lower | Upper | ||||
Model 1 | |||||
Unconstrained | |||||
Intercept | −2.7625 | 0.09 | <0.0001 | −2.9497 | −2.5753 |
Model 2 | |||||
(Random Intercept Fixed Slope Covariates at Level 1) | |||||
Level 1 predictor | |||||
Intercept | −2.1994 | 0.15 | <0.0001 | −2.5041 | −1.8948 |
TT inj | −0.7198 | 0.16 | <0.0001 | −1.0227 | −0.4169 |
Literacy | −0.3848 | 0.17 | 0.0196 | −0.7080 | −0.0617 |
Marriage to First IBI | 0.1643 | 0.15 | 0.2855 | −0.1373 | 0.4658 |
Model 3 | |||||
(Random Intercept Fixed Slope Outcome dependent on Level 2 Covariate (connectivity) | |||||
Level 1 predictor | |||||
Intercept | −2.1151 | 0.32 | <0.0001 | −2.7721 | −1.4582 |
TT inj | −0.7238 | 0.16 | <0.0001 | −1.0281 | −0.4195 |
Literacy | −0.3934 | 0.17 | 0.0173 | −0.7174 | −0.0694 |
Marriage to First IBI | 0.1647 | 0.15 | 0.2846 | −0.1370 | 0.4663 |
Level 2 predictor | |||||
Community availability of doctor or village attendant/guide or mobile health unit (hwmu) | −0.1023 | 0.32 | 0.7544 | −0.7605 | 0.5560 |
Level 1
η i j = β 0 j
Level 2
β 0 j = γ 00 + μ 0 j where μ 0 j N ( 0 , τ 00 )
Combined
η i j = γ 00 + μ 0 j (1)
From the above model 1 which is intercept only model we get the value of ICC as
ICC = τ 00 τ 00 + σ 2 = 2 .7625 2 .7625 + ( π 2 3 ) = 2 .7625 2 .7625 + 3.289 = 0.4565 (2)
From the Equation (2) we get the value of ICC as 0.4565 or the 46 percent of the total variance in neonatal deaths is due to the between community differences. As a next step in the model building covariates at level 1 (individual level) were introduced in the model and the intercept was allowed to vary across the community. For model to converge only those covariates at level were included in the model those that were relatively significant such as mothers literacy (0 = illiterate & 1 = literate), tetanus toxide injection during pregnancy (0 = no & 1 = yes) and marriage to first birth interval (BI ≤ 33 months = 0 & >33 months = 1). The model specification is given by Equation (3) below
Level 1
η i j = β 0 j + β 1 j ( l i t e r a c y i j ) + β 2 j ( T T i n j e c i o n i j ) + β 3 j ( B I i j )
Level 2
β 0 j = γ 00 + μ 0 j where μ 0 j N ( 0 , τ 00 ) β 1 j = γ 10 β 2 j = γ 20 β 3 j = γ 30
Combined
η i j = γ 00 + γ 10 ( l i t e r a c y i j ) + γ 20 ( T T i n j e c t i o n i j ) + γ 30 ( B I i j ) + μ 0 j (3)
In the model 3 the intercepts are allowed to vary across communities but the slope is fixed. The findings present in
Level 1
η i j = β 0 j + β 1 j ( l i t e r a c y i j ) + β 2 j ( T T i n j e c t i o n i j ) + β 3 j ( B I i j )
Level 2
β 0 j = γ 00 + γ 01 ( h w m u j ) + μ 0 j where μ 0 j N ( 0 , τ 00 ) β 1 j = γ 10 β 2 j = γ 20 β 3 j = γ 30
Combined
η i j = γ 00 + γ 10 ( l i t e r a c y i j ) + γ 20 ( T T i n j e c t i o n i j ) + γ 30 ( B I i j ) + γ 01 ( h w m u j ) + μ 0 j (4)
Here ‘hwmu’ in Equation (4) is a combined variable at level-2 created from the responses of five different set of covariates at level-2 listed in Tabe-1. The hwmu is a dichotomous predictor variable at level-2. The main effect of the predictors at level-1 after the introduction of the level-2 predictor does not change. Mothers literacy and the TT injection during pregnancy were still the important predictors explaining the variations in the neonatal mortality at individual level. The communities were either private doctor, or a visiting doctor or a traditional health attendant, or a village guide, or a mobile health unit was present had less chance of neonatal deaths compared to those villages/communities where the private doctor, or a visiting doctor or a traditional health attendant, or a village guide, or a mobile health unit were not present. The risk of neonatal death in such communities were [exp(−0.1093) = 0.9028] 10 percent less compared to the communities where the private doctor, or a visiting doctor or a traditional health attendant, or a village guide, or a mobile health unit were not present. [exponential of the effect (−0.1093) due to community variable is taken to get odds ratio estimates] However, this is not statistically significant. So it is not definite to conclude that the presence of community doctor/health worker or mobile unit improves individual neonatal survival chances.
At individual level mother’s literacy is the most important variable as the risk of neonatal deaths among illiterate mothers is high compared to literate mothers. Government intervention of providing tetanus toxide injection to mothers during pregnancy is another statistically significant variable. Those mothers who received no TT injections during pregnancy have much higher risk of neonatal deaths than those who received any number of TT injections during pregnancy. The village or the community level availability of either the private doctor or a visiting doctor or a village health guide or traditional attendant or a mobile van reduces the neonatal deaths by 10 percent however it’s statistically not significant. As 40 percent of the neonatal deaths occur on day one and given the fact that high proportion of deliveries in EAG states still occur at home, it is imperative that the deliveries that happen at home happen under the supervision of SBA trained ANM. The success of HBNC program initiated in 2011/2012 largely depends on quality training provided to ASHAs in identifying danger signs among neonates and guiding the family members to get speedy referral and admission into state of art Sick New Born Care Units set up at district hospitals. Future demographic surveys in India should capture more and more community level indicators in order to assess the success of governmental interventions within the communities. Government’s community based health intervention programs are largely made after the launch of National Health Mission programs in 2005 after the NFHS-2 data was collected and this being the limitation of the study.
Pandey, A., Nath, D.C. and Sharma, R. (2017) Does the Availability of Community Health Worker/Mobile-Health Van Unit within the Community Impacts Neonatal Survival? Open Journal of Statistics, 7, 803-814. https://doi.org/10.4236/ojs.2017.75057