Indian J Med Res 125, January 2007, pp 31-42

Health infrastructure & immunization coverage in rural India Ashlesha Datar, Arnab Mukherji* & Neeraj Sood Economics & Statistics Group & *Pardee RAND Graduate School, RAND Corporation Santa Monica, CA, USA

Received September 1, 2005

Background & objectives: Immunization coverage in India is far from complete with a disproportionately large number of rural children not being immunized. We carried out this study to examine the role of health infrastructure and community health workers in expanding immunization coverage in rural India. Methods: The sample consisted of 43,416 children aged 2-35 months residing in rural India from the National Family Health Surveys (NFHS) conducted in 1993 and 1998. We estimated separate multinomial logit regression models for polio and non polio vaccines that estimated the probability that a child would receive “no cover,” “some cover” or “full age-appropriate cover.” The key measure of health infrastructure was a hierarchical variable that assigned each child to categories (no facility, dispensary or clinic, sub-centre, primary health care centre, and hospital) based on the best health facility available in the child’s village. We also included variables capturing the availability of various types of community health workers in the village and other health infrastructure. Results: While there was under-provision of rural health infrastructure, our results showed that the availability of health infrastructure had only a modest effect on immunization coverage. Larger and better-equipped facilities had bigger effects on immunization coverage. The presence of community health workers in the village was not associated with increased immunization coverage. Interpretation & conclusion: Our findings suggest that expanding the availability of fixed health infrastructure is unlikely to achieve the goal of universal coverage. Reforming community outreach programmes might be better strategy for increasing immunization coverage.

Key words Community health workers - health infrastructure - immunization - India - polio

most cost-effective public health initiatives2, yet the cover against VPDs remains far from complete; recent estimates suggest that approximately 34 million children are not completely immunized with

Roughly 3 million children die each year of vaccine preventable diseases (VPDs) with a disproportionate number of these children residing in developing countries1. Vaccines remain one of the 31

32

INDIAN J MED RES, JANUARY 2007

almost 98 per cent of them residing in developing countries3. Vaccination coverage in India is also far from complete despite a longstanding commitment to universal coverage. A recent evaluation of VPD coverage in India found that 18 million children did not receive any coverage in 2001-20024. In India, immunization has been a central goal of the health care system from the 1970s, first through the Expanded Programme on Immunization (EPI) in 1978, and later with the universal immunization programme (UIP) since 1985. Constitutionally, health care is on the State’s list of responsibilities and is financed by the State. The UIP is an exception; it is one of the few 100 per cent centrally sponsored family welfare programmes and provides support for vaccine storage, training of medical and paramedical staff, and all infrastructure needs specific to delivering immunization to infants at the village level 5. Throughout the 1980s, gains in immunization coverage proved to be rapid for all VPDs; some VPDs showed gains from below 20 per cent coverage to over 60 per cent coverage by the early 1990s6. Pulse polio immunization (PPI) campaign initiated in 1995, was successful in significantly increasing first-dose polio immunization coverage, however, there were limited gains in complete coverage for polio vaccines7. Moreover, coverage of non-polio vaccines seemed to have remained unaffected by the PPI campaign. This limited success in expanding full coverage for VPDs has renewed the search of ways to expand coverage effectively. A natural place to start was by studying the strengths and weakness of the current vaccine delivery system through the rural health infrastructure in India. India’s rural health care system has a strong dependence on community health infrastructure and outreach, particularly in remote villages. Community health infrastructure has been shown to be an important correlate of health outcomes in other developing countries8-11. In the Indian context, two

studies have examined the role of health infrastructure on the utilization of maternal health care using national data 12,13 while others have examined the relationship between health infrastructure and child and maternal health at the State level14,15. These studies have mixed findings. For example, one study used multivariate regressions to control for individual and household characteristics and found that distance to a health facility did not predict health services utilization13. Others have used a similar analytic strategy, but different data, and the results indicated that presence of a health facility in the village significantly increased the likelihood of maternity care utilization10,14. In this study, we examined the role of rural health infrastructure and community health workers in expanding immunization coverage in rural India. We built on this prior research in four important ways. First, by characterizing the availability of rural health infrastructure in more detail than prior studies we were able to discuss the effectiveness of different levels of health infrastructure in expanding coverage. Important differences in both the nature of services provided and the staffing at different levels of care are likely to have different effects on immunization coverage (Table I). We constructed a hierarchical measure of health infrastructure that captured the highest level of health facility available in the village, and thus we were able to examine whether it matters if the best available health facility in the village is a hospital, or a less equipped, smaller facility. Second, we also examined whether the availability of community health workers (CHWs) such as village health guides (VHG), Anganwadi workers (AWW), and trained birth attendants (TBA) affects immunization coverage. Third, we classified the set of vaccines under the UIP into polio and non-polio vaccines to identify the effect of different mechanisms of immunization delivery. Since the PPI campaign was initiated

DATAR et al: RURAL HEALTH INFRASTRUCTURE IN INDIA

between the two National Family Health Survey (NFHS) waves, it is important to examine changes in coverage for polio and non-polio immunizations separately. Finally, in addition to the usual household and individual covariates we included village level variables as well as indicator variables for each State (state fixed effects) in our regression models to control for variations across areas in unobserved factors such as policy environment and governance that are correlated with health infrastructure as well as immunization coverage. Given that States differ markedly in terms of development and health related measures, it was important to adequately control for these differences. Material & Methods Rural health care system: The health care delivery system in rural India relies on a combination of primary health care infrastructure and community outreach. Table I provides details about various levels of government funded rural health infrastructure, the intended and actual populations that each level

33

serves, the medical staff available at each level, and the shortfall for each level. Community health centres (CHC) are the largest facility and are the most endowed in terms of medical staff and equipment, followed by primary health centres (PHC), and subsequently, subcentres (SC). CHCs are designed to serve a much larger population, followed by PHCs, and then SCs. However, there are substantial shortfalls at each level of health infrastructure; as of 2005, there were 10 per cent fewer PHCs and SCs than needed and 50 per cent fewer CHCs than needed (Table I). From an immunization delivery perspective, one key problem in expanding coverage is the demanding, temperature-controlled environment needed to store vaccines before they are administered. Most CHCs and many PHCs, though not all, are currently a part of the Cold Chain and are technically able to maintain a stock of vaccines at recommended temperature zones. While stock-outs at these levels have been reported, they are rare15. Thus, the key area of concern is how effectively the immunization delivery system administers vaccines to infants in villages away from these cold chain facilities.

Table I. State of rural health care system in India

Population capacity: Plain area Hilly/tribal area Actually serving: Average population Average area (sq. km) Specified manpower and asset strength

Sub centre (SC)

Primary health centre (PHC)

Community health centre (CHC)

5,000 3,000

30,000 20,000

1,20,000 80,000

4,579 22.81 2 multi-purpose workers (1 male & 1 female (ANM*) + 1 lady health visitor supervising 6 SCs

27,364 136.31 1 Medical officer + 14 Para medical staff; 4-6 beds; acts as a referral unit to 6 SCs

214,000 1,067.10 4 Medical officers (surgeon, medicine, gynecologist medicine, gynecologist and paediatrician) + 21 paramedical staff; 30 beds + 1 OT + X-ray facilities + Labour room and Lab facilities; serves 4 PHCs 6,491 3,222 50.36

Requirement (Census 2001) Number functioning as of 2005 Shortfall (%)

1,58,792 1,42,655 10.16

26,022 23,109 11.19

Sources: Ref. 5 : (1) Chapter 3, Section 8, Annual Report 1997-98, Ministry of Health and Family Welfare, http://mohfw.nic.in/ reports/1997-98Er/Contents.pdf (accessed January 31, 2006). (2) Ministry of Health and Family Welfare website. http://mohfw.nic.in/dofw%20website/about%20us/infrastructure%20frame.htm#a2 (accessed January 31, 2006) and Ref. 15 *ANM, Auxiliary nurse midwife

34

INDIAN J MED RES, JANUARY 2007

In villages with just a SC, vaccination delivery is done with the help of para-medical staff and community outreach workers. The para-medical staff consists of the lady health worker (LHV) stationed at the PHC, and the auxiliary nurse midwife (ANM) stationed at the SC. The community health and outreach workers consist of the male health workers stationed at the SC, Anganwadi workers (AWW) stationed at an Anganwadi centre that provides day care for children, village health guides (VHG) who provide first-aid and act as outreach workers, and trained birth attendants (TBA) who help with child birth as well as outreach. The key role in immunization delivery is that of the ANM; she is responsible for not only administering vaccines, but also for monitoring immunization coverage4. Her primary duty is to conduct monthly immunization sessions at the SC, and at other villages in the SC’s catchment area that have more than a 1000 individuals. In smaller villages, she is expected to coordinate immunization sessions with the help of the AWW, once every 3 months. The AWW is responsible for bringing the children to the Angwanwadi where the immunization session is held. The ANM is also responsible for (i) maintaining data on immunization delivery; (ii) writing monthly reports on the state of immunization in her catchment area with input from the LHV and the Medical Officer at the PHC; (iii) physically carrying vaccines in from the nearest CHC or PHC on a weekly basis; and (iv) scheduling each of her immunization sessions at each of the villages in the SC’s catchment area. The male health worker helps the ANM with some of the logistics, while the AWW plays an important support role in the scheduled immunization session, many of which take place in an Anganwadi centre. Thus, by and large, parents are expected to bring their children to immunization sessions run by the ANM, if fixed facilities are not available, or directly to PHCs or CHCs4,5,15. Study hypotheses: For our statistical analyses, we used this knowledge about the health infrastructure and the ways in which each facility and individual

contributed to the immunization effort to develop our study hypotheses. We expected to see the presence of each level of health care infrastructure, such as the CHC, or the PHC, or the SC, or the availability of community health workers like the TBA, VHG or AWW would expand immunization coverage, particularly when compared to villages without access to any of these. We also expected that availability of health facilities that are directly linked to the cold chain (CHCs & PHCs) to have larger effects. Among community health workers, we expected AWWs to have the largest effect on immunization coverage due to their relatively welldefined role in immunization delivery. In addition to the public health infrastructure, the UIP provides vaccines free of cost to private sector medical practitioners who can use them on their patients to further expand immunization coverage4. Therefore, apart from public health facilities, we also expected the private health facilities such as dispensaries, clinics, and pharmacies to play a positive role in expanding coverage. Approximately 7 per cent of mothers in our data reported using a private facility for their child’s vaccination. Data: We used data from the 1993 and 1998 waves of the NFHS. The NFHS surveyed a representative sample of households from 26 major States in India. Data were collected using structured interviews with women in the 13-49 yr age group who are, or have been, married. Interviews were conducted using a questionnaire designed specially for mothers, and this was also used to collect data on immunization for children. In addition, for women residing in rural areas information was also collected on the availability of village infrastructure through a village questionnaire. Our analysis sample consisted of 43,416 children sampled from rural primary sampling units, including 22,473 children in wave 1 and 20,943 children in wave 2. By matching the mother level data files with the village level data files we got information on each child’s immunization status, their personal attributes, their maternal and household attributes, and the village infrastructure that they have access to.

DATAR et al: RURAL HEALTH INFRASTRUCTURE IN INDIA

12-25 months in a subsample of States in India7,13. Our characterization of immunization coverage differed from such studies by defining immunization cover on the basis of its appropriateness at every age of the infant. This approach had two advantages over prior studies. First, it allowed us to include all children in the 2 to 35 months age group in our analysis. Second, it allowed us to distinguish children who received age-appropriate coverage from children who were immunized at an older (or younger) age and were therefore exposed to the risk of VPDs for a longer duration of time (or receive vaccinations prior to being physiologically ready).

Immunization: Mothers were asked about immunizations received by each of her eligible children, and when possible, this information was verified by cross-checking against the child’s vaccination card. Specifically, the survey asked whether the child had received BCG, DPT (all doses), polio (all doses) and measles vaccinations. Since we do not observe the frequency with each dose is administered, a child was classified as receiving “all doses” of the polio vaccine if the survey reported that the child had received at least one dose (oral or otherwise) of each of polio1, 2, and 3 vaccines (“all doses” is similarly defined for DPT). Our definition of immunization distinguished across vaccine types to identify immunization coverage rates for two categories of VPDs - polio and non-polio. This categorization was considered useful since the two waves of NFHS straddled the PPI.

Measures of rural health infrastructure and community health workers: The NFHS collected village level information in each wave regarding the presence of rural health infrastructure and community health workers relevant for immunizations. First, we constructed a categorical variable that captured the hierarchy of primary health infrastructure, where a dispensary or a clinic was the smallest facility, followed by SC, PHC and the largest facility being a hospital. Specifically, we assigned each village to one of five mutually exclusive categories that captured the highest level of public or private health facility available in the village (i) no health facility present, (ii) best facility was a dispensary or a clinic, (iii) best facility was a SC, (iv) best facility was a PHC, and (v) best facility was a hospital. While it was rare for SC, PHC, and hospitals to be co-located, we found that NFHS data did not report these to be mutually exclusive (about 2% of all children lived in villages which had both a hospital and a PHC, and 9 % of all children lived in villages with both a PHC

In each of the NFHS waves we categorized a child as having either “no coverage,” “some coverage” or “full age appropriate coverage” using Government of India’s Recommended Immunization Schedule (Table II). Thus, for example, a child who is 3 months old, and has BCG, DPT1 and 2, and Polio1 vaccines would be classified as having “fully age-appropriate coverage” under the non-polio vaccine category, but would only be classified as having “some cover” for the polio vaccine. Since many diseases require multiple doses to provide full coverage across a range of strains of the disease (e.g., polio) we used this three-fold classification of immunization status. While some prior studies have used a similar threefold classification (no coverage, partial coverage, full coverage), they restricted their sample to children

Table II. Recommended immunization schedule Age (wk) Birth 6 10 14 36 Source: Ref. 16

Vaccine (months) 0 1.5 2.5 3.5 9.0

BCG

DPT

Polio

X X X

X X X X

X

35

Measles

X

Age appropriate coverage for all India BCG BCG + DPT1 + Polio1 BCG + DPT1-2 + Polio1-2 BCG + DPT1-3 + Polio1-3 BCG + DPT1-3 + Polio1-3 + Measles

36

INDIAN J MED RES, JANUARY 2007

and SC). We were unable to distinguish between private and public hospitals due to the nature of the survey questions. As a result, our definition of hospital included CHCs, government hospitals, NGO hospitals, and private hospitals. Second, we constructed indicator variables for various community health workers present in the village who did not provide immunizations to children but played an important role in community mobilization. These included (i) whether a VHG was present in the village, (ii) whether a TBA was present in the village, and (iii) whether an AWW was present in the village. We also included an indicator for whether there was a pharmacy or medical shop present in the village. We expected this variable to proxy for the availability of private health care in the village. The survey also asked about visit from a MHU. We included a separate indicator for whether a MHU visited the village. Many households did not have a health facility in their village, but a facility was available in a neighboring village. Therefore, we constructed alternative measures of infrastructure availability that captured the highest level of facility available within 2 and 5 km from each child’s village of residence. Analysis approach: For each category of immunizations (polio vaccines and non polio vaccines), we estimated a multinomial logit regression model, which estimates the conditional probability that a child with a specific set of characteristics (at the child, household and village levels) was likely to receive “no cover,” “some cover” or “full age-appropriate cover.” The key explanatory variables of interest in our models were the village health infrastructure and health personnel variables. Since placement of health facilities and community health workers was likely to be influenced by other population and village characteristics that might also be correlated with a child’s likelihood of being immunized, controlling

for these factors was important in order to assess the independent impact of health infrastructure on immunization coverage. The NFHS survey data allowed us to include a rich set of covariates, at various levels, that related to a child’s likelihood of being immunized - child-level (age, sex, birth order), mother-level (age, education, cohabiting, work status, and if working for a salary), husband-level (age and occupation), household-level (religion, tribal status, household size, wealth index17). In addition, we also included a number of village-level variables that were likely to influence the placement of health facilities (access to roads, distance to a major town, availability of post office, schools, and electricity). We included dummy variables for each State in our models to control for time invariant unobserved differences across States that were related to immunization coverage. We also included a dummy variable for wave, which controlled for any general time trend in immunization coverage (e.g., improved efficiency in providing health care). We estimated our models using Stata for Windows, Version 8 (Stata Corp, College Station, Texas, USA). All estimates were weighted to adjust for the multistage sampling design. Robust standard errors were estimated to adjust for clustering at the village level. The variables in our analysis had extremely low rates of missing data; on average the variables had a less than 1 per cent missing rate, with the maximum missing rate for any one variable being 3 per cent. Results Polio and non-polio vaccine coverage: Data on distribution of age-appropriate immunization coverage for polio and non-polio vaccines in each wave in both urban and rural India showed that in 1993, a significant proportion of children in rural India did not have any vaccination coverage. Between 1993 and 1998 there was a significant decline in the proportion of children with no coverage for both polio and non-polio vaccines (Table III). However, the decline in no-coverage was much more pronounced for polio vaccines (21 percentage points) compared

DATAR et al: RURAL HEALTH INFRASTRUCTURE IN INDIA

to non-polio vaccines (11 percentage points). The reduction in no-coverage for polio vaccines translated into a roughly equal increase in partial and full coverage. In contrast, the decline in no-coverage for non-polio vaccines increased full coverage almost twice as much as partial coverage. Immunization rates for polio and non polio vaccines in urban areas were much higher than corresponding rates in rural areas (Table III). Availability of rural health infrastructure and community health workers: Data on the distribution of rural health infrastructure in each wave of the NFHS showed three salient trends (Table IV). First was the limited availability of health infrastructure in rural India in the early 1990s. For example, in 1993 as many as 43 per cent of children lived in villages with no health facility and roughly half did not have a PHC or hospital within a 5 km radius. Second, the data showed substantial improvements in the availability of health infrastructure across the two waves that mimiced the trend of higher immunization coverage. Third, the data suggested that the improvement in health infrastructure was Table III. Age appropriate coverage (%) for polio and nonpolio vaccination, By wave NFHS I (1993) NFHS II (1998) Rural sample Polio age appropriate coverage: No cover 47.3 Some cover 13.8 Full cover 38.9 Non-polio age appropriate coverage: No cover 45.3 Some cover 27.4 Full cover 27.4 Urban sample Polio age appropriate coverage: No cover 29.2 Some cover 11.5 Full cover 59.3 Non polio age appropriate coverage: No cover 25.5 Some cover 28.9 Full cover 45.6

26.0 25.7 48.3 34.0 31.5 34.5

14.3 22.5 63.2 14.8 27.6 57.6

NFHS, National Family Health Survey Estimates are based on weighted NFHS I and II data

37

concentrated in areas that already had a health facility. For example, across all our hierarchical measures of health infrastructure, the proportion with no health facility remained relatively stable while the proportion with PHC or hospital increased across the two waves. Finally, as expected, the best available facility in the village improved as we examined availability within larger distances. Most notably, there was a significant drop in the percentage of children who had no dispensary/clinic, SC, PHC or hospital in their village (only 28% within 2 km, and just under 9% within 5 km), and an increase in the percentage of children living in villages where the best health facility was a hospital (Table IV). Table IV. Availability of rural health infrastructure in the child’s village, By Wave NFHS I (1993) NFHS II (1998) (%) (%) Best health facility in the village: None 42.9 Dispensary or clinic 20.7 Subcentre 20.2 Primary health centre 5.1 Hospital 11.2

46.6 10.0 21.9 6.5 14.9

Best health facility within 2 km of the village: None 28.1 Dispensary or clinic 21.8 Subcentre 24.3 Primary health centre 7.3 Hospital 18.5

29.5 9.1 28.2 11.0 22.2

Best health facility within 5 km of the village: None 8.8 Dispensary or clinic 18.9 Subcentre 21.8 Primary health centre 11.8 Hospital 38.8

9.7 4.8 26.1 18.7 40.7

Other health infrastructure in the village: Mobile health unit in the village 16.2 Pharmacy or medical shop in the village 26.9 Community health workers in the village: Village health guide 45.0 Trained birth attendant 50.1 Anganwadi worker 46.2

11.3 23.9 33.2 57.8 62.1

Estimates were based on weighted NFHS I and II data Sample consists of children between the age of 2-35 months of age whose immunization records were complete for each category of immunization NFHS, National Family Health Survey

38

INDIAN J MED RES, JANUARY 2007

The availability of other health infrastructure was also relatively scarce in wave 1. Less than one in three children lived in villages that had a pharmacy or medical shop. Visits by a MHU were even less common. In contrast, the availability of community health workers in the village was relatively more common. As many as one in two children lived in villages that had a TBA, 45 per cent lived in villages with a VHG, and 46 per cent lived in villages that had an AWW. Regression results: In regression analyses, we estimated a series of multinomial logit models for non-polio and polio vaccine coverage. Panels A to C (Table V) reported estimates of the effect of village health infrastructure and community health workers on non polio vaccination coverage for the three hierarchical measures of health infrastructure. The estimates in all three panels showed that availability of a dispensary/clinic, SC, PHC or hospital in the village reduced no-coverage for non polio vaccines. The estimates also showed that larger and betterequipped facilities such as PHCs or hospitals had a larger effect on non polio vaccine coverage compared to dispensaries or SCs. For example, children living in villages where the best available health facility was a hospital are 4 percentage points less likely to have no-cover for non-polio vaccines compared to children that have no health facility in their village. This decrease translated into an almost equal increase in some-cover and full-cover. By contrast, children living in villages where the best available health facility was a dispensary or clinic are 3 percentage points less likely to have no-cover for non polio vaccines and this decreased translates mostly into an increase in some-cover. The effect of a MHU on immunization coverage appeared to be weak. One of the reasons could be that MHUs typically served very remote populations of whom we saw very little of in the data. In villages that reported a MHU visit the mean distances to the nearest SC, PHC, and hospital were 4, 10 and 11 km, respectively. Data showed that more restrictive measures of the lack of health infrastructure had a stronger effect

on vaccination coverage (Table V). For example, children who resided in villages with no health facility within 2 km were 4.8 percentage points more likely to have no vaccination coverage compared to children where the best facility within 2 km radius was a hospital. By contrast, children who resided in villages with no health facility within 5 km were 7.1 percentage points more likely to have no vaccination coverage compared to children in villages where the best facility within 5 km was a hospital. Thus, the results suggested that the likelihood of having some or full immunization coverage decreased with increasing distance to rural healthcare infrastructure. While these gains were important, and statistically significant, the increase in coverage from health infrastructure was at best a small fraction of the gap between current coverage levels and universal immunization. The results for other health facilities and community health workers variables showed that these factors had little or no influence on immunization coverage. The availability of MHU was not associated with changes in non-polio vaccine coverage. However, the availability of a pharmacy or medical shop in the village was associated with an increase in some-cover (2.2 percentage points), although the effect was smaller than that of hospitals and PHCs. Among community health workers, presence of VHGs and TBAs in the village was not associated with increased immunization coverage. However, as expected, the presence of an AWW reduced the likelihood of no-coverage (-1.8 percentage points). In contrast to the results for non polio coverage, the association between the best health facility available in the village and polio vaccine coverage was smaller, and not statistically significant (Table VI). For the more restrictive measures, such as having a PHC or hospital within 2 or 5 km of the respondent’s village, the likelihood of full-cover for polio vaccine increased and was statistically significant but was also substantively small in terms of realizing full immunization coverage

Table V. Marginal effects of rural health infrastructure and community health workers on non-polio vaccine coverage Panel A: Non-polio vaccine coverage response to availability in village

Panel C: Non-polio vaccine coverage response to availability within 5 km

Full AAC

No cover

Full AAC

No cover

0.4 1.8 0.9 1.8

-3.7** -4.0*** -5.8*** -4.8***

1.5 1.6 1.8 1.1

2.3* 2.4** 4.0** 3.6***

-6.6*** -6.1*** -7.0*** -7.1***

3.1* 2.5* 2.7 2.5*

3.5* 3.6** 4.3** 4.6***

-0.4 -1.6

0.1 -0.7

0.2 2.5***

-0.2 -1.8*

-0.1 -1.0

0.3 2.5***

-0.2 -1.5

-0.7 0.7 1.0

1.4 0.1 -1.9**

-0.6 -0.8 0.9

-0.8 0.6 1.1

1.4 -0.2 -2.0**

-0.6 -0.6 0.9

-0.8 0.8 1.1

Some cover

Some cover

Full AAC

*Significant at 10%; **significant at 5%; ***significant at 1%. Estimates shown in the Table reflect percentage point changes in “No cover”; “Some cover”; and “Full cover” in response to changes in availability of health infrastructure and community health workers. All estimates come from weighted multinomial logit models that include individual, household and village level variables as well as State and wave fixed effects.†Information on distances for the “other health infrastructure” and “community health worker” variables was not collected in wave 2. Therefore, estimates for these variables in panels B and C capture the effect of availability in the village. AAC, age-appropriate cover

Table VI. Marginal effects of rural health infrastructure and community health workers on polio vaccine coverage Panel A: Polio vaccine coverage response to availability in village No cover Some cover Best health facility (reference=none) Dispensary or clinic -0.9 1.6* Subcentre -0.4 0.5 Primary health centre -2.2 1.5 Hospital -1.3 -0.1 Other health infrastructure in the village† Mobile health unit visit 3.2** -1.7** Pharmacy or medical shop -1.3 0.6 Community health workers in the village † Village health guide -0.7 0.2 Trained birth attendant 0.7 -1.7*** Anganwadi worker -1.3 0.6

Full AAC

Panel B: Polio vaccine coverage response to availability within 2 km

Panel C: Polio vaccine coverage response to availability within 5 km

No cover

No cover

Some cover

-2.4 -0.6 -3.3* -3.4**

0.6 -1.5 -2.4** -1.0

Some cover

Full AAC 2.2 1.8 4.6** 4.3***

Full AAC

-0.8 0.0 0.7 1.5

-1.5 -0.9 -3.7* -2.5*

-0.8 -1.0 -0.9 -1.9**

1.8 2.1 5.7*** 4.4***

-1.5 0.7

3.0* -1.2

-1.6* 0.8

-1.4 0.4

2.9* -1.3

-1.5* 0.7

-1.4 0.6

0.5 0.9 0.7

-0.7 0.8 -1.4

0.3 -1.5** 0.7

0.3 0.7 0.7

-0.7 0.5 -1.4

0.3 -1.5** 0.5

0.4 1.0 0.8

39

*Significant at 10%; **significant at 5%; ***significant at 1%. Estimates shown in the table reflect percentage point changes in “No cover”; “Some cover”; and “Full cover” in response to changes in availability of health infrastructure and community health workers. All estimates come from weighted multinomial logit models that include individual, household and village level variables as well as State and wave fixed effects. †Information on distances for the “other health infrastructure” and “community health worker” variables was not collected in wave 2. Therefore, estimates for these variables in panels B and C capture the effect of availability in the village

DATAR et al: RURAL HEALTH INFRASTRUCTURE IN INDIA

No cover Some cover Best health facility (reference=none): Dispensary or clinic -3.0* 2.6** Subcentre -3.2** 1.4 Primary health centre -4.7** 3.9** Hospital -4.0** 2.2 Other health infrastructure in the village†: Mobile health unit visit 0.2 0.2 Pharmacy or medical shop -0.6 2.2** Community health workers in the village †: Village health guide 1.4 -0.7 Trained birth attendant 0.1 -0.9 Anganwadi worker -1.8* 0.8

Panel B: Non-polio vaccine coverage response to availability within 2 km

40

INDIAN J MED RES, JANUARY 2007

(4.6 percentage points and 5.7 percentage points increase in full-cover for PHC within 2 and 5 km; 4.3 percentage points and 4.4 percentage points increase in full-cover for hospital within 2 and 5 km). In contrast to what we found for non-polio coverage, MHU visits in the village were associated with a statistically significant decline in no-cover for polio vaccination (-3.2 percentage points). The availability of a pharmacy or a medical shop, VHG, or AWW in the village was not associated with coverage for polio vaccines. Presence of a TBA in the village, while not associated with changes in non polio coverage, reduced the likelihood of somecover. Finally, the results for other co-variates in the models (not presented in the Tables) showed that there were a number of other important predictors of childhood immunization. The most prominent predictors in our models included sex of the child, maternal literacy and whether the child belonged to a scheduled caste/tribe (SC/ST) household. For example, a child born to a mother who was literate, but not completed middle school was 8.6 percentage points less likely to have no-cover for non polio vaccines compared to an illiterate mother. Similarly, a child born in an SC/ST household was 3.0 percentage points more likely to have no-cover for non polio vaccines when compared to a child born in a non-SC/ST family; and a girl child was 4.0 percentage points more likely to have no-cover for non polio vaccines when compared to a boy child. Thus, immunization coverage levels varied systematically across distributions of socio-economic and demographic variables and some sub-groups were more at risk of not receiving immunization coverage than others. Discussion Our results highlighted two salient facts. First, the immunization coverage in rural India was far from universal. Second was that expanding the availability of fixed health infrastructure would

result in only modest gains in immunization coverage and was unlikely to achieve the goal of universal coverage. Our results also showed that immunization coverage in urban India although higher than in rural India was far from universal despite presumably better access to fixed health infrastructure. In principle, community health workers should play an important role in expanding immunization coverage to areas that do not have a well-developed health infrastructure. However, we found that the effectiveness of community health workers in expanding immunization coverage was at best small. It was not immediately obvious from our study if this poor association was due to weak mobilization efforts of these community workers, or due to the ANM’s inadequate coverage of the SC catchment area where these community health workers operated. Unfortunately, the NFHS does not provide data on either the location or the time allocation of the ANM across the many duties she has. Anecdotal and case-study evidence suggested that the current job requirements of the ANM may not be optimally allocating the ANM’s time if the goal is to expand immunization coverage 4,18,19. At the same time, providing better training to and monitoring of the community health workers has been identified as an important area for improvement 18,19. The results also showed that the availability of a higher level facility like a hospital or PHC in the village, or within 2 or 5 km, tended to have a larger effect on immunization coverage than lower level facilities (SCs, dispensaries). This was not surprising since hospitals and PHCs were not only better staffed but were also a part of the cold chain and therefore had a regular supply of vaccines. In addition, they conducted weekly scheduled immunization sessions as opposed to the monthly sessions, which were run at SCs. But the point of concern was that the spread of India’s rural health network was the thinnest at the levels of health infrastructure that we found to be most effective.

DATAR et al: RURAL HEALTH INFRASTRUCTURE IN INDIA

Finally, even though the study focused on evaluating the role of health infrastructure and community health workers in expanding immunization coverage, the analysis suggested other potential avenues for expanding coverage. For example, maternal literacy was strongly correlated with immunization coverage and thus policies that encourage female education might lead to significant gains in immunization. Also there were vulnerable groups like the girl child, or children born to SC/ST families who were less likely to receive coverage. Special interventions tailored for such vulnerable children may also significantly improve coverage. In conclusion, systematic expansion of the availability of fixed infrastructure requires financial support and a strong political commitment from State governments. Yet, such a commitment is unlikely to attain universal coverage. Critical improvements in delivery of care are more important as a key failure has been the low effectiveness of community outreach activities in remote villages. There is evidence that the ANM has huge responsibilities with few resources. Significant improvement in the functionality of the existing rural infrastructure may be possible by rationalizing the role of the ANMs and providing them with adequate (financial and manpower) resources to realistically meet the goals. Other interventions such as female literacy and those targeted towards the needs of vulnerable populations would help expand the coverage of immunization in India. Acknowledgment Authors acknowledge the financial support from the RAND Corporation, a non-partisan and not for profit think tank based in Santa Monica, California, USA. Authors thank Dr P. Biswal (Assistant Commissioner, Universal Immunization Programme, Government of India), Dr Naresh Goel (Monitoring and Surveillance Officer, WHO-NPSP India), Dr R.S. Bakshi (Chief Medical Officer, Ranbaxy Community Healthcare Society), and Dr Vandana Joshi (Regional Director, CARE India) for discussion about the rural health system in India. Neeraj Sood and Ashlesha Datar are economists at the RAND Corporation and Arnab Mukherji is a doctoral fellow at the Pardee RAND Graduate School.

41

References 1. Kane M, Lasher H. The case for childhood immunization. Occasional Paper No. 5. Children’s vaccine program at PATH. Seattle, WA; 2002. 2. Jamison DT, Mosley WH, Measham AR, Bobadilla JL. Disease control priorities in developing countries. New York: Oxford University Press; 1993 3. Frenkel LD, Nielsen K. Immunization issues for the 21st century. Ann Allergy Asthma Immunol 2003; 90 (Suppl 3) : 45-52. 4. WHO India. Universal immunization program national review report (2004). [Accessed on February 25, 2006]. Available from: http://www.whoindia.org/EN/Section6/ Section284/Section286_507.htm 5. Government of India. Annual Report of the Ministry of Health and Family Welfare, various issues (1992-3, 19967, 1997-8). [Accessed on January 31, 2006]. Available from: http://mohfw.nic.in/reports/index.htm 6. WHO. Review of national immunization coverage 19802003: India, WHO/UNICEF Report. 2004 August. [Accessed on August 1, 2005]. Available from: www.who.int/vaccinessurveillance/ WHOUNICEF_Coverage_Review/pdf/ india.pdf 7. Bonu S, Rani M, Baker TD. The impact of the national polio immunization campaign on levels and equity in immunization coverage: Evidence from rural North India. Soc Sci Med 2003; 57 : 1807-19. 8. Maitra P, Ray R. The impact of resource inflows on child health: evidence from Kwazulu-Natal, South Africa, 1993-98. J Dev Stud 2004; 40 : 78-114. 9. Valdivia M. Poverty, health infrastructure and the nutrition of Peruvian children. Econ Hum Biol 2004; 2 : 489-510. 10. Frankenburg E, Suriastini W, Thomas D. Can expanding access to basic health care improve children’s health status? Lessons from Indonesia’s ‘Midwife in the Village’ Programme. Popul Stud 2005; 59 : 5-19. 11. Sastry N. Community characteristics, individual and household attributes, and child survival in Brazil. Demography 1996; 33 : 211-29. 12. Stephenson R, Tsui AO. Contextual influences on reproductive wellness in northern India. Am J Public Health 2003; 93 : 1820-6.

42

INDIAN J MED RES, JANUARY 2007

13. Das N, Mishra V, Saha P. Does community access affect the use of health and family welfare services in rural India? National Family Health Survey Subject Reports. Number 18; May 2001; Indian Institute of Population Sciences, Mumbai, India. 14. Shariff A, Singh G. Determinants of maternal health care utilization in India: Evidence from a recent household survey. 2002. Working Paper Series No. 85. National Council of Applied Economic Research. New Delhi, India. 15. Government of India. Strengthening of public institution for health delivery, Technical Report, Ministry of Health and Family Welfare. 2005. [Accessed on January 31, 2006]. Available from: http://mohfw.nic.in/phi%20 strengthening%2015th%20march.pdf. 16. Universal Immunization Program Division, Department of Family Welfare, Ministry of Health & Family Welfare, http:/ /cbhidghs.nic.in/hii2003/12.01.htm [Accessed April 18, 2004].

17. Filmer D, Prichett L. Estimating wealth effects without expenditure data- or tears: An application to educational enrollments in states of India. Demography 2001; 38 : 115-32. 18. Bajpai N, Dholakia RH, Sachs JD. Scaling up primary health services in rural India. CSDG Working Paper No.29. The Earth Institute at Columbia University. 2005 [Accessed on January 30, 2006]. Available from: http:// www.earthinstitute.columbia.edu/cgsd/documents/ RuralHealthPaper.pdf. 19. Rao KS. Delivery of health services in the pubic sector. In : PG. Lal and Byword Editorial Consultants, editor. Background papers of the National Commission on Macroeconomics and Health. New Delhi: Ministry of Health and Family Welfare, Government of India. 2005. [accessed January 30, 2006] Available from: http://mohfw.nic.in Report%20on%20NCMH/Background %20Papers% 20report.pdf.

Reprint requests: Dr Neeraj Sood, Associate Economist, RAND Corporation 1776 Main Street, Santa Monica, CA 90407, USA e-mail: [email protected]