Trend and inequity in infant vaccination coverage: Analysis from three recent Demographic Health Surveys in Nepal


 Background Despite policy intention to reach disadvantaged populations, inequities in child health care use and health outcomes persist in Nepal. The current study aimed to investigate the trend of full vaccination coverage among infants and its equity gaps between 2002 and 2016.Methods Using data from demographic health surveys conducted in 2006, 2011 and 2016, we investigated the trend of coverage of six antigens: Bacille Calmette Guerin (BCG), Diptheria, Pertussis, Tetanus (DPT), Polio, and Measles) between 2002 to 2016. Rich-poor difference, Rich: Poor ratio and concentration index were calculated to measure income inequity. Lorentz curve was drawn to show the change in income-related inequity over time. Bivariate and multivariate logistic regression analyses were conducted to investigate socio-demographic correlates of full vaccination coverage.Results Full immunization coverage was slightly increased from an average of 83% during 2002-2006 to 87% during 2007-2011, but it decreased to 78% during 2012-2016. There was a significant increase in full vaccination coverage among infants from the poorest income quintile and a simultaneous decrease among infants from richer income quintiles. Province 2 saw the largest drop, from 79.2% (95%CI 64.8-88.8) during 2002-2006 to 65.2% (95%CI 56.4-73.0) during 2012-2016. In Province 2, maternal education was the independent predictor of full vaccination coverage; the mother with secondary education was over three times more likely to fully immunize their children compared to mothers with no formal education (AOR 3.2; 95% CI:1.5-6.7).Conclusion Full vaccination coverage in Nepal saw significant decrement away from the national target after 2011. A sharp decrease in coverage of full vaccination among infants from wealthier income quintiles and an increase in coverage among infants from the poorest income quintile between 2002 and 2016 created a pro-poor equity gain. While a national effort to improve full vaccination coverage is overdue, children from province 2, specifically those born to mothers with no or primary education need particular programmatic focus. Further research is needed to understand the reasons behind decrement in full vaccination coverage, particularly among rich income quintiles.


Introduction
Government of Nepal (GoN) aims to reach all children under-ve years of age with vaccines to prevent vaccinepreventable diseases. Immunization services are being provided through a network of static clinics located at health facilities, and outreach clinics located at the community (1). Private and Non-government organization run health facilities have been increasingly mobilized to provide immunization services. These services can be availed from routine immunization clinics as well as during special campaigns (1).
Reaching all children with full immunization services is vital to meet Nepal's commitment to Sustainable Development Goals. The success of the National Immunization Program (NIP) depends on vaccination coverage, quality of vaccination reporting, and strategies to effectively reach Nepal's diverse and geographically dispersed population (2).Nepal Health Sector Strategy 2015-2020 and its implementation plan have a target are to achieve more than 90% full vaccination coverage for children (3). Previous research has shown that national or regional statistics on coverage can mask the underlying inequalities among population groups (2). Therefore, it is essential to present disaggregated data on immunization coverage to identify unreached population groups.
GoN, along with its development partners, have taken various strategies to strengthen immunization services, including full immunization declaration of wards, municipalities, and districts. However, concerns have been raised regarding the sustainability of full immunization declaration program as being a one-off activity rather than a continuous improvement.
A wealth of evidence suggests that inequalities in child survival outcomes exist mainly by wealth quintiles(4), mother's educational status (5) and geographical region (6). However, whether these inequalities are driven by the differences in coverage and service utilization remains poorly understood. A recent study by KC et al. showed an equity gap to have been narrowed in Nepal with increased coverage of immunization service between 2001 and 2014 (7). However, they used data from different surveys -Demographic Health Surveys and Multiple indicators Cluster Survey (MICS) with different methods to compare immunization coverage. Furthermore, the ndings might have been subjected to recall bias since they included all eligible children under-ve years of age in their analyses, while much of the literature includes only 12-23 months children to measure immunization coverage (8,9). Therefore, current study reports trend and disparities in full vaccination coverage of six antigens available during most recent three rounds of Nepal demographic Health Surveys conducted in 2006, 2011 and 2016 among children of age 12-23 months. The ndings will help to understand the recent trend in full immunization coverage along with the trend in equity gap among different population groups. This could help policymakers identify disadvantaged groups in terms of immunization service utilization and facilitate policymaking to reduce inequalities and achieve universal health coverage in line with sustainable development goals.

Data Source
This study uses data from three Nepal Demographic and Health Surveys (NDHS), conducted in 2006, 2011, and 2016, which obtained comparable nationally representative samples of children under ve years. DHS is an extensive survey which covers a variety of indicators including vaccination coverage among children 12-23 months, and has a rigorous design, (10). The DHS sample typically is selected in two stages. The rst stage involves selecting clusters with probability proportional to size from a national master sample frame. At the second stage, a systematic sample of households is drawn from a listing of households in each of the DHS clusters. Both 2006 and 2011 DHS were selected in two stages while 2016 NDHS sample was strati ed and selected in two stages in rural areas and three stages in urban areas (11)(12)(13).

Measures
Full vaccination refers to having a vaccination against tuberculosis ,i.e. Bacille Calmette Guerin (BCG), three doses of combination vaccine including at least diphtheria, pertussis, and tetanus (DPT), three doses of oral polio vaccines, and a dose of measles-containing vaccine by the age of 12 months. The full immunization rate is calculated based on vaccination data collected from mothers of children of 12-23 months. If vaccination cards were not available at selected households at the time of the survey, interviewers relied on mother's reports to determine vaccination record. Predictor variables used to assess disparities included: mothers' education (i. no education, ii. primary, iii. secondary, and iv. higher education), wealth quintiles (i. poorest, ii. Poorer, iii. Middle, iv. Richer, and v. richest), sex of the child (i. male and ii. female), caste/ethnicity (i. Brahmin/Chettri, ii. Terai/Madhesi other, iii.Dalits, iv. Newar/Janajatis and v. Muslims/others), place of residence (i. rural and ii. urban), province (1to7) and ecological zone (i. Mountain, ii.Hills, and iii. Terai). These variables were taken from the available literature (7,14,15) and were used to demonstrate the disaggregated rates of full immunization coverage in Nepal.

Measurements of inequalities
Given the di culty of collecting income and expenditure data in developing countries, DHS surveys collect data on household ownership of consumer goods, dwelling materials, source of drinking water, types of sanitation facilities, and other characteristics that relate to economic status. With this data, an index score is computed for each household, using principal component analysis. The entire sample is then ranked according to this score and is divided into quintiles, from the rst quintile (Q1)-the poorest 20 percent of the household population-to the fth quintile (Q5)-the wealthiest 20 percent(16). Wealth quintile ranking indicates relative rather than the absolute economic status of the household. The bottom 20 percent measured in the 2006 NDHS may not have the same absolute level of wealth as the bottom 20 percent measured in the 2011 and 2016 NDHS. In other words, wealth status is not comparable across surveys and countries. This study, however, is not affected by this limitation because it focused on the disparities in vaccination coverage between the wealthy and the poor only within each survey.
We calculated three inequality indicators-the ratio between Q5 and Q1 (ratio of Q1 to Q5 for health indicators and the ratio of Q5 to Q1 for indicators of health care) and, Rich Poor difference and the concentration index. The ratio indicator compares the level of health or use of health services between the wealthiest and the poorest quintiles. Rich-poor difference provides an absolute difference in coverage between the wealthiest quintiles and the poorest quintiles. To some extent, these two indicators provide information on the disparities between the wealthy and the poor. However, it is only based on the information of the two extremes among the wealth groups but ignores the other three quintiles between the top and bottom, and therefore cannot provide a picture of inequalities across the entire population.
The second indicator, the concentration index, quanti es the degree of economic inequality using information from all ve quintiles. Therefore, it is a composite summary of inequality across the entire population. The concentration index is calculated in reference to the Lorenz curve. On a Lorenz curve, the x-axis represents the cumulative percentage of the sample, ranked by wealth status from low to high (i.e., from the poorest to the wealthiest); the y-axis plots the cumulative percentage of the outcome variable (i.e., immunization coverage) corresponding to each wealth group.
Analyses were performed initially using the full national sample. Bivariate and multivariate logistic regression models were developed to obtain crude and adjusted odds ratios for full immunization coverage taking into account survey design (sampling weights and strata). Time trends were examined by background characteristics to assess if immunization coverage was changing at the same rate during the study period. Given the signi cant decline in full immunization coverage in province 2, further analysis on determinants of full vaccination coverage was carried out in province 2sample. Combining data from three rounds of DHS resulted in a sample of 667 children aged 12-23 months in province 2. Variables that showed p<0.10 in bivariate models were included in the multivariate model. All analyses were carried out in STATA 15.0 Stata corporation. Ethics DHS surveys were conducted after receiving ethical approval from the Nepal Health Research Council (NHRC) ethical review committee. All of the respondents were informed of the purpose of the survey and were informed that participation was not compulsory and that if they did choose to participate, were assured of the con dentiality of the information. Verbal consent was sought before beginning the interview as per NHRC ethical review guidelines. Separate ethical approval was not required for this analysis.

Results
Trend data showed that there had been considerable socio-demographic changes in Nepal over the last fteen years ( Table 1)   Although there was more than twice likelihood of receiving full immunization by children born in the wealthiest quintile households compared to the poorest, our analysis suggests that the inequity is reducing over time in   There was a striking and statistically signi cant trend of annual decrease in full immunization coverage throughout the study period in province 2 while no signi cant difference was observed in other provinces. On further analysis of 667 children in province 2, the odds of receiving full immunization was 60% less for children aged 12-23 months during 2012-2016 compared to those during 2002-2006 (Table 4). Educational status of mothers showed a positive association with the likelihood of receiving full immunization. Children born to mothers with secondary education were over three times more likely to receive full immunization compared to children born to mothers with no formal education (AOR 3.2; 95% CI:1.5-6.7). Similarly, children born to mothers with higher education were over two times more likely to receive full immunization compared to women with no formal education, although the difference did not reach statistical signi cance (AOR 2.4; 95% CI: 0.9-6.4). Other variables, namely, sex of the child, wealth quintile, caste/ethnicity, urban/rural place of residence, were not associated with full vaccination among children of province 2 from 2002-2016. immunization coverage reducing over time in terms of rich: poor ratio, rich-poor difference and concentration indices and con rmed by Lorentz curve. The decrease in pro-rich inequalities from 2002 to 2016 was partly attributable to a substantial decrease in full immunization coverage among the children from the wealthier quintiles and a slight increase among children from the lowest income quintile. A signi cant declining trend in full immunization coverage in Nepal, albeit retention of equity gain among the children from households with poorest wealth quintile, puts Nepal's immunization program far behind GoN's target to fully immunize more than 90% of children by 2020 (3). Nepal's experience of the signi cant simultaneous decline in overall full immunization coverage together with a decline in pro-rich inequity is similar to that experienced by the Central African Republic(8).
A similar study by KC et al.  (7). Phase-out of community health workers' position such as Village Health Workers might have affected immunization service utilization since direct communication through the household visit by health workers had a positive impact on immunization service utilization (15,19). However, the decline in full vaccination coverage, mainly driven by the decline in DPT3 coverage is di cult to interpret. The decline may be due to health system factors such as changes in national immunization program and the introduction of new vaccines (PCV, IPV, MR, JE). Logistics and supply chain management issues after the introduction of new vaccines contributed to the decline in DPT3 coverage in South Africa in 2009 (20). Other possible explanation could be due to mothers' and their newborns' temporary move to her mother's house(21) around 10-14 weeks of childbirth and less familiarity with vaccination schedule, and place in the new place. Another reason may be people's perception of not feeling the importance of three doses of vaccine after one or two doses of the same vaccine has already been received. It may also be associated with a reduced feeling of threat against polio(7) since DPT and polio are administered simultaneously at 6,10 and 14 weeks. However, this needs further investigation.
Further analysis of NDHS 2016 showed the main reason for the signi cant decline in full vaccination to be the decline in the percentage of children who received the third dose of DPT containing vaccine from 91.7% in 2011 to 85.9% in 2016 (14). The drop out in the third dose of DPT is an increase from 5% (95% CI 3.1-7.6) in 2011 to 11% (95% CI 8.8-13.8) in 2016. However, the coverage of measles-containing vaccine, which is provided after the third dose of DPT, has increased from 88.0% in 2011 to 90.4% in 2016, suggesting a missed opportunity for vaccination of DPT3 at nine months. Also, vaccination card was retained by just over half of the mothers (52.3%) (13). Card retention was one of the factors associated with full immunization coverage, along with the place of delivery (14). In their analysis, geographic and urban/rural place of residence were not associated with differences in full immunization coverage. However, current analysis using recent three NDHS demonstrated the decreasing trend in full immunization coverage over last 15 years to be signi cant for children living in rural areas, Terai region and in province 2 (Table 3). While an increase was observed among most disadvantaged Caste/ethnicity 'Dalits,' other relatively disadvantaged castes/ethnicity groups such as " Terai/Madhesi and Muslim/others" saw a decreasing trend. The more substantial decline in caste groups primarily residing in the Terai region requires further ethnographic exploration. In addition, the reason for the statistical difference in annual coverage among speci c population groups may be partly due to a more substantial relative change in full vaccination coverage (increasing and then decreasing) between 2002 to 2016. It could also be due to health service or health belief related factors not included in DHS data.
We analyzed factors associated with a continuously decreasing trend in immunization coverage in province 2. Surprisingly, only maternal education was the independent predictor of full immunization coverage in province 2, and wealth quintile did not make any difference. Maternal education was found to be a strong predictor of immunization uptake in Nepal (7) and other settings (22).
The overall decline in full immunization coverage, coupled with a signi cant declining trend among the rich income quintiles has multiple implications. In one hand, as the coverage of full immunization starts to decline, the momentum of pro-poor equity gain may be reversed, and poor and disadvantaged groupswill be most likely to be missed out. On the other, a signi cant declining trend of full immunization coverage among rich income quintiles needs an urgent investigation and timely action. A study from Nepal has shown that improving the quality of the vaccination program maybe even more important than improving access to it (19). While improving access is essential to reach some sections of the populations, improving service quality is equally important.

Strengths and Limitations
We used nationally representative data from the DHS survey. Furthermore, we merged dataset from 3 recent DHS surveys conducted in 2006, 2011, and 2016, which increased the power of our regression analyses. We also used three inequality indicators, namely the ratio between Q5 and Q1, the rich-poor difference, and the concentration index for the analysis of equity gaps in full immunization coverage. However, the study has some limitations. Only six antigens administered during the infancy were considered for comparison of coverage over the years because some of the antigens recently introduced were not available during the period covered by earlier surveys conducted in 2006 and 2011.
Additionally, no supply-side factors were included in the models since DHS data lacks health service-related data. Variables related to socio-cultural practices, social norms, and beliefs regarding immunization were not available. Furthermore, we included income quintiles as a proxy measure of socioeconomic status. However, multiple aspects of poverty might re ect SES better than income quintiles. When vaccination cards were not available, interviewers relied on mothers' reports to determine receipt of immunization. Therefore, misclassi cation could have arisen if mothers did not correctly recall the name and receipt of the vaccine.

Conclusion
This study, utilizing data from three demographic health surveys from Nepal, demonstrates a recent decreasing trend in full immunization coverage in infants, moving away from the national target, while equity gain among the poor is still preserved. Besides, province 2 saw a signi cant drop in coverage, with maternal education being the independent predictor. This implies a need to redirect the focus of the national immunization program towards an overall increase in immunization coverage in the country, with additional measures in Province 2, particularly among motherswith no or primary education. Effectively reaching all sections of the populations is critical to attaining the goal of universal coverage at a fast pace.

Competing interests
The authors declare that they have no competing interests.

Funding
The authorsreceive no speci c grant from any funding agency in the public, commercial or not-for-pro t sectors. Lorentz curve showing income inequality in vaccination coverage over time