The European Journal of Public Health Advance Access originally published online on October 5, 2005
The European Journal of Public Health 2006 16(2):137-142; doi:10.1093/eurpub/cki194
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Infectious Diseases |
Understanding variation in measles–mumps–rubella immunization coverage—a population-based study
James A. Wright1 and Clare Polack2
1 Department of Geography, University of Southampton, Highfield, Southampton SO17 1BJ, UK
2 MRCGP, Division of Medical Education, University of Southampton, UK
Correspondence: James A. Wright, tel: +44 2380 594619, fax: +44 2380 593295, e-mail: j.a.wright{at}soton.ac.uk
Received February 3, 2005, accepted August 30, 2005
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionDisclaimerReferences |
|---|
Background: Coverage of the Measles–Mumps–Rubella combined vaccine (MMR) has declined in recent years in the UK, following adverse publicity about possible links between the vaccine, autism, and Crohn's disease. The objectives of this study were to assess geographical variation in trends in MMR coverage and to identify the factors affecting MMR uptake at population level. Methods: We conducted an ecological study of immunization coverage by second birthday, based on routinely collected data from 1993–2004 for England. Trends in MMR uptake were assessed in 95 District Health Authorities in England over the study period. We investigated the relationship between MMR immunization uptake and deprivation, ethnicity, education, population density, rurality, and socioeconomic class. Results: Since 2000, MMR coverage has declined significantly in virtually all areas of England. Population density and deprivation were both strongly correlated with low MMR uptake. The decline in coverage since 1993–94 was significantly related to the proportion of educated population and was greater in densely populated areas. Conclusion: Decline in MMR coverage now affects most areas of England. The lowest rates of MMR coverage remain in urban areas, particularly in inner cities, which also tend to show high levels of deprivation. Public health resources should continue to target inner city areas, as well as focus on the concerns of the better educated about vaccine safety.
Keywords: censuses, geographic information systems, immunization, measles
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDisclaimerReferences |
|---|
Overview of the controversy
A combined vaccine for measles, mumps, and rubella (MMR) was introduced in 1988.1 In recent years, public concern about a possible link between the vaccine and autism and/or Crohn's disease has led to reduced MMR coverage. The MMR vaccine is usually administered to children aged 12–15 months. As an alternative, however, concerned parents may opt for separate mumps, measles, and rubella vaccines or opt for no vaccination. Since separate vaccines are not available through the National Health Service (NHS), those parents who choose separate vaccines must pay privately for this option. Uptake of single vaccines through the private sector is not recorded in routine statistics. Thus, it is not possible at the population level to distinguish between those who opt for single vaccines and those who choose not to vaccinate their children whatsoever.
During the mid-1990s, a link was postulated between measles vaccination and Crohn's disease.2 Subsequently, in February 1998, a paper was published3 suggesting a link between the vaccine and autistic spectrum disorders (ASDs), which subsequently received substantial media coverage. Several subsequent studies and reviews concluded that there was no evidence of such a link.4–7 Such evidence led the Institute of Medicine's Immunization Safety Review Committee and the World Health Organisation's Global Advisory Committee on Vaccine Safety to conclude that there was no evidence for a causal relationship between MMR and ASD at population level and that there was no plausible biological mechanism for such a link.8–9 Despite these recommendations and attempts by the British government to reverse this decline by publishing health promotion materials, public concern about an MMR–autism link remains high in the UK.10
‘Herd immunity’ considerations
‘Herd immunity’ is the idea that provided that a sufficiently large proportion of a population has been vaccinated, the remaining unvaccinated population members will be protected from infection. At least 92–95% immunization coverage is necessary to prevent measles outbreaks from occurring in a given population through ‘herd immunity’.11,12 At least 90% immunization coverage is required to prevent mumps outbreaks12,13 and 83–90% coverage to prevent rubella outbreaks.12,14 In every population, some children (such as those undergoing chemotherapy) will be unable to receive the MMR vaccine, and in a small percentage the vaccine will fail to confer immunity. Provided a sufficiently large proportion of the population is vaccinated, however, such children are protected from infection by ‘herd immunity’. MMR coverage for England and Wales now stands at 86% and is <80% in London.15 This implies that there is a risk of increased MMR cases locally, but this has not yet been demonstrated statistically.
Overview of this study
This study examines the factors affecting MMR uptake at population level through an investigation of published vaccination coverage data in England and related census data.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDisclaimerReferences |
|---|
Data
Detailed statistics are publicly available for the percentage of children given the MMR vaccine by their second birthday (e.g. for 1999–200016). The statistics are available annually for children whose birthdays fall between April 1st and March 31st of the following year. The data are compiled from the Coverage of Vaccination Evaluated Rapidly (COVER) system and a summary KC50 form completed by individual health authorities.
Data sources for the other variables included in our analysis were derived from the 2001 census and the routinely collected governmental data on housing and vehicle ownership.
Hypotheses
The factors affecting the change in MMR coverage between 1997–98 and 2003–04 and affecting percentage coverage in 1997–98 and 2003–04 were examined separately. Geographic variation in percentage coverage was investigated in relation to the following factors:
- Vaccination coverage for other diseases. Previous studies17 suggest that MMR vaccination uptake is greater among parents whose children have previously received other vaccines. We therefore examined MMR percentage coverage in relation to vaccination coverage for other diseases.
- Ethnicity. Studies outside the UK18 have suggested that vaccination uptake may be lower in ethnic minority groups. Ethnicity was measured by the percentage of non-white population, according to the 2001 census.
- Social class. A previous study of measles immunization found that uptake was lower among social classes IV and V.19 We also postulated here that the reduction in MMR coverage would be greater in areas with a large proportion of high income families. Socioeconomic class was therefore measured through two variables. First, high income households were identified based on the proportion of households that contained higher managerial or professional workers, with dependent children aged 0–4 years. In addition to this census-derived measure, the percentage of High Income Families, according to the Experian MOSAIC classification,20 was identified. This classification draws on routinely collected housing and vehicle ownership data in addition to the 2001 census.
- Deprivation. Vaccination coverage has previously been shown to be marginally lower in deprived areas than elsewhere, once other factors have been accounted for.19 Deprivation was measured using the Townsend index of deprivation,21 derived from 2001 census data on unemployment, home ownership, overcrowding, and car ownership. A modified version of this index was calculated based on households with dependent children aged 0–4 years, rather than the whole population. The Townsend index was used in preference to other alternative measures because of its known relationship with health outcomes.22 Unlike some other deprivation measures, the index also excludes aspects of deprivation not directly related to young children (e.g. old people living alone). Its use enabled other factors that might be responsible for uptake, such as ethnicity, to be examined separately.
- Educational level. One previous study postulated that immunization uptake was greater among better educated parents.23 Educational level was therefore included here and measured through the percentage of the economically active population with no formal qualifications, again based on the 2001 census data.
- Population density. We calculated population density from the 2001 census data to account for this possible confounding factor.
- Rurality. We calculated the percentage of rural population according to a standard classification24 to account for rurality as a possible confounder.
Measles immunization rates have also been found to decline with increase in family size, both in the UK19 and elsewhere.25 However, we were unable to identify family size on the basis of census data.
Data collation
The geographical units used to collate immunization statistics were modified over the study period. Immunization figures between April 1993 and March 2002 were collated by District Health Authority (DHA) and within this period several smaller DHAs were merged together. From April 2002, immunization figures were collated for Primary Care Organisations (PCOs). We therefore estimated immunization rates by DHA for 2003 and 2004 using a technique known as areal interpolation. Specifically, we calculated an area-weighted average of the PCO immunization rates in 2003–04 and 2002–03 for each DHA.
Similarly, the geographical units used to collect census data, Experian MOSAIC data, and health data differ. To overcome this problem, we calculated the centroids of 2001 census wards and post-code sectors. We were then able to calculate aggregate figures for all wards or post-code sectors whose centroids lay within a given DHA or PCO.
Statistical analysis
Trend lines were fitted to the immunization figures for each DHA using stepwise regression by the method of ordinary least squares. Two alternative models were assessed within the stepwise regression: a quadratic trend and a linear trend. The quadratic trend term was included if significant at the P = 0.10 level. Two possible alternative approaches were discarded: piecewise regression and autoregression, on the grounds that there was insufficient data for any one DHA to use these techniques.
We used backwards stepwise Ordinary Least Squares (OLS) regression to assess the possible causes of variation in MMR coverage by DHA for 1996–97 (the year immediately preceding the publicizing of the link between MMR and autism) and by PCO for 2003–04. We also assessed the change in coverage between 1996–97 and 2003–04 by DHA. We tested for spatial autocorrelation of regression residuals via the robust Lagrange multiplier test26, using Stata 7.0. Where spatial autocorrelation was present, we used spatial lag and spatial error regression models by the method of maximum likelihood27 to correct for autocorrelation in the regression residuals. A row-standardized binary spatial weights matrix with a threshold of 80 km was used for this analysis.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDisclaimerReferences |
|---|
Analysis of trends in MMR coverage
Three types of trends in MMR coverage were identified among English DHAs:
- A significant quadratic trend (e.g. Oxfordshire—see Figure 1a), where coverage rose during the period of study, peaked, and then declined. A total of 53 DHAs exhibited this trend.
- A significant linear trend (e.g. Herefordshire—see Figure 1b), where coverage had already peaked by the start of the study and declined linearly thereafter. A total of 40 DHAs exhibited this trend.
- No trend (e.g. Lincolnshire), where unusually high or low estimates of MMR coverage in a particular year precluded the identification of any trends (2 DHAs). On inspection, it was apparent that this pattern was related to changes in reporting procedures and therefore exceptional. Thus, the 2 DHAs concerned were excluded from the regression analysis given below.
|
Based on these trends, we estimated the year of peak MMR coverage for each DHA. In all DHAs, the year of peak MMR coverage occurred in the year 2000 or earlier.
Assessment of factors affecting MMR coverage
The most recent percentage of MMR coverage data are shown in Figure 2.
|
Table 1 summarizes the results of the regression-based analysis of MMR coverage. Higher MMR coverage among DHAs in 1997–97 was significantly related to lower Townsend deprivation scores and to a lower proportion of households working in higher managerial or professional positions. There was no evidence of spatial autocorrelation in the regression residuals, and so a spatial regression analysis was not undertaken.
|
In 2003–04, higher MMR coverage among PCOs was significantly related to higher population densities and to a higher proportion of the active population with no qualifications. A cubic transformation was applied to 2003–04 MMR coverage data before fitting the regression model because of the non-linear nature of its relationship with deprivation. Residuals from an OLS regression showed spatial autocorrelation of residuals, and so a spatial lag model was fitted to the data. The spatial lag (coverage in neighbouring areas) was also statistically significant and this held true whether the threshold for identifying neighbouring PCOs was those within 80, 100, 120, or 140 km. A spatial error model produced similar results.
The change in MMR coverage between 1997–8 and 2003–4 among DHAs was also analysed. MMR coverage declined more slowly where the proportion of the active population without qualifications was high. A decline in MMR coverage was also significantly associated with high population densities. The OLS regression model showed significant spatial autocorrelation of residuals and a spatial lag model was fitted. The spatial lag (change in coverage in neighbouring areas) was a statistically significant predictor of MMR coverage decline.
The percentage MMR coverage by PCO in 2003–04 was significantly correlated with immunization for other major childhood diseases. The percentage immunization coverage for diphtheria, polio, tetanus, and pertussis was significantly related to MMR coverage (r = 0.76, P = 0.00, n = 294).
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDisclaimerReferences |
|---|
Findings in relation to earlier work
Little work has been undertaken at the population level to assess geographical variation in MMR uptake. Previous work28 has assessed MMR uptake among different GP practices in rural Scotland in terms of practice immunization procedures. However, these authors did not consider the possible effects of socioeconomic and demographic practice characteristics on the uptake.
Several previous studies have explored the reasons behind the decision to vaccinate for MMR among individual parents, either quantitatively through surveys or qualitatively through focus groups or interviews with individuals. Most of these studies have focused on parental beliefs,17,29–31 information flows,30 and/or previous immunization history.17
Although most of these UK-based studies have not assessed the influence of socioeconomic or demographic household characteristics on the uptake, studies in other countries have linked such variables to MMR uptake. In the USA, for example, parents with more children and African-American ethnicity were found to be less probable to immunize their children at the recommended time.18 In France, a sample survey of parents suggested that the number of children and the mother's age were significantly related to the decision to vaccinate,25 whilst in Australia neither educational level, economic status, nor ethnicity were related to MMR immunization status in children.23
Previous population-based studies of health have been criticized for their exclusive reliance on census data.32 In this study, in addition to the census we made use of Experian Mosaic data, which draws on housing and vehicle transactions when measuring the social class. We were also able to take account of the statistical problems caused by spatial autocorrelation (the tendency for neighbouring districts to share similar characteristics), through the use of spatial regression techniques.
Policy implications
To our knowledge, this study is the first to examine the socioeconomic factors influencing MMR uptake at population level in the UK. These results suggest that all of England is affected by the decline in MMR coverage. There do not appear to be any geographical ‘pockets’ that defy the national trend of declining coverage. This decline in coverage is of concern given the high measles threshold for ‘herd immunity’ of 92–95%. National serological surveillance data based on 
7000 serum samples a year has been undertaken in England in Wales since 1987. In 1997–98, the proportion of serum samples testing negative for measles antibody was 6.5% in the 3- to 20-year age cohort.33 More recent surveillance data are not yet publicly available and so the impact of decline in MMR coverage on herd immunity is presently unclear.
The areas with the lowest MMR coverage tend to be the inner cities with high population density. Many of these inner city areas exhibit high levels of deprivation. This is again a matter of concern, since young children in these deprived areas are likely to be in poorer health status and therefore at a greater risk of developing measles complications.
There is some evidence here that the decline in MMR coverage has been less marked in areas with more poorly qualified populations. For example, Kensington in central London with a small unqualified population has shown a much more rapid decline in MMR coverage than the East London DHA, which has a higher proportion of unqualified population. There may thus be some justification in targeting MMR health promotion materials at the more educated sectors of the population. Nonetheless, the problem of low MMR coverage remains greatest in the most deprived areas and targeting of these areas should remain a priority.
Limitations in the study design
These conclusions may be affected by several limitations in the study design. Because of possible confounding, we are unable to discern whether low MMR uptake is due to deprivation or some other related but unmeasured characteristic of inner city areas. It is also well known that the geographic units used to summarize the data can influence study findings: had we used areal units other than DHAs, our findings might have been different. Our findings are also subject to the ‘ecological fallacy’ in that the characteristics of those who do not immunize their children within a given DHA may differ from the aggregate characteristics of the district as a whole.
Because of the limited time series available for each DHA, the method used to fit trends does not take account of serial autocorrelation in the time series. The use of centroids to aggregate census and post-code data assumes that all spatial units are entirely contained within a single PCO or DHA. In reality this is not the case. Similarly, the method of areal interpolation used to aggregate PCO data assumes that population is uniformly distributed within each DHA. In reality this is unlikely to be true. However, neighbouring DHAs had similar MMR coverage, which would lessen the impact of this problem.
Future work
The same techniques could be applied to immunization for other diseases or extended to earlier MMR coverage in 1988–93. Furthermore, the possible linkages between deprivation and educational level require a more detailed investigation at the household level rather than the population level.
|
Disclaimer |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDisclaimerReferences |
|---|
Clare Polack also works as a general practitioner in Hampshire, UK.
Key points
|
|
Acknowledgments |
|---|
The authors acknowledge the suppliers of the data used in this study, including the National Digital Archive of Data (NDAD), Experian Ltd, and MIMAS. This work is based on the data provided through EDINA UKBORDERS, with the support of the ESRC and JISC, and uses boundary material which is copyright of the Crown. Source of support: University of Southampton internal funding.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDisclaimerReferences |
|---|
1 Bedford H, Elliman D. Childhood immunisation. London: Health Education Authority, 1998.
2 Thompson NP, Montgomery SM, Pounder RE, Wakefield AJ. Is measles vaccination a risk factor for inflammatory bowel disease? Lancet 1995;345:1071–74.[CrossRef][ISI][Medline]
3 Wakefield AJ, Murch SH, Anthony A, et al. Ileal-lymphoid nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children. Lancet 1998;351:637–41.[CrossRef][ISI][Medline]
4 Madsen KM, Hviid A, Vestergaard M, et al. A population-based study of measles, mumps, and rubella vaccination and autism. N Engl J Med 2002;347:1477–82.
5 Honda H, Shimizu Y, Rutter M. No effect of MMR withdrawal on the incidence of autism: a total population study. J Child PsycholPsychiatr 2005;46:572–79.
6 Wilson K, Mills E, Ross C, McGowan J, Jadad A. Association of autistic spectrum disorder and the measles, mumps, and rubella vaccine—A systematic review of current epidemiological evidence. Arch Pediatr Adolesc Med 2003;157:628–34.
7 Parker SK, Schwartz B, Todd J, Pickering LK. Thimerosal-containing vaccines and autistic spectrum disorder: A critical review of published original data. Pediatrics 2004;114:793–804.
8 Stratton S, Gable A, Shetty P, McCormick M (eds). Immunization safety review: measles-mumps-rubella vaccine and autism. Washington D.C.: National Academies Press, 2001.
9 Folb PI, Bernatowska E, Chen R, et al. A global perspective on vaccine safety and public health: The global advisory committee on vaccine safety.Am J Public Health 2004;94:1926–31.
10 Bellaby P. Has the UK government lost the battle over MMR? Br Med J 2005;330:552–53.
11 Hutchins SS, Baughman AL, Orr M, et al. Vaccination levels associated with lack of measles transmission among preschool-aged populations in the United States, 1989–1991. J Infect Dis 2004;189:S108–15.
12 Anderson RM, May RM. Infectious diseases of humans: dynamics and control. Oxford: Oxford University Press, 1992.
13 Gupta RK, Best J, MacMahon E. Mumps and the UK epidemic 2005. Br Med J 2005;330:1132–35.
14 Stevenson J, Murdoch G, Riley A, et al. Implementation and evaluation of a measles/rubella vaccination campaign in a campus university in the UK following an outbreak of rubella. Epidemiol Infect 1998;121:157–64.[Medline]
15 Public Health Laboratory Services. Measles outbreak in south London. CommunDis Rep CDR Wkly 2002;12: news.
16 Government Statistical Service. National Immunisation Statistics, England: 1999–2000. GSS Bulletin 2000, 2000/26.
17 Flynn M, Ogden J. Predicting uptake of MMR vaccination: a prospective questionnaire study. Br J Gen Pract 2004;54:526–30.[Medline]
18 Trauth JM, Zimmerman RK, Musa D, et al. ‘Do beliefs of inner-city parents about disease and vaccine risks affect immunization?’. J Natl MedAssoc 2002;94:820–32.
19 Peckham C, Bedford H, Senturia Y, Ades A. National immunisation study: factors influencing immunisation uptake in childhood. London: Action Research for the Crippled Child, 1989.
20 University of Edinburgh. About Great Britain MOSAIC. Available at: http://datalib.ed.ac.uk/EUDL/GBmosa.html (accessed on December 7, 2004).
21 Townsend P, Phillimore P, Beattie A. Health and deprivation: inequality and the North. London: Croom Helm, 1988.
22 Hoare, J. Comparison of area-based inequality measures and disease morbidity in England, 1994–1998. Health Stat Q 2003;18:18–24.
23 Skinner SR, Nolan T, Bowes G. Measles-mumps-rubella and hepatitis B vaccination uptake in adolescents: a survey in metropolitan Melbourne.Med J Aust 1998;168:546–49.[Medline]
24 Bibby P, Shepherd J. Developing a new classification of urban and rural areas for policy purposes—the methodology. London: Office for National Statistics, 2004.
25 Rotily M, Guagliardo V, Fontaine D, et al. Evaluation of measles mumps, and rubella vaccine coverage in 3 year old children in 12 French counties. Time-trends and related factors. Rev Epidemiol Sante Publique 2001;49:331–41.[Medline]
26 Anselin L, Bera AK, Florax R, Yoon MJ. Simple diagnostic tests for spatial dependence. Reg Sci Urban Econ 1996;26:77–104.
27 Anselin L, Hudak S. Spatial econometrics in practice: a review of software options. Reg Sci Urban Econ 1992;22:509–36.[CrossRef][ISI]
28 Henderson R, Oates K, Macdonald H, et al. Factors influencing the uptake of childhood immunisation in rural areas. Br J Gen Pract 2004;54:114–18.[Medline]
29 McMurray R, Cheater FM, Weighall A, et al. Managing controversy through consultation: a qualitative study of communication and trust around MMR vaccination decisions. Br J Gen Pract 2004;54:520–25.[Medline]
30 Smailbegovic MS, Laing GJ, Bedford H. ‘Why do parents decide against immunization? The effect of health beliefs and health professionals.’ Child Care Health Dev 2003;29:303–11.[CrossRef][ISI][Medline]
31 Evans M, Stoddart H, Condon L, et al. Parents' perspectives on the MMR immunisation: a focus group study. Br J Gen Pract 2001;51:904–10.[ISI][Medline]
32 Roux AVD. People, places and health. Am J Epidemiol 2002;156:516–19.
33 Vyse AJ, Gay NJ, White JM, et al. Evolution of surveillance of measles, mumps, and rubella in England and Wales: providing the platform for evidence-based policy. Epidemiol Rev 2002;24:125–36.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Pearce, C. Law, D. Elliman, T. J Cole, H. Bedford, and the Millennium Cohort Study Child Health Group Factors associated with uptake of measles, mumps, and rubella vaccine (MMR) and use of single antigen vaccines in a contemporary UK cohort: prospective cohort study BMJ, April 5, 2008; 336(7647): 754 - 757. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Elliman and H. Bedford MMR: where are we now? Arch. Dis. Child., December 1, 2007; 92(12): 1055 - 1057. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||