Religious subgroups influencing vaccination coverage in the Dutch Bible belt: an ecological study

An ecological study at municipality level was performed. All 458 municipalities in the Netherlands (reference date 01-01-2006) were included. As in the Netherlands municipal merging is an ongoing process and as in small municipalities churches may attract believers from neighbouring municipalities, in the provinces Zuid-Holland, Utrecht and Gelderland municipalities were aggregated for this study. In these provinces municipalities with less than 15.000 inhabitants were aggregated according to existing plans for municipal merger or according to geographical entities like (former) islands and polders. In this way 36 municipalities were aggregated to 10 geographical entities. Thus the study includes 432 municipalities and geographical entities, comprising all inhabitants of the Netherlands.

In this study the Bible belt is defined as all municipalities and geographical entities where one or more OPDs are established (irrespective of percentage of votes for SGP).

Some variables had a somewhat skewed distribution, which leaves the use of parametric tests open to discussion. Therefore we performed non-parametric tests as well (N.B. both tests led to similar conclusions).

First, for all municipalities and geographical entities in the Netherlands, the effect of presence or absence of OPDs on vaccination coverage was assessed by comparing mean vaccination coverage using the independent samples t-test. As Levene's test for equality of variances was significant (p < 0.05) homogeneity of variance was not assumed, and therefore we used the corrected independent samples T-test in SPSS. Since the distribution of some variables was somewhat skewed we performed the Mann-Whitney test as well.

Second, for municipalities and geographical entities where one or more OPDs were present, the relationship between vaccination coverage and the explaining variables (membership ratios of the orthodox protestant denominations, urbanization, socio-economical status and immigration) was analysed by bivariate correlation (Pearson's r) and multiple regression analysis, using a backward selection method (removal criterion p > = 0.1). Since the distribution of some variables was somewhat skewed, Spearman's rho test was performed as well. Multiple regression analysis was repeated without the outliers responsible for skewed distribution of some variables. The residuals were all independent and normally distributed, there was no heteroscedacity and no collinearity.

Finally, to compare the influence of the membership ratios of various OPDs to the influence of over 5% votes for SGP the bivariate and multiple regression analyses were repeated with the variable >5% votes for SGP replacing the membership ratios of the OPDs.

Overall the OPDs in the Netherlands had almost 220,000 members. This means that 1.3% of the Dutch population is member of one of the OPDs. The membership numbers of the various OPDs on national level are shown in table 1.

The 432 municipalities and geographical entities in our study had a mean population of 36,781 inhabitants. In 135 of these municipalities and geographical entities one or more OPDs were established (Table 2). Their geographical distribution is shown in Figure 1.

In only 41 (30%) of these 135 municipalities and geographical entities the percentage votes for SGP was over 5%. Almost a quarter of the orthodox Protestants is living outside municipalities and geographical entities with more than 5% votes for SGP.

Including all municipalities and geographical entities, mean vaccination coverage was 95.8% (SD 3.5). In the 297 municipalities without OPDs mean vaccination coverage was 96.9% (SD 2.1) whereas in the 135 municipalities and geographical entities where one or more OPDs were established mean vaccination coverage was 93.5% (SD 4.7). The mean vaccination coverage of municipalities with at least one OPD (93.5%) is statistically significant lower than the mean vaccination of municipalities without OPDs (96.9%) (P < 0.001). As the number of OPDs established in a municipality or geographical entity increases, mean vaccination coverage decreases (Table 2).

In municipalities and geographical entities where one or more OPDs were established we assessed the influence of the individual OPDs on vaccination coverage, as well as the influence of urbanization, immigration and socio-economical status.

In table 3 for the 135 municipalities and geographical entities with OPDs, the bivariate correlations, using Pearson's r, between the vaccination coverage and the independent variables are shown. As expected, membership ratios of all OPDs have negative correlations with vaccination coverage, meaning that higher membership ratios are related with lower vaccination coverage. Level of urbanization showed the expected positive relation: meaning that in urban areas the vaccination coverage was higher than in rural areas. At first sight the proportions of non-western immigrants and of people dependent on income support (indicating socio-economical status) showed unexpected positive relations with vaccination coverage. This can be explained, however, because non-western immigrants and people dependent on income support mainly live in urbanized areas where the OPDs are underrepresented. Repeating bivariate correlation using Spearman's rho gave comparable results, except for the Christian Reformed Churches (rho = - 0.17, p = 0.053) and for level of urbanization (rho = 0.13, p = 0.131)

Table 3 also shows the result of a multiple regression analysis, using a backward selection method (removal criterion p > = 0.1). Level of urbanization and socio-economical status did not have any explanatory effect. A percentage of 84 of the variance in vaccination coverage was explained by membership of the various OPDs. The b-values all showed the expected negative sign but varied for the various denominations. The largest denominations -the Reformed Congregations and Restored Reformed Church- both had b-values around -0.40. This implies that 1 per cent point increase in membership ratio is associated with only 0.40 per cent point decrease in vaccination coverage. For the Reformed Congregations in the Netherlands and the Old Reformed Congregations, b-values exceeded minus 1, which implies that 1 per cent point increase in membership ratio is associated with even more than 1 per cent point decrease in vaccination coverage. Immigration had a significant, but very small explanatory effect; the total explanation only increased to 85%. The b-value now showed the expected negative sign.

Seven municipalities or geographical entities, all strongholds of certain OPDs, were recognized as outliers. Compared to the other municipalities and geographical entities they had an extremely high membership ratio for one OPD, which might have had an undue influence on the analysis. Repeating the analysis in 128 municipalities or geographical entities, leaving out the strongholds, 73% of variance in vaccination coverage could be explained. Again level of urbanization and socio-economical status had no significant effect, and here the membership ratio of Christian Reformed Churches was removed as well. All other results are comparable with the analyses on the basis of 135 geographical entities (Table 3).

In order to check if indeed the various OPDs were influential or merely the total membership ratio of all OPDs in the municipality, we repeated the analyses replacing the membership ratios of the various OPDs by the total membership ratio of all OPDs together. However, in this way the explanation of variance in vaccination coverage was 77% (versus 85%) in the group of 135 municipalities and geographical entities, and 64% (versus 73%) in the group of 128.

As the Bible belt is often defined as municipalities with more than 5% votes for SGP, we repeated the bivariate correlation and multiple regression analysis replacing the membership ratios of the OPDs by the dichotomized variable more than 5% votes for SGP. In bivariate correlation more than 5% votes for SGP was, as expected, negatively correlated to vaccination coverage. Multiple regression analyses showed a negative b-value as well, however the explanation of variance in vaccination coverage was only 45% in the group of 135 geographical entities, and 43% in the group of 128, (Table 4).

To assess the influence of the membership ratios of the various OPDs on vaccination coverage in municipalities with less than 5% votes for SGP, we repeated the bivariate correlation and multiple regression analyses in these municipalities and geographical entities where OPDs were present (n = 94). In that case 26% of variance in vaccination coverage could be explained, the membership ratios of the two most conservative OPDs (the Reformed Congregations in the Netherlands and the Old Reformed Congregations) and the percentage non-western immigrants had a significant influence. The b-values for both conservative OPDs exceeded again minus 1.

In multiple regression the largest denominations - the Reformed Congregations and Restored Reformed Church - both had b-values around -0.40. Although our analysis was not at an individual level, this finding is a strong indication that a substantial part of the members of these denominations is vaccinated.

Two smaller denominations - the Reformed Congregations in the Netherlands and the Old Reformed Congregations - had the strongest negative relation with vaccination coverage. This is in line with the negative statements on vaccination in their church periodicals [11]. In multiple regression analysis for both denominations the b-value exceeds -1. This could be explained assuming that the members of these two denominations not only reject vaccination for themselves but that they influence others (from their own and other denominations) to reject vaccination as well. Another possible explanation for the b-values exceeding minus 1, however, is the age distribution within these two denominations. In the Netherlands as well as in the United States orthodox protestants refrain from family planning, their families are large and the young members of the denomination outnumber the older members [6, 21]. As in our study vaccination coverage is assessed at two years age and membership proportion is according to the total population, the proportion members of these denominations in the two years age cohort might exceed the proportion in the total population. Nevertheless, the findings strongly suggest that vaccination coverage in these denominations is very low.

The orthodox protestant subgroup of the Christian Reformed Churches is the smallest denomination in our study. The b-value in the multiple regression analysis suggests that a substantial part of the members are vaccinated. In our second analysis, leaving out among others two strongholds of the Christian Reformed Churches, their influence on vaccination coverage was not significant anymore.

As in current policy the Bible-belt is often defined as municipalities with more than 5% votes for SGP, we replaced the membership ratios of the OPDs by the variable more than 5% votes for SGP. However, in this way the explanatory effect regarding variance in vaccination coverage is considerably lower.

The present study has some limitations. In this ecological study analyses could only take place on the municipality level, which hinders drawing conclusions on the individual level. Moreover municipalities were included regardless of their population number, and multiple regression analyses were not weighted for municipal population size. This means that small and large municipalities have an equal impact on the results. The aim of our study is, however, to assess the relationship between the membership ratios of the various OPDs (measured as percentage) and the municipal vaccination coverage (measured as percentage as well). The relationship itself (measured as b-value) is not dependent on the size of the municipality and we corrected for possible confounding factors such as urbanization, immigration and low socio-economical standard. Thus multiple regression analyses weighted for municipal population size would not affect the conclusions regarding the b-values of the various OPDs.

Misclassification may occur if members of OPDs do not live in the municipality where their church is seated. We partly corrected for this problem by aggregation of small and medium sized municipalities to geographical entities like (former) islands and polders. However, municipalities where an OPD is seated might have been allocated more members of that OPD than really live in this municipality. Vaccination coverage is always measured according to home address. Since the negative correlations between OPD and vaccination coverage probably would have been even more pronounced when both variables would have been measured according to home address, this inconsistency does not interfere with our conclusions.

Finally, variance in vaccination coverage could not be completely explained by the variables in our model. In international literature low vaccination coverage is related to lack of health insurance, lack of reimbursement and lack of long term preventive care [17, 22, 23]. In the Netherlands this is not expected to be a problem as vaccinations according to the NVP are provided free of charge. Travelling communities, like Roma and Irish travellers are associated with outbreaks of vaccine preventable diseases [24]. However, in the Netherlands travelling communities are not expected to influence municipal vaccination coverage as these travelling communities are small (an estimated 20,000 persons) and vaccination coverage among children travelling with fairs was comparable to the general Dutch population[25]. Finally, apart from religious objections to vaccination, parents may have philosophical objections or a critical attitude towards vaccination because of perceived side effects [26, 27]. These non-religious objections are often confined to MMR-vaccination as measles, mumps and rubella are considered to be useful diseases to strengthen the immune system [28]. As these objections are not registered we could not take them into account. However, we tried to minimize their influence by choosing DTPP-vaccination coverage as the dependent variable. Orthodox protestant objections concern all vaccinations.

Until now preparedness for epidemics of vaccine preventable diseases in the Netherlands has been focused on the geographical area of the Bible belt, often defined as municipalities with over 5% votes for SGP. However, 23% of the OPD-members is living outside this area and our study showed that in municipalities with less than 5% votes for SGP, where OPDs are present, the membership ratios of the most conservative OPDs still have a significant influence on vaccination coverage. In current policy, orthodox Protestants living outside the area defined as Bible belt are not addressed by health promotion and vaccination campaigns during epidemics. The orthodox Protestants constitute a closed community maintaining almost all social contacts within their own group [6] and orthodox Protestants living outside the municipalities with more than 5% votes for SGP are at considerable risk for infection during epidemics. Therefore we suggest to include them in prevention and control measures. As the orthodox Protestants have a strong social infrastructure, public health workers may seek cooperation with orthodox protestant intermediaries like schools or patients' associations, in order to prevent and control outbreaks of vaccine preventable diseases.