Measles transmission from an anthroposophic community to the general population, Germany 2008

At the end of March 2008, physicians notified 15 persons with measles disease residing in two neighbouring Bavarian counties to the respective local health authorities. Of those, 14 attended an Austrian anthroposophic school with an ongoing measles outbreak since the beginning of March [3, 11]. The affected German counties share a border with Austria, hence regular school attendance in areas across the border is not unusual. The Austrian outbreak investigation team found a 34% (101 of 294) coverage with at least one dose of measles vaccine among students attending the anthroposophic school in Salzburg city [3]. Other investigations in European countries had already yielded, that immunisation against measles, mumps, and rubella (MMR) was lower in students attending anthroposophic schools than in those attending non-anthroposophic schools [12][13]. Hence, these students may form a pocket of susceptible persons.

Until mid-April 2008, the number of measles cases possibly related to the outbreak in the Austrian school increased to 55, affecting the population of four neighbouring Bavarian counties. Between 2001 and 2007, a median of seven measles cases per year had been notified in the four counties. The average vaccination coverage of first graders for two doses of measles vaccine in the four counties was 69% (range 64-72%) in 2006/2007. We report about the outbreak investigation, which was initiated to describe the public health implications of the measles outbreak in Bavaria, to explore the reasons for non-vaccination and to compare the effect of implemented control measures in order to guide future public health action.

Measles is a notifiable disease in Germany. Physicians must report any suspected case, clinical cases of and deaths from measles within 24 hours to the responsible local health authority. Heads of laboratories are obliged to report any direct or indirect evidence of measles virus --- if the evidence suggests an acute infection - within 24 hours to the responsible local health authority. The German local health authorities routinely collect case based data, e.g. on demographics, clinical symptoms, hospitalisation, vaccination status, travel history and laboratory testing. Contact tracing is also performed as a standard procedure. Local health authorities of two Bavarian counties accepted self reporting of the vaccination status by patients while the other two Bavarian counties requested verification by physicians or health inspectors. In this outbreak investigation, we defined a case as any resident of the four affected Bavarian counties (Berchtesgadener Land, Rosenheim, Traunstein, Mühldorf am Inn), diagnosed with clinical measles and disease onset between February 1 and August 30, 2008. Clinical measles was defined as a generalised maculopapular rash persisting for at least 3 days, fever ≥ 38.5°C, and at least one of the four following symptoms: conjunctivitis, cough, runny nose or Koplik's spots. Cases were further categorized as having a link with the Austrian outbreak or not. We considered "attending school in Austria", "visiting Salzburg city" or "having contact with a measles case from Salzburg city" during the incubation period as potential links to the Austrian outbreak.

Cases were notified by physicians and laboratories to the respective local health authority (LHA) which entered the information into the electronic surveillance system. Additionally, the LHA collected data on reasons for non-immunization and complications from all notified cases through a questionnaire administered by telephone or post. If patients were under age 16, parents were queried. Study participation was voluntary. We assumed implicit informed consent when a completed questionnaire was returned. Information on non-responders was available.

We used EpiData software version 3 (EpiData Association, Odense, Denmark) for double data entry, and Stata (Stata Statistical Software: Release 10, Texas, USA) for descriptive analysis.

Laboratory confirmation using serology was performed by regional laboratories. The National Reference Centre for Measles, Mumps and Rubella in Berlin (NRC) used the Enzygnost Anti-Measles Virus IgM ELISA (Siemens, Germany) for the detection of anti-measles IgM in serum. A subset of samples was genotyped by the NRC according to the WHO recommendation for measles virus [14]. As described in Santibanez et al., 2002 [15], viral RNA was extracted from urine or throat swab specimens and reversely transcribed into cDNA. A C-terminal 450 nt fragment corresponding to the hypervariable 150 aa of the N-protein of measles virus was amplified by PCR and subsequently sequenced. The sequences were phylogenetically analysed.

Cases with disease onset a maximum of 18 days apart and spatial contact (e.g. same household, same school) were grouped into clusters.

In Germany, the LHAs are responsible for the implementation of control measures. In order to assess the effectiveness of implemented control measures, the Robert Koch Institute interviewed the LHAs by telephone. Our interviews of the local health authorities revealed that they applied two different control measures in schools and kindergartens (preschool) during the outbreak: the trigger for exclusion of non-immune persons from the respective school or kindergarten for 14 days after the last contact to an infectious case was either the notification of at least one measles case (intervention A) or at least two measles cases (intervention B) in the respective school or kindergarten. Immune persons were those with at least one documented vaccination against measles a minimum of three weeks before disease onset, immunity confirmed by serology or anamnestic measles. Intervention A and B were implemented compulsory. Compliance was not controlled for. Vaccination cards in schools and kindergartens were checked by staff from the LHAs in three of the affected counties, and by staff from the respective school or kindergarten in the fourth county.

Attack rates were calculated and negative binomial regression was used to estimate the average attack rate ratio with 95% confidence intervals (95% CI) according to the implemented measure. Teachers and staff of schools and kindergartens were not included in the denominators of the specific institutions.

In order to analyse the effect of the respective intervention we excluded the case with the first disease onset from all clusters in which intervention A was applied and the two cases with the first disease onsets from all clusters where intervention B was used. The duration of clusters by intervention was calculated starting with the onset of the respective intervention. The school cluster in the anthroposophic school in Salzburg city (since Austrian health authorities were responsible for the intervention) and clusters in kindergartens and schools with all cases occurring on the same day (0 days apart, mode of intervention did not differ) were excluded from analysis.

We compared the median number of cases per cluster and the median duration of clusters in days according to the implemented intervention using the Mann-Whitney-U-Test. Using a Poisson regression model we estimated the average decline of cases per 7-day interval. We categorized the days of disease onsets in 7-day intervals, starting with the onset of the second case in clusters with intervention A and of the third case in clusters with intervention B as day 1.

From 14 March to 15 July (calendar weeks 11-29) 2008, we identified 217 measles cases in the four counties. No case was found in teachers or staff of schools and kindergartens. In the outbreak period, the four counties had an average measles incidence of 32 cases per 100,000 population (range: 12-71) which was the highest among all Bavarian counties (overall: 2 cases per 100,000 population, Figure 1). Six weeks before and after the outbreak period, no other measles case was reported within the four counties. For the 215 measles cases with available information on date of onset and date of reporting the mean reporting delay was 6 days (range: 0-28 days); the reporting delay did not differ among the four counties.

Fifty-one cases (24%) could be linked to the ongoing measles outbreak in Austria. Among those, 28 attended the same anthroposophic school in Salzburg city. All linked cases occurred in the first seven weeks of the outbreak. Case numbers peaked in May, with 28 and 27 cases respectively in week 19 and 21. After that, case numbers decreased continuously (Figure 2).

The median age of cases was 11 years (range: 0-57), 47% were male. The incidence rate was highest in persons aged 10-14 years (170 cases/100,000 person-years), followed by those 0-4 years (150/100,000) and 5-9 years (140/100,000) of age (Figure 3).

Overall, 25 of 217 (11%) cases were hospitalised for a median of four days (range: 1-13 days). The hospitalised patients had a median age of 16 years in contrast to 10 years for the non-hospitalized group (p = 0.001).

Of the 217 cases, 161 (74%) responded to the questionnaire. The non-responder analysis revealed geographic differences: while one county managed a response of 100%, the others achieved 56%, 60% and 62% (P < 0.001). The non-responders did not differ regarding age, sex, vaccination status or epidemiological link to Austria.

Twenty-nine cases (18%) of 161 developed complications; most frequently, otitis media was reported (n = 14). A 13-year-old girl, who was hospitalised for five days, suffered from seizures (Table 1).

Overall, 156 (97%) of 161 measles cases filling in the questionnaire were unvaccinated. Four persons had received one measles vaccine dose, one patient two doses. All vaccinations had been applied at least one year before symptom onset.

The main reasons for not being vaccinated against measles were "fear of vaccine-related adverse events" (33%), "opposing measles vaccination in general" (30%) and the opinion "measles is not a severe disease" (18%) (Table 2).

In total, 54 (25%) of 217 cases were laboratory confirmed by serology. Two of the five vaccinated persons, including the patient vaccinated twice were among the confirmed cases. Virus isolates from 28 cases were genotyped. In two cases, sequencing of the viral genome was not successful, while in the remaining 26 cases, identical sequences revealing genotype D5 were detected. This virus variant was found in all four counties and included patients with and those without links to Austria.

A total of 184 (85%) cases could be epidemiologically grouped to 59 clusters. These could be linked to 41 households, 13 schools or kindergartens, four hospitals or medical practices and one firemen's festival (Table 3). A total of 32 cases were linked to the use of a school bus which carried children and teenagers to and from three of the affected schools. However, all cases linked with the school bus were included in the respective school clusters.

Overall, we included 10 clusters, three in kindergartens and seven in schools, with 85 cases altogether from all four counties for the comparison of the effect of intervention A and B (Table 4). Two schools had anthroposophic teaching methods. Local health authorities recommended intervention A during six clusters (28 cases) and intervention B during four clusters (57 cases), based on their own decision.

Three counties implemented solely intervention A. The fourth county used intervention B until calendar week 22 (12 weeks) and intervention A since calendar week 23 (7 weeks, Figure 2). The county using intervention B informed as early as the occurrence of the first incident measles case in a kindergarten or school all contact persons about the incident measles case and the necessity of protection by vaccination.

Both, the attack rates as well as the average attack rate ratios were calculated but should be interpreted with caution since the population at risk (number of susceptible contacts per school or kindergarten) was not available: The mean attack rate in institutions was 2.7%. Stratification by intervention led to a mean attack rate of 1.9% (range: 1-16%) if intervention A was used and 3.3% (range: 1-21%) if intervention B was applied. Average attack rate ratios in institutions where intervention A was implemented were 1.6 times lower (95% CI: 0.4-6.1) than those of institutions where intervention B was applied.

For the following analyses - as described in the methods section - we neither counted the first case of clusters A nor the first two cases of clusters B; furthermore, the calculation of the cluster duration in days started with the first day of the respective intervention.

Clusters in kindergartens and schools had a median number of 3 (intervention A; range: 1-7) versus 13 (intervention B; range: 2-21) cases (p = 0.05), and lasted a median of 3 (intervention A; range: 0-36) versus 26 (intervention B; range: 3-66) days (p = 0.13).

The estimated mean decrease of cases per 7-day interval was 58% using intervention A and 29% intervention B, indicating that the average number of cases per 7-day interval decreased 1.7 times (95% confidence interval: 1.1-2.7; p = 0.03) slower using intervention B compared to intervention A (Figure 4).