In this study we followed pneumococcal carriage in 59 day care attendees, their family members and the employees in three Finnish day care centres for 9 months. We showed that exposure to pneumococci both within day care centre (DCC) and within family are important risk factors for acquisition of carriage in children when measured with the same accuracy. Exposure confined to the DCC or family only signified a 5-fold rate of acquisition compared to no exposure in the family nor DCC. Simultaneous exposure both in DCC and in family further increased the relative rate to almost 30-fold. Pneumococcal carriage was more prevalent in DCCs than in families, and consequently the majority of acquisitions in the day care attendees was associated with carriage of the homologous type in the DCC. As a whole, pneumococcal carriage appeared highly clustered. Each of the three day care centres, together with the cohort of individuals linked to it, was dominated by a serotype of its own (19F, 18C, 9V).
We have formerly shown that attendance to day care, as a proxy for exposure, was not associated with risk of carriage in young children when simultaneous type-specific carriage by family members was known . This lack of association was probably due to the non-specificity of day care attendance as a measure of exposure in comparison to explicit knowledge of simultaneous within-family carriage at the serotype level. The children in the earlier study were also younger and attended communal day care less than the children in the present one. The main motivation of the present study was to gather data about exposure to pneumococci on equal footing from families and day care. With these new data, we could indeed show that acquisition of carriage in children is strongly associated with previous carriage of homologous serotypes in both mixing groups.
Although exposure to pneumococci within families increased the risk of pneumococcal acquisition in the day care attendee, exposure was highly more prevalent in day care facilities than families. Consequently, acquisition of pneumococcal carriage by the day care attendee was strongly associated with previous exposure to a homologous type in the DCC: in the 36 acquisitions with known exposure within the family or the DCC, the child had been exposed in the DCC in 35 cases and in the family in 9 cases (Table 4).
The serotype distribution in this study clearly deviates from the expectation based on contemporaneous or preceding studies of pneumococcal carriage in the same location[10, 22, 23]. In general, most studies in Finland and elsewhere have revealed a surprisingly stable serotype distribution, with the same, so called paediatric serotypes appearing prevalent in young children (6B, 6A, 19F, 23F). Most strikingly, serotype 23F was completely absent in day care cohorts in our material and types 6A and 6B were rare. The skewed serotype distribution in the present study is due to the fact that we sampled day care cohorts rather than individuals. The observed pattern of carriage can be interpreted as manifestations of micro-epidemics, that is, clustering of serotype specific carriage in the mixing groups.
The almost complete absence of many common serotypes in our material actually strengthens its micro-epidemic interpretation, with low background (community) rates of acquisition for single serotypes and much higher within-group rates (see e.g.). In particular, the fact that a follow-up of 59 children for 9 months did not result in a single observation of 23F carriage implies that the acquisition rate from the community even for this type is low. In fact, the rate of 0.0044 per month (Table 4) indicates a mean waiting time of 3.9 months until appearance of a new pneumococcal serotype in a cohort of 59 children, which is not in gross contradiction with the observed absence of 9 months for 23F.
In most family studies, the reported serotype distribution has resembled that from a random sample of individuals, the "common" serotypes being most prevalent [2, 6, 10, 25, 26] even though clear pneumococcal clusters have been seen in large families . An obvious explanation is the fact that a family/household is too small a unit to exhibit extended micro-epidemics. By contrast, some previous studies based on day care centres have specifically reported clustering of pneumococcal carriage among children attending day care [28, 29]. A larger group size and a bigger proportion of children, who are more prone to carry, allow pneumococcal microepidemics to be seen in studies reporting carriage in day care. In fact, school studies, especially cross sectional ones, with class size 20 and roughly 10% carriage, would be able to present only clusters of size 2 on average. This was actually seen also in our data (2 serotype 10 isolations in one class), which intuitively did not appear to show pneumococcal clusters.
The frequency of carriage in children was lower (mean 25%, range at the ten sampling rounds to the DCCs 14–44%) in our material than in many other published studies, taking into account that children attending day care are particularly assumed to carry pneumococcus more than children cared for at home. The day care attendees in our study were on average 4 years old, thus being older than in most of the studies concentrating on the first pneumococcal acquisition and hence sampled during the first 2 years of life, a period of heavier carriage. When comparing to other carriage data collected from the same location previously or simultaneously, the frequency of carriage in the present study is of the same magnitude: 25% vs. 28%  and 30% although somewhat lower than expected as all children in the present study were attending day care.
There are obvious limitations to our study. First, a considerable proportion of day care attendees were not enrolled in the study and the information about exposure to pneumococci within day care is therefore not complete. With the assumption that sampling and transmission were not associated, however, the un-sampled group of day care attendees should have harboured similar serotypes as the sampled one. Moreover, the risk of acquiring pneumococcus was clearly lowest in the 'non-exposure' class (Table 4), implying that large amount of exposure cannot have remained hidden in the un-sampled children of the DCC. The strong clustering of carriage within DCC obviously means that most circulating serotypes were found.
Second, a study based on only three day care centres is inevitably small to fully characterize the micro-epidemic potential of particular pneumococcal serotypes. The serotypes identified as dominant in the present study, i.e., those found to cause micro-epidemics, are not necessarily the most transmissible ones. It is probable that e.g. serotypes 23F, 6A and 6B transmit in a similar fashion in the child population. In fact, we conjecture that practically all relatively common serotypes can form clusters. Therefore, a larger sample of DCCs would have presented clusters of the "usual" serotypes. Although serotype 3 appeared somewhat different, this study did not address differences across serotypes in transmissibility. It also remains a question whether micro-epidemic patterns are determined solely by heterogeneous transmission in a structured population of interconnected "patches" (day care cohorts), or if the dynamics of transmission needs to be augmented by competition in which one serotype hinders the presence, or at least compromises the sensitivity of finding others.
In conclusion, we hypothesize that pneumococci transmit in a population of "day care cohorts" and is regulated by the strength of within-cluster micro-epidemics. Specifically, because transmission is more intense within than across clusters, the ensuing micro-epidemic behaviour enhances pneumococcal transmission. This is similar to the role of core groups of disease transmission . In a homogeneously mixing population, the vaccination effort to eradicate an SIS (susceptible-infectious-susceptible) type infection is typically of the order of the endemic prevalence of carriers. Due to within-cluster transmission, the critical efficacy for a vaccine to eradicate pneumococcal transmission may actually be larger than in a homogeneously mixing population with the same prevalence of carriage (cf. ). Microepidemic behaviour of carriage within host clusters may also have other bearings on pneumococcal epidemiology. For example, it is reflected in the genetic composition of the pneumococcal population, thus being an essential factor when attempting to elucidate the evolutionary mechanisms of pneumococci .