We identified substantial transmission within an educational institution located in an area with a low background rate of TB in the community. We hypothesise this was predominately due to a single infectious case with smear positive disease and a high bacterial load. This is illustrated by the higher odds of LTBI associated with sharing a course with the suspected index case compared to the second infectious case, and how specific communities were associated with substantially greater odds of LTBI. The data indicates that the second case propagated the outbreak but to a lesser extent, however, we are unable to discount that the suspected index case was the source of all transmission events. The presence of a single SNP differences between isolates which underwent WGS suggests contemporaneous transmission to or between cases. It is important to note that none of the other active cases were infectious whilst attending the educational institution. Close contact screening was undertaken as per NICE guidelines; these persons were not included in the analysis [8].
The multinomial analysis found students to have the highest odds of LTBI acquisition and this is likely due to the intensive social mixing of students with the two infectious cases compared to staff who worked at the educational institution. Educational institutions are semi-closed settings where students have close and prolonged contact with each other, there is a high degree of social mixing and classrooms can be poorly ventilated making them an ideal environment for transmission [5, 7]. In this outbreak, we observed that transmission can extend beyond direct contact in the classroom due to LTBI cases in persons who did not share a course with either infectious case. This emphasises the importance of shared airspace in classrooms and places of social aggregation within the educational institution [6]. Consequently, the public health response to TB outbreaks in educational settings should look at shared airspace and opportunities for social mixing in addition to direct course sharing. This is in addition to current NICE guidelines [8].
Following this outbreak additional cases with the same MIRU-VNTR profile that are not students or members of staff at the educational institution have been identified. Three had epidemiological links to the educational institution, leaving three cases with no epidemiological links but with links to the same geographical area. This shows that the transmission network has extended into the wider community where ongoing routine surveillance is important to ensure case ascertainment and local TB control.
Delay in diagnosis has frequently been cited as an important factor in previous outbreaks and played a role in this outbreak. In low incidence areas, clinical exposure and awareness surrounding the signs and symptoms of TB contributes to delays [1, 2, 5, 6]. Awareness around TB in both educational staff and primary care doctors is important as delays can result in prolonged opportunities for transmission and a greater public health impact on the population at risk.
Health protection measures were implemented alongside the epidemiological investigation being undertaken. This outbreak was originally investigated using traditional epidemiological and microbiological methods which were later supplemented by WGS and SNA. Our study demonstrates that WGS provides added inference regarding temporality of TB transmission over and above MIRU-VNTR reference typing, which supported the multiagency incident control team to describe key aspects of the putative TB transmission network and focus public health action accordingly. WGS is currently being introduced as a routinely available service for TB isolates across England and provides numerous advantages compared to MIRU-VNTR reference typing. In addition to enhanced strain discrimination and relatedness, laboratory turnaround times are substantially reduced and inferences regarding transmission between cases can guide control measures in real-time [20]. The discriminatory power of WGS precisely defines clusters of genetically related isolates and allows for highly sensitive and specific case definitions [9, 12, 20]. This decreases misclassification bias in epidemiological studies, improves resource allocation and public health risk assessment. The ability to estimate time since transmission and the corresponding directionality provides information to support contact tracing, understand transmission networks and identify at-risk persons. MIRU-VNTR typing cannot accurately capture this genomic diversity and is associated with lower precision and misclassification [9]. The advantages of WGS been demonstrated in previous studies and outbreak investigations and were similarly observed within our investigation [9, 10, 12].
The SNA analysis allowed visualisation of course sharing in greater detail and demonstrated the nature of transmission in this setting. We found that certain courses or combinations of courses (communities) had a greater number of LTBI positive students, suggesting enhanced transmission within key course sharing groups. We found that SNA can add greater precision and accuracy to traditional ‘stone in the pond’ contact tracing in an outbreak setting [9]. These findings combined with the factors identified in the multinomial analysis provided a high level of detail and these methods can be used to prioritise screening interventions in future outbreaks. For example, we identified eight persons within the high risk community who were not screened (as they had left the educational institution and were lost to follow up). Traditional contact tracing methods identified these cases but prioritised them alongside a large number of lower risk cases. Social networks can be produced based upon course sharing, attendance of extracurricular activities in relation to the index case, or any other unifying activities or shared exposures. Further iterations of this network can be produced as screening results are obtained and can be used to both refine and validate the network as well as direct future screening activities.
Strengths of this study include its large and comprehensive sample. There was a high response and a large number of persons were screened. This limited sampling bias and provided sufficient data for analysis. The innovative methods used, WGS and SNA, provided further insight to the outbreak and potential areas to guide public health intervention over traditional methods. We made use of multiple sources of data and combined them to give insights into TB epidemiology in an educational setting; furthermore it confirmed that the action the incident control team had taken was correct.
The nature of an outbreak provides strong ecological validity however the challenges posed by precision of data collection and resulting issues with completeness will undoubtedly affect the findings of this study. The limitations were that the SNA was done following the initial outbreak investigation and therefore could not be used to prioritise cases for screening. Prior administration of BCG was self-reported and liable to recall bias which could affect the estimates in either direction and likely explains the incongruous association found in our data. In particular, we understand that some people believed that it was part of routine immunisation and responded affirmatively if their children were otherwise fully immunised. The investigation occurred over two different years which meant that people left and enrolled in the educational institution. This complicated contact tracing and altered the data collected and the exposure status of these groups.
This manuscript and others have shown that introduction of an infectious student to an educational setting can result in a large number of both latent and active TB cases [1,2,3,4,5,6,7]. This puts a logistical and financial strain on local clinical and public health services, and threatens TB control efforts. The suspected index case in this outbreak was an international student from a high burden country. Those entering the UK since May 2012 undergo active TB screening as part of UK entry regulations and whilst new migrant LTBI screening was introduced in 2016, funding has not been made routinely available to low incidence areas of the country [8, 9]. A large number of boarding schools and universities are located in rural low incidence areas and therefore students, including within our study, will not be eligible for LTBI screening whereas they were previously [21]. A study in Canada showed that when LTBI therapy was considered as a marginal cost (included as part of an existing screening programme) it was highly cost-effective [21]. We recommend health economic studies are undertaken in England, modelling the impact of screening all persons applying for a student visa for LTBI as part of their routine pre-entry active TB screening assessment. Finally, we recommend that SNA and WGS are considered when investigating TB outbreaks in educational settings, and potentially other closed and semi-closed settings.