Pinpointing clusters of apparently sporadic cases of Legionnaires' disease

Epidemiological analysis of Legionnaires' disease in Scotland: a genomic study

BACKGROUND

Legionella pneumophila is the main cause of a severe pneumonic illness known as Legionnaires' disease and is a global public health threat. Whole-genome sequencing (WGS) can be applied to trace environmental origins of L pneumophila infections, providing information to guide appropriate interventions. We aim to explore the evolutionary and epidemiological relationships in a 36-year Scottish L pneumophila reference isolate collection.

METHODS

We investigated the genomic epidemiology of Legionnaires' disease over 36 years in Scotland, comparing genome sequences for all clinical L pneumophila isolates (1984-2020) with a sequence dataset of 3211 local and globally representative isolates. We used a stratified clustering approach to capture epidemiological relationships by core genome Multi-locus Sequence Typing, followed by high-resolution phylogenetic analysis of clusters to measure diversity and evolutionary relatedness in context with epidemiological metadata.

FINDINGS

Clustering analysis showed that 111 (57·5 %) of 193 of L pneumophila infections in Scotland were caused by ten endemic lineages with a wide temporal and geographical distribution. Phylogenetic analysis of L pneumophila identified hospital-associated sublineages that had been detected in the hospital environment up to 19 years. Furthermore, 12 (30·0%) of 40 community-associated infections (excluding a single, large outbreak) that occurred over a 13 year period (from 2000 to 2013) were caused by a single widely distributed endemic clone (ST37), consistent with enhanced human pathogenicity. Finally, our analysis revealed clusters linked by national or international travel to distinct geographical regions, indicating several previously unrecognised travel links between closely related isolates (fewer than five single nucleotide polymorphisms) connected by geography.

INTERPRETATION

Our analysis reveals the existence of previously undetected endemic clones of L pneumophila that existed for many years in hospital, community, and travel-associated environments. In light of these findings, we propose that cluster and outbreak definitions should be reconsidered, and propose WGS-based surveillance as a critical public health tool for real-time identification and mitigation of clinically important endemic clones.

FUNDING

Chief Scientist Office, Biotechnology and Biological Sciences Research Council (UK), Medical Research Council Precision Medicine Doctoral Training Programme, Wellcome Trust, and Medical Research Council (UK).

Environmental sources of community-acquired legionnaires' disease: A review

BACKGROUND

Most Legionnaires' disease in the US and abroad is community-acquired and believed to be sporadic, or non-outbreak associated. Most patients are exposed to numerous water sources, thus making it difficult to focus environmental investigations. Identifying known sources of sporadic community-acquired Legionnaires' disease will inform future sporadic Legionnaires' disease investigations as well as highlight directions for research. The objective is to summarize and rank sporadic Legionnaires' disease sources based on the level of linkage between the environmental source and cases.

METHODS

A PubMed search was conducted using the search terms legion* and (origins or source or transmission) and (sporadic or community-acquired). Studies of nosocomial and/or outbreak-associated disease were excluded from this review. Definite, probable, possible and suspect ranks were assigned to sources based on evidence of linkage to sporadic Legionnaires' disease.

RESULTS

The search yielded 196 articles and 47 articles were included in the final review after application of exclusion criteria. A total of 28 sources were identified. Of these, eight were assigned definite rank including residential potable water and car air-conditioner water leakage. Probable rank was assigned to five sources including solar-heated potable water and soil. Possible rank was assigned to nine sources including residential potable water and cooling towers. Suspect rank was assigned to 20 sources including large building water systems and cooling towers.

CONCLUSION

Residential potable water, large building water systems and car travel appear to contribute to a substantial proportion of sporadic Legionnaires' disease. Cooling towers are also a potentially significant source; however, definitive linkage to sporadic cases proves difficult. The sources of sporadic Legionnaires' disease cannot be definitively identified for most cases.

A case-association cluster detection and visualisation tool with an application to Legionnaires' disease

Statistical methods used in spatio-temporal surveillance of disease are able to identify abnormal clusters of cases but typically do not provide a measure of the degree of association between one case and another. Such a measure would facilitate the assignment of cases to common groups and be useful in outbreak investigations of diseases that potentially share the same source. This paper presents a model-based approach, which on the basis of available location data, provides a measure of the strength of association between cases in space and time and which is used to designate and visualise the most likely groupings of cases. The method was developed as a prospective surveillance tool to signal potential outbreaks, but it may also be used to explore groupings of cases in outbreak investigations. We demonstrate the method by using a historical case series of Legionnaires' disease amongst residents of England and Wales.

Meteorological conditions and incidence of Legionnaires' disease in Glasgow, Scotland: application of statistical modelling

This study investigated the relationships between Legionnaires' disease (LD) incidence and weather in Glasgow, UK, by using advanced statistical methods. Using daily meteorological data and 78 LD cases with known exact date of onset, we fitted a series of Poisson log-linear regression models with explanatory variables for air temperature, relative humidity, wind speed and year, and sine-cosine terms for within-year seasonal variation. Our initial model showed an association between LD incidence and 2-day lagged humidity (positive, P = 0·0236) and wind speed (negative, P = 0·033). However, after adjusting for year-by-year and seasonal variation in cases there were no significant associations with weather. We also used normal linear models to assess the importance of short-term, unseasonable weather values. The most significant association was between LD incidence and air temperature residual lagged by 1 day prior to onset (P = 0·0014). The contextual role of unseasonably high air temperatures is worthy of further investigation. Our methods and results have further advanced understanding of the role which weather plays in risk of LD infection.

Knox meets Cox: adapting epidemiological space-time statistics to demographic studies

Many important questions and theories in demography focus on changes over time, and on how those changes differ over geographic and social space. Space-time analysis has always been important in studying fertility transitions, for example. However demographers have seldom used formal statistical methods to describe and analyze time series of maps. One formal method, used widely in epidemiology, criminology, and public health, is Knox 's space-time interaction test. In this article, we discuss the potential of the Knox test in demographic research and note some possible pitfalls. We demonstrate how to use familiar proportional hazards models to adapt the Knox test for demographic applications. These adaptations allow for nonrepeatable events and for the incorporation of structural variables that change in space and time. We apply the modified test to data on the onset offertility decline in Brazil over 1960-2000 and show how the modified method can produce maps indicating where and when diffusion effects seem strongest, net of covariate effects.

Geographical variation of sporadic Legionnaires' disease analysed in a grid model

The aim was to analyse variation in incidence of sporadic Legionnaires' disease in a geographical information system in three time periods (1990–2005) by the application of a grid model and to assess the model's validity by analysing variation according to grid position. Coordinates of the addresses at time of disease of 606 confirmed cases with Legionnaires' disease were obtained. The incidence was calculated in cells of 10×10 km in 25 different grids superimposed on a map of Denmark. A 95% and 99% threshold was applied to identify cells with excess incidence representing potential clusters. Four cells had excess incidence in all three time periods. The analysis in 25 different grid positions indicated a low risk of overlooking cells with excess incidence in a random grid. The coefficient of variation ranged from 0·08 to 0·11 independent of the threshold. By application of a random grid model we demonstrated that it was possible to detect small areas with excess incidence that were not detected in the present surveillance system.

Outbreak detection and secondary prevention of Legionnaires’ disease: A national approach

BACKGROUND

To stop a possible outbreak of Legionnaires' disease (LD) at an early stage an outbreak detection programme was installed in The Netherlands.

METHODS

The programme consisted of sampling and controlling of potential sources to which LD patients had been exposed during their incubation period. Potential sources were considered to be true sources of infection if two or more LD patients (cluster) had visited them, or if available patients' isolates and environmental Legionella spp. were indistinguishable by amplified fragment length polymorphism genotyping.

RESULTS

Rapid sampling and genotyping as well as cluster recognition helped to target control measures. Despite these measures, two small outbreaks were only stopped after renewal of the water system. The combination of genotyping and cluster recognition lead to 29 of 190 (15%) patient-source associations.

CONCLUSION

Systematic sampling and cluster recognition can contribute to outbreak detection and lead to cost-effective secondary prevention of Legionnaires' disease.

Scales of environmental justice: combining GIS and spatial analysis for air toxics in West Oakland, California

This paper examines the spatial point pattern of industrial toxic substances and the associated environmental justice implications in the San Francisco Bay Area, California, USA. Using a spatial analysis method called Ripley's K we assess environmental justice across multiple spatial scales, and we verify and quantify the West Oakland neighborhood as an environmental justice site as designated by the US Environmental Protection Agency. Further, we integrate the ISCST3 air dispersion model with Geographic Information Systems (GIS) to identify the number of people potentially affected by a particular facility, and engage the problem of non-point sources of diesel emissions with an analysis of the street network.

Using GIS-Based Approaches to Support Research on Neurotoxicants and Other Children's Environmental Health Threats

Environmental threats to children's health are complex and multifaceted; consequently, children's environmental health research strives to identify areas of elevated exposure, determine whether particular demographic groups are inequitably exposed, and link exposures to incidence of disease. Many environmental health researchers use geographic information systems (GIS) to ex post display the results of their data collection and analysis. This methodological paper shows some ways by which the ex ante integration of GIS into environmental exposure and epidemiological research can significantly enhance: research design; sampling, recruitment, and retention strategies; data management and analysis; and community translation.

Users’ perspectives on epidemiological, GIS and point pattern approaches to analysing environment and health data

Despite examples showing the usefulness of geographical information systems (GIS) and spatial point pattern analysis in health research, there remain barriers to their widespread use within health service settings. This paper explores potential users' views on the relative usefulness of such approaches for analysing spatially referenced environmental health data. Our findings indicate that researchers and practitioners do not always prefer the approach with which they are most familiar. In addition, there is a need for higher levels of understanding of, and confidence in, GIS and point pattern analysis techniques amongst health service professionals. The greatest need is for multi-disciplinary research which uses the most appropriate approach for each investigation, rather than that with which researchers are most familiar.

Analysing spatially referenced public health data: a comparison of three methodological approaches

In the analysis of spatially referenced public health data, members of different disciplinary groups (geographers, epidemiologists and statisticians) tend to select different methodological approaches, usually those with which they are already familiar. This paper compares three such approaches in terms of their relative value and results. A single public health dataset, derived from a community survey, is analysed by using 'traditional' epidemiological methods, GIS and point pattern analysis. Since they adopt different 'models' for addressing the same research question, the three approaches produce some variation in the results for specific health-related variables. Taken overall, however, the results complement, rather than contradict or duplicate each other.

Space-time clustering of childhood leukaemia in Greece: evidence supporting a viral aetiology

The method introduced by Knox for evaluation of space-time clustering has been applied to 872 cases of childhood (0-14 year old) leukaemia diagnosed in Greece over the 10 year period 1980-89. Greek towns are characterised by substantial population mixing due to internal migration, whereas there is relative isolation in mountainous rural areas. Predetermined space (5 km) and time (1 year) limits were used on the basis of previous reports in order to define the clustering cell. There is highly significant evidence for clustering of childhood leukaemia in Greece as a whole, the observed number of pairs that are close in both spaces and time exceeding the expected number by 5.2% (P = 0.004). The excess is particularly evident for leukaemia cases in 0 to 4-year-old children, among whom the observed number of pairs that are close in both space and time exceeded the expected number by 9.4% (P = 0.004). There is no evidence of space-time clustering for leukaemia cases older than 5 years. The overall pattern is descriptively similar in urban and semiurban areas and is especially marked for acute lymphoblastic leukaemia at the childhood peak ages (2-4 years) with an excess of 19% (P = 0.0006). In the rural population there is evidence for clustering of cases belonging to older and broader age groups, a phenomenon compatible with a delay in the development of herd immunity against putative infectious aetiological agents. The findings of the present study provide support for the hypothesis that a substantial proportion of cases of childhood leukaemia may arise as a rare sequel to exposure to an agent or agents, most probably viral in nature.

Legionellosis linked with a hotel car park--how many were infected?

An outbreak of legionellosis associated with a hotel in Sydney, Australia, and the subsequent epidemiological and environmental investigations are described. Four cases of Legionnaires' disease were notified to the Public Health Unit. A cross-sectional study of 184 people who attended a seminar at the hotel was carried out. Serological and questionnaire data were obtained for 152 (83%) of these. Twenty-eight (18%) respondents reported symptoms compatible with legionellosis. Thirty-three subjects (22%) had indirect fluorescent antibody (IFA) titres to Legionella pneumophila serogroup 1 (Lp-1) of 128 or higher. The only site which those with symptoms of legionellosis and IFA titre > or = 128 were more likely to have visited than controls was the hotel car park (adjusted odds ratio [OR] 14.7, 95% confidence interval [CI]: 1.8-123.1). Those with symptoms compatible with legionellosis, but whose IFA titres were < 128 were also more likely to have visited the hotel car park (adjusted OR 4.4, 95% CI: 1.5-12.9). Seroprevalence of Lp-1 antibodies was higher in those who attended the seminar than in a population sample of similar age. Findings suggested that the 4 cases represented a small fraction of all those infected, and highlighted difficulties in defining illness caused by Lp-1 and in interpreting serology.

Interactive spatial data analysis in medical geography

Interactive spatial data analysis involves the use of software environments that permit the visualization, exploration and, perhaps, modelling of geographically-referenced data. Such systems are of obvious value in epidemiological research, both of an environmental and geographical nature. There is an increasing number of such software environments available on a variety of platforms and operating systems. This paper considers the use of the proprietary Geographical Information System, ARC/INFO, in a spatial analysis context, showing how the spatial analytic tools that may be added to it can be exploited by geographical epidemiologists; such tools include those for modelling possible raised incidence of disease around suspected sources of pollution. The paper also reviews the use of systems such as S-Plus and XLISP-STAT, statistical programming environments to which spatial analysis functions or libraries may be added. The use of INFO-MAP, a system designed to aid in the teaching of interactive spatial data analysis, is also highlighted. The various software environments are illustrated with reference to examples concerned with: clustering of childhood leukaemia in part of Lancashire, England; Burkitt's lymphoma in Uganda; larynx cancer in Lancashire; and childhood mortality in Auckland, New Zealand.

MCMC methods for putative pollution source problems in environmental epidemiology

This paper demonstrates the use of the Gibbs Sampler and other Markov Chain Monte Carlo (MCMC) methods in two applications in environmental epidemiology. The first example concerns the application of a Metropolis-Hastings/Gibbs sampler to a Cox process with a direction-dependent cluster variance parameter. The second example consists of the estimation of the posterior (spatial) distribution of a putative location.

Asthma and factory emissions in northern England: addressing public concern by combining geographical and epidemiological methods

STUDY OBJECTIVE

The prevalence of asthma was assessed to investigate respiratory health effects of airborne emissions from a factory. A geographical information system allowed flexible definition of study areas in terms of their size, distance, and location in relation to the factory. The value of the approach for this type of investigation is focussed on.

SETTING

The factory is located in the south western part of a small market town in County Durham.

MEASUREMENTS AND MAIN RESULTS

A total of 1573 asthma cases were identified from general practitioner computerised repeat prescribing systems. Population denominators were defined from family health services authority patient registers. The area within 1 km and immediately to the north east of the factory had an age and sex standardised asthma prevalence 24% (confidence interval 4, 44%) in excess of the expected rate (p = 0.01). The increased prevalence was confined to middle aged and elderly adults living in the area between 0.5 and 1 km to the north east of the factory.

CONCLUSIONS

The value of combining the skills of geographers and epidemiologists in addressing public health issues is shown, particularly through the use of geographical information systems which proved powerful and effective.

Second-order analysis of space-time clustering

We consider the problem of detecting and describing space-time interaction in point process data. We extend existing second-order methods for purely spatial point process data to the spatial-temporal setting. This extension allows us to estimate space-time interaction as a function of spatial and temporal separation, and provides a useful reinterpretation of a popular test, due to Knox, for space-time interaction. Applications to simulated and real data indicate the method's potential.

The geographical epidemiology of ocular diseases: some principles and methods

With the increasing availability of geographically referenced data in health research the time is ripe to review the use of particular geographical and spatial analysis techniques in ophthalmic research. Analysis of the geographical distribution of ocular diseases, particularly in Britain, has not had a high profile, but there are certain diseases, such as congenital eye malformations in children, where such analysis methods are particularly appropriate. We review the data requirements and then a variety of analytical techniques, some of which partition geographical space into areal units (such as counties or electoral wards), others of which treat space as continuous. We conclude with some comments on software that is available for such analyses.

Mortality mapping in Hong Kong, 1979-83 and 1984-88: the patterns of major non-malignant diseases

We examined the spatial patterns of mortality from various non-malignant diseases in Hong Kong during the two quinquennia, 1979-83 and 1984-88. Population data and parameters reflecting socioeconomic factors, including ethnic backgrounds, were selected from census data. Mortality data were obtained from death registration files. The standardized mortality ratios (SMRs) for major diseases were calculated for 27 census districts. The rankings of the districts' SMRs were shown in map form. Correlations were calculated between the districts' SMRs for the diseases, between them and the SMRs for cancers, and between them and socioeconomic and ethnic parameters. Many spatial patterns and correlations showed consistency and were biologically plausible. These results showed that mapping for a rapidly growing city such as Hong Kong could be a valuable exercise for detecting "at risk" populations where causal factors for non-malignant diseases can be investigated and identified for possible elimination.

Legionnaires' disease outbreak, Fairfield 1992: public health aspects

An investigation of an outbreak of Legionnaires' disease in 1992 in Fairfield, a municipality of Sydney, was carried out to determine the source of the outbreak. Cases of Legionnaires' disease with onset of symptoms between 11 and 20 April 1992 were included. Definite cases were individuals with a history consistent with Legionnaires' disease, confirmed by direct fluorescent antibody testing plus serology or culture. There were two control groups: patients admitted to the same hospital as the cases, matched for age and sex, and patients admitted to hospital with a presumptive diagnosis of legionnaires' disease, in whom the diagnosis was subsequently excluded. There were 26 definite cases with onset of symptoms between 11 and 20 April 1992. Six (23 per cent) died. Twenty-two cases (85 per cent) reported visiting the Fairfield business district during the ten days prior to the onset of symptoms. They were 20 times more likely to have visited Fairfield than were matched controls. Matching of Legionella pneumophila serogroup 1 from environmental and clinical samples was achieved by cytogenetic fingerprinting. Fourteen cases were linked to a single environmental sample. The epidemiological findings were consistent with a point source of Legionella in the Fairfield business district. It is most likely that the exposure occurred on 10 April 1992.