Estimating the role of casual contact from the community in transmission of Bordetella pertussis to young infants

Emerging Themes in BioMed Central Open Access Analytic perspective Estimating the role of casual contact from the community in transmission of Bordetella pertussis to young infants Aaron M Wendelboe*1, Michael G Hudgens2, Charles Poole1 and Annelies Van Rie1 Address: 1Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA and 2Department of Biostatistics, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA Email: Aaron M Wendelboe* - ; Michael G Hudgens - ; Charles Poole - ; Annelies Van Rie - * Corresponding author Published: 19 October 2007 Emerging Themes in Epidemiology 2007, 4:15 doi:10.1186/1742-7622-4-15 Received: 3 May 2007 Accepted: 19 October 2007 This article is available from: http://www.ete-online.com/content/4/1/15 © 2007 Wendelboe et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Abstract The proportion of infant pertussis cases due to transmission from casual contact in the community has not been estimated since before the introduction of pertussis vaccines in the 1950s. This study aimed to estimate the proportion of pertussis transmission due to casual contact using demographic and clinical data from a study of 95 infant pertussis cases and their close contacts enrolled at 14 hospitals in France, Germany, Canada, and the U.S. between February 2003 and September 2004. A complete case analysis was conducted as well as multiple imputation (MI) to account for missing data for participants and close contacts who did not participate. By considering all possible close contacts, the MI analysis estimated 66% of source cases were close contacts, implying the minimum attributable proportion of infant cases due to transmission from casual contact with community members was 34% (95% CI = 24%, 44%). Estimates from the complete case analysis were comparable but less precise. Results were sensitive to changes in the operational definition of a source case, which broadened the range of MI point estimates of transmission from casual community contact to 20%–47%. We conclude that casual contact appears to be responsible for a substantial proportion of pertussis transmission to young infants. Medical subject headings (MeSH): multiple imputation, pertussis, transmission, casual contact, sensitivity analysis, missing data, community. Introduction Pertussis disease is poorly controlled among infants, adolescents, and adults in developed countries despite high immunization coverage rates [1-3] of ≥ 93 percent for both the primary infant series [4-6] and the booster at school entry [7]. Bordetella pertussis is reported to be among the most contagious pathogens in humans as an average of 15 secondary infections arise from a single case in a susceptible population [8]. Public health messages focus on the importance of transmission from close contacts [9-11] implying that pertussis transmission due to casual contact from community members is not appreciable. Several studies investigated the disease dynamics of B. pertussis, especially as they relate to the transmission of the bacteria to young infants. This has been done by collecting diagnostic information on close contacts and assignPage 1 of 7 (page number not for citation purposes) Emerging Themes in Epidemiology 2007, 4:15 ing, where possible, one person as the most probable source of infection (where the difficulty in identifying the source case usually lies in identifying even one symptomatic source case as opposed to choosing from multiple potential cases). These studies identified close/household contacts as a source of infection for 40–53 percent of young infants with pertussis [12-16]. However, none of these studies investigated whether the remaining 47–60 percent of transmission was due to casual contact in the community or caused by transmission from unidentified close contacts as no attempt was made to rule out transmission from all identifiable household and other close contacts. In order to infer transmission from a casual contact source, it is necessary to conclusively determine transmission did not occur from a close contact source. Several obstacles have hindered previous studies' ability to eliminate the possibility that transmission came from a close contact source. First, missing data due to non-participation of close contacts and participants' refusal to provide specimens for laboratory diagnostic testing has been high [12] or unreported, limiting the ability to determine whether a given contact would have been identified as the source of the index case's infection had the data not been missing. Second, diagnosing pertussis is often problematic since many adolescent and adult pertussis cases do not present with the typical symptoms of whooping cough [17-19]. This is further complicated by the lack of a highly sensitive and specific laboratory diagnostic method [10,20]. Third, inter-person variability in the incubation and infectious periods [21,22] may result in failure to identify source cases if their incubation or infectious periods lie in the tails of the distributions not captured by standard definitions [16,23-25]. Finally, it is uncertain whether individuals with asymptomatic infection can transmit pertussis [13,26,27]. In the absence of sound evidence for or against the infectiousness of asymptomatic infections, the systematic exclusion of asymptomatically infected individuals as possible source cases (as done in all previous studies) may bias the results. In this paper we estimated the minimum proportion of infant pertussis cases due to transmission from casual contact sources using information from a study designed to identify the source of infection in young infants. Results from multiple diagnostic tests (including polymerase chain reaction [PCR], culture, and paired serology) were available on household contacts and non-household persons with close contact with the infant. To adjust for missing data arising from non-enrollment or failure to provide diagnostic specimens, multiple imputation (MI) analysis was used. MI is a widely accepted method to account for missing data and is superior to complete case analyses for two reasons. http://www.ete-online.com/content/4/1/15 First, as the amount of missing data increases, the results from complete case analyses suffer a greater loss in precision than results obtained by MI analyses [28,29]. Second, when data are not missing completely at random (MCAR) and the missing data mechanism is appropriately specified, MI will produce less biased estimates than a complete case analysis to the extent that missingness is associated with the observed data [28,30]. This is particularly germ (...truncated)