Our analyses indicate that IGI outbreaks are common occurrences aboard U.S. Navy ships in this region. To the best of our knowledge, this is the first study to estimate the incidence of IGI outbreaks among deployed U.S. Navy ships. We found an overall incidence of 33.2 outbreaks per 1000 ship-weeks (Poisson 95% CI 16.6 – 59.5). A recent epidemiologic study published on IGI outbreaks among vacation cruise ships (median cruise duration of one week) found an outbreak frequency of 6.3 outbreaks per 1000 cruises (ship-weeks) during 1990 – 1995, and was reduced to 3.7 outbreaks per 1000 cruises during 1996–2000 [20]. This decrease may have been attributable to vigilance by public health and industry officials in control of person-to-person spread of illness among crew and passengers and environmental disinfection [21]. Cruise ships and Navy ships are different environments with different populations; therefore, comparisons of incidence may be misleading and should be avoided. The high rate of outbreaks among U.S. Navy ships may be explained by younger age (higher risk-taking population), higher risk ports visited by U.S. Navy ships, much longer time at sea, and/or the methods we used to define outbreaks.
The cumulative ARs in this study are similar to rates previously reported (SA Thornton, personal communication), but much lower than rates described in other IGI outbreak investigations among Navy ships [3, 12, 13, 15]. Among 15 Norovirus-confirmed outbreaks, Thornton found a mean cumulative AR of 5.3%, slightly higher than what we found. This is likely due to the fact that his study used active surveillance methodology, while we relied on passive reporting. We also found cumulative attack rates that were lower than those reported among cruise ships. In a recent account of six reported outbreaks among five cruise ships during July 1 2002 to December 2002, attack rates based on sick-call visits ranged between 2.5 – 11.5% of all shipboard persons (median 6.1) [22]. A number of reasons could explain the differences in attack rates including, differences among shipboard health seeking behavior and the limitation of the surveillance window within this study possibly resulting in under-capture of all visits associated with a given outbreak. Despite finding lower attack rates, eight of eleven of the possible outbreaks we described equaled or exceeded the 3% attack rate that is used to trigger an outbreak in the cruise ship industry. As others have pointed out, the AR based on medical clinic visits likely underestimates the total number of cases associated with these outbreaks [21]. For example in one cruise ship outbreak it was found that while 8% of the passengers reported to sick call with acute gastroenteritis (AGE) symptoms, 41% met a case definition for AGE during a subsequent epidemiologic investigation (75% of passengers surveyed) [22]. In our study, we were unable to ascertain the full extent of IGI due to the passive surveillance methods used by the DNBI reporting system.
The outbreak definition we used is novel and may or may not compare equally to the Vessel Sanitation Program (VSP) classification. The VSP defines outbreaks if = 3% of the passengers and/or crewmembers develops AGE symptoms. This cutoff, developed by the CDC, was based on previous data of ship-associated outbreaks where the incidence of gastrointestinal illness was found to be = 1% of 92% of cruises and = 3% on 3% of cruises [23]. Our outbreak cut-point was based on retrospective analysis of a cohort of U.S. Navy ships deployed to the region over a one-year period. Only prospective evaluation of such a method with attempts to identify etiologic causes of these outbreaks can inform whether this method is sensitive and specific.
Outbreaks were found throughout the year, which is inconsistent with previous reports of IGI outbreaks among cruise ships which have demonstrated a winter-spring predominance and where Norovirus is implicated in 69% of these outbreaks [24]. It is an assumption that the outbreaks in our study are due to Norovirus-related disease, since a surveillance report among investigated GI outbreaks aboard large U.S. Navy ships identified Norovirus in 4 out of 4 which submitted stool specimens for testing Norovirus-confirmed outbreaks [25]. In addition, at least one of the eleven possible outbreaks identified in our study was found to be associated with Norovirus based on a concurrent surveillance study being conducted (SA Thornton, personal communications). The lack of a seasonality of outbreaks in our study could be explained by a number of reasons. In this equatorial region, while a season of warm and cold can be delineated, differentiation between seasons are very different than temperate climates, thus a seasonal distribution for these viruses may not be found [26–28]. Furthermore, these ships often acquire their outbreaks during visits to other equatorial parts of the world and arrive in the CENTCOM AOR with an outbreak already underway.
We also found that larger ships had more frequent outbreaks than smaller ships (6.6 per 100 ship-weeks vs. 2.6 per 100 ship-weeks, respectfully). One explanation is that larger ships often transport operational forces during deployments, with multiple embarkations and debarkations, thereby possibly introducing new enteric pathogens to shipboard personnel. It is also possible that there were differences in port visit schedules (e.g., increased number, longer duration, etc.) between the different ship classes that could account for increased risk of outbreaks. A common theme among a number of case reports has been that crowding is likely to play a role in these outbreaks [3, 12, 13]. One recent report published in Navy Medicine details a shipboard outbreak possibly due to viral gastroenteritis and suggests that communal toileting facilities and population density may contribute the increased risk of outbreaks [29]. Lastly, the case definition that was utilized may have resulted in fewer outbreaks associated with smaller ships, as perhaps with smaller ships outbreaks would have been recognized sooner and control measures more easily implemented. While cumulative ARs appear to be similar among the different ship classes, the duration of outbreaks appear to be longer on the smaller ships. The reason for this finding is not clear and may be due to chance or different population density and transmission dynamics among different shipboard architectures and/or populations. Differential duration of surveillance windows between large and small ships does not explain this finding either as most ships types were under observation on average 7 – 9 weeks, with only destroyers (small ship) with a shorter average surveillance window of 6.7 weeks.
While our analysis of morbidity associated with these possible outbreaks was hindered due to the aggregate nature of the data, we did find an increase in the amount of SIQ given during weeks when there were outbreaks, compared to weeks when there were no outbreaks. However, there was less hospitalization. This finding suggests that there may be a difference in severity of illnesses during outbreaks, which results in more days lost (SIQ), but requires fewer hospitalizations. The higher rate of SIQ given during outbreak weeks may also be reflective of a control strategy for IGI outbreaks, whereby ill persons are effectively removed from the workplace for 48–72 hours during the period where infectious agent shedding may be very high. Thus, the assertion of higher morbidity during outbreaks as measured by rates of SIQ may be confounded. The estimation of work days lost may be an underestimate as Whittaker et al. reported that during a large IGI outbreak aboard an aircraft carrier, only patients requiring treatment in the clinic were given SIQ status [14]. Thus, an undercounting of work-time lost due to illness may occur during large outbreaks. Due to the aggregate nature of this data and low numbers of hospitalizations, efforts should be made to confirm this differential morbidity before drawing any definite conclusions.
The methodology utilized in this study to define IGI outbreaks is novel and has several limitations. DNBI surveillance data collected and reported at the aggregate level is used for the purpose of tracking trends and identifying possible clusters across a broad range of clinical syndromes grouped into non-specific disease and injury categories. While this design appears useful for its intended purpose, the aggregate nature, both across individuals of each reporting ship, and across specific clinical syndromes within broad disease categories (e.g. catch-all IGI category), limits the inferences one can draw compared to more traditional individual unit of observation based epidemiological studies. Therefore, the results of this study should be put in context, and the derived estimates of particular outbreak attributes (e.g. outbreak rates, population attack rates, outbreak durations, and disease morbidity comparisons) may not be comparable to other epidemiological studies which use more traditional outbreak investigation designs. Specifically, the definition of IGI visits exceeding an upper 95% confidence (among all ships reporting) for two consecutive weeks would likely result in missing shorter outbreaks (decreased sensitivity) which are probably more commonly attributable to other common bacterial causes and a point source mode of transmission. The choice of this definition was for purposes of specificity to detect viral gastroenteritidies which can appear from a point-source introduction but tend to demonstrate a person-to-person mode of transmission with longer outbreak durations. In addition, this choice of definition may bias towards increasing the duration estimates of the described outbreaks. However, to balance this is the observation that many outbreaks were not followed to complete resolution due to the surveillance window for a particular ship ending due to its movement out of the surveillance system, thus, resulting in a possible bias towards shortening the length of a given outbreak. Furthermore, it was a limitation of our data that the incidence measurements made were only based on denominator ship-time within the CENTCOM AOR. Future studies directed at the entire deployment period should be conducted to accurate describe the true rate of these possible outbreaks.