Enter the Enterovirus
The recent outbreak of human enterovirus 68 (EV-D68) poses several challenges to syndromic surveillance. Its presentation can resemble any number of other upper respiratory infections of varying severity. The timing coincided with the end of school summer vacation, a time of year when respiratory illnesses are always spread. Differentiation based on key symptoms is theoretically possible, but also highlights where the assumptions of surveillance on chief complaints can break down. And while the outbreak received national media coverage, the lack of a memorable consensus on what to call the pathogen may have limited patients’ and registrars’ ability to ask for it by name.
According to the CDC, the most common symptoms of EV-D68 “include fever, runny nose, sneezing, cough, and body and muscle aches.” This is both good and bad news for chief complaint surveillance. These are symptoms that patients encounter regularly and are likely able to name specifically and accurately when seeking healthcare. Unfortunately, that is because they are common symptoms of many ailments, making EV-D68 difficult to distinguish at the chief complaint level. In other words, it is possible to get fairly good sensitivity when detecting EV-D68 cases from syndromic surveillance data, but specificity will probably be pretty low.
Exacerbating the specificity challenge in this particular event was the timing. As the chart of cough counts from the past 3 years shows, late August or early September, when children return to school, always brings an increase in respiratory symptom complaints. Whatever factors–colder temperatures, more time indoors, more interactions between children, more sharing of supplies and facilities, etc.–lead to this annual pattern likely contributed to the spread of EV-D68 at this time. So it’s not just an unfortunate coincidence that this outbreak blends into the annual pattern of respiratory disease, but the causal connection doesn’t mitigate the detection challenge. The chart shows some evidence that the counts increase more rapidly than in prior years, which would increase the timeliness of detection but not necessarily help distinguish the underlying pathogen.
More severe symptoms of EV-D68 include wheezing and exacerbation of asthma; these features were mentioned in media coverage of the outbreak. Thus it might be possible to use these symptoms to distinguish areas where EV-D68 cases were occurring from those seeing common colds or other illnesses that are less likely to have those symptoms. Comparing early September counts for wheezing and asthma specifically revealed that both showed an increase in regions where EV-D68 cases had been confirmed, but not in regions without confirmed cases. Wheezing in particular looked promising as a differentiator. Further drilldown revealed that most of those wheezing complaints came from a single facility, a large children’s hospital. This raised the question of whether they were the only facility seeing wheezing patients, or were just more likely to record wheezing in a complaint.
Analysis of historical data prior to the present EV-D68 outbreak revealed some interesting patterns. The likelihood of wheezing appearing in a chief complaint declines as the patient age increases. After adjusting for this distribution of wheezing by age, children’s hospitals were still more likely to have wheezing in a complaint than other hospitals. Even more striking, the particular children’s hospital relevant to this event had a still higher likelihood for including wheezing in a chief complaint. Either the facility is more likely to ask about that symptom, knowing that it can be particularly relevant to differential diagnosis and assessing severity in childhood illness; their registration system has predefined chief complaint options that include wheezing specifically while other facilities do not; or patients with wheezing in the region are more likely to seek treatment at this particular hospital. All three could contribute to the observed increase in wheezing complaints at that facility.
This finding highlights the tradeoff of sensitivity and specificity in syndromic surveillance. As we noted, a broad respiratory category will be fairly sensitive in detecting EV-D68 cases in a nonspecific way. We might hope to increase specificity by narrowing to one particular symptom, but now we are at the mercy of hospital procedures. Some EDs use registration systems with predetermined complaint options, some enforce the chiefness of a chief complaint, and some permit extensive narratives. Consequently, the exact same patient reporting the exact same condition will have their chief complaint recorded differently at different facilities (without even considering differences between registrars).
Wheezing and asthma classifications were added to EpiCenter to assist with monitoring this particular event. However, for the reasons detailed here, they should not be considered definitive for this outbreak. They should still be used in conjunction with other, more sensitive classifications to get as complete a picture as is possible from chief complaint data.
A patient concern classification was also added specifically for EV-D68. Despite wide media coverage, however, there have thus far been very few complaints to specifically reference it. This include mentions of “enterovirus,” “EV-D68,” “D68” and “respiratory syndrome” as some media reports named it (such as this NY Times report). The diversity of names may contribute to the lack of direct references, and none of them are as memorable as, for example, “swine flu,” which appeared in over 100 chief complaints in the days following initial media reports–when there were far fewer confirmed cases in the US than there were of EV-D68 when it was reported widely.
Fortunately, outbreaks of this sort are relatively rare in the US, making it all the more important to learn as much as possible from those events that do occur. We were able to respond quickly to this outbreak by working with our partners in public health and adding new, relevant classifications to EpiCenter. The insights provided by this event will inform future developments so that we can continue make the most of emergency department chief complaint surveillance.