Clusters of Human Infections With Avian Influenza A(H7N9) Virus in China, March 2013 to June 2015

Abstract

From March 2013 through June 2015, a total 657 cases of human infection with novel avian influenza A(H7N9) virus were reported; most case patients presented with severe pneumonia, and 269 persons (41%) died. Most cases occurred in 3 waves; the initial wave was in the spring of 2013, followed by a second wave from approximately November 2013 through May 2014; a third, smaller wave occurred from November 2014 through the spring of 2015. Findings of virologic and epidemiological studies indicate that poultry appear to be the predominant source of infection [1–5]. However, investigators have identified clusters of epidemiologically linked cases of human infection with H7N9, as well as cases in persons who had no known poultry contact, and investigators have concluded that person-to-person transmission may have occurred [2, 3, 6–18].

Investigations of clusters of human H7N9 infection are critically important to understanding potential person-to-person transmission and monitoring for any changes in the fitness of viruses to infect humans. In many situations, however, it remains unclear whether clusters of infection are due to exposure to a common infectious source or person-to-person transmission, and there is currently no standard method for assessing likely mode of transmission within clusters of human H7N9 infection. Several studies have used data from clusters of human H7N9 infections and statistical modeling to assess the transmissibility of the H7N9 virus [15, 17].

Our study includes individual assessments of all clusters of human H7N9 cases identified by the Chinese Center for Disease Control and Prevention as of June 2015. In addition to describing the case clusters, we developed a framework to assess the likelihood that person-to-person transmission occurred in these clusters. This framework uses available information on the timing of exposures to help distinguish between infections that were likely to have resulted from person-to-person transmission rather than a common infectious source. This simple framework was used to determine the likely number of reported person-to-person transmission events occurring in the study period; it can also be used as a standardized tool to aid in the rapid assessment of future clusters as cases of H7N9 continue to occur.

METHODS

After Chinese public health authorities were notified of a possible human H7N9 case, national, provincial, and local public health authorities conducted an investigation. The Chinese Center for Disease Control and Prevention received information about all confirmed H7N9 cases in mainland China and subsequent epidemiological investigations, and public health officials interviewed case patients and relatives about possible exposures, obtained information from medical records and medical staff interviews, and identified close contacts. Health authorities conducted daily telephone monitoring of contacts for fever and respiratory symptoms. If symptoms developed, a throat swab sample was collected from the contact and sent for real-time reverse-transcription polymerase chain reaction (RT-PCR) testing.

A cluster was defined as ≥2 persons with laboratory-confirmed H7N9 infection who had interacted with each other. The index case patient was defined as the first person in a cluster to develop symptoms; a contact case patient, a person with exposure to the index case patient with subsequent development of symptoms and laboratory-confirmed H7N9 infection. We did, however, include the first cluster in China, for which the index case was a suspected case without laboratory confirmation. Clusters were assigned numbers based on the index case patient’s date of symptom onset.

To determine whether person-to-person transmission or a common exposure was the most likely source of infection for the contact case, we compared the timing of potential exposure of the contact case patient to both the index case patient and other possible infection sources (eg, live poultry). We assumed that the index case patient was infectious starting 1 day before the onset of symptoms [19] and did not include earlier exposure to the index patient . We defined “other infection sources” based on previous literature showing that exposure to poultry or live-bird markets [6], and possibly to other birds [20], could be risk factors for H7N9 infection.

For an individual contact case patient, the incubation period distribution for H7N9 implies a range of dates over which he or she was likely to have been exposed to the source of infection; we examined the exposures during that period. Based on biological plausibility and a study that estimated a longer incubation period [21], we used the 12 days before symptom onset in the contact case patient as a conservative estimate of the maximum incubation period. Days within that range were not equally likely to lead to transmission, and we estimated an unconditional probability that exposure to the actual source of infection took place on each day during the possible incubation period. To determine these probabilities, we examined previously published research that estimated the distribution of the incubation period of A(H7N9) virus [3, 21–25]. We used an incubation period distribution from a report that examined the incubation period for 229 cases; that report, by Virlogeux et al [26], found that the mean incubation period was 3.4 days and a Weibull distribution was the best fit for the observed distribution of incubation periods [26]. Using this distribution, we derived the probability that transmission occurred on a given day during a contact case patient’s possible incubation period; these daily probabilities are shown in Supplementary Figure 1.

Exposure timing data and the incubation period distribution for each day on which a potential exposure occurred were added to compute an incubation period–weighted score for exposure to the index case patient. A score was computed separately for exposure to other infection sources. These exposure scores range from 0 to 1. By comparing the incubation period–weighted exposure score for the index case patient with that for other infection sources, we categorized each contact case by whether person-to-person transmission was probable, possible, or unlikely. “Probable” person-to-person exposure indicated that this was the most likely route of infection based on an exposure score for likely transmission from the index case patient that was ≥50% higher than the exposure score for other possible infectious sources. “Possible” person-to-person exposure indicated that person-to-person transmission could reasonably have occurred but that a common exposure to another infected source was also likely based on exposure scores that did not differ by >50% of the lower score. “Unlikely” indicates that although person-to-person transmission may have occurred, a common exposure to an infectious source was the most likely route of infection given an exposure score for other infection sources that was ≥50% higher than the index case exposure score.

The category thresholds were chosen both because 50% is understood intuitively and also because when exposure scores were compared graphically, with lines representing category thresholds (Figure 1), there was clustering into “probable” and “unlikely” categories above and below these thresholds. An example showing how an exposure score and the likelihood of person-to-person transmission is calculated for an individual cluster (cluster 16) is shown in the Supplementary Table 1.

Figure 1.

Comparison of exposure scores for likelihood of transmission of influenza A(H7N9) virus infection from exposure to the index case patient or to another infectious source (eg, infected poultry). Clusters were assigned numbers based on the index case patient’s date of symptom onset. The likelihood of person-to-person transmission is characterized as probable, possible, or unlikely. The score for exposure by a contact case patient to a cluster’s index case patient indicates the probability that such exposure during the former’s potential incubation period resulted in person-to-person transmission of influenza A(H7N9) virus infection. The score for exposure by a contact case patient to poultry or other infectious sources indicates the probability that such exposure resulted in transmission of influenza A(H7N9) virus infection.

Open in new tabDownload slide

Because other reports described longer estimated mean incubation periods, we performed a sensitivity analysis to estimate the probability of transmission based on a mean incubation of 5 days [3, 21]. Moreover, given that specific types of exposure may have been more likely to lead to virus transmission, we performed another sensitivity analysis that graded exposures on the likelihood of transmission (low, medium, or high). Our grading for other infectious sources incorporated results of environmental and animal testing conducted during the epidemiological investigations and also used a previous study that quantified the risk of infection from particular exposures [6]. These exposure grades were incorporated into the probability assessment for this sensitivity analysis; methodological details of the weighting used for various exposures are described in Supplement 1.

RESULTS

Twenty-one clusters of human infection, including 22 contact cases, were identified in 8 provinces between March 2013 and June 2015 (Table 1 and Supplementary Figure 2). One cluster had 2 contact cases (cluster 8). Four clusters were identified in the first wave of H7N9 infections (2.9% of all first wave cases; 4 of 134), 11 in the second wave (3.9%; 12 of 306), and 6 in the third wave (2.8%; 6/217). The median age was 49 years (interquartile range, 29–57 years) for the index case patients and 26.5 years (4–57 years) for the contact case patients. Fourteen of 21 index case patients (66%) were male, as were 11 of 22 contact case patients (51%). Eleven index case patients (52%) and 6 contact case patients (27%) died.

(Note: There were 2 other published reports of asymptomatic contacts of confirmed case patients who tested positive for H7N9, both by RT-PCR, but because these contacts were asymptomatic and their incubation periods could not be estimated, they were not included in this analysis [10, 27]. A suspicious nosocomial cluster including 1 index and 2 possible H7N9 contact case patients, both physicians, was identified in Shantou, Guangdong province, in January/February 2015 and reported in 2 publications [28, 29]. Because the index case and 1 contact case lacked RT-PCR confirmation of H7N9 infection, this cluster was not included in our analysis.)

A description of the contact cases patients’ exposures are shown in Table 1, and a timeline for each cluster is shown in Supplementary Figure 2. Clinical characteristics each of cluster’s patients are described in Supplement 3.

The days of exposure, exposure scores, and risk assessment for all contact cases are shown in Table 2. Based on the exposure score, the infections in 12 of 22 contact case patients were deemed to have probably resulted from person-to-person transmission (Table 2 and Figure 1). Another 4 infection were considered possible person-to-person transmission events, although exposure timing information did not allow us to distinguish between the likelihood of transmission from the index case patient or from a common poultry source. Infections in 6 contact case patients were unlikely to have resulted from person-to-person transmission and more likely to have been transmitted from a common infectious source. Among those clusters in which person-to-person transmission was deemed probable, all but 2 involved family members. In clusters 7 and 21, nosocomial transmission probably occurred when the index case patients, hospitalized with what was later confirmed to be H7N9 virus infection, shared hospital wards with the respective contact case patients. The latter had no known exposure to live poultry, and all 4 patients died [18, 30].

When we conducted a sensitivity analysis using a mean incubation of 5 days, the probability assessments for person-to-person transmission were unchanged for 21 of 22 contact cases; cluster 14 changed from probable to possible person-to-person transmission (Table 2). In the sensitivity analysis that also graded exposures as high, medium, or low risk, cluster 10 changed from possible to unlikely, and cluster 12 from possible to probable (Supplementary Table 2).

DISCUSSION

Detailed epidemiological investigations and analysis of 21 clusters of human infection with H7N9 virus suggests that limited person-to-person transmission of H7N9 virus was likely to have occurred multiple times since the virus emerged in spring 2013. The simple framework developed here used the timing of exposures to assess the likelihood that person-to-person transmission of H7N9 infection had occurred in identified clusters. Using this framework, we found that between March 2013 and June 2015, probable person-to-person transmission occurred in 12 instances, including an instance of probable nosocomial transmission (clusters 7 and 21) [18], with possible person-to-person transmission occurring in another 4 instances. Although these cases represent only a small fraction of human cases of H7N9 infection identified in the study period, these findings underscore the need for continued vigilant H7N9 surveillance and for the early detection and isolation of infected persons to minimize the possibility of person-to-person transmission, including in hospital settings.

Detailed investigations of human clusters provide us with valuable information about the epidemiology of H7N9 virus. If the H7N9 virus gains the ability to transmit easily in a sustained way between persons, it has the potential to cause a global influenza pandemic. Several studies used case data and statistical modeling to assess the transmissibility of the H7N9 virus and estimate reproduction number of H7N9, including 2 that examined H7N9 cluster data during the first 2 waves of H7N9 infections [15–17]. Yang et al [17] used information on household contacts, including cases identified as part of household clusters, to develop a statistical model that found evidence of nonsustainable person-to-person transmission during the first 2 waves. This is similar to the findings of another study that used cluster data to compare the epidemiology of H5N1 viruses to that of H7N9 [15].

Our study adds to this body of work by presenting detailed exposure and epidemiological data for the clusters of human infection, as well as assessments for each individual cluster regarding the likelihood that person-to-person transmission occurred. The simple method developed to assess clusters can be used as a standard way to rapidly evaluate future clusters and to monitor for important changes in the risk posed by the virus. The distinction between those cases in which person-to-person transmission was deemed “probable” compared with “possible” may be somewhat arbitrary, and even small changes in case finding and contact tracing methods could affect how clusters are identified and classified. However, if we found an increase in the proportion of contacts whose infection was deemed probable or possible person-to-person transmission, this method could be used to identify possible increased transmissibility among humans requiring further investigation.

We found that person-to-person transmission was likely to have occurred on several occasions during the first 3 waves of infection; our analysis depended on an estimated incubation period of 3.4 days, with probability distributed in a Weibull distribution [26]. Our results were largely unchanged in a sensitivity analysis using a longer incubation period. Estimates of the median incubation period for H7N9 infections vary from 2 [24] to 7.5 days (range, 2–12.5 days) [21]. Other influenza viruses have shorter estimated incubation periods. A systematic review estimated the incubation period for seasonal influenza A to be 1.4 days, with symptoms developing in 95% of case patients by 2.8 days after infection [31]. Previous estimates of the median incubation for avian influenza A(H5N1) vary from 3 days in Thailand [32] and Vietnam [33] to 5 days (range 2–9.5 days) in China [34].

Our estimates of the probability of person-to-person transmission could be inaccurate if the incubation period distribution was misestimated. We also assumed that there were no differences in incubation periods between those who acquired their infection from exposure to poultry and the environment and those who acquired it from person-to-person transmission. Nevertheless, as more cases of human infection with H7N9 occur and estimates of its incubation period are refined, this model is easily adjusted to incorporate this updated information.

Our method of assessing person-to-person transmission in a given cluster has a number of limitations. If an individual has exposure to both infected humans and other possible sources of infection, judgments regarding the likelihood that person-to-person transmission occurred will necessarily be subjective. We attempted to limit subjectivity in our analysis by focusing solely on exposure timing. However, by incorporating only exposure timing, we ignore a large amount of potentially useful information, including the presence or absence of laboratory-confirmed H7N9 virus at a site of potential exposure, information about molecular epidemiology of the recovered viruses, and information regarding exposure details.

In a sensitivity analysis, we did attempt to incorporate additional data, including the type of exposure that the contact had to the index case patient and to poultry; this was included it as a secondary analysis presented in Supplement 2. Certain exposures are more likely than others to increase the risk of transmission (eg, direct handling of infected birds vs going to an infected live-bird market but having no direct bird contact) [6]. However, it is difficult to assess the risk of H7N9 infection after an exposure to infected poultry compared with that after exposure to an infected person; it is thus challenging to correctly define relative weights for these types of exposure. If cases of human infection with H7N9 continue to occur and more information becomes available about the type and quantity of exposures that result in infection, our methods can be refined, and more variables can reliably be included in future cluster assessments.

This study had several other limitations. Even though we limited our assessment to exposure timing, we were still unable to distinguish between poultry and human sources of infection in a number of cases. We also used information from detailed epidemiological investigations and aimed to be inclusive in this article in discussing possible exposures that could have resulted in infection. However, judgments about what constitutes a potentially relevant exposure are subjective, and potential exposures may have been overlooked or excluded. In addition, many patients died or were very ill, and exposure information could not always be corroborated. Potential sources of infection may have been overlooked, and information on timing may have been incorrect.

In conclusion, we demonstrate that limited person-to-person transmission is likely to have occurred on multiple occasions since the H7N9 virus was first identified, although these transmission events represented a small fraction of all identified cases of H7N9 human infection and sustained person-to-person transmission was not documented. Our findings underscore the importance of thorough investigation of all human cases of novel influenza A.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official views of the US Centers for Disease Control and Prevention.

Financial support. This work was supported by the Chinese Center for Disease Control and Prevention.

Supplement sponsorship. This work is part of a supplement sponsored by the Centers for Disease Control and Prevention.

Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References