Beijing, People’s Republic of China

The attack rate was low, and having contact with an ill household member and younger age were the major risk factors.

I n early April 2009, human cases of infection with a novel infl uenza virus of swine origin, pandemic (H1N1) 2009 virus, were identifi ed in the United States and Mexico, and this virus spread rapidly across the world (1-3). On June 11, 2009, the World Health Organization raised the pandemic level to 6, the highest level for pandemic alert (4).
Estimating attack rates is a major task in characterizing pandemic (H1N1) 2009. Some studies have reported attack rates of pandemic (H1N1) 2009 among household members and aircraft passengers (5)(6)(7). These studies suggested that the transmissibility of pandemic (H1N1) 2009 virus was low. These studies were conducted in outbreak settings, and attack rates were calculated on the basis of clinical diseases that included infl uenza-like illness (ILI) or acute respiratory illness (ARI) of close contacts rather than confi rmed infection with pandemic (H1N1) 2009 virus. In addition, in these studies only symptomatic index and secondary cases were included. Although most infections of pandemic (H1N1) 2009 infl uenza virus produce ILI or ARI symptoms (8)(9)(10)(11)(12), subclinical infection can occur and can change the estimate of attack rate. In addition, the infectivity of asymptomatic case-patients has not been clearly defi ned (13).
Because of the high rates of illness and death among the initial case-patients with pandemic (H1N1) 2009 (14), the Chinese government decided to prevent and contain the rapid spread of disease through tracing and quarantine of persons who had close contact with persons with confi rmed cases of pandemic (H1N1) 2009. Beijing, the capital of the People's Republic of China, took strict containment and control measures through October 2009. The Beijing municipal government implemented border entry screening, ILI screening in hospitals, health follow-up of travelers from overseas, and quarantine and testing of close contacts to identify new introduction of cases and local transmission. Public health workers conducted epidemiologic investigation of all index case-patients (including those with subclinical infections) and traced and quarantined close contacts whose residence was within the jurisdiction of Beijing. We estimated the attack rate of pandemic (H1N1) 2009 virus infection and assessed risk factors or correlates for infection among different types of close contacts, including household members and aircraft passengers.

Confi rmation of Index Cases
In 2009, under the guidance of the Beijing Center for Disease Prevention and Control (Beijing CDC), a network of 55 collaborating laboratories was established to perform reverse transcription PCR testing to confi rm cases of pandemic (H1N1) 2009 (15). The confi rmed cases included symptomatic and asymptomatic cases, and these cases were detected mainly by border entry screening, ILI screening in hospitals, health follow-up of travelers from overseas, and quarantine and testing of close contacts. Once confi rmed, index case-patients were immediately quarantined in designated hospitals to receive treatment while in isolation. All the confi rmed cases were required by law to be reported to Beijing and local CDCs. From May through October 2009, a detailed epidemiologic investigation was conducted for each confi rmed case of pandemic (H1N1) 2009 (including symptomatic and asymptomatic cases) by Beijing and local CDCs within 6 hours after confi rmation of infection. Patients with confi rmed cases were interviewed about demographic characteristics, course of illness, travel and contact history, and information about close contacts. Patients with confi rmed cases were categorized as having imported cases (travelers) and locally acquired cases (no travel history) on the basis of where the infection was acquired.

Defi nition of Close Contacts
Close contacts were defi ned as anyone who ever came within 2 meters of an index case-patient without the use of effective personal protective equipment (PPE) (including masks and gloves, with or without gowns or goggles) during the presumed infectious period. Trained staff from local CDCs made the determinations on the basis of fi eld investigation. The relationships of close contacts to index case-patients were categorized as 1) spouses, 2) other household members, 3) nonrelated roommates, 4) contacts at workplace or school, 5) nonhousehold relatives, 6) passengers on the same fl ight, 7) friends, and 8) service persons met at public places. A close contact on an aircraft was defi ned as a passenger sitting within 3 rows in front and 3 rows behind the index case-patient.
All close contacts were traced and quarantined for 7 days after the most recent exposure to the index casepatient. All index case-patients detected between May 16 (the fi rst case, the date of confi rmation) and September 15, 2009 (before widespread transmission in Beijing), and their close contacts were included in this study. We excluded cluster or outbreak cases for which close contacts could not be determined clearly by epidemiologic investigation (the transmission chain was obscure).

Laboratory Screening
For each close contact, before quarantine, a pharyngeal swab specimen was collected for reverse transcription PCR testing, regardless of symptoms. A second pharyngeal swab specimen was collected for testing for pandemic (H1N1) 2009 virus if any of the following symptoms developed in a close contact during quarantine: axillary temperature >37.3°C, cough, sore throat, nasal congestion, or rhinorrhea.

Statistical Analysis
Data were analyzed by using SPSS version 11.5 (SPSS Inc., Chicago, IL, USA). Median and range values were calculated for continuous variables, and percentages were calculated for categorical variables. Differences in attack rates were compared between subgroups of close contacts by using the χ 2 test. For the signifi cant difference found in multiple subgroups, this test does not enable identifi cation of which multiple subgroups are signifi cantly different, only that across all the subgroups there are differences. The variables with p<0.10 in χ 2 test were included in multivariate analysis. Multivariate unconditional logistic regression analysis was conducted to determine risk factors associated with infection in close contacts. Backward logistic regression was conducted by removing variables with p>0.10. Odds ratios (ORs) and 95% confi dence intervals were calculated for potential risk factors of infection. The Hosmer-Lemeshow goodness-of-fi t test was used to assess the model fi t for logistic regression. All statistical tests were 2-sided, and signifi cance was defi ned as p<0.05.

Timeliness and Intensity of Index Case Detection and Contact Tracing
A total of 613 eligible index case-patients, detected from May 16 through September 15, 2009, were included in this study. Through fi eld epidemiologic investigations, 7,099 close contacts were traced and quarantined in Beijing. The median number of close contacts per index case per day was 7.0 persons (range 2.0-95.0 persons); the median number for an imported index case was 7.0 persons (range 1.7-95.0 persons) and for a locally acquired index case was 5.3 persons (range 1.0-25.0 persons). For the 601 symptomatic index case-patients, the median interval between illness onset and sample collection was 1.0 days (range −1.9 to 7.0 days).
Among close contacts with symptomatic infection, the median interval between illness onset and sample collection was 0.5 days. More than 85% of close contacts were quarantined within 72 hours after interview of the index case-patients. The median interval between fi rst exposure and quarantine was 3.4 days for the close contacts, and it was shorter, on average, for fl ight passenger contacts than nonpassenger contacts (1.7 days vs. 3.8 days). For symptomatic close contacts infected with pandemic (H1N1) 2009, the median of generation time (i.e., the time from illness onset in an index case to illness onset in a secondary case) were 2.4 days; it was shorter for fl ight passenger contacts than nonpassenger contacts (1.6 days vs. 2.5 days) ( Table 1).

Characteristics of Index Case-Patients and Close Contacts
Approximately 43% of the index case-patients were women; the median age was 20 years, and 38% likely contracted pandemic (H1N1) 2009 virus locally because they had not traveled recently. Among the index casepatientss, 2% had subclinical infection. Only 18% of index case-patients had close contacts with confi rmed pandemic (H1N1) 2009 (Table 2), and the total number of close contacts who were infected by the virus from 110 index case-patients was 167.
Fifty percent (3,514 of 7,032) of close contacts were women, and the median age was 27 years. Approximately 12% of close contacts were household member of index case-patients (spouse or other household member), and aircraft passengers accounted for 44% of close contacts. Approximately 61% of close contacts were exposed to symptomatic index case-patients during their symptomatic phase. About 70% were quarantined in a quarantine station ( Table 2).

Attack Rate
The overall attack rate for infection among close contacts (positive test result) was 2.4% (167 of 7,099), indicating that 1 index case-patient transmitted infection to 0.27 close contacts (167 of 613) on average (reproduction number = 0.27). Among those close contacts with a positive test result, 14.4% (24 of 167) had subclinical infection; among the close contacts with positive test results at the start of quarantine, 17.2% (20 of 116) had subclinical infection.
Attack rates did not differ by index case-patient's sex (p = 0.225). However, attack rates differed signifi cantly by index case-patient's age (p = 0.022), and the lower attack rate was found for older index case-patients. There was no signifi cant difference in attack rates between close contacts of patients with imported cases and those with locally acquired cases (p = 0.282). No infection was found in close contacts exposed to index case-patients with subclinical infection, and the attack rate observed in close contacts exposed to symptomatic index case-patients during their symptomatic phase was higher (p<0.001). Almost identical attack rates were found among male and female close contacts (p = 0.808). However, attack rates were signifi cantly different among different age groups of close contacts (p<0.001), and the lowest attack rate was found for those >50 years of age. The attack rates were signifi cantly different across 8 contact types (p<0.001). The attack rate was 5.3% among spouses and 6.6% among other family members in the household, and was lower among other types of close contacts ( Table 3).The attack rate among passengers on the same fl ight was low, 0.9% overall, and 1 index case-patient transmitted infection to 0.19 close contacts on a fl ight on average (28 of 147), and the attack rate was higher among the passengers with longer fl ight times (>12 hours, p = 0.001). The attack rate among close contacts of service persons at public places was 0.2%, and 1 index case-patient transmitted infection to 0.01 close contacts of service persons on average (1 of 113). Nonpassenger close contacts with longer exposure duration (>12 hours), compared with those with shorter duration (>12 hours), recorded the higher attack rate (p<0.001) ( Table 3).

Risk Factors
By multivariate analysis, age and type of contact were the major predictors of infection (  (OR 3.42; p = 0.002) and 0-19 years of age (OR 7.76; p< 0.001) were at higher risk for infection. Other signifi cant independent risk factors associated with infection included being a household member of a person with an index case (OR 3.83; p<0.001), being exposed to index case-patients during their symptomatic phase (OR 1.86; p = 0.003), and exposure duration >12 hours (OR 1.83; p = 0.002). Similar risk factors were observed among aircraft passengers.

Discussion
We estimated that pandemic (H1N1) 2009 virus was transmitted by 18% of index case-patients to their close contacts and that 2.4% (167 of 7,099) of close contacts we traced were infected. Our data indicate that pandemic (H1N1) 2009 virus has low transmissibility in nonoutbreak settings.
We found that 1 index case-patient transmitted infection to 0.27 close contacts on average, i.e., reproduction number = 0.27. This fi nding suggests that among those quarantined index case-patients, the number of persons with secondary cases who could be traced through rigorous fi eld investigation was small and far less than the number needed for the sustainable transmission of infectious disease in the population (reproduction number >1). However, the fact that the pandemic eventually spread in Beijing indicates that contact and case tracing were far from complete, especially later in the summer and early fall of 2009. The strict control measures may have worked to some extent at the beginning but were outpaced by local transmission (16); the percentage of locally acquired infections ranged from <10% in June 2009 to >80% in September 2009 (data not shown).
In this study, the median number of close contacts per index case-patient per day was 7.0 persons. Although locating and quarantining these close contacts was done quickly, and stringent quarantine measures were used, which hindered implementation of control measures, the real number of close contacts was unknown and probably exceeded this number. Many close contacts were persons met in public places, including public transportation, theaters or cinemas, and shopping malls, and it is nearly impossible to trace all of the contacts. In addition, some persons who had worn PPE during contact with index casepatients were excluded from close contacts management (i.e., they were not quarantined), but because wearing PPE might not protect (or fully protect) against infection, some persons excluded might have become infected. In addition, many persons with mild and asymptomatic cases cannot be detected, but they may transmit the virus. Furthermore, the short generation time of pandemic (H1N1) 2009 shown in this study and in a previous study (13) could lead to the rapid accumulation of infection sources and close contacts. This rapid compounding could overwhelm response capacity and would have resulted in compromised effectiveness of containment measures. It should also be mentioned that we did not include persons with cluster or outbreak cases for whom close contacts could not be determined clearly by epidemiologic investigation to examine the basic feature of pandemic (H1N1) 2009 (e.g., generation time), and the reproduction number obtained from our data is an underestimate.
Attack rates of infection differed signifi cantly by contact type. Among household members of index casepatients, the attack rate was the highest, as shown in the multivariate analysis after controlling for age and other factors. The most likely reason for this fi nding is that household members are more likely to have come into closer contact with index case-patients for a longer period with shorter distance and longer duration. Another possible reason is that household members may have some certain linkage with index cases in genetic susceptibility or living habits that would cause higher predisposition in household members than in other close contacts. This fi nding is similar to fi ndings in other investigations of respiratory infectious disease (17).
Close contacts on fl ights accounted for the highest proportion of all the close contacts, in part because of how the index cases were detected and the broad defi nition we used for close contacts. However, the attack rate was much lower than that for other close contacts; 1 index case infected only 0.19 close contacts on fl ights on average. This fi nding indicated that the possibility of transmission of pandemic (H1N1) 2009 virus on fl ights was low, and the yield of tracing and quarantining of close contacts on fl ights was limited. Tracing contacts of service persons at public places was more diffi cult than tracing other categories of contacts, and the lowest attack rate (0.2%) was recorded in this category. Despite extensive measures, on average, only 0.01 infected close contacts per index case-patient were identifi ed among service persons. Tracing the contacts of service persons at public places seems far less cost-effective. Criteria for close contacts on fl ights and those of service persons should be refi ned with respect to exposure duration and age of those exposed. Exposure to index case-patients for >12 hours was a signifi cant independent risk factor for infection in fl ight passenger contacts. This fi nding suggests that limiting the time of contact with persons with ILI on aircraft can reduce risk for transmission, and a long duration of exposure may be necessary for transmission to occur on aircraft.
Younger close contacts were at higher risk for infection than older ones. The possible reason was that younger persons had much closer contact with index case-patients than did older persons; another reason may be that younger persons were more susceptible to infection with pandemic (H1N1) 2009 virus (18). This fi nding was consistent with fi ndings reported in other studies (5,6). No secondary cases were found among close contacts exposed to index case-patients with subclinical infection. The attack rate among close contacts who were exposed to symptomatic index case-patients during their symptomatic phase was much higher than that among those exposed to these case-patients before their illness onset. Exposure to index case-patients during the symptomatic phase was a signifi cant independent risk factor for infection among close contacts. These fi ndings indicate that the infectivity of pandemic (H1N1) 2009 virus was higher after illness onset, and that the infectivity of symptomatic pandemic (H1N1) 2009 case-patients before illness onset was higher than that of persons with subclinical cases, although persons in each group were asymptomatic when in contact with other persons.
In general, the earliest infectious time for pandemic (H1N1) 2009 was considered as 1 day before illness onset (19). We found that index case-patients and infected close contacts shed pandemic (H1N1) 2009 virus <1 day before illness onset, which suggests that the infectious period of symptomatic persons with pandemic (H1N1) 2009 might be <1 day before illness onset.
Among close contacts with pandemic (H1N1) 2009, ≈14.4% were asymptomatic. It is noteworthy that specimens from some close contacts tested negative for pandemic (H1N1) 2009 virus before quarantine, but those persons could shed the virus during quarantine without symptoms. Such infection could not be detected, and the proportion of subclinical infection was underestimated. Therefore, we calculated the proportion of subclinical infection by cross-sectional analysis of the subclinical infection of close contacts before quarantine, and we found that ≈17% of case-patients with pandemic (H1N1) 2009 were asymptomatic.
This study has several limitations. We could not fi nd all close contacts of persons with pandemic (H1N1) 2009 and did not know their infection status, so the infection parameters of pandemic (H1N1) 2009 that we found in this study might not be precise, especially for reproduction number, which may be underestimated to some extent. Furthermore, we could not exclude the possibility that the infected close contacts had been infected from another unknown source before quarantine started, which might infl uence our conclusion to some extent.
In summary, the attack rate among close contacts was low, even among household contacts. Household member and younger age were the major risk factors for infection with pandemic (H1N1) 2009 virus among close contacts. Approximately 17% of cases of pandemic (H1N1) 2009 were asymptomatic.  Table 2 were included in multivariate unconditional logistic regression analysis. Hosmer-Lemeshow goodness-of-fit test was used to assess the model fit for logistic regression. OR, odd ratio; CI, confidence interval; NA, not available, indicating not included in the final model. †One dependent variable (infection with pandemic [H1N1] 2009 virus) and 5 independent variables (age of index case-patient, type of exposure to index case-patients, age of close contacts, relationships to index case-patients, and exposure duration of close contacts) were included in multivariate analysis. One independent variable (age of index case-patient) was removed in the stepwise regression equation. The goodness-of-fit test suggested that the logistic regression model fitted well (p = 0.631).
‡One dependent variable (infection with pandemic [H1N1] 2009 virus) and 4 independent variables (age of index case-patient, type of exposure to index case-patient, age of close contacts, and exposure duration of close contacts) were included in multivariate analysis. Two independent variables (age of index case-patient and type of exposure to index case-patient) were removed in the stepwise regression equation. The goodness-of-fit test suggested that the logistic regression model fitted well (p = 0.982).
§One dependent variable (infection with pandemic [H1N1] 2009 virus) and 5 independent variables (age of index case-patient, type of exposure to index case-patient, age of close contacts, relationships to index case-patient, and exposure duration of close contacts) were included in multivariate analysis. Two independent variables (age of index case-patient and exposure duration of close contacts) were removed in the stepwise regression equation. The goodness-of-fit test suggested the logistic regression model fitted well (p = 0.751).
¶Exposed to symptomatic index case-patients before their illness onset or exposed to index case-patients who had subclinical infections.