New H7N9 Data: An Epidemic Rising?

By Neil Seeman, JD, MPH* CEO, The RIWI Corporation

Senior Resident, Massey College, University of Toronto

CEO, Health Strategy Innovation Cell

Adjunct Lecturer, Dalla Lana School of Public Health Policy

Adjunct Professor, Institute for Health Policy, Management & Evaluation

Adjunct Professor, Ryerson University School of Health Services Management

As of the time of writing, it has been suggested that the new bird flu strain in China, H7N9 – with a higher case fatality rate (CFR) at present than SARS – bears possible signs of human-to-human transmission. Our global data capture company, The RIWI Corporation, tracked 7,016 Chinese “fresh” (i.e. non-panel based) Internet users – with a 24.08% response rate – over 20 hours. Our tracking data are visualized here with mouse-over technology revealing different levels of intensity of public awareness of the contagion (varying by region and by city) across China. At the 95% confidence interval, the margin of error is 1.8%, 19 times out of 20. See how RIWI’s early predictive analysis compares to visuals of the case contagion published in Nature on April 24, 2013.

The RIWI data tracker ran in all parts of China from Sunday April 14th at 22:49 to Monday April 15th at 19:50 China Standard Time.

The findings were based on The RIWI Corporation’s patented, privacy-compliant and peer-reviewed data algorithms and technology that reach Internet users with equal random probability in any targeted region of the Internet-enabled world. Through the patented process of someone making a manual type-in error in the browser bar (URL bar, or navigation bar) on any browser on any Web-enabled device, such as a tablet or desktop; and thereafter stumbling upon the thousands of random, rotating so-called ‘non-sense’ domain names on any top-level or country-specific domain (e.g., in this case, the top-level country domain is .cn) where RIWI Nano-Surveys™ are hosted, respondents were asked, in Mandarin, very short, rapid questions with low-latency and the replies were immediately and automatically tallied and geo-segmented on multiple RIWI servers, which then link IP-based respondents to privacy-compliant identifiers. RIWI software ensures that respondents can only answer once, that no Internet ‘bots’ can spam the question pages, and that all respondents are real, human respondents.

More on RIWI’s unique methodology may be found here and here.

For this project, the questions (in Mandarin) were: “Do you know someone who has contracted the new Avian Flu in your local area?” “What is your age?” “What is your gender?” The data were segmented according to Province and city. The first question was deliberately phrased to allow public health leaders to move scarce resources to affected areas based on the assumption that a growing awareness of contagion in an area precedes actual deaths. If there are areas where awareness is high, this could be an early signal that there are more deaths looming, especially when the incubation period for this particular strain appears to be 5-7 days.

Almost 30,000 – 29,136 “fresh” Internet users (i.e. that is, people who do not ‘opt in’ to incentivized survey panels or regularly take surveys) – were randomly exposed to the questions over 20 hours. At 24.08%, this response rate eclipses typical response rates to panel-based Internet surveys (ranging from 4.7% for personalized approaches, and 2.2% for a generic approach using panels). Scholarly literature on this topic concurs that a first-time “fresh” responder (not on a panel) to a survey provides more reliable replies than does a practiced, incented responder who has previously been through many Internet surveys.

Our data findings are below.

H7N9 Awareness Across China

H7N9 Awareness vs. Expected Awareness in Different Chinese Cities

The red-highlighted cities are Chinese cities where, as compared to their Census (2010) population, the level of contagion awareness statistically exceeds the expected level of awareness. These cities, as explained below in the analysis, tend to be those where H7N9 deaths have been reported. This analysis is based on the Chi Square Test of Independence at the 0.05 level of significance.

Geographic Representativeness of the Data

Removing outliers (notably, six cities, Zhejiang, Shanghai, Jiangsu, Guangdong, Fujian, and Beijing) where Internet penetration is at least two standard deviations above the mean, a regression analysis and correlation coefficient (below) illustrates that our data are geographically representative of the Chinese data (applying 2010 Chinese census data, the most recent publicly available for all cities measured).

Analysis of Major Findings

  1. Elevated awareness of someone having contracted the new Avian Flu appears to correspond with the prevalence of deaths, except in Anhui Province, where there had been one death (and three cases) as of the time of writing. Official data (at 6pm ET) from the Chinese National Health and Family Planning Commission noted that there had been nine deaths (and 24 cases) in Shanghai; two deaths (and 17 cases) in Jiangsu Province; two deaths (and 16 cases) in Zhejiang Province; one case in Beijing; and two cases in Henan Province. That means the fatality rate is 14 out of 63, or 22.2%. By comparison, the case fatality rate from SARS was one out 10, or 9.6%.
  2. Awareness levels of someone knowing that someone in one’s local area has contracted the bird flu strain vary across provinces dramatically, from 0% in Xizang to 39.82% in Shanghai.
  3. There is no statistical dependency or correlation between gender and awareness.
  4. There is some dependency between age group and awareness. The under-18 population and those over 65 reported more often that they were aware of the possible contagion.
  5. Once removing outliers, we have statistical geographic representation of China in terms of respondents and population (as noted in the correlation line graph above).
  6. City and region location are overwhelmingly correlated with knowledge of someone having contracted the illness.

By way of comparison, the H5N1 bird flu strain killed at least 371 people in Asia and Africa over the past ten years. Moreover, in 2009, the H1N1 flu virus caused the first influenza pandemic in over forty years.

This article provides an early view of how the current H7N9 bird flu appears to be spreading. Due to the emergency nature of this possible pandemic, we chose not to pursue publication in traditional research journals, even so-called ‘fast-track’ publications, since the importance of releasing the data immediately is essential.

Knowledge of apparent bird flu in the local areas of the major cities provides a clue as to where additional public health services may be needed in the near future. In addition to the H7N9 virus spreading from birds to humans, it is unknown at this time whether there is, in fact, human-to-human spread. Moreover, the prevalence of bird flu in the local area naturally depends on the 5 to 7 day incubation period of the virus.

While it is acknowledged that the RIWI Nano-Survey™ data shown here were collected from Internet users who were only informed indirectly or by hearsay that someone in their local area had apparent bird flu, the data do provide a basis for a possible epidemic, as based on the rapid spread of the self-reported information. The very high response rate (24.08%) far exceeds that of other Internet surveys, suggesting an intense relevancy of interest in the subject matter, another signal of the urgency of this matter in the minds of average Chinese citizens.

Notes on methodology:

  • No privacy identifiers were collected. IP addresses in China are immediately and anonymously switched to location-specific privacy compliant identifiers.
  • No information on the health of the respondents was collected.
  • Each respondent was made aware that he or she does not need to answer the question and is presented with information that the data will not be disclosed to any third parties or used in any way for commercial purposes.
  • There are no browser ‘cookies’ that “follow” a respondent after he or she has answered the questions, or that collect information of any kind that is disclosed to the RIWI Corporation or to other third parties.
  • The RIWI Corporation performed this research in the public interest, with no outside financial support from anyone or from any organization.