Open Access Research BMJ Open: first published as 10.1136/bmjopen-2015-009900 on 30 December 2015. Downloaded from Influenza newspaper reports and the influenza epidemic: an observational study in Fukuoka City, Japan Akihito Hagihara,1 Daisuke Onozuka,1 Shougo Miyazaki,2 Takeru Abe3 To cite: Hagihara A, ABSTRACT et al Strengths and limitations of this study Onozuka D, Miyazaki S, . Objectives: We examined whether the weekly number Influenza newspaper reports of newspaper articles reporting on influenza was ▪ and the influenza epidemic: This is the first study to quantify the association related to the incidence of influenza in a large city. an observational study in between media attention and the incidence of Fukuoka City, Japan. BMJ Design: Prospective, non-randomised, observational influenza in a large city. Open 2015;5:e009900. study. ▪ Although pertinent information related to influ- doi:10.1136/bmjopen-2015- Setting: Registry data of influenza cases in Fukuoka enza might have been broadcast through other 009900 City, Japan. media except for newspaper, these effects were Participants: A total of 83 613 cases of influenza not evaluated. ▸ Prepublication history for cases that occurred between October 1999 and March ▪ We need to be careful when generalising our this paper is available online. 2007 in Fukuoka City, Japan. results to other cities with the same population To view these files please Main outcome measure: A linear model with size in industrialised countries. visit the journal online autoregressive time series errors was fitted to time (http://dx.doi.org/10.1136/ series data on the incidence of influenza and the bmjopen-2015-009900). accumulated number of influenza-related newspaper Vaccination has been shown to be a prevent- articles with different time lags in Fukuoka City, Japan. Received 3 September 2015 ive measure for influenza; vaccines contain- In order to obtain further evidence that the number of Revised 28 October 2015 ing antigens that match the predominant Accepted 24 November 2015 newspaper articles a week with specific time lags is related to the incidence of influenza, Granger causality circulating strains have been reported to – was also tested. prevent illness in 70 90% of healthy adult 125 Results: Of the 16 models including ‘number of volunteers. Approximately 50 countries newspaper articles’ with different time lags between 2 have government-funded national influenza fl http://bmjopen.bmj.com/ and 17 weeks (xt-2 to t-17), the β coefficients of ‘number immunisation programmes, and the in u- of newspaper articles’ at time lags between t-5 and t- enza vaccine is available in many other coun- 13 were significant. However, the β coefficients of tries as well.1 ‘number of newspaper articles’ that are significant with According to the Elaboration Likelihood respect to the Granger causality tests (p<0.05) were Model (ELM) of Petty et al7 people are the weekly number of newspaper articles at time lags most likely to pay careful attention to infor- between t-6 and t-10 (time shift of 10 weeks, β= mation when it is relevant to them and when −0.301, p<0.01; time shift of 9 weeks, β=−0.200, p<0.01; time shift of 8 weeks, β=−0.156, p<0.01; time it can be processed without assistance. shift of 7 weeks, β=−0.122, p<0.05; time shift of Concomitant with this theoretical expect- on October 2, 2021 by guest. Protected copyright. 1 Department of Health 6 weeks, β=−0.113, p<0.05). ation, it has been reported that media activ- Services Management and Conclusions: We found that the number of ities can directly affect immunisation take-up, Policy, Kyushu University 8 newspaper articles reporting on influenza in a week positively and negatively. A recent Cochrane Graduate School of Medical – Sciences, Fukuoka, Japan was related to the incidence of influenza 6 10 weeks review also concluded that mass media 2Department of Acupuncture after media coverage in a large city in Japan. should be considered as a tool for encour- and Moxibustion, Faculty of aging the use of effective health services and Health Care, Teikyo Heisei discouraging the use of those of unproven University, Tokyo, Japan effectiveness.9 Specifically, media reports 3Advanced Critical Care and have affected the vaccination rates for Emergency Center, 10 Yokohama City University INTRODUCTION measles, mumps and rubella, whooping 11–13 8 14 15 Medical Center, Kanagawa, During annual influenza epidemics, approxi- cough, rotavirus and influenza. Japan mately 5–15% of the population is affected Specifically, Ma et al14 reported that the by upper respiratory tract infections.1 paediatric influenza vaccination rates from Correspondence to fl Professor Akihito Hagihara; Accordingly, in uenza is a serious health 2003 to 2004 were similar to those from 2002 [email protected] threat and a significant cause of mortality in to 2003 until the onset of intense media – u.ac.jp elderly people in industrialised countries.2 6 coverage, at which time the number of Hagihara A, et al. BMJ Open 2015;5:e009900. doi:10.1136/bmjopen-2015-009900 1 Open Access BMJ Open: first published as 10.1136/bmjopen-2015-009900 on 30 December 2015. Downloaded from influenza vaccinations per week increased more than influenza-related newspaper articles. The systematic sur- threefold. According to a study whose participants were veillance of influenza as a notifiable disease, as defined community-dwelling elderly people in the USA insured under the Law concerning the Prevention of Infectious through Medicare, media reports in television and news- Diseases and Medical Care for Patients Suffering papers were associated with earlier and increased influ- Infectious Diseases, began in 1999 in Japan and allowed enza vaccination, shifting the mean vaccination timing the determination of the number of influenza cases in earlier by 1.8–4.1 days (p<0.01) or increasing the annual any given week. Each influenza case was defined by the vaccination rate by 2.3–3.7% (p<0.01).15 following symptoms: sudden fever (temperature >38°C), With respect to media coverage and influenza, previ- respiratory symptoms, general malaise and myalgia. The ous findings can be summarised as follows: (1) media number of patients infected with influenza was reported coverage is focused on a specific high-risk group such as on a weekly basis by 198 sentinel paediatric and general children or senior citizens; (2) the outcome measure for physicians within the Fukuoka Prefecture.17 The senti- the effectiveness of media coverage is influenza vaccin- nels were selected so that they were evenly distributed in ation, not the number of cases of influenza. As for terms of population, number of medical facilities and these points, however, in view of ELM and previous geographical conditions. Relevant data for Fukuoka City – findings,7 91415it is possible that influenza-related were used in this study. Data related to the influenza media reports might be related to the incidence of influ- cases in Fukuoka City were obtained from the National enza cases. To date, however, the association between Epidemiological Surveillance of the Infectious Diseases media attention and the incidence of influenza in a System, which monitors infectious disease events among large city has not been evaluated. the roughly 1.2 million residents of Fukuoka City We obtained information concerning influenza from through 50 sentinel physicians.17 various media sources including TV, radio, internet and A total of six newspapers (Yomiuri, Asahi, Mainichi, magazines. Among the various types of mass media avail- Nikkei, Sankei and Nishinippon), including the five most able, information in the newspapers is coming to the widely circulated newspapers in Japan and the most readers, whereas in the internet people have to search widely circulated newspaper in the Fukuoka area for this information. Newspapers are read by more than (Nishinippon), were examined for the publication of 90% of the population and represent the most trusted influenza-related articles.18 The Nikkei Telecom 21 data- source of information in Japan.16 In addition, it is rela- base covers newspaper articles from 1980 to the tively easy to evaluate the amount of information on a present.19 In this study, we determined the number of specific topic in newspapers, whereas it is extremely diffi- articles published per week containing the word ‘influ- cult to evaluate the amount of information conveyed via enza’ in the headline. An article on ‘bird flu’ was TV, radio, internet and magazines. Thus, we used the excluded. In addition, when the content of an article number of influenza-related newspaper articles report- identified by ‘influenza’ was not in fact related to influ- ing on influenza. In this study, we examined whether enza, the article was excluded. Figure 1 shows the con- http://bmjopen.bmj.com/ the number of influenza-related newspaper articles in a tents of newspaper articles between October and the week was related to influenza incidence by using recent following March during the study period. Articles were data gathered on the number of relevant newspaper arti- categorised into the following six subgroups: ‘influenza’ cles and the incidence of infection in Fukuoka City, and ‘epidemic’, ‘influenza’ and ‘symptoms’, ‘influenza’ Japan. and ‘prevention’, ‘influenza’ and ‘immunisation’, a com- bination of the above key words, and ‘others’. The ‘others’ category included articles identified by key METHODS words such as ‘elderly’ and ‘influenza’, ‘Tamiflu’ and on October 2, 2021 by guest. Protected copyright. Setting and study period ‘influenza’, the brand name of a drug and ‘influenza’, fl We studied the number of newspaper articles on in u- and ‘new type’ and ‘influenza’. During the study period, fl enza and the incidence of in uenza in Fukuoka, located more than 70% of newspaper articles included the in southwest Tokyo. The population of Fukuoka, which search term ‘influenza’.