Mass media informatics and disease surveillance Page 3 of 9 retaining subcultural differences. However, local interests often cross national boundaries, and the focus of this study is on a disease epidemic ravaging a country but with the potential to globalise in our increasingly interconnected world. caso confirmado de Zika’ (‘Rio has the first confirmed case of Zika’) shows health officials still regarded Zika as benign and advised clinicians to treat as dengue when in doubt. Reference to both diseases did not occur again until November 2015 with Zika later reaching its peak in the second week of February 2016 when WHO declared a global emergency. The Kendall’s tau_b correlation coefficient from 13 December 2015 to 31 December 2016 shows a significant association between Zika and dengue (r=0.70; p<0.001) but not between Zika and chikungunya (r=0.54; p<0.08). Data for this study were obtained from O Estado de São Paulo (O Estado) and O Globo, two leading Brazilian newspapers; The Times of London and The Telegraph published in the UK and the US-based USA Today and New York Times. Coverage data were downloaded from the NEXIS database at various times between March 2017 and February 2018. Google Trends was used for online searches in Brazil, UK, USA and worldwide to compare with newspaper trends. Figure 1b shows that dengue and chikungunya had been in the media discourse in Brazil, in O Estado, from 1 January 2015 when the sampling commenced, with coverage of dengue rising rapidly thereafter with a major peak in the first week of May 2015. This peak shows an unusual pattern – an indication of rising public anxiety – which should raise an alarm for further clinical investigation. May 2015 was also the month Zika appeared in O Globo (Figure 1a), indicating it was flagged by one local newspaper and not the other, also showing the need for multiple media sources. The experience of doctors was with dengue and Zika was thought, at this time, to be benign. This may explain the comment by a Brazilian official reported in the New York Times of 31 December 2015 that ‘Zika virus doesn’t worry us’ because it is a ‘benign disease’. The appearance of Zika peaked at about the same time in the second week of February 2016 in both Brazilian newspapers. There is a significant association between Zika and dengue (r=0.69; p<0.001) and between Zika and chikungunya (r=0.68; p<0.001) between 13 December 2015 and 31 December 2016. The US newspaper was chosen because of the presence of the mosquito carrier in some parts of the USA and the high risk of spread compared with the UK where the carriers are absent. The UK and US newspapers also provide indicators of when the media detected the events to report and the attention on two important events in Brazil (the Rio Carnival and the Olympics) that may have affected foreign visitors from these countries. The Brazilian newspapers were chosen to identify trends and aberrations to media attention in the local press within the review period. For O Globo and O Estado, the NEXIS search was conducted using the keywords dengue or Zika or chikungunya between 1 January 2015 and 31 December 2016. Using these keywords, 503 articles for O Globo and 813 for O Estado were downloaded. For the UK and US newspapers, the keyword used was ‘Zika’. There were no search results returned for ‘Zika’ between 1 January 2015 and 31 October 2015. Blogs, duplicates and articles with high levels of similarity were discarded. In total, 246 articles from The Times of London and 414 from the New York Times were used for the trend analysis. Google Trends21 was used for online searches. For Google Brazil, the keywords used for the search were ‘Zika’, ‘Dengue’ and ‘Chikungunya’ to map the co-occurrence of searches about the three diseases from 1 January 2015 to 31 December 2016. The same dates were used for ‘Zika’ on Google worldwide, Google UK and Google USA. The NEXIS output is in the daily format of newspapers and downloads both headlines and text. The daily format was re-coded to weekly to match Google weekly outputs. Figure 1c compares Google searches for dengue, Zika and chikungunya. The Google search number represents an index of popularity. It is used here as a variable indicating an index of appearances. The graph shows that those seeking online information had a much higher and sustained interest in dengue in the review period, except in early December 2015 when it was surpassed by Zika. The December surge coincides with similar surges in O Estado and O Globo. Searches for Zika were noticeable from the beginning of the review period in very low numbers, mostly one, with the highest peak recorded in the second week of February 2016, as with the newspapers. For Google and newspaper searches between 13 December 2015 and 31 December 2016, the correlation was significant for Zika in Google Brazil and O Estado (r=0.65; p<0.001) and O Globo (r=0.59; p<0.001). The correlation was also significant for dengue in Google Brazil and O Estado (r=0.69; p<0.001) and O Globo (r=0.53; p=0.001). The QDA/Miner WordStat program, which runs in both English and Portuguese, was used to analyse the data. Limitations of research Secondary data were used and are limited in time to 2 years and to selected newspapers in Brazil, UK and USA. The study of the Brazilian media was limited to detecting aberrations and trends and comparing with Google searches. Other authors24 have examined the framing of the debate in the Brazilian media. It would be interesting to see how a wider longitudinal study of newspapers worldwide could be used to trace the movement of the disease in the Americas and when the debates about its presence emerged in other Latin American newspapers. This would provide more data to support growing research on mass media as surrogate data for epidemiological studies. Coverage by both newspapers was compared over the period. The correlation for Zika in O Globo and O Estado was significant (r=0.54; p<0.001) as was the correlation for dengue (r=0.61; p<0.001). Research question 1.2 This question addressed the levels of attention in the UK and US media. How did attention levels compare with Google Trends UK, USA and worldwide? For the UK, a comparison was made between Google UK, Google worldwide and The Times of London. Figure 2a shows that coverage in Google UK and The Times of London peaked at the same time, the third week of January 2016, a week before Google worldwide, and remained high until March 2016. The Rio Carnival was held in February 2016 and there is a noticeable peak in coverage in Google UK at this time. Coverage in the Times of London peaked again in the last week of July, just before the commencement of the Rio Olympics held from 5 August to 21 August 2016. The July to August surge in coverage in The Times of London, higher than the February peak, was not replicated in Brazilian newspapers42, indicating differences in the perception of risk and anxiety. There was a significant association for the keyword Zika between The Times of London and Google UK (r=0.37; p<0.001) and between The Times of London and Google worldwide (r=0.30; p<0.008). Findings Research question 1.1 This question addressed how early the Zika virus discourse appeared in the Brazilian media. How did the trend compare with similar diseases such as dengue and chikungunya? What were the levels of attention that indicate a rise in public anxiety and an alarm for further clinical analysis? How did these indicators compare with Google Trends Brazil? Figure 1 shows the co-occurrence of Zika and dengue, both with similar symptoms, in May 2015 in O Globo. May 2015 is the earliest appearance of the word Zika in the two Brazilian newspapers; the articles indicate the level of unfamiliarity with the disease at that time. The first report, on 14 May 2015, titled ‘O novo vírus do ‘aedes’ (‘The new ‘aedes’ virus’) said (translated) ‘It [Zika] seems less serious, but it is possible to acquire the three viruses at the same time, and we do not know the clinical course of this co-infection with the Zika virus, for which we do not have immunity’. A 3 June 2015 follow-up article titled ‘Rio tem o primeiro Research Article https://doi.org/10.17159/sajs.2019/5290 The New York Times data show a similar trend with those of The Times of London. The first peak in coverage in the New York Times was in late January 2016, coinciding with that in Google worldwide but a week after coverage in Google USA peaked. Coverage in the New York Times reached another height 2 weeks later, followed by several highs and lows mimicking the UK data in the run up to the Olympics with a noticeable 77 Volume 115| Number 3/4 March/April 2019