Assessing patterns of global interest in hypertension using internet search engine data

Abstract

Background and objectives:Multiple studies have shown that ambulatory and office blood pressures (BP) have seasonal variations with a winter peak. A previous analysis demonstrated search volumes on Google to be predictive of influenza outbreaks that was on a par with surveillance data collected by the CDC. We sought to ascertain if search volumes on HTN show significant seasonal trends and whether this trend remained consistent across the Northern and Southern hemispheres. Methods: We searched Google Trends for the search terms ‘‘Hypertension’’ and ‘‘High Blood Pressure’’. Weekly normalized search volumes (adjusted search volume index reported by Google, ranging from 0-100 to standardize volumes across geographic locations) were obtained for the United States and Australia for the duration January 2007-November 2013. We compared highest with lowest monthly search volumes in each country. Locally weighted regression of time series graphs was used to visually assess annual periodicity in the data. We fit a cosinor model to assess the statistical significance of seasonal variation and to calculate phase difference between the Northern and Southern hemispheres. Results: The mean search volume in the US was 57.11 6.97, highest in February (64.18 6.16) and lowest in July (51.77 5.00) (p<0.001). The mean search volume in Australia was 42.72 7.63, highest in August (48.42 6.97) and lowest in December (33.77 5.34) (p<0.001). Search volumes in winter were higher than summer in the US (59.03 7.30 vs. 53.54 5.42, p<0.001) and Australia (46.93 6.35 vs. 37.18 6.29, p<0.001). Visual examination of locally weighted, smoothed scatterplot showed marked periodicity in search volumes that appeared near opposite for the US and Australia (Fig). Cosinor analysis showed statistically significant seasonal variation in US and Australian search volumes (p<0.05)