Spanish Influenza in Spain: Re-examining strength and timing differences across space Laura Cilek, Gerardo Chowell, Beatriz Echeverri Davila, Diego Ramiro Fari˜ nas * 1 Background Considerable research has examined how spatial differences and temporal patterns affect flu mortality. While climatic differences between tropical and northern countries play a role in seasonal influenza activity, the extreme world-wide mortality of the 1918 Spanish flu pandemic indicates these climactic patterns did not play a major role in regional death outcome [1, 2]. Yet even within countries, the presence and severity of each successive epidemic wave, beginning in spring 1918, differed by location, creating a debate regarding transmission mechanisms and the role acquired immunity in consecutive breakouts may have played in the tempering of each successive wave [3, 4, 5]. For example, a virus will spread more slowly in a population with some immunity (i.e. the reproduction number R effective in a partially-immune population will be lower than R 0 ), 1 assuming the virus has not yet mutated and evolved to be significantly different from the initial strain [7]. Additionally, the total length of pandemic wave is important, as it describes the heightened risk of others to contract the virus, allowing it to continue to spread. A comprehensive background on the Spanish Influenza virus in Spain qualitatively describes the four unique, but not universally experienced waves of influenza in Spain beginning with a herald wave in May 1918 and ending with a large echo wave in the winter of 1919-1920 [8, 9]. Attempts to quantify the spatial- temporal influenza patterns throughout Spain found significant variations in excess mortality among 49 provinces of Spain, finding cumulative (across all waves) excess rates as high as 212.2 (per 10,000) in Zamora and 6.2 in the Canary Islands [10]. Slightly over half of the provinces experienced fall and or winter waves, but with the exception of Madrid, this mortality paled in comparison to the dramatic fall wave of 1918. In older and more qualitative research the echo wave in 1919-1920 is well documented, but the impact of this outbreak in contemporary literature is muted [11, 8, 12]. Moreover, the data used in this prior work has several issues limiting the overall interpretation of the results. Not only does the “Bolet´ ın mensual de estad´ ıstica demogr´ afica sanitaria” data end in 1919, making the study of the aforementioned echo wave impossible, the death counts are presented as monthly statistics, which make it difficult to pinpoint the onset timing of each wave. The quality of the data in these monthly bulletins also varies by each Spanish province, as the data was aggregated based on reported statistics from each town in the province and do not have universal coverage based on various response rates [13]. 2 Prior research focusing on entire cities examines transmission mechanisms and the role acquired im- munity in consecutive breakouts may have played in the tempering of each successive wave of Spanish Influenza [3, 4, 5]. Other recent research explores the correlation between district level demographic and * laura.cilek@cchs.csic.es Centro de Ciencias Humanas y Sociales (CCHS) Consejo Superior de Investigaciones Cientificas (CSIC) School of Public Health, Division of International Epidemiology and Population Studies, Georgia State University Division of International Epidemiology and Population Studies, Fogarty International Center, National Institutes of Health Grupo de Estudios de Poblaci` on y Sociedad, Universidad Complutense de Madrid 1 The Reproduction Number (R 0 ), can be interpreted as the number of additional cases a single case will cause [6]. In an epidemic period, the Reproduction number will be greater than one, meaning that the number of infected individuals increases as time progresses. 2 For example, in the province of Cuenca, the median percentage of reported statistics (based on a weighted average of the total population in the province) is only 38 percent (low of 21% in July/September and a high of 53% in August), but in the province of Toledo, generally, there is 89% coverage of statistics. 1