Spanish Influenza in Madrid: Contextualizing Timing and Strength Differences in Spatial and Social Variations
Diego Ramiro-Fariñas, Spanish National Research Counsel (CSIC)
Laura Cilek, Centro de Ciencias Humanas y Sociales - CSIC
Beatriz Echeverri, Grupo de Estudios de Poblacio`n y Sociedad, Universidad Complutense de Madrid
SpanishInfluenza in Madrid: Contextualizing wave differencesthrough spatial and social variations
Echeverri presents a comprehensivebackground on the Spanish Influenza virus in Madrid, whichexperienced four distinct waves, beginning in May of 1918. Likecities such as New York and Copenhagen, the spring wave in Madrid isclearly defined[2,3], probably explaining moderate mortality insubsequent fall and winter waves. An additional wave of influenza inDecember and January 1919-20 produced the highest mortality peak ofthe period.
Rapidly urbanizing, theneighborhoods of Madrid and their inhabitants varied considerably bydemographic and socioeconomic indicators[4,5], perhaps playing asignificant role in the spread of influenza. In this paper, wefocus our analysis on two of the city’s districts, Inclusa andCentro. Located in the South of the city, Inclusa had particularlyhigh mortality. The population of the district was generally poorerthan the rest of the city, and it was also home to a large migrantpopulation and the city’s Foundling Hospital. Centro had muchlower mortality rates than Inclusa and a more permanent population,despite the ongoing development in the city.
Prior research focusing on entirecities examines transmission mechanisms and the role acquiredimmunity in consecutive breakouts may have played in the tempering ofeach successive wave of Spanish Influenza[7,3,8]. Some efforts havebeen made to look at compositional effects of local districts orcensus tracts on overall pandemic mortality, but far less attentionhas been paid to the interaction of individual and neighborhoodcharacteristics with regard to mortality over several successiveinfluenza waves.
Clear evidence points toward asocial gradient in the transmission and strength of seasonalinfluenza outbreaks and preventative vaccination campaigns in the USand across the world, especially in elderly and minoritypopulations[9,10]. However, many social scientists believe the viralstrain present in the 1918-920 pandemic events was so virulent thataside from affecting all age groups, the airborne nature of thedisease outweighed the potential of any other social variables tocreate class mortality differentials [11,12,13]. Mamelund providesexamples of these studies, including a 1920 Great Britain Ministry ofHealth survey about fatality and social status, in his ownpaper[14,12].
Yet other recent research hasfound a social gradient. These analyses also consider how geographicvariation, including underlying differences in neighborhoodsocioeconomic environment, affected mortality. For example, Grantzet al find that during the strong fall wave in Chicago (September toNovember 1918), influenza and related mortality was greater in censustracts with higher illiteracy. Mamelund used individual andhousehold level data to find that both neighborhood of residence andapartment size, as a proxy for household wealth, had effects oninfluenza survival in Kristiana during 1918.
Given the lack of significantgeo-spatial change within Madrid, influenza transmission shouldaffect each district equally, yet current understanding of socialinequalities suggests mortality differences between the two districtsdoes exist[15,9,16].In this analysis, we aim to disentangle how, beyond the demographicmakeup of a district, social and economic differences played a rolein influenza-related mortality and if this changed over time.
To establish wave and timingstrength, we use 103,500 detailed Madrid Civil Register death recordsbetween 1917-1922 (Inclusa=12,452 and Centro=5,286). We also useindividual level demographic and social data for each household inMadrid as at December 31,1920 for the districts of Inclusa (n=68,707)and Centro (n=47,448).Additional information for Madrid comes from yearly population books,which provide a summary of population and demographic events in eachpart of the city.
After fitting a mortality baselineusing a modified Serfling regression model[18,19], we mathematicallyestimated unique parameters of each wave and district in Madrid usingSegmented Regression[20,21,22]. Given the nature of oursingle-point-in-time population data,we face small limitations to thetypes of employable analyses. We plan to use either multi-levelnegative binomial regression to estimate predicted mortality for theneighborhoods in Inclusa and Centro oruse logistic regressionexamining the likelihood of a given individual to die from influenza.
In the fall of 1918, the city-wideascending phase lasts much longer than the strong wave from May-June,though R0is lower and does not reach epidemic levels in alldistricts. The weakness of the winter wave is evident at the districtlevel, where most districts (including Inclusa and Centro) do notreach epidemic levels. While the fourth wave of influenza isgenerally strong across the city, the R issuspiciously low in Centro.
Presented in Session 1232: Posters