key: cord-102377-n57hoty4
authors: Egli, Adrian; Goldman, Nina; Müller, Nicola F.; Brunner, Myrta; Wüthrich, Daniel; Tschudin-Sutter, Sarah; Hodcroft, Emma; Neher, Richard; Saalfrank, Claudia; Hadfield, James; Bedford, Trevor; Syedbasha, Mohammedyaseen; Vogel, Thomas; Augustin, Noémie; Bauer, Jan; Sailer, Nadine; Amar-Sliwa, Nadezhda; Lang, Daniela; Seth-Smith, Helena M.B.; Blaich, Annette; Hollenstein, Yvonne; Dubuis, Olivier; Nägele, Michael; Buser, Andreas; Nickel, Christian H.; Ritz, Nicole; Zeller, Andreas; Stadler, Tanja; Battegay, Manuel; Schneider-Sliwa, Rita
title: High-resolution influenza mapping of a city reveals socioeconomic determinants of transmission within and between urban quarters
date: 2020-04-04
journal: bioRxiv
DOI: 10.1101/2020.04.03.023135
sha: 
doc_id: 102377
cord_uid: n57hoty4

With two-thirds of the global population projected to be living in urban areas by 2050, understanding the transmission patterns of viral pathogens within cities is crucial for effective prevention strategies. Here, in unprecedented spatial resolution, we analysed the socioeconomic determinants of influenza transmission in a European city. We combined geographical and epidemiological data with whole genome sequencing of influenza viruses at the scale of urban quarters and statistical blocks, the smallest geographic subdivisions within a city. We observed annually re-occurring geographic clusters of influenza incidences, mainly associated with net income, and independent of population density and living space. Vaccination against influenza was also mainly associated with household income and was linked to the likelihood of influenza-like illness within an urban quarter. Transmissions patterns within and between quarters were complex. High-resolution city-level epidemiological studies combined with social science surveys such as this will be essential for understanding seasonal and pandemic transmission chains and delivering tailored public health information and vaccination programs at the municipal level.

6 as a surrogate marker for the population density and available living space in a particular area 122 and thereby also be an indicator of influenza burden ( Figure 1D ).

To account for this potential ecologic fallacy, we corrected the influenza incidence rates per 1,000 125 inhabitants for each statistical block. We still observed similar dense influenza case patterns at 8 higher the socioeconomic score, the higher the vaccine rate. Income showed the highest 189 correlation with vaccination rates, followed by living space and population density, respectively 190 (R 2 0.622, p=0.0067; R 2 0.618, p=0.007; R 2 0.532, p=0.0167). Self-reported vaccination rates 191 may serve as a surrogate marker for herd immunity 27, 28 

However, in years with a low vaccine effectiveness (Table S4 ) this association may be weaker.

In order to monitor antibody titres over time in a healthy population across the city, we recruited Table S1 ). Before the 2016/2017 influenza season, we observed 211 that across all urban quarters a median of 21% (IQR 17-28.5%) had seroprotective antibody levels

(defined as hemagglutination inhibition titres equal or more than 1:40 31 ) ( Figure 3A) . Again, urban 213 quarters with lower socioeconomic scores also showed low seroprotection rates (e.g. Matthaeus,

Breite, Kleinhueningen and Klybeck) ( Figure S6A ). Urban quarters with higher socioeconomic 215 scores showed a median seroprotection rate of 26.1%, whereas those with lower socioeconomic 216 scores showed a median seroprotection rate of 14.6% (p=0.05). Blood donors with influenza 217 vaccination showed significant higher H3N2 specific HI titers in comparison to people who were 218 not vaccinated (p<0.0001; Figure S6B ). Similar to the survey, in this cohort net income was Table S5 ). This visualization can be 

(WE) were more identical in comparison to other urban quarters (p<0.005, Figure 3C ).

Transmission events within the same urban quarter were explored. Interestingly, two urban 284 quarters -Gundeldingen (GU) and Vorstaedte (VO) -showed influenza isolates that were 285 significantly more related to other isolates from within the same urban quarter than to isolates 286 from other quarters or outside of Basel (p<0.001, Figure 3C ). These two urban quarters show a 287 low socioeconomic score and lower pre-seasonal seroprotection rate. Phylogenetic cluster size 288 did not correlate with any socioeconomic factors (p=n.s.).

Each year influenza infects millions of people around the globe 36, 37 

This would allow us to address more and account for a greater variety of population segments 347 and help to identify potential drivers of transmission.

The overall study design has been previously published 20 . Briefly, the study had retrospective 

To account for socioeconomic differences related to each urban quarter and potentially 100. Sites with a coverage of less than 100 were assumed to be unknown and were denoted as 436 N, that is any possible nucleotide. Sequences that showed a read depth of 100 for at least 80% 437 of the positions in at least four segments were used for the analysis. Non double infections with 438 two strains were noted. Using these parameters, we continued our analysis using 663 samples.

The consensus sequences from these strains were deposited in GenBank (numbers will be 440 available upon acceptance of the manuscript). We aligned the consensus sequences using 

We also acknowledge the contributing colleagues and centers to GISAID (see Data availability statement: All sequencing data (raw reads) will be made available at NCBI. As 536 well tables with anonymized PCR-confirmed cases and anonymized survey results will be made 537 available in a public data repository.

Code availability statement: All codes used to process the viral genomic data will be made 540 available on github.

Tables 542 543 

Patients with PCR-confirmed influenza are compared against patients with negative PCR testing. 

We next reconstructed the phylogenetic trees of all initial clusters by using the full genomes of all 751 samples from the initial clusters. We fixed the evolutionary rates to be equal to the mean 752 evolutionary rates as estimated using the methods above. As a population model, we used a 753 constant coalescent model with an estimated effective population size that was shared among all 754 initial clusters. We then estimated a distribution of phylogenies for each initial cluster, assuming 755 that all segments share the same phylogeny. As estimated in the previous analysis, reassortment

will not bias evolutionary rates.

Local cluster identification. To identify sets of sequences from Basel that were likely to have been 759 transmitted locally, we reconstructed the geographic origins of lineages that were introduced into 760

Basel. Therefore, we used the phylogenetic tree distributions for each initial cluster to reconstruct 761 the ancestral states using parsimony. We made some modifications to the standard algorithm for 762 parsimony ancestral state reconstruction to reflect our prior assumption, that Basel is unlikely to 

Continental synchronicity of human influenza virus epidemics