Spatially explicit modeling of schistosomiasis transmission dynamics: applications to Senegal Lorenzo Mari Dipartimento di Elettronica, Informazione e Bioingegneria Politecnico di Milano mailto: lorenzo.mari@polimi.it web: home.deib.polimi.it/mari
Outline - introducing schistosomiasis fact sheet spatial distribution transmission cycle schisto in Senegal 500mm - a country-wide model for schisto transmission modeling approach data for model calibration the role of human mobility implications for disease control - a regional-scale model for schisto transmission data and modeling approach the role of biological dispersal - the MASTR-SLS project Schistosomiasis prevention posters, US War Department 1945 (US National Library of Medicine)
Introducing schistosomiasis Schistosomiasis is a disease caused by parasitic worms of the genus Schistosoma. Depending on their species, parasites may colonize the urinary tract or the intestines. Worldwide, schistosomiasis is the most 500mm deadly NTD, and it is second only to malaria as the parasitic disease with the greatest economic impact. Schistosoma mansoni (Science Photo Library) Signs and symptoms abdominal pain, diarrhea, blood in stool or urine Long-term effects liver damage, kidney failure, infertility, bladder cancer + poor growth and learning difficulty in children Schistosomiasis in a child (flickr.com)
Spatial distribution of schistosomiasis - 700+ M people in 70+ countries live in endemic areas - 200+ M people affected worldwide (90% cases in sub-saharan Africa) - 12 200 k deaths yearly 500mm Colley et al. Lancet 2014
Schistosomiasis transmission cycle (cdc.gov) and the role of movement - people can be exposed to infested water and/or contaminate water (if infected) while traveling - miracidia and cercariae can be transported by the water flow - snails can move along rivers and canals
Schistosomiasis in Senegal Schistosomiasis is widespread in Senegal, especially in its urinary form. It still represents a major health problem in the country, where it is the third disease (after malaria and lymphatic filariasis) in terms of years lived with disability (according to WHO data). Prevalence of urinary schistosomiasis [% infected people within health districts] In the past 20 years, infection prevalence in Senegal has increased from approximately 11% (as estimated from a 1996 national survey) to about 21% (as estimated from current epidemiological data available at the Senegalese Ministry of Health).
Modeling country-wide transmission stratified parasite burden in humans + mobility The human population is subdivided into local communities following e.g. mh H administrative boundaries Force of infection F can account for mobility-driven exposure to cercariae H 0 Within each community, the resident population is divided into infection classes characterized by increasing parasite burden F(C) g1 mh H 1 F(C) g2 mh + ah1 F(C) gp 1 H P 1 F(C) gp mh + ahp 1 HP mh + ahp Gurarie et al. Parasitology 2010
Modeling country-wide transmission stratified parasite burden in humans + mobility ms N SI-like model for snails S ms M I ms + as
Modeling country-wide transmission stratified parasite burden in humans + mobility SI-like model for snails balance equations for parasite larval stages + human mobility Human contamination rate G can account for non-local shedding of eggs/miracidia
Data for model set-up High-resolution population density map [inhabitants km-2 ] (data from WorldPop project, worldpop.org.uk) population abundance in each arrondissement (third-level administrative units) Rurality index [% people living in rural conditions] (data from Global Atlas of Helminth Infections, thiswormyworld.org) + Rivers of Senegal (data from diva-gis.org/gdata) local exposure and contamination rates (exposure risk includes snail abundance)
Human mobility Human mobility patterns are estimated from Call Detail Records (CDRs), i.e. anonymous mobile-phone users' communication metadata. CDRs from Sonatel (the largest Senegalese telco provider, with a user base of ~9 M people, > 60% of the Senegalese population) subscribers were made available by Orange for the year 2013 in the aftermath of the 2014 D4D-Senegal challenge (d4d.orange.com). 1;781584154;2013-10-01 2;782080423;2013-10-01 1;782517857;2013-10-01 1;782554884;2013-10-01 23:59:59;11;608-01-00125-01206;1;774350871;35726705;52 23:59:59;11;608-01-00141-11891;2;34635163404;35860304;45 23:59:59;22;608-01-00124-05145;1;777189407;35168905;1 23:59:59;22;608-01-00152-01388;1;776099804;35714505;1 anonymous ID time stamp antenna ID > 15 G records
Human mobility - exposure/contamination risk ~ time spent in each arrondissement - time spent in j by people living in i ~ number of phone calls made by people from i while being in j - users' home ~ arrondissement where the most night calls (7pm to 7am) are made from mobility matrix Q = [Qij] with Qij = From arrondissement (i) Human mobility matrix [-] 1 # calls made by users living in i while being in j # calls made by users living in i Human mobility fluxes [# people day-1] 104 10-2 103 10-4 102 To arrondissement (j)
Model calibration For the sake of robustness, calibration has been performed against regional prevalence data. However, the model is run at the arrondissement scale, therefore it can provide predictions at a relatively fine spatial scale. Coefficient of determination R2 = 0.76 Mean absolute data-model deviation = 6.0% Prevalence of urinary schistosomiasis [% infected people within arrondissements]
The role of human mobility Artificial manipulations of the mobility matrix show that movement may have a predominantly protective role at large (regional) spatial scales. Also, disease prevalence is predicted to attain a well-defined minimum for intermediate levels of mobility remarkably, close to the actual estimate of mobility obtained from CDRs.
The model as a decision support tool WASH: WAter, Sanitation and Hygiene decreasing the share of people living in rural conditions IEC: Information, Education and Communication decreasing the baseline exposure and contamination rates
A regional transmission model At a smaller spatial scale, it is possible to build a more accurate schistosomiasis transmission model including some details about: - snail ecological and epidemiological dynamics density-dependent demography SEI-like transmission - spatial coupling mechanisms other than human mobility hydrological transport of parasite larvae snail dispersal along rivers and canals
A regional transmission model Saint-Louis Epidemiological data courtesy of the Upstream Alliance (theupstreamalliance.org)
A regional transmission model At a smaller spatial scale, it is possible to build a more accurate schistosomiasis transmission model including some details about: - snail ecological and epidemiological dynamics density-dependent demography SEI-like transmission - spatial coupling mechanisms other than human mobility hydrological transport of parasite larvae snail dispersal along rivers and canals In particular, the model can consider two layers of spatial connectivity (multidimensional network model): - human-to-water contact patterns CDR analysis (village-to-antenna mobility matrix) proximity analysis (antenna-to-water adjacency matrix) - water-mediated dispersal (biased) random walk along the hydrological network Boccaletti et al. Physics Reports 2014
A regional transmission model macroparasitic model for human infection + mobility Force of infection F accounts for mobility-driven exposure to cercariae Parasite load in humans ~ negative binomial distribution
A regional transmission model macroparasitic model for human infection + mobility SEI-like model for snails + dispersal L(S,E,I) S ms M E d ms + h I ms + h
A regional transmission model macroparasitic model for human infection + mobility SEI-like model for snails + dispersal balance equations for parasite larval stages + human mobility + larval transport Human contamination rate G accounts for non-local shedding of eggs/miracidia
Model simulation model data lakes rivers canals
The role of biological dispersal The model shows that endemic disease transmission is strongly affected by hydrological transport of parasite larvae. In fact, transmission may not be maintained if larval transport is too low. Conversely, average infection prevalence seems to be less strongly affected by snail dispersal, with relatively higher infection intensities being found for low and/or predominantly upstream snail dispersal.
Wrapping up A quantitative description of large-scale schistosomiasis transmission dynamics requires a modeling approach that is - spatially explicit, because of the importance of spatial coupling mechanisms in the definition of epidemiological patterns - integrative, because of the need of tapping different data sources and blending approaches from different disciplines Models with these characteristics seem to be able to reproduce the observed spatial patterns of schistosomiasis infection intensity, and can thus be used to - better understand disease transmission mechanisms - draw meaningful epidemiological predictions - support informed decision making
The way forward We will be working on schistosomiasis transmission in the Saint-Louis region of Senegal for the next two years within the project MASTR-SLS MApping Schistosomiasis Trasmission Risk in Saint-Louis, Senegal Principal Investigator: Renato Casagrandi, DEIB Project Manager: Lorenzo Mari, DEIB Project Partners: Chiara Francalanci, DEIB; Maresa Bertolo, DESIGN funded through the 2015-16 PoliSocial Award competition (polisocial.polimi.it). Aim of the project is to mitigate the spread of schistosomiasis in this endemic region through the development of reliable risk maps based on in situ data and state-of-the art modeling techniques. Risk maps will - guide people in avoiding water sites where infection risk is highest - provide decision-makers with operational tools to take actions against disease transmission in the most-at-risk areas
The MASTR-SLS project Work Package Expected result WP1. Data collection R1. Mutual capacity building R2. ICT for data collection WP2. Data- and model- R3. Maps of transmission risk guided mapping R4. Analysis of intervention strategies WP3. Communication R5. Decision support R6. Information campaign Are you interested in - collecting/analyzing in situ data? - developing apps for data collection? - `playing' with spatially explicit models? - developing decision-support tools? - planning communication strategies? Drop us a line, we may have work for you... Lorenzo Mari lorenzo.mari@polimi.it Renato Casagrandi renato.casagrandi@polimi.it
Thanks to Renato Casagrandi Manuela Ciddio Marino Gatto Giulio De Leo Sanna Sokolow References Ciddio, Mari et al (2016). The spatial spread of schistosomiasis: a multidimensional network model applied to Saint-Louis region, Senegal. Advances in Water Resources, in press. Mari et al. (2017). Big-data-driven modeling unveils country-wide drivers of endemic schistosomiasis. Submitted