Featured Publications


Filter options:

Hannah C. Slater, Amanda Ross, Ingrid Felger, Natalie E. Hofmann, Leanne Robinson, Jackie Cook, Bronner P. Gonçalves, Anders Björkman, André Lin Ouédraogo, Ulrika Morris, Mwinyi Msellem, Cristian Koepfli, Ivo Mueller, Fitsum Tadesse, Endalamaw Gadisa, Smita Das, Gonzalo Domingo, Melissa Kapulu, Janet Midega, Seth Owusu-Agyei, Cécile Nabet, Renaud Piarroux, Ogobara Doumbo, Safiatou Niare Doumbo, Kwadwo Koram, Naomi Lucchi, Venkatachalam Udhayakumar, Jacklin Mosha, Alfred Tiono, Daniel Chandramohan, Roly Gosling, Felista Mwingira, Robert Sauerwein, Eleanor M Riley, Nicholas J White, Francois Nosten, Mallika Imwong, Teun Bousema, Chris Drakeley, Lucy C Okell


Malaria infections occurring below the limit of detection of standard diagnostics are common in all endemic settings. However, key questions remain surrounding their contribution to sustaining transmission and whether they need to be detected and targeted to achieve malaria elimination. In this study we analyse a range of malaria datasets to quantify the density, detectability, course of infection and infectiousness of subpatent infections. Asymptomatically infected individuals have lower parasite densities on average in low transmission settings compared to individuals in higher transmission settings. In cohort studies, subpatent infections are found to be predictive of future periods of patent infection and in membrane feeding studies, individuals infected with subpatent asexual parasite densities are found to be approximately a third as infectious to mosquitoes as individuals with patent (asexual parasite) infection. These results indicate that subpatent infections contribute to the infectious reservoir, may be long lasting, and require more sensitive diagnostics to detect them in lower transmission settings.

Wedlock PT, Mitgang EA, Oron AP, Hagedorn BL, Leonard J, Brown ST, Bakal J, Siegmund SS, Lee BY



The lack of specific policies on how many children must be present at a vaccinating location before a healthcare worker can open a measles-containing vaccine (MCV) - i.e. the vial-opening threshold - has led to inconsistent practices, which can have wide-ranging systems effects.


Using HERMES-generated simulation models of the routine immunization supply chains of Benin, Mozambique and Niger, we evaluated the impact of different vial-opening thresholds (none, 30% of doses must be used, 60%) and MCV presentations (10-dose, 5-dose) on each supply chain. We linked these outputs to a clinical- and economic-outcomes model which translated the change in vaccine availability to associated infections, medical costs, and DALYs. We calculated the economic impact of each policy from the health system perspective.


The vial-opening threshold that maximizes vaccine availability while minimizing costs varies between individual countries. In Benin (median session size = 5), implementing a 30% vial-opening threshold and tailoring distribution of 10-dose and 5-dose MCVs to clinics based on session size is the most cost-effective policy, preventing 671 DALYs ($471/DALY averted) compared to baseline (no threshold, 10-dose MCVs). In Niger (median MCV session size = 9), setting a 60% vial-opening threshold and tailoring MCV presentations is the most cost-effective policy, preventing 2897 DALYs ($16.05/ DALY averted). In Mozambique (median session size = 3), setting a 30% vial-opening threshold using 10-dose MCVs is the only beneficial policy compared to baseline, preventing 3081 DALYs ($85.98/DALY averted). Across all three countries, however, a 30% vial-opening threshold using 10-dose MCVs everywhere is the only MCV threshold that consistently benefits each system compared to baseline.


While the ideal vial-opening threshold policy for MCV varies by supply chain, implementing a 30% vial-opening threshold for 10-dose MCVs benefits each system by improving overall vaccine availability and reducing associated medical costs and DALYs compared to no threshold.

Zhe Bai, Eurika Kaiser, Joshua L. Proctor, J. Nathan Kutz, and Steven L. Brunton


Dynamic mode decomposition has emerged as a leading technique to identify spatiotemporal coherent structures from high-dimensional data, benefiting from a strong connection to nonlinear dynamical systems via the Koopman operator. In this work, two recent innovations that extend dynamic mode decomposition to systems with actuation and systems with heavily subsampled measurements are integrated and unified. When combined, these methods yield a novel framework for compressive system identification. It is possible to identify a low-order model from limited input–output data and reconstruct the associated full-state dynamic modes with compressed sensing, adding interpretability to the state of the reduced-order model. Moreover, when full-state data are available, it is possible to dramatically accelerate downstream computations by first compressing the data. This unified framework is demonstrated on two model systems, investigating the effects of sensor noise, different types of measurements (e.g., point sensors, Gaussian random projections, etc.), compression ratios, and different choices of actuation (e.g., localized, broadband, etc.). In the first example, this architecture is explored on a test system with known low-rank dynamics and an artificially inflated state dimension. The second example consists of a real-world engineering application given by the fluid flow past a pitching airfoil at low Reynolds number. This example provides a challenging and realistic test case for the proposed method, and results demonstrate that the dominant coherent structures are well characterized despite actuation and heavily subsampled data.

N. M. Mangan , T. Askham , S. L. Brunton , J. N. Kutz and Joshua L. Proctor


Hybrid systems are traditionally difficult to identify and analyse using classical dynamical systems theory. Moreover, recently developed model identification methodologies largely focus on identifying a single set of governing equations solely from measurement data. In this article, we develop a new methodology, Hybrid-Sparse Identification of Nonlinear Dynamics, which identifies separate nonlinear dynamical regimes, employs information theory to manage uncertainty and characterizes switching behaviour. Specifically, we use the nonlinear geometry of data collected from a complex system to construct a set of coordinates based on measurement data and augmented variables. Clustering the data in these measurement-based coordinates enables the identification of nonlinear hybrid systems. This methodology broadly empowers nonlinear system identification without constraining the data locally in time and has direct connections to hybrid systems theory. We demonstrate the success of this method on numerical examples including a mass–spring hopping model and an infectious disease model. Characterizing complex systems that switch between dynamic behaviours is integral to overcoming modern challenges such as eradication of infectious diseases, the design of efficient legged robots and the protection of cyber infrastructures.

Travis C. Porco, Catherine E. Oldenburg, Ahmed M. Arzika, Khumbo Kalua, Zakayo Mrango, Catherine Cook, Elodie Lebas, Robin L. Bailey, Sheila K. West, Assaf P. Oron, Jeremy D. Keenan, Thomas M. Lietman and for the MORDOR Study Group


Mass azithromycin distribution has been shown to reduce all-cause mortality in preschool children in sub-Saharan Africa. However, substantial heterogeneity in the apparent effect has been noted across geographic settings, suggesting a greater relative benefit in higher mortality settings. Here, we evaluated the relationship between the underlying mortality rate and the efficacy of azithromycin for the prevention of child mortality using data from multiple sites in Ethiopia, Malawi, Niger, and Tanzania. Between regions, we find no strong evidence of effect modification of the efficacy of azithromycin distribution for the prevention of child mortality by the underlying mortality rate (P = 0.12), although a modest effect is consistent with our findings. Higher mortality settings could be prioritized, however, because of the larger number of deaths which could be averted with azithromycin distribution.


Assaf P. Oron, Roy Burstein, Laina D. Mercer, Ahmed M. Arzika, Khumbo Kalua, Zakayo Mrango, Sheila K. West, Robin L. Bailey, Travis C. Porco and Thomas M. Lietman


We examined whether baseline mortality risk, as a function of child age and site, modified the azithromycin mortality-reduction effect in the Macrolide Oraux pour Réduire les Décès avec un Oeil sur la Résistance (MORDOR) clinical trial. We used the Cox proportional hazards model with an interaction term. Three models were examined representing three sources for the baseline-risk covariate: two using sources external to MORDOR and the third leveraging data within MORDOR. All three models provided moderate evidence for the effect becoming stronger with increasing baseline mortality (P = 0.02, 0.02, and 0.07, respectively) at the rate of approximately 6–12% additional mortality reduction per doubling of baseline mortality. Etiological and programmatic implications of these findings are discussed.


Background: Pediatric diarrhea can be caused by a wide variety of pathogens, from bacteria to viruses to protozoa. Pathogen prevalence is often described as seasonal, peaking annually and associated with specific weather conditions. Although many studies have described the seasonality of diarrheal disease, these studies have occurred predominantly in temperate regions. In tropical and resource-constrained settings, where nearly all diarrhea-associated mortality occurs, the seasonality of many diarrheal pathogens has not been well characterized. As a retrospective study, we analyze the seasonal prevalence of diarrheal pathogens among children with moderate-to-severe diarrhea (MSD) over three years from the seven sites of the Global Enteric Multicenter Study (GEMS). Using data from this expansive study on diarrheal disease, we characterize the seasonality of different pathogens, their association with site-specific weather patterns, and consistency across study sites. Methodology/Principal Findings: Using traditional methodologies from signal processing, we found that certain pathogens peaked at the same time every year, but not at all sites. We also found associations between pathogen prevalence and weather or "seasons", which are defined by applying modern machine-learning methodologies to site-specific weather data. In general, rotavirus was most prevalent during the drier "winter" months and out of phase with bacterial pathogens, which peaked during hotter and rainier times of year corresponding to "monsoon", "rainy", or "summer" seasons. Conclusions/Significance: Identifying the seasonally-dependent prevalence for diarrheal pathogens helps characterize the local epidemiology and inform the clinical diagnosis of symptomatic children. Our multi-site, multi-continent study indicates a complex epidemiology of pathogens that does not reveal an easy generalization that is consistent across all sites. Instead, our study indicates the necessity of local data to characterizing the epidemiology of diarrheal disease. Recognition of the local associations between weather conditions and pathogen prevalence suggests transmission pathways and could inform control strategies in these settings.

Steve J. Kroiss, Maiwand Ahmadzai, Jamal Ahmed, Muhammad Masroor Alam, Guillaume Chabot-Couture, Michael Famulare, Abdirahman Mahamud, Kevin A. McCarthy, Laina D. Mercer, Salman Muhammad, Rana M. Safdar, Salmaan Sharif, Shahzad Shaukat, Hemant Shukla, Hil Lyons


The polio environmental surveillance (ES) system has been an incredible tool for advancing polio eradication efforts because of its ability to highlight the spatial and temporal extent of poliovirus circulation. While ES often outperforms, or is more sensitive than AFP surveillance, the sensitivity of the ES system has not been well characterized. Fundamental uncertainty of ES site sensitivity makes it difficult to interpret results from ES, particularly negative results.


Ambitious global goals have been established to provide universal access to affordable modern contraceptive methods. The UN's sustainable development goal 3.7.1 proposes satisfying the demand for family planning (FP) services by increasing the proportion of women of reproductive age using modern methods. To measure progress toward such goals in populous countries like Nigeria, it's essential to characterize the current levels and trends of FP indicators such as unmet need and modern contraceptive prevalence rates (mCPR). Moreover, the substantial heterogeneity across Nigeria and scale of programmatic implementation requires a sub-national resolution of these FP indicators. However, significant challenges face estimating FP indicators sub-nationally in Nigeria. In this article, we develop a robust, data-driven model to utilize all available surveys to estimate the levels and trends of FP indicators in Nigerian states for all women and by age-parity demographic subgroups. We estimate that overall rates and trends of mCPR and unmet need have remained low in Nigeria: the average annual rate of change for mCPR by state is 0.5% (0.4%,0.6%) from 2012-2017. Unmet need by age-parity demographic groups varied significantly across Nigeria; parous women express much higher rates of unmet need than nulliparous women. Our hierarchical Bayesian model incorporates data from a diverse set of survey instruments, accounts for survey uncertainty, leverages spatio-temporal smoothing, and produces probabilistic estimates with uncertainty intervals. Our flexible modeling framework directly informs programmatic decision-making by identifying age-parity-state subgroups with large rates of unmet need, highlights conflicting trends across survey instruments, and holistically interprets direct survey estimates.

Bradley H. Wagenaar, Orvalho Augusto, Kristjana Ásbjörnsdóttir, Adam Akullian, Nelia Manaca, Falume Chale, Alberto Muanido, Alfredo Covele, Cathy Michel, Sarah Gimbel, Tyler Radford, Blake Girardot, Kenneth Sherr



Lack of accurate data on the distribution of sub-national populations in low- and middle-income countries impairs planning, monitoring, and evaluation of interventions. Novel, low-cost methods to develop unbiased survey sampling frames at sub-national, sub-provincial, and even sub-district levels are urgently needed. This article details our experience using remote satellite imagery to develop a provincial-level representative community survey sampling frame to evaluate the effects of a 7-year health system intervention in Sofala Province, Mozambique.


Mozambique’s most recent census was conducted in 2007, and no data are readily available to generate enumeration areas for representative health survey sampling frames. To remedy this, we partnered with the Humanitarian OpenStreetMap Team to digitize every building in Sofala and Manica provinces (685,189 Sofala; 925,713 Manica) using up-to-date remote satellite imagery, with final results deposited in the open-source OpenStreetMap database. We then created a probability proportional to size sampling frame by overlaying a grid of 2.106 km resolution (0.02 decimal degrees) across each province, and calculating the number of buildings within each grid square. Squares containing buildings were used as our primary sampling unit with replacement. Study teams navigated to the geographic center of each selected square using geographic positioning system coordinates, and then conducted a standard “random walk” procedure to select 20 households for each time a given square was selected. Based on sample size calculations, we targeted a minimum of 1500 households in each province. We selected 88 grids within each province to reach 1760 households, anticipating ongoing conflict and transport issues could preclude the inclusion of some clusters.


Civil conflict issues forced the exclusion of 8 of 31 subdistricts in Sofala and 15 of 39 subdistricts in Manica. Using Android tablets, Open Data Kit software, and a remote RedCap data capture system, our final sample included 1549 households in Sofala (4669 adults; 4766 children; 33 missing age) and 1538 households in Manica (4422 adults; 4898 children; 33 missing age).


Other implementation or evaluation teams may consider employing similar methods to track population distributions for health systems planning or the development of representative sampling frames using remote satellite imagery.