Publications

In 2020, the WHO launched its first global strategy to accelerate the elimination of cervical cancer, outlining an ambitious set of targets for countries to achieve over the next decade. At the same time, new tools, technologies, and strategies are in the pipeline that may improve screening performance, expand the reach of prophylactic vaccines, and prevent the acquisition, persistence and progression of oncogenic HPV. Detailed mechanistic modelling can help identify the combinations of current and future strategies to combat cervical cancer. Open-source modelling tools are needed to shift the capacity for such evaluations in-country. Here, we introduce the Human papillomavirus simulator (HPVsim), a new open-source software package for creating flexible agent-based models parameterised with country-specific vital dynamics, structured sexual networks, and co-transmitting HPV genotypes. HPVsim includes a novel methodology for modelling cervical disease progression, designed to be readily adaptable to new forms of screening. The software itself is implemented in Python, has built-in tools for simulating commonly-used interventions, includes a comprehensive set of tests and documentation, and runs quickly (seconds to minutes) on a laptop. Performance is greatly enhanced by HPVsim’s multiscale modelling functionality. HPVsim is open source under the MIT License and available via both the Python Package Index (via pip install) and GitHub (hpvsim.org).

Somalia is a complex and fragile setting with a demonstrated potential for disruptive, high-burden measles outbreaks. In response, since 2018, Somalian authorities have partnered with UNICEF and the WHO to implement measles vaccination campaigns across the country. In this paper, we create a Somalia-specific model of measles transmission based on a comprehensive epidemiological dataset including case-based surveillance, vaccine registries, and serological surveys. We use this model to assess the impact of these campaign interventions on Somalian’s measles susceptibility, showing, for example, that across the roughly 10 million doses delivered, 1 of every 5 immunized a susceptible child. Finally, we use the model to explore a counter-factual epidemiology without the 2019–2020 campaigns, and we estimate that those interventions prevented over 10,000 deaths.

Access to medical treatment for fever is essential to prevent morbidity and mortality in individuals and to prevent transmission of communicable febrile illness in communities. Quantification of the rates at which treatment is accessed is critical for health system planning and a prerequisite for disease burden estimates. In this study, national data on the proportion of children under five years old with fever who were taken for medical treatment were collected from all available countries in Africa, Latin America, and Asia (n = 91). We used generalised additive mixed models to estimate 30-year trends in the treatment-seeking rates across the majority of countries in these regions (n = 151). Our results show that the proportions of febrile children brought for medical treatment increased steadily over the last 30 years, with the greatest increases occurring in areas where rates had originally been lowest, which includes Latin America and Caribbean, North Africa and the Middle East (51 and 50% increase, respectively), and Sub-Saharan Africa (23% increase). Overall, the aggregated and population-weighted estimate of children with fever taken for treatment at any type of facility rose from 61% (59–64 95% CI) in 1990 to 71% (69–72 95% CI) in 2020. The overall population-weighted average for fraction of treatment in the public sector was largely unchanged during the study period: 49% (42–58 95% CI) sought care at public facilities in 1990 and 47% (44–52 95% CI) in 2020. Overall, the findings indicate that improvements in access to care have been made where they were most needed, but that despite rapid initial gains, progress can plateau without substantial investment. In 2020 there remained significant gaps in care utilisation that must be factored in when developing control strategies and deriving disease burden estimates.

Much of the world’s population had already been infected with COVID-19 by the time the Omicron variant emerged at the end of 2021, but the scale of the Omicron wave was larger than any that had come before or has happened since, and it left a global imprinting of immunity that changed the COVID-19 landscape. In this study, we simulate a South African population and demonstrate how population-level vaccine effectiveness and efficiency changed over the course of the first 2 years of the pandemic. We then introduce three hypothetical variants and evaluate the impact of vaccines with different properties. We find that variant-chasing vaccines have a narrow window of dominating pre-existing vaccines but that a variant-chasing vaccine strategy may have global utility, depending on the rate of spread from setting to setting. Next-generation vaccines might be able to overcome uncertainty in pace and degree of viral evolution.

Sciris aims to streamline the development of scientific software by making it easier to perform
common tasks. Sciris provides classes and functions that simplify access to frequently used
low-level functionality in the core libraries of the scientific Python ecosystem (such as numpy
for math and matplotlib for plotting), as well as in libraries of broader scope (such as
multiprocess for parallelization and pickle for saving and loading objects). While low-level
functionality is valuable for developing robust software applications, it can divert focus from
the scientific problems being solved. Some of Sciris’ key features include: ensuring consistent
dictionary, list, and array types (e.g., enabling users to provide inputs as either lists or arrays);
enabling ordered dictionary elements to be referenced by index; simplifying datetime arithmetic
by allowing date input in multiple formats, including strings; simplifying the saving and loading
of files and complex objects; and simplifying the parallel execution of code. With Sciris, users
can often achieve the same functionality with fewer lines of code, avoid reinventing the wheel,
and spend less time looking up recipes on Stack Overflow. This can make writing scientific code
in Python faster, more pleasant, and more accessible, especially for people without extensive
training in software development.

Background
Circulating vaccine-derived poliovirus outbreaks are spreading more widely than anticipated, which has generated a crisis for the global polio eradication initiative. Effectively responding with vaccination activities requires a rapid risk assessment. This assessment is made difficult by the low case-to-infection ratio of type 2 poliovirus, variable transmissibility, changing population immunity, surveillance delays, and limited vaccine supply from the global stockpile. The geographical extent of responses have been highly variable between countries.

Methods
We develop a statistical spatio-temporal model of short-term, district-level poliovirus spread that incorporates known risk factors, including historical wild poliovirus transmission risk, routine immunization coverage, population immunity, and exposure to the outbreak virus.

Results
We find that proximity to recent cVDPV2 cases is the strongest risk factor for spread of an outbreak, and find significant associations between population immunity, historical risk, routine immunization, and environmental surveillance (p < 0.05). We examine the fit of the model to type 2 vaccine derived poliovirus spread since 2016 and find that our model predicts the location of cVDPV2 cases well (AUC = 0.96). We demonstrate use of the model to estimate appropriate scope of outbreak response activities to current outbreaks.

Conclusion
As type 2 immunity continues to decline following the cessation of tOPV in 2016, outbreak responses to new cVDPV2 detections will need to be faster and larger in scope. We provide a framework that can be used to support decisions on the appropriate size of a vaccination response when new detections are identified. While the model does not account for all relevant local factors that must be considered in the overall vaccination response, it enables a quantitative basis for outbreak response size.

Background: Primary healthcare systems require adequate staffing to meet the needs of their local population. Guidelines typically use population ratio targets for healthcare workers, such as Ethiopia’s goal of two health extension workers for every five thousand people. However, fixed ratios do not reflect local demographics, fertility rates, disease burden (e.g., malaria endemicity), or trends in these values. Recognizing this, we set out to estimate the clinical workload to meet the primary healthcare needs in Ethiopia by region.

Methods: We utilize the open-source modeling package PACE-HRH for our analysis. This is a stochastic Monte Carlo simulation model, which samples annually from distributions for fertility, mortality, disease burden, and the trends in these rates. Inputs were drawn from literature, DHS, and UN population estimates. We model seven regions and two charted cities of Ethiopia, based on data availability and the anticipated reliability of historical trends into the future.

Results: All regions and charted cities are expected to experience increased workload between 2021 and 2035 for a starting catchment of five thousand people. The expected (mean) clinical workload varied from 2,930 hours (Addis) to 3,752 (Gambela) and increased by 19-28% over fifteen years. This results from a decline in per capita workload (due to declines in fertility and infectious diseases), overpowered by total population growth. Pregnancy, non-communicable diseases, sick child care, and nutrition remain the largest service categories, but their priority shifts substantially in some regions by 2035. Sensitivity analysis shows that fertility assumptions have major implications for workload. We incorporate seasonality and estimate monthly variation of up to 8.9% (Somali), though most services with high variability are declining.

Conclusions: Regional variation in demographics, fertility, seasonality, and disease trends all affect the workload estimates. This results in differences in expected clinical workload, the level of uncertainty in those estimates, and relative priorities between service categories. By showing these differences, we demonstrate the inadequacy of a fixed population ratio for staffing allocation. Policy-makers and regulators need to consider these factors in designing their healthcare systems, or they risk sub-optimally allocating workforce and creating inequitable access to care.

Policymakers must make management decisions despite incomplete knowledge and conflicting model projections. Little guidance exists for the rapid, representative, and unbiased collection of policy-relevant scientific input from independent modeling teams. Integrating approaches from decision analysis, expert judgment, and model aggregation, we convened multiple modeling teams to evaluate COVID-19 reopening strategies for a mid-sized United States county early in the pandemic. Projections from seventeen distinct models were inconsistent in magnitude but highly consistent in ranking interventions. The 6-mo-ahead aggregate projections were well in line with observed outbreaks in mid-sized US counties. The aggregate results showed that up to half the population could be infected with full workplace reopening, while workplace restrictions reduced median cumulative infections by 82%. Rankings of interventions were consistent across public health objectives, but there was a strong trade-off between public health outcomes and duration of workplace closures, and no win-win intermediate reopening strategies were identified. Between-model variation was high; the aggregate results thus provide valuable risk quantification for decision making. This approach can be applied to the evaluation of management interventions in any setting where models are used to inform decision making. This case study demonstrated the utility of our approach and was one of several multimodel efforts that laid the groundwork for the COVID-19 Scenario Modeling Hub, which has provided multiple rounds of real-time scenario projections for situational awareness and decision making to the Centers for Disease Control and Prevention since December 2020.

Vaccines against the SARS-CoV-2 virus were developed in record time, but their distribution has been highly unequal. With demand saturating in high-income countries, many low- and middle-income countries (LMIC) finally have an opportunity to acquire COVID-19 vaccines. But the pandemic has taken its toll, and a majority of LMIC populations have partial immunity to COVID-19 disease due primarily to viral infection. This existing immunity, combined with resource limitations, raises the question of how LMICs should prioritize COVID-19 vaccines relative to other competing health priorities. We modify an established computational model, Covasim, to address these questions in four diverse country-like settings under a variety of viral evolution, vaccine delivery, and novel immunity scenarios. Under continued Omicron-like viral evolution and mid-level immunity assumptions, results show that COVID-19 vaccines could avert up to 2 deaths per 1,000 doses if administered to high-risk (60+) populations as prime+boost or annual boosting campaigns. Similar immunization efforts reaching healthy children and adults would avert less than 0.1 deaths per 1,000 doses. Together, these modeling results can help to support normative guidelines and programmatic decision making towards objectively maximizing population health.

Background
Effective healthcare systems need adequate numbers of well-trained human resources for health (HRH). To support evidence-based strategic planning, modeling is sometimes used to estimate the number of required health workers and to allocate them appropriately. However, despite the demonstrated utility of models, there are several limitations to existing tools, including the inability to reflect the stochastic nature of workload and parameter uncertainty, or to incorporate seasonal variations. Additionally, some tools are proprietary or no longer supported, which makes them difficult for decision makers to adopt.

Methods
To address these issues, we have created an open-source, freely available modeling tool called the Population-Aware Capacity Estimator for Human Resources for Health (PACE-HRH). The modeling platform has two components: an Excel-based workbook for data input and scenario management, and a stochastic Monte Carlo simulation package and analysis pipeline written in R. PACE-HRH has a demographics model that projects future populations, a task time model that estimates workload from both variable responsibilities and overhead, an optional seasonality model, and an optional cadre allocation model.

Results
To establish the utility of PACE-HRH, we run a demonstrative model based on a subset of eight clinical service categories, populated with Ethiopian data. The projections show an increase in weekly workload for a baseline population from 37.8 (36.0, 39.7) hours in 2021 to 44.0 (37.9, 49.8) hours in 2035. The ability to calculate a confidence interval is unique to PACE-HRH, as is the option to calculate the monthly variation in workload, which in this case amounts to seasonal amplitude of 6.8%. These results are demonstrative only and more curated input assumptions would be needed in order for the results to support decision making.

Conclusions
Modeling HRH requirements is valuable to planning processes. The PACE-HRH modeling package takes a novel approach to generating these estimates and is designed to be an easy-to-use platform that reduces barriers to use. There is a shortage of observational data on task times, which are key model assumptions, and time and motion studies are needed. However, even without improved data, PACE-HRH is an advancement in the field of HRH modeling and can be used to support evidence-based planning processes.