arXiv subjects

Nicholas Geard

Publications and source records attributed to Nicholas Geard.

13 recordsLinked to original sources

arXiv2026

Generating synthetic electronic health record data using agent-based models to evaluate machine learning robustness under mass casualty incidents

ML models in healthcare are typically evaluated using curated real-world EHR data. A key limitation of such evaluations is that they may fail to assess the robustness of ML models to changes in the data at deployment, which is a common issue because EHR data used for ML model development cannot capture all such changes. Mass casualty incidents (MCIs) caused by disasters are critical instances where this will be an issue, as they induce rare, uncertain, and novel changes to routine system conditions. Because real-world EHR data from MCIs are often limited or unavailable, assessing ML robustness under such conditions before deployment remains challenging. Here, we propose an agent-based modelling approach for generating synthetic EHR data to evaluate the robustness of ML models under MCI scenarios. We use real-world EHR data to develop and calibrate an agent-based model (ABM) of an emergency department (ED) that explicitly models patient arrivals, resource capacity, and clinical workflow. By changing these system conditions to reflect plausible MCI scenarios, the ED model generates synthetic versions of the real-world EHR data that exhibit shifts in system behaviour. Using these synthetic data, we test ML models for predicting length of stay. We observed consistent declines in recall under MCI conditions relative to baseline system conditions, resulting in an increase in the number of patients with prolonged length of stay that were missed by the ML models. These results highlight the impact of changes in system conditions on patient outcomes, EHR data, and ML model performance. Our work establishes ABM-based synthetic EHR data generation as a proactive and systematic approach for evaluating the robustness of ML models under MCI or other system conditions not captured in real-world EHR data, supporting the safer and more effective deployment of ML models in healthcare systems.

cs.LG↗

arXiv2025

Modelling the impact of improving access to healthcare on Hepatitis B prevalence in the Thai-Myanmar border region

Introduction: In Thailand, Hepatitis B is still endemic despite a strong program to eliminate the disease. A higher prevalence is reported in the border region and among migrants due to physical, financial and cultural barriers. Policies and programs targeting the border region and migrant communities have been suggested. Models can be used to understand and quantify the impact of these policies, given they can capture the heterogeneity within the population. Methods: In this study, we developed an Agent-based model that captures the differences between the Thai and migrant populations living in this region, notably the higher level of mobility, lower access to healthcare, and the higher prevalence of Hepatitis B among migrants, by modelling the origin of each individual explicitly. We used the model to estimate future trends of Hepatitis B prevalence in Thailand near the border with Myanmar under different scenarios of intervention. Results: Our study shows that although the current intervention level is effective in the Thai population, it is insufficient to reach national elimination targets due to high prevalence in migrants. Improving access to healthcare for migrants and the border region could potentially help to reach elimination targets, and we quantified the level of improvement needed to achieve elimination. Conclusion: Although there already exist policies to make healthcare more accessible to migrants and the border regions, they are still not yet effective due to financial and cultural barriers. Bringing down those barriers could reduce Hepatitis B prevalence in those communities and regions and contribute to reaching elimination targets in a reasonable timeline.

q-bio.OT↗

arXiv2025

Simulating the impact of perception bias on social contact surveys for infectious disease modelling

Social contact patterns are a key input to many infectious disease models. Contact surveys, where participants are asked to provide information on their recent close and casual contacts with others, are one of the standard methods to measure contact patterns in a population. Surveys that require detailed sociodemographic descriptions of contacts allow for the specification of fine-grained contact rates between subpopulations in models. However, perception biases affecting a surveyed person's ability to estimate sociodemographic attributes (e.g., age, race, socioeconomic status) of others could affect contact rates derived from survey data. Here, we simulate contact surveys using a synthetic contact network of New Mexico to investigate the impact of these biases on survey accuracy and infectious disease model projections. We found that perception biases affecting the estimation of another individual's age and race substantially decreased the accuracy of the derived contact patterns. Using these biased patterns in a Susceptible-Infectious-Recovered compartmental model lead to an underestimation of cumulative incidence among older people (65+ years) and individuals identifying as races other than White. Our study shows that perception biases can impact contact patterns estimated from surveys in ways that systematically underestimate disease burden in minority populations when used in transmission models.

q-bio.PE↗

arXiv2024

Should public health policy exempt cases with low viral load from isolation during an epidemic?: a modelling study

During the COVID-19 pandemic, case isolation emerged as a key non-pharmaceutical intervention in pandemic response. Its effectiveness hinges on the timing of isolation, which is often dictated by the onset of contagiousness. While minimizing isolation delays is crucial, the impact of evolving epidemic dynamics, such as changes in viral load distributions among cases, is less understood. These dynamics could inform more efficient isolation strategies. We developed a multi-scale agent-based model to assess isolation policies that consider viral loads. Our model compares the outcomes of universal isolation with strategies that exempt low viral load cases post-peak. We found that most low viral load cases identified after the peak are less contagious, raising the question of their need for isolation. Our analysis reveals that exempting these individuals slightly increases new infections and, unexpectedly, reduces efficiency, defined as infections averted per isolation. These results advise caution with adaptive isolation policies. Our flexible model can evaluate complex strategies, aiding future public health planning.

q-bio.PE↗

arXiv2024

Apparent structural changes in contact patterns during COVID-19 were driven by survey design and long-term demographic trends

Social contact patterns are key drivers of infectious disease transmission. During the COVID-19 pandemic, differences between pre-COVID and COVID-era contact rates were widely attributed to non-pharmaceutical interventions such as lockdowns. However, the factors that drive changes in the distribution of contacts between different subpopulations remain poorly understood. Here, we present a clustering analysis of 33 contact matrices generated from surveys conducted before and during the COVID-19 pandemic, and analyse key features distinguishing their topological structures. While we expected to identify aspects of pandemic scenarios responsible for these features, our analysis demonstrates that they can be explained by differences in study design and long-term demographic trends. Our results caution against using survey data from different studies in counterfactual analysis of epidemic mitigation strategies. Doing so risks attributing differences stemming from methodological choices or long-term changes to the short-term effects of interventions.

physics.soc-ph↗

arXiv2023

A model-based assessment of social isolation practices for COVID-19 outbreak response in residential care facilities

Residential aged-care facilities (RACFs, also called long-term care facilities, aged care homes, or nursing homes) have elevated risks of respiratory infection outbreaks and associated disease burden. During the COVID-19 pandemic, social isolation policies were commonly used in these facilities to prevent and mitigate outbreaks. We refer specifically to general isolation policies that were intended to reduce contact between residents, without regard to confirmed infection status. Such policies are controversial because of their association with adverse mental and physical health indicators and there is a lack of modelling that assesses their effectiveness. We developed an agent-based model of COVID-19 transmission in a structured population, intended to represent the salient characteristics of a residential care environment. Using our model, we generated stochastic ensembles of simulated outbreaks and compared summary statistics of outbreaks simulated} under different mitigation conditions. Our study focuses on the marginal impact of general isolation (reducing social contact between residents), regardless of confirmed infection. In the absence of any asymptomatic screening, general isolation of residents to their rooms reduced median cumulative cases by approximately 27%. However, when conducted concurrently with asymptomatic screening and isolation of confirmed cases, general isolation reduced the median number of cumulative infections by only 12% in our simulations. Our simulations showed that general isolation of residents did not provide substantial benefits beyond those achieved through screening, isolation of confirmed cases, and deployment of PPE. Our conclusions are sensitive to assumptions about the proportion of total contacts in a facility accounted for by casual interactions between residents.

q-bio.PE↗

arXiv2023

AI driven B-cell Immunotherapy Design

Antibodies, a prominent class of approved biologics, play a crucial role in detecting foreign antigens. The effectiveness of antigen neutralisation and elimination hinges upon the strength, sensitivity, and specificity of the paratope-epitope interaction, which demands resource-intensive experimental techniques for characterisation. In recent years, artificial intelligence and machine learning methods have made significant strides, revolutionising the prediction of protein structures and their complexes. The past decade has also witnessed the evolution of computational approaches aiming to support immunotherapy design. This review focuses on the progress of machine learning-based tools and their frameworks in the domain of B-cell immunotherapy design, encompassing linear and conformational epitope prediction, paratope prediction, and antibody design. We mapped the most commonly used data sources, evaluation metrics, and method availability and thoroughly assessed their significance and limitations, discussing the main challenges ahead.

q-bio.QM↗

arXiv2022

Correlation of viral loads in disease transmission chains could bias early estimates of the reproduction number

Early estimates of the transmission properties of a newly emerged pathogen are critical to an effective public health response, and are often based on limited outbreak data. Here, we use simulations to investigate a potential source of bias in such estimates, arising from correlations between the viral load of cases in transmission chains. We show that this mechanism can affect estimates of fundamental transmission properties characterising the spread of a virus. Our computational model simulates a disease transmission mechanism in which the viral load of the infector at the time of transmission influences the infectiousness of the infectee. These correlations in transmission pairs produce a population-level decoherence process during which the distributions of initial viral loads in each subsequent generation converge to a steady state. We find that outbreaks arising from index cases with low initial viral loads give rise to early estimates of transmission properties that are subject to large biases. These findings demonstrate the potential for bias arising from transmission mechanics to affect estimates of the transmission properties of newly emerged viruses.

q-bio.PE↗

arXiv2022

MPVNN: Mutated Pathway Visible Neural Network Architecture for Interpretable Prediction of Cancer-specific Survival Risk

Survival risk prediction using gene expression data is important in making treatment decisions in cancer. Standard neural network (NN) survival analysis models are black boxes with lack of interpretability. More interpretable visible neural network (VNN) architectures are designed using biological pathway knowledge. But they do not model how pathway structures can change for particular cancer types. We propose a novel Mutated Pathway VNN or MPVNN architecture, designed using prior signaling pathway knowledge and gene mutation data-based edge randomization simulating signal flow disruption. As a case study, we use the PI3K-Akt pathway and demonstrate overall improved cancer-specific survival risk prediction results of MPVNN over standard non-NN and other similar sized NN survival analysis methods. We show that trained MPVNN architecture interpretation, which points to smaller sets of genes connected by signal flow within the PI3K-Akt pathway that are important in risk prediction for particular cancer types, is reliable.

q-bio.QM↗

arXiv2021

Modelling the effect of within-host dynamics on the diversity of a multi-strain pathogen

Multi-strain pathogens such as Group A Streptococcus, Streptococcus pneumoniae, and Staphylococcus aureus cause millions of infections each year with a substantial health burden. Control of multi-strain pathogens can be complicated by the high strain diversity often observed in endemic settings. It is not well understood how high strain diversity is maintained in populations, given that they compete with each other both directly (within an individual host) and indirectly (via host immunity). Previous modelling studies have investigated how indirect competition affects the prevalence and diversity of strains. However, these studies often make simplifying assumptions about the direct competition that occurs within hosts. Currently, little data is available to validate these assumptions, hence there is a need to clarify how sensitive model outputs are to these assumptions. In this study, we compare the dynamics of multi-strain pathogens under different assumptions about direct competition between strains using an agent-based model. We find that the assumptions made about direct competition can affect the epidemiological dynamics, particularly when there is no long-term immunity following infections and a low rate of importation of non-circulating strains. Our results suggest that while direct and indirect competition can each decrease strain diversity when they act in isolation, they may increase strain diversity when they act together. This finding highlights the importance of examining sensitivity to assumptions about strain competition. In particular, omitting consideration of direct competition can lead to inaccurate estimates of the likely effectiveness of control strategies as changes in strain diversity shift the level of direct strain competition.

q-bio.PE↗

arXiv2021

Global and local epidemiology of Group A Streptococcus indicates that naturally-acquired immunity is enduring and strain-specific

The bacterium Group A Streptococcus (Streptococcus pyogenes, GAS) is a human-specific pathogen and a major cause of global morbidity and mortality. Despite decades of research our knowledge of GAS infection and immunity is incomplete, hampering vaccine design and other efforts to reduce disease prevalence. Epidemiological studies indicate positive associations between the prevalence of GAS-related disease, the diversity of circulating strains and the degree of poverty in host populations. However, the infection and immune mechanisms underlying these associations are not clear. In this work, we use an agent-based model to demonstrate that observed diversity and prevalence are best accounted for by the hypothesis that GAS infection confers enduring strain-specific immunity, with reduced or absent cross-protection against infection by other strains. Our results suggest that the success of GAS vaccines will depend on their ability to elicit long-lasting cross-protective immunity over multiple strain types.

q-bio.PE↗

arXiv2021

COVID-19 in low-tolerance border quarantine systems: impact of the Delta variant of SARS-CoV-2

In controlling transmission of COVID-19, the effectiveness of border quarantine strategies is a key concern for jurisdictions in which the local prevalence of disease and immunity is low. In settings like this such as China, Australia, and New Zealand, rare outbreak events can lead to escalating epidemics and trigger the imposition of large scale lockdown policies. Here, we examine to what degree vaccination status of incoming arrivals and the quarantine workforce can allow relaxation of quarantine requirements. To do so, we develop and apply a detailed model of COVID-19 disease progression and transmission taking into account nuanced timing factors. Key among these are disease incubation periods and the progression of infection detectability during incubation. Using the disease characteristics associated with the ancestral lineage of SARS-CoV-2 to benchmark the level of acceptable risk, we examine the performance of the border quarantine system for vaccinated arrivals. We examine disease transmission and vaccine efficacy parameters over a wide range, covering plausible values for the Delta variant currently circulating globally. Our results indicate a threshold in outbreak potential as a function of vaccine efficacy, with the time until an outbreak increasing by up to two orders of magnitude as vaccine efficacy against transmission increases from 70% to 90%. For parameters corresponding to the Delta variant, vaccination is able to maintain the capacity of quarantine systems to reduce case importation and outbreak risk, by counteracting the pathogen's increased infectiousness. To prevent outbreaks, heightened vaccination in border quarantine systems must be combined with mass vaccination. The ultimate success of these programs will depend sensitively on the efficacy of vaccines against viral transmission.

q-bio.PE↗

arXiv2020

Risk mapping for COVID-19 outbreaks in Australia using mobility data

COVID-19 is highly transmissible and containing outbreaks requires a rapid and effective response. Because infection may be spread by people who are pre-symptomatic or asymptomatic, substantial undetected transmission is likely to occur before clinical cases are diagnosed. Thus, when outbreaks occur there is a need to anticipate which populations and locations are at heightened risk of exposure. In this work, we evaluate the utility of aggregate human mobility data for estimating the geographic distribution of transmission risk. We present a simple procedure for producing spatial transmission risk assessments from near-real-time population mobility data. We validate our estimates against three well-documented COVID-19 outbreak scenarios in Australia. Two of these were well-defined transmission clusters and one was a community transmission scenario. Our results indicate that mobility data can be a good predictor of geographic patterns of exposure risk from transmission centres, particularly in scenarios involving workplaces or other environments associated with habitual travel patterns. For community transmission scenarios, our results demonstrate that mobility data adds the most value to risk predictions when case counts are low and spatially clustered. Our method could assist health systems in the allocation of testing resources, and potentially guide the implementation of geographically-targeted restrictions on movement and social interaction.

physics.soc-ph↗