Having enough doctors to provide healthcare services is a concern internationally. In the UK, significant resources for education and training have been devoted to medical workforce management. Nevertheless, some areas of the country still struggle to recruit and retain staff compared with others. Solutions to this problem have focused on attracting students from backgrounds not traditionally represented in medicine to choose it as a career, and opening new medical schools in different areas of the country. The main objective of this paper is to examine medical student and doctor distribution in order to contribute to understanding the distribution of health and health service inequalities. We used a modelling approach to understand characteristics of medical students, medical schools and foundation schools to interpret and identify the relationship between geographic distribution and socio-economic deprivation. This geographical and statistical analysis aims to identify patterns in workforce distribution, layering these with data on deprivation and inequality. Analysis shows that there are fewer students who come from from more deprived areas, and that different patterns can be observed in geographic locations of training when considering gender and ethnicity. While there is greater diversity of the future workforce in terms of gender and ethnicity, there is evidence that fewer students from more deprived backgrounds are attending medical schools. This has implications for the future workforce, and medical schools may need to play a greater role in increasing access to medical education to overcome observed inequalities.

This work develops a block aggregation approach to spatial estimation and prediction when the response is observed at a coarse spatial scale, for example as counts of events in administrative areas, or blocks, while covariates are available at a finer spatial resolution, typically as raster images. Our approach specifies a linear predictor at the finer resolution as a combination of covariate effects and a latent, spatially continuous Gaussian process. This linear predictor then determines the distribution of the response through an inverse link function and spatial integration. We use a simulation study to evaluate the performance of the proposed approach in comparison to two industry standard approaches: a traditional geostatistical model that associates each response with the centroid of its block; and a Markov random field (MRF) approach that aggregates covariate data to block-level. As expected, the differences in performance among the three approaches are small with respect to block-level prediction. The rationale for, and advantage of, the block aggregation approach lies in its delivery of reliable inferences at whatever spatial resolution is required in a particular application. We describe two applications: a linear Gaussian sampling model of wastewater virus concentrations in England, using population density as covariate; and log-linear Poisson model of cardiovascular hospitalisations in England using socio-demographic variables at fine-scale administrative units as covariates.

Objectives The rapid spread of SARS-CoV-2 infection caused high levels of hospitalisation and deaths in late 2020 and early 2021 during the second wave in England. COVID-19 disease during this period was associated with marked health inequalities across ethnic and sociodemographic subgroups. In this paper, we aim to investigate and quantify inequalities in the risk of SARS-CoV-2 infection across ethnic and sociodemographic subgroups during a key period before widespread vaccination, thereby identifying the populations that bore a disproportionate burden of risk.Methods We analysed risk factors for test-positivity for SARS-CoV-2, based on self-administered throat and nose swabs in the community during rounds 5-10 of the REal-time Assessment of Community Transmission-1 (REACT-1) study between 18 September 2020 and 30 March 2021.Results Compared with the White ethnicity, people of Asian and the Black ethnicity had a higher risk of infection during rounds 5-10, with odds of 1.45 (1.27, 1.66) and 1.34 (1.11, 1.63), respectively, adjusted for demographic factors including age, sex, region, key work status, ethnicity and deprivation. Among ethnic subgroups, the highest and the second-highest odds were found in Bangladeshi and Pakistani participants at 3.28 (2.24, 4.80) and 2.12 (1.70, 2.64), respectively, when compared with the White British participants. People in larger (compared with smaller) households had higher odds of infection. Healthcare workers with direct patient contact and care home workers showed higher odds of infection compared with other essential/key workers. Additionally, the odds of infection among participants in public-facing activities or settings were greater than among those not working in those activities or settings.Conclusion Our findings highlight the differences in the risk of SARS-CoV-2 infection in a global north population during a period when the risk of infection was high, and there were substantial levels of social mixing. Planning for future waves of severe respiratory pathogens should include policies to reduce inequality in the risk of infection by ethnicity, household size and occupational activity in order to reduce inequality in disease.

Prevalence surveys are routinely used to monitor the effectiveness of mass drug administration (MDA) programmes for controlling neglected tropical diseases (NTDs). We propose a decay-adjusted spatio-temporal (DAST) model that explicitly accounts for the time-varying impact of MDA on NTD prevalence, providing a flexible and interpretable framework for estimating intervention effects from sparse survey data. Using case studies on soil-transmitted helminths and lymphatic filariasis, we show that DAST offers a practical alternative to standard geostatistical models when the objective includes quantifying MDA impact and supporting short-term programmatic forecasting. We also discuss extensions and identifiability challenges, advocating for data-driven parsimony over complexity in settings where the available data are too sparse to support the estimation of highly parameterised models.

Key challenges in the analysis of highly multivariate large-scale spatial stochastic processes, where both the number of components (p) and spatial locations (n) can be large, include achieving maximal sparsity in the joint precision matrix, ensuring efficient computational cost for its generation, accommodating asymmetric cross-covariance in the joint covariance matrix, and delivering scientific interpretability. We propose a cross-MRF model class, consisting of a mixed spatial graphical model framework and cross-MRF theory, to collectively address these challenges in one unified framework across two modelling stages. The first stage exploits scientifically informed conditional independence (CI) among p component fields and allows for a step-wise parallel generation of joint covariance and precision matrix, enabling a simultaneous accommodation of asymmetric cross-covariance in joint covariance matrix and sparsity in joint precision matrix. The second stage extends the first-stage CI to doubly CI among both p and n and unearths the cross-MRF via an extended Hammersley-Clifford theorem for multivariate spatial stochastic processes. This results in the sparsest possible representation of the joint precision matrix and ensures its lowest generation complexity. We demonstrate with 1D simulated comparative studies and 2D real-world data.

Background Snakebite is a priority neglected tropical disease, but incidence data are lacking; current estimates rely upon incomplete health facility reports or ad hoc surveys. Spatial analysis methods harness statistical associations between case incidence and spatially varying factors to improve estimates. This systematic review aimed to identify variables associated with snakebite risk in spatial and temporal analyses for inclusion in geospatial studies to improve risk estimation accuracy.Methods We searched MEDLINE, Global Health, PubMed and Web of Science in January 2023 for studies published since 1980 assessing snakebite outcomes and spatially varying factors at the setting level. Study quality was assessed using an adapted Joanna Briggs Institute tool. The results are presented by narrative synthesis.Results Thirty-five studies were eligible; the majority were from Central and South America (18), then Asia (11). Climate and environment were most frequently assessed, with temperature, humidity and tree cover predominantly positively associated with snakebite risk, drought negatively associated and altitude negative/mixed. Crop and livestock variables mostly showed positive associations; population density and urban residence overwhelmingly displayed negative associations.Conclusions This review identifies key variables that should be considered in future snakebite risk research. Limitations include low research availability from the highest risk regions. There is an evident need for greater research into snakebite risk variation, particularly in sub-Saharan Africa.

Background Snakebite envenoming is a medical emergency that requires rapid access to essential medicines and well-trained personnel. In resource-poor countries, mapping snakebite incidence can help policymakers to make evidence-based decisions for resource prioritisation. This study aimed to characterise the spatial variation in snakebite risk, and in particular to identify areas of relatively high and low risk, in Eastern Province, Rwanda. Methods Snakebite surveillance of people bitten in 2020 was conducted in Eastern Province through household visits and case verification. Geostatistical modelling and predictive mapping were applied to data from 617 villages in six districts to develop sector-level and district-level risk maps. Results There were 1217 individuals bitten by snakes across six districts. The estimated population-weighted snakebite incidence in Eastern Province was 440 (95% predictive interval 421 to 460) cases per 100 000 people, corresponding to 13 500 (95% predictive interval 12 950 to 14 150) snakebite events per year. Two sectors in the southwest, Gashanda and Jarama, showed >1500 snakebite events per 100 000 annually. The lowest incidence was observed in the north. Conclusions Considerable differences exist in snakebite risk between sectors in Eastern Province, with the highest risk concentrated in the southwest. Policymakers should consider prioritising resources related to snakebite prevention, essential medicines and health worker training in this region.

To explore how model-based geostatistics (MBG) could support trachoma elimination efforts, a technical consultation was held on March 4 and 5, 2024 by the Centre for Health Informatics, Computing, and Statistics at Lancaster University, United Kingdom, a WHO Collaborating Centre on Geostatistical Methods for Neglected Tropical Disease Research. The meeting aimed to foster collaboration for sharing insights on using MBG for decision-making; showcase its applications in assessing trachoma elimination status; address challenges, such as setting the probability threshold for elimination and resolving conflicts between survey and MBG evidence; and discuss considerations for integrating MBG into Tropical Data. Participants, including trachoma program managers, experts, academics, donors, and statisticians, reviewed MBG applications, discussed ongoing studies, identified knowledge gaps, and planned future work. This article summarizes the meeting's presentations, discussions, and outcomes, highlighting current conclusions on and research priorities to evaluate MBG's feasibility and utility in trachoma elimination programs.

There is a long-held hypothesis that multiple sclerosis (MS) affects the central nervous system in a length dependent way reflecting the propensity of longer central axonal projections to accumulate damage, but evidence for this is lacking. To determine the prevalence of body part involvement in MS and relate this to the putative axonal length innervating each body part, we asked people with MS to indicate affected body parts on a somatic diagram. Axonal length for each body part was calculated from neuroanatomical literature. The survey was part of the TONiC-MS study. Records from 5925 respondents were analysed for involvement of eleven distinct body parts (either hand/ upper limb /lower limb, urinary bladder, neck, speech, vision, swallowing), and also balance. Participants had a wide range of age, disease duration, disease subtypes and disability levels. Body part involvement in the whole sample was highly correlated with axonal length (rho 0.933). At an individual level, a logistic regression including covariates of age, disease type and disability level demonstrated that the probability of body part involvement was substantially dependent on axonal length across all disease types. Our study supports the hypothesis that MS disability reflects a length-dependent central axonopathy.

BackgroundChikungunya virus (CHIKV) is an arbovirus with a significant global public health burden. Delineating the specific contributions of individual behaviour, household, natural and built environment to CHIKV transmission is important for reducing risk in urban informal settlements but challenging due to their heterogeneous environments. The aim of this study was to quantify variation in CHIKV seroprevalence between and within four urban communities in a large Brazilian city, and identify the respective contributions of individual, household, and environmental factors for seropositivity.Methodology/principal findingsA cross-sectional serological survey was conducted in four low-income communities in Salvador, Brazil in 2018 to collect individual, household and CHIKV IgG serology data for 1318 participants. Fine-scale community mapping of high-risk environmental features and remotely sensed environmental data were used to improve characterisation of the microenvironment close to the household. We categorised risk factors into three domains - individual, household, and environmental and used binomial mixed-effect models to identify associations with CHIKV seropositivity. CHIKV seroprevalence was 4.8%, 6.1% and 4.3% in three communities and 22.6% in one community which had a distinct topographical profile. The only individual domain variable associated with seropositivity was male sex (OR 1.67, 95% CI 1.11 - 2.36), but several environmental variables, including living in a house on a steep hillside, at medium to high elevations, and with surface water nearby, were associated with higher seropositivity.Conclusions/significanceOur findings indicate that CHIKV exposure risk can vary significantly between nearby communities and at fine spatial scales within communities and is likely to be driven more strongly by the availability of mosquito breeding sites rather than individual exposure patterns. They suggest that environmental deficiencies and topography, a proxy for several environmental processes including the degree of urbanisation and flooding risk, may play an important role in driving risk at both of these scales.

BACKGROUND:Soil-transmitted helminthiasis (STH) are a parasitic infection that predominantly affects impoverished regions. Model-based geostatistics (MBG) has been established as a set of modern statistical methods that enable mapping of disease risk in a geographical area of interest. We investigate how the use of remotely sensed covariates can help to improve the predictive inferences on STH prevalence using MBG methods. In particular, we focus on how the covariates impact on the classification of areas into distinct class of STH prevalence. METHODS:This study uses secondary data obtained from a sample of 1551 schools in Kenya, gathered through a combination of longitudinal and cross-sectional surveys. We compare the performance of two geostatistical models: one that does not make use of any spatially referenced covariate; and a second model that uses remotely sensed covariates to assist STH prevalence prediction. We also carry out a simulation study in which we compare the performance of the two models in the classifications of areal units with varying sample sizes and prevalence levels. RESULTS:The model with covariates generated lower levels of uncertainty and was able to classify 88 more districts into prevalence classes than the model without covariates, which instead left those as "unclassified". The simulation study showed that the model with covariates also yielded a higher proportion of correct classification of at least 40% for all sub-counties. CONCLUSION:Covariates can substantially reduce the uncertainty of the predictive inference generated from geostatistical models. Using covariates can thus contribute to the design of more effective STH control strategies by reducing sample sizes without compromising the predictive performance of geostatistical models.

Recent work has reported that AI classifiers trained on audio recordings can accurately predict severe acute respiratory syndrome coronavirus 2 (SARSCoV2) infection status. Here, we undertake a large scale study of audio-based deep learning classifiers, as part of the UK governments pandemic response. We collect and analyse a dataset of audio recordings from 67,842 individuals with linked metadata, including reverse transcription polymerase chain reaction (PCR) test outcomes, of whom 23,514 tested positive for SARS CoV 2. Subjects were recruited via the UK governments National Health Service Test-and-Trace programme and the REal-time Assessment of Community Transmission (REACT) randomised surveillance survey. In an unadjusted analysis of our dataset AI classifiers predict SARS-CoV-2 infection status with high accuracy (Receiver Operating Characteristic Area Under the Curve (ROCAUC) 0.846 [0.838, 0.854]) consistent with the findings of previous studies. However, after matching on measured confounders, such as age, gender, and self reported symptoms, our classifiers performance is much weaker (ROC-AUC 0.619 [0.594, 0.644]). Upon quantifying the utility of audio based classifiers in practical settings, we find them to be outperformed by simple predictive scores based on user reported symptoms.

Rats are major reservoirs for pathogenic Leptospira, the bacteria causing leptospirosis, particularly in urban informal settlements. However, the impact of variation in rat abundance and pathogen shedding rates on spillover transmission to humans remains unclear. This study aimed to investigate how spatial variation in reservoir abundance and pathogen pressure affect Leptospira spillover transmission to humans in a Brazilian urban informal settlement. A longitudinal eco-epidemiological study was conducted from 2013 to 2014 to characterize the spatial distribution of rat abundance and Leptospira shedding rates in rats and determine the association with human infection risk in a cohort of 2,206 community residents. Tracking plates and live-trapping were used to measure rat abundance and quantify rat shedding status and load. In parallel, four sequential biannual serosurveys were used to identify human Leptospira infections. To evaluate the role of shedding on human risk, we built three statistical models for: (1) the relative abundance of rats, (2) the shedding rate by individual rats, and (3) human Leptospira infection, in which “total shedding”, obtained by multiplying the predictions from those two models, was used as a risk factor. We found that Leptospira shedding was associated with older and sexually mature rats and varied spatially and temporally—higher at valley bottoms and with seasonal rainfall (December to March). The point estimate for “total shedding” by rat populations was positive, i.e., Leptospira infection risk increased with total shedding, but the association was not significant [odds ratio (OR) = 1.1; 95% confidence interval (CI): 0.9, 1.4]. This positive trend was mainly driven by rat abundance, rather than individual rat shedding (OR = 1.8; 95% CI: 0.6, 5.4 vs. OR = 1.0; 95% CI: 0.7, 1.4]. Infection risk was higher in areas with more vegetative land cover (OR = 2.4; 95% CI: 1.2, 4.8), and when floodwater entered the house (OR = 2.4; 95% CI: 1.6, 3.4). Our findings indicate that environmental and hydrological factors play a more significant role in Leptospira spillover than rat associated factors. Furthermore, we developed a novel approach combining several models to elucidate complex links between animal reservoir abundance, pathogen shedding and environmental factors on zoonotic spillover in humans that can be extended to other environmentally transmitted diseases.

Background Soil-transmitted helminths (STH) and schistosomiasis comprise the most wide-spread NTDs globally. Preventative chemotherapy is a cost-effective approach to controlling morbidity of both diseases, but relies on large scale surveys to determine and revise treatment frequency. Availability of detailed information on survey costs is limited despite recent methodological surveying innovations. We micro-costed a survey of STH and schistosomiasis in Kenya, and linked results to precision estimates of competing survey methods to compare cost-efficiency. Methods Costs from a 2017 Kenyan parasitological survey were retrospectively analyzed and extrapolated to explore marginal changes when altering survey size, defined by the number of schools sampled and the number of samples taken per school. Subsequent costs were applied to simulated precision estimates of model-based geostatistical (MBG) and traditional survey designs. Cost-precision was calculated for a range of survey sizes per method. Four traditional survey design scenarios, based around WHO guidelines, were selected to act as reference cases for calculating incremental cost-effectiveness ratios (ICERs) for MBG design. Findings MBG designed surveys showed improved cost-precision, particularly if optimizing number of schools against samples per school. MBG was found to be more cost-effective under 87 of 92 comparisons to reference cases. This comprised 14 situations where MBG was both cheaper and more precise, 42 which had cost saving with precision trade off (ICERs; $8,915-$344,932 per percentage precision lost); and 31 more precise with increased cost (ICERs; $426-$147,748 per percentage precision gained). The remaining 5 comparisons represented extremes of MBG simulated site selection, unlikely to be applied in practice. Interpretation Efficiency gains are possible for deworming surveys when considering cost alone, such as through minimizing sample or analysis costs. However further efficiency maximization is possible when designing surveys using MBG given its improved precision and ability to optimize the balance between number of schools and sample size per school.

During the COVID-19 pandemic, studies in a number of countries have shown how wastewater can be used as an efficient surveillance tool to detect outbreaks at much lower cost than traditional prevalence surveys. In this study, we consider the utilisation of wastewater data in the post-pandemic setting, in which collection of health data via national randomised prevalence surveys will likely be run at a reduced scale; hence an affordable ongoing surveillance system will need to combine sparse prevalence data with non-traditional disease metrics such as wastewater measurements in order to estimate disease progression in a cost-effective manner. Here, we use data collected during the pandemic to model the dynamic relationship between spatially granular wastewater viral load and disease prevalence. We then use this relationship to nowcast local disease prevalence under the scenario that (i) spatially granular wastewater data continue to be collected; (ii) direct measurements of prevalence are only available at a coarser spatial resolution, for example at national or regional scale. The results from our cross-validation study demonstrate the added value of wastewater data in improving nowcast accuracy and reducing nowcast uncertainty. Our results also highlight the importance of incorporating prevalence data at a coarser spatial scale when nowcasting prevalence at fine spatial resolution, calling for the need to maintain some form of reduced-scale national prevalence surveys in non-epidemic periods. The model framework is disease-agnostic and could therefore be adapted to different diseases and incorporated into a multiplex surveillance system for early detection of emerging local outbreaks.

BACKGROUND:Control of schistosomiasis (SCH) relies on the regular distribution of preventive chemotherapy (PC) over many years. For the sake of sustainable SCH control, a decision must be made at some stage to scale down or stop PC. These "stopping decisions" are based on population surveys that assess whether infection levels are sufficiently low. However, the limited sensitivity of the currently used diagnostic (Kato-Katz [KK]) to detect low-intensity infections is a concern. Therefore, the use of new, more sensitive, molecular diagnostics has been proposed. METHODS:Through statistical analysis of Schistosoma mansoni egg counts collected from Burundi and a simulation study using an established transmission model for schistosomiasis, we investigated the extent to which more sensitive diagnostics can improve decision making regarding stopping or continuing PC for the control of S. mansoni. RESULTS:We found that KK-based strategies perform reasonably well for determining when to stop PC at a local scale. Use of more sensitive diagnostics leads to a marginally improved health impact (person-years lived with heavy infection) and comes at a cost of continuing PC for longer (up to around 3 years), unless the decision threshold for stopping PC is adapted upward. However, if this threshold is set too high, PC may be stopped prematurely, resulting in a rebound of infection levels and disease burden (+45% person-years of heavy infection). CONCLUSIONS:We conclude that the potential value of more sensitive diagnostics lies more in the reduction of survey-related costs than in the direct health impact of improved parasite control.