Geographic Covariates Research Articles

Determining the risk factors of malaria and anemia in children between 6 and 59 months using the joint generalized linear mixed model on the 2021 Nigeria Malaria Indicator Survey dataset.

Malaria and anemia are significant public health concerns that contribute to child mortality in African. Despite global efforts to control the two diseases, their prevalence in high-risk regions like Nigeria remains high. Understanding socioeconomic, demographic, and geographical factors associated with malaria and anemia, is critical for effective intervention strategies. The study aims to assess the risk factors for malaria and anemia in children under 5 years using the joint generalized linear mixed model (JGLMM). The Nigeria Malaria Indicator Survey (NMIS) 2021 dataset was used, with a sample of 10,120 children aged 6-59 months. A two-stage sampling method was applied. Descriptive statistics and chi-square tests examined factors associated with malaria and anemia prevalence. Missing data were handled using multiple imputations with chained equations (MICE). Lastly, the adjusted odds ratio was interpreted for the current study. SAS was used in analyzing the data and statistical significance was set at 5% significance level. The prevalence of malaria and anemia was 36.81 and 67.66%, respectively, in children between 6 and 59 months old in Nigeria. The JGLMM was used to examine malaria RDT and anemia findings in conjunction with demographic, geographic, and socioeconomic covariates. The following underlying risk factors for malaria and anemia in children were discovered in the study: region, altitude, age of child in months, toilet facility of the household, main wall material used for the house, main roof material used for the house, children under five who slept under a mosquito net, whether the child had fever in last 2 weeks before the survey took place, place of residence where the child resides, household wealth index, sex of child, and mother's education level. However, whether the mother knew of ways to prevent malaria was not statistically significant regarding anemia. This study highlights the importance of addressing fever as a key factor for anemia and improving housing conditions to reduce malaria and anemia prevalence. Policymakers should prioritize mosquito net distribution and healthcare access, particularly in rural areas. The study's novelty lies in its handling of missing data through imputation techniques, enhancing the reliability of findings.

The 2023 fatal dengue outbreak in Bangladesh highlights a paradigm shift of geographical distribution of cases.

In 2023, Bangladesh experienced its largest and deadliest outbreak of the Dengue virus (DENV), reporting the highest-ever recorded annual cases and deaths. Historically, most of the cases were recorded in the capital city, Dhaka. We aimed to characterize the geographical transmission of DENV in Bangladesh. From 1 January-31 December 2023, we extracted and analyzed daily data on dengue cases and deaths from the Management Information System of the Ministry of Health and Family Welfare. We performed a generalized linear mixed model to identify the associations between division-wise daily dengue counts and various geographical and meteorological covariates. The number of dengue cases reported in 2023 was 1.3 times higher than the total number recorded in the past 23years (321,179 vs. 244,246), with twice as many deaths than the total fatalities recorded over the past 23years (1705 vs. 849). Of the 1,705 deaths in 2023, 67.4% (n=1,015) died within one day after hospital admission. The divisions southern to Dhaka had a higher dengue incidence/1000 population (2.30 vs. 0.50, p<0.01) than the northern divisions. Festival-related travel along with meteorological factors and urbanization are likely to have contributed to the shift of dengue from Dhaka to different districts in Bangladesh.

Satellite-informed smart sensor placement framework for near-optimal PM2.5 monitoring in urban areas.

Air pollution is a global public health concern, particularly due to PM2.5, which can cause respiratory and cardiovascular diseases. Accurate placement of monitoring sensors is essential to effectively monitor and mitigate PM2.5 effects. However, the complex nature of air pollution, including factors like traffic density, population density, and weather conditions, poses challenges for sensor placement. Additionally, cost and resource constraints further complicate the process. In this study, we propose a novel algorithm that utilizes a multi-criteria optimization approach to identify optimal locations and distribution of PM2.5 monitoring sensors. The algorithm integrates various geographical covariates, such as roads, population density, terrain elevation, and satellite observations of surface PM2.5. By applying the Non-dominated Sorting Genetic Algorithm II (NSGA-II), we optimize sensor placement. Our algorithm is validated through a case study in a metropolitan area, demonstrating its ability to identify optimal sensor locations while reducing their number and maintaining high accuracy. Furthermore, we highlight the value of satellite observations for initial PM2.5 estimates and aiding sensor placement. Our comprehensive algorithm optimizes air quality monitoring, enabling the identification of pollution hotspots, assessment of health risks, and informing policy and mitigation strategies.

Global estimates of ambient NO2 concentrations and long-term health effects during 2000–2019

High concentrations of ambient NO2 causes serious air pollution and could also pose great threats to human health. However, the long-term trends (20-year) and potential health effects of ambient NO2 exposure globally still shows high uncertainties. In this work, the field measurements, satellite dataset, GEOS-Chem output, and multiple geographical covariates were incorporated into the multi-stage model to investigate the global evolutions of ambient NO2 during 2000–2019. The results indicated that the cross-validation (CV) R2 values of ambient NO2 based on multi-stage model displayed satisfied performance (R2 = 0.78), which was superior to the individual model. Besides, the out-of-bag R2 was 0.75, which suggested the multi-stage model showed the better transferability. At the spatial scale, the NO2 concentrations followed the order of China (16.9 ± 9.0 μg/m3) > India (15.5 ± 5.6 μg/m3) > United States (10.7 ± 5.6 μg/m3) > Europe (7.7 ± 4.5 μg/m3), which was in consistent with the anthropogenic NOx emission. At the temporal scale, the ambient NO2 levels in China experienced persistent increases (0.29 μg/m3/year) during 2000–2013, whereas they showed slight decreases (−0.23 μg/m3/year) during 2013–2019. The ambient NO2 levels in the United States experienced continuous decreases during 2000–2019 (−0.20 μg/m3/year), while both of India and Europe remained relatively stable. Long-term NO2 exposure inevitably increased premature mortalities. The global premature all-cause mortalities associated with the excessive NO2 exposure increased from 288,169 (95% CI: 43,650, 527,971) to 461,301 (95% CI: 69,973, 843,996) in the past 20 years. This study would provide sufficient policy support for future ambient NO2 mitigation.

Satellite-Based Reconstruction of Atmospheric CO2 Concentration over China Using a Hybrid CNN and Spatiotemporal Kriging Model

Although atmospheric CO2 concentrations collected by satellites play a crucial role in understanding global greenhouse gases, the sparse geographic distribution greatly affects their widespread application. In this paper, a hybrid CNN and spatiotemporal Kriging (CNN-STK) model is proposed to generate a monthly spatiotemporal continuous XCO2 dataset over China at 0.25° grid-scale from 2015 to 2020, utilizing OCO-2 XCO2 and geographic covariates. The validations against observation samples, CAMS XCO2 and TCCON measurements indicate the CNN-STK model is effective, robust, and reliable with high accuracy (validation set metrics: R2 = 0.936, RMSE = 1.3 ppm, MAE = 0.946 ppm; compared with TCCON: R2 = 0.954, RMSE = 0.898 ppm and MAE = 0.741 ppm). The accuracy of CNN-STK XCO2 exhibits spatial inhomogeneity, with higher accuracy in northern China during spring, autumn, and winter and lower accuracy in northeast China during summer. XCO2 in low-value-clustering areas is notably influenced by biological activities. Moreover, relatively high uncertainties are observed in the Qinghai-Tibet Plateau and Sichuan Basin. This study innovatively integrates deep learning with the geostatistical method, providing a stable and cost-effective approach for other countries and regions to obtain regional scales of atmospheric CO2 concentrations, thereby supporting policy formulation and actions to address climate change.

Rising risk and localized patterns of Indian Summer Monsoon rainfall extremes

Extreme precipitation during the Indian Summer Monsoon period often leads to heavy flooding, posing risks to infrastructure and human life across India. Modelling such phenomena requires developing spatio-temporal models that incorporate spatial dependence for a more realistic representation of spatial variation over a large domain. In this study, we employed a multivariate max-stable process-based model to capture the joint spatial dependence and to estimate the impact of geographic and temporal covariates on the risk associated with anomalous rainfall. Data pooling and local optimization at the cluster level help to reduce the uncertainty in the parameter estimates, whereas the incorporation of El Niño Southern Oscillations (ENSO) enhances the predictive power of the models. The results indicate that approximately 75% of India's land area is at risk of experiencing heavy rainfall with a 50-year or more return period. The proposed max-stable process model adequately captured the nonstationary behaviour of Indian Summer Monsoon Rainfall (ISMR) extremes. A post-hoc analysis was conducted to assess the practical significance of differences in estimated parameters. This is essential for understanding the real-time variations in the extremal behaviour of ISMR across the country. In addition, the phase effect of ENSO and spatial scale (cluster/region) changes were quantified using a linear mixed-effects model. The results suggest that the impact of ENSO becomes more pronounced for longer return periods and intensification of ISMR extremes occurs due to the shift from El Niño to La Niña phases. Moreover, treating spatial scales as a random component in the modelling reveals that ISMR extremes are highly localized in nature. Specifically, around 80% of the variation in ISMR extremes for shorter return levels is attributed to the nested spatial variation.

Health Care Access and COVID-19 Vaccination in the United States: A Cross-Sectional Analysis.

Although federal legislation made COVID-19 vaccines free, inequities in access to medical care may affect vaccine uptake. To assess whether health care access was associated with uptake and timeliness of COVID-19 vaccination in the United States. A cross-sectional study. 2021 National Health Interview Survey (Q2-Q4). In all, 21,532 adults aged≥18 were included in the study. Exposures included 4 metrics of health care access: health insurance, having an established place for medical care, having a physician visit within the past year, and medical care affordability. Outcomes included receipt of 1 or more COVID-19 vaccines and receipt of a first vaccine within 6 months of vaccine availability. We examined the association between each health care access metric and outcome using logistic regression, unadjusted and adjusted for demographic, geographic, and socioeconomic covariates. In unadjusted analyses, each metric of health care access was associated with the uptake of COVID-19 vaccination and (among those vaccinated) early vaccination. In adjusted analyses, having health coverage (adjusted odds ratio [AOR] 1.60; 95% CI: 1.39, 1.84), a usual place of care (AOR 1.58; 95% CI: 1.42, 1.75), and a doctor visit within the past year (AOR 1.45, 95% CI: 1.31, 1.62) remained associated with higher rates of COVID-19 vaccination. Only having a usual place of care was associated with early vaccine uptake in adjusted analyses. Receipt of COVID-19 vaccination was self-reported. Several metrics of health care access are associated with the uptake of COVID-19 vaccines. Policies that achieve universal coverage, and facilitate long-term relationships with trusted providers, may be an important component of pandemic responses.

Leveraging low-cost sensors to predict nitrogen dioxide for epidemiologic exposure assessment.

Statistical models of air pollution enable intra-urban characterization of pollutant concentrations, benefiting exposure assessment for environmental epidemiology. The new generation of low-cost sensors facilitate the deployment of dense monitoring networks and can potentially be used to improve intra-urban models of air pollution. Develop and evaluate a spatiotemporal model for nitrogen dioxide (NO2) in the Puget Sound region of WA, USA for the Adult Changes in Thought Air Pollution (ACT-AP) study and assess the contribution of low-cost sensor data to the model's performance through cross-validation. We developed a spatiotemporal NO2 model for the study region incorporating data from 11 agency locations, 364 supplementary monitoring locations, and 117 low-cost sensor (LCS) locations for the 1996-2020 time period. Model features included long-term time trends and dimension-reduced land use regression. We evaluated the contribution of LCS network data by comparing models fit with and without sensor data using cross-validated (CV) summary performance statistics. The best performing model had one time trend and geographic covariates summarized into three partial least squares components. The model, fit with LCS data, performed as well as other recent studies (agency cross-validation: CV- root mean square error (RMSE) = 2.5 ppb NO2; CV- coefficient of determination ( ) = 0.85). Predictions of NO2 concentrations developed with LCS were higher at residential locations compared to a model without LCS, especially in recent years. While LCS did not provide a strong performance gain at agency sites (CV-RMSE = 2.8 ppb NO2; CV- = 0.82 without LCS), at residential locations, the improvement was substantial, with RMSE = 3.8 ppb NO2 and = 0.08 (without LCS), compared to CV-RMSE = 2.8 ppb NO2 and CV- = 0.51 (with LCS). We developed a spatiotemporal model for nitrogen dioxide (NO2) pollution in Washington's Puget Sound region for epidemiologic exposure assessment for the Adult Changes in Thought Air Pollution study. We examined the impact of including low-cost sensor data in the NO2 model and found the additional spatial information the sensors provided predicted NO2 concentrations that were higher than without low-cost sensors, particularly in recent years. We did not observe a clear, substantial improvement in cross-validation performance over a similar model fit without low-cost sensor data; however, the prediction improvement with low-cost sensors at residential locations was substantial. The performance gains from low-cost sensors may have been attenuated due to spatial information provided by other supplementary monitoring data.

A framework to evaluate the impact of a hazard chain and geographical covariates on spatial extreme water levels: A case study in the Pearl River Delta

The interactions and collective impacts of different types of hazards within a compound hazard system, along with the influence of geographical covariates on flooding are presently unclear. Understanding these relationships is crucial for comprehending the formation and dynamic processes of the hazard chain and improving the ability to identify flood warning signals in complex hazard scenarios. In this study, we presented a multivariate spatial extreme value hierarchical (MSEVH) framework to assess the spatial extreme water levels (EWL) at different return levels under the influence of a hazard chain and geographical covariates. The Pearl River Delta (PRD) was selected as a research example to assess the effectiveness of the MSEVH framework. Firstly, we identified a hazard chain (extreme streamflow from the Xijiang River (XR) - extreme streamflow from the Beijiang River (BR) - extreme sea level) and three geographical covariates influencing EWL in the PRD. Then, we compared four hazard scenarios in the MSEVH framework to evaluate the spatial EWL at different return levels under the influence of the hazard chain in the PRD. The final step involves assessing spatial EWL with the effect of the hazard chain and geographical covariates. The results indicate that when extreme streamflow from XR and BR occurs concurrently, the extreme streamflow from BR weakens the influence of extreme streamflow from XR on EWL in the PRD. However, it cannot fully offset the overall impact of extreme streamflow from XR on EWL. In addition, when extreme streamflow from XR, extreme streamflow from BR, and extreme sea level occur simultaneously, the extreme sea level enhances the influence of concurrent extreme streamflow from XR and BR on EWL in the PRD. The proposed MSEVH is not only applicable to the PRD but also shows promising potential for evaluating extreme hydrometeorological variables under the influence of other hazard chains.

Integrating traffic pollution dispersion into spatiotemporal NO2 prediction

Accurately predicting ambient NO2 concentrations has great public health importance, as traffic-related air pollution is of major concern in urban areas. In this study, we present a novel approach incorporating traffic contribution to NO2 prediction in a fine-scale spatiotemporal model. We used nationally available traffic estimate dataset in a scalable dispersion model, Research LINE source dispersion model (RLINE). RLINE estimates then served as an additional input for a validated spatiotemporal pollution modeling approach. Our analysis uses measurement data collected by the Multi-Ethnic Study of Atherosclerosis and Air Pollution in the greater Los Angeles area between 2006 and 2009. We predicted road-type-specific annual average daily traffic (AADT) on road segments via national-level spatial regression models with nearest-neighbor Gaussian processes (spNNGP); the spNNGP models were trained based on over half a million point-level traffic volume measurements nationwide. AADT estimates on all highways were combined with meteorological data in RLINE models. We evaluated two strategies to integrate RLINE estimates into spatiotemporal NO2 models: 1) incorporating RLINE estimates as a space-only covariate and, 2) as a spatiotemporal covariate. The results showed that integrating the RLINE estimates as a space-only covariate improved overall cross-validation R2 from 0.83 to 0.84, and root mean squared error (RMSE) from 3.58 to 3.48 ppb. Incorporating the estimates as a spatiotemporal covariate resulted in similar model improvement. The improvement of our spatiotemporal model was more profound in roadside monitors alongside highways, with R2 increasing from 0.56 to 0.66 and RMSE decreasing from 3.52 to 3.11 ppb. The observed improvement indicates that the RLINE estimates enhanced the model's predictive capabilities for roadside NO2 concentration gradients even after considering a comprehensive list of geographic covariates including the distance to roads. Our proposed modeling framework can be generalized to improve high-resolution prediction of NO2 exposure – especially near major roads in the U.S.

Geese migrating over the Pacific Ocean select altitudes coinciding with offshore wind turbine blades

Abstract Renewable energy facilities are a key part of mitigating climate change, but can pose threats to wild birds and bats, most often through collisions with infrastructure. Understanding collision risk and the factors affecting it can help minimize impacts on wild populations. For wind turbines, flight altitude is a major factor influencing collision risk, and altitude‐selection analyses can evaluate when and why animals fly at certain altitudes under certain conditions. We used GPS tags to track Pacific Flyway geese (Pacific greater white‐fronted goose, tule greater white‐fronted goose and lesser snow goose) on transoceanic migrations between Alaska and the Pacific Coast of the contiguous United States, an area where offshore windfarm development is beginning. We evaluated how geographic and environmental covariates affected (1) whether birds were at rest on the water versus in flight (binomial model) and (2) altitude selection when in flight (similar to a step‐selection framework). We then used a Monte Carlo simulation to predict the probability of flying at each altitude under various conditions, considering both the fly/rest decision and altitude selection. In both spring and fall, geese showed strong selection for altitudes within the expected rotor‐swept zone (20–200 m asl), with 56% of locations expected to be within the rotor‐swept zone under mean daylight conditions and 28% at night. This indicates a high possibility that migrating geese may be at risk of collision when passing through windfarms. Although there was some variation across subspecies, geese were most likely to be within the rotor‐swept zone with little wind or light tailwinds, low clouds, little to no precipitation and moderate to cool air temperatures. Geese were unlikely to be in the rotor‐swept zone at night, when most individuals were at rest on the water. Synthesis and applications. These results could be used to inform windfarm management, including decisions to shut down turbines when collision risk is high. The altitude‐selection framework we demonstrate could facilitate further study of other bird species to develop a holistic view of how windfarms in this area could affect the migratory bird community as a whole.

Quantifying the availability of seasonal surface water and identifying the drivers of change within tropical forests in Cambodia.

Surface freshwater is a vital resource that is declining globally, predominantly due to climate and land use changes. Cambodia is no exception and the loss threatens many species, such as the giant ibis a Critically Endangered waterbird. We aimed to quantify the spatial and temporal (2000-2020) change of surface water availability across northern and eastern Cambodia and to assess the impact of this on the giant ibis. We used a Random Forest Classifier to determine the changes and we tested the impact of land use and geographical covariates using spatially explicit regression models. We found an overall reduction of surface water availability of 4.16%. This was predominantly driven by the presence of Economic Land Concessions and roads which increased the probability of extreme drying and flooding events. The presence of protected areas reduced these probabilities. We found changes in precipitation patterns over the wider landscape did not correlate with changes in surface water availability, supporting the overriding influence of land use change. 98% of giant ibis nests recorded during the time period were found within 25m of surface water during the dry season, highlighting their dependency on surface water. The overall surface water decline resulted in a 25% reduction in dry season suitable habitat for the giant ibis. Although absolute changes in surface water over the whole area were relatively small, the impact on the highest quality habitat for ibis is disproportionate and therefore threatens its populations. Defining the threats to such an endangered species is crucial for effective management.

Assessing the impact of COVID-19 lockdown on fine-scale air quality across a heavy-pollution city using low-cost sensors

COVID-19 pandemic led to restricted anthropogenic activities and drastic emission reduction, while air quality in response to strict control measures at a fine spatial scale still remained unknown. At first, we estimated the high-resolution air quality based on low-cost sensors, assimilated meteorology, and other geographical covariates. The result suggested that PM2.5, NO2, and SO2 across Handan decreased by 21%, 55%, and 12%, respectively. In contrast, the 8-h O3 level over Handan increased by 87%. Then, the random forest (RF) model was applied to remove the meteorology in order to capture the response of air quality to emission reduction at the fine spatial scale. The deweathered concentrations of some species varied greatly at different roads and points of interest (POI). The deweathered NO2 level suffered from more drastic reduction in some industrial points and roads compared with other POI, which was closely associated with the dense nitrogen oxide (NOx) emissions in these sites/roads. However, the deweathered O3 concentrations in some main lines (e.g., highway and sidewalk) showed more significant increases, which might be due to more rapid decline of NOx emission compared with volatile organic compounds (VOC) emission in a VOC-limited region.

FRI526 Geographic Trends In Thyroid Hormone Prescribing In The US: 2010 - 2020

Abstract Disclosure: M.D. Ettleson: None. N. Laiteerapong: None. W. Wan: None. A.C. Bianco: Consulting Fee; Self; Abbvie, Acella, Synthonics, Thyron, Madrigal. Introduction: Levothyroxine (LT4) is the first line treatment of hypothyroidism. However, some clinical guidelines suggest a trial of adding triiodothyronine (T3) therapy if LT4 monotherapy does not adequately resolve symptoms despite a normal thyrotropin (TSH) level. It is unknown how patterns of thyroid hormone (TH) prescribing (LT4 vs T3-containing therapy) differ across the US. Methods: We conducted a cross-sectional study of TH prescription data from a large, US-based claims database from January 1, 2010 to December 31, 2020. TH type (LT4, liothyronine [LT3], or desiccated thyroid extract [DTE]), geographic, and clinical data were collected. Only patients with a new TH prescription and a diagnosis of hypothyroidism were included. The study population was subdivided into 4 TH treatment classes: 1) LT4 monotherapy, 2) LT3 therapy with or without LT4, 3) DTE therapy with or without LT4, and 4) other. Statewide physician density and urban-rural classification were the primary geographic covariates analyzed. Other covariates included in the analytical models were age, sex, comorbidities, and co-prescribed medications. Results: A total of 524,818 adult patients accounted for 537,594 new TH prescriptions over the 11 year study period. Of the total population, 89.0% received LT4 monotherapy, 8.8% received DTE therapy, and 2.0% received LT3 therapy. The proportion of patients receiving DTE increased from 5.4% to 10.2% during the study period. DTE and LT3 prescriptions were not uniformly distributed throughout the US, with higher rates in the West and South regions. Residence in a state with higher endocrinologist density (3.0/100k population) was associated with a decreased likelihood of receiving LT3 or DTE therapy (adjusted OR 0.33, 95% CI [0.20 – 0.55], p &amp;lt;0.001; adjusted OR 0.18, 95% CI [0.08 – 0.41], p &amp;lt;0.001, respectively). Residence in large central metro zones was associated with an increased likelihood of receiving LT3 or DTE therapy (adjusted OR 1.32, 95% CI [1.22 – 1.42]; p &amp;lt;0.001; adjusted OR 1.05, 95% CI [1.01 – 1.10], p = 0.008, respectively). Conclusions: There do appear to be geographic disparities in the patterns of TH prescribing in the US. Differences in access to endocrine care may contribute to these disparities. Also, patients in more urban zones were more likely to receive LT3 or DTE therapies. Further investigation is necessary to understand the patient and physician-level factors underlying the geographic disparities identified. Presentation: Friday, June 16, 2023

Role of geographic characteristics in the spatial cluster detection of cancer: Evidence in South Korea, 1999–2013

BackgroundIdentification of high-risk areas of cancer, referred to as spatial clusters, can inform targeted policies for cancer control. Although cancer cluster detection could be affected by various geographic characteristics including sociodemographic and environmental factors which impacts could also vary over time, studies accounting for such influence remain limited. This study aims to assess the role of geographic characteristics in the spatial cluster detection for lung and stomach cancer over an extended period. MethodsWe obtained sex-specific age-standardized incidence and mortality rates of lung and stomach cancer as well as geographic characteristics across 233 districts in South Korea for three five-year periods between 1999 and 2013. We classified geographic characteristics of each district into four categories: demography, socioeconomic status, behaviors, and physical environments. Specifically, we quantified physical environments using measures of greenness, concentrations of particulate matter and nitrogen dioxide, and air pollution emissions. Finally, we conducted cluster detection analyses using weighted normal spatial scan statistics with the residuals from multiple regression analyses performed with the four progressive sets of geographic attributes. ResultsWe found that the size of clusters reduced as we progressively adjusted for geographic covariates. Among the four categories, physical environments had the greatest impact on the reduction or disappearance of clusters particularly for lung cancer consistently over time. Whereas older population affected a decrease of lung cancer clusters in the early period, the contribution of education was large in the recent period. The impact was less clear in stomach cancer than lung cancer. ConclusionOur findings highlight the importance of geographic characteristics in explaining the existing cancer clusters and identifying new clusters, which jointly provides practical guidance to cancer control.

A hierarchical Bayesian non‐asymptotic extreme value model for spatial data

AbstractSpatial maps of extreme precipitation are crucial in flood prevention. With the aim of producing maps of precipitation return levels, we propose a novel approach to model a collection of spatially distributed time series where the asymptotic assumption, typical of the traditional extreme value theory, is relaxed. We introduce a Bayesian hierarchical model that accounts for the possible underlying variability in the distribution of event magnitudes and occurrences, which are described through latent temporal and spatial processes. Spatial dependence is characterized by geographical covariates and effects not fully described by the covariates are captured by spatial structure in the hierarchies. The performance of the approach is illustrated through simulation studies and an application to daily rainfall extremes across North Carolina (USA). The results show that we significantly reduce the estimation uncertainty with respect to state of the art techniques.

Optimizing Substance Use Treatment Selection Using Reinforcement Learning

Substance use disorder (SUD) exacts a substantial economic and social cost in the United States, and it is crucial for SUD treatment providers to match patients with feasible, effective, and affordable treatment plans. The availability of large SUD patient datasets allows for machine learning techniques to predict patient-level SUD outcomes, yet there has been almost no research on whether machine learning can be used to optimize or personalize which treatment plans SUD patients receive. We use contextual bandits (a reinforcement learning technique) to optimally map patients to SUD treatment plans, based on dozens of patient-level and geographic covariates. We also use near-optimal policies to incorporate treatments’ time-intensiveness and cost into our recommendations, to aid treatment providers and policymakers in allocating treatment resources. Our personalized treatment recommendation policies are estimated to yield higher remission rates than observed in our original dataset, and they suggest clinical insights to inform future research on data-driven SUD treatment matching.

Multivariate assessment of vaccine equity in Nigeria: A VERSE tool case study using demographic and health survey 2018

Nigeria experiences wide heterogeneity in vaccination rates by vaccine and region. However, inequities in vaccination status extend beyond just geographic covariates. Traditionally, inequity is represented by a single metric pertaining to socioeconomic status. A growing body of literature suggests that this view is limiting, and a multi-factor approach is necessary to comprehensively evaluate relative disadvantage between individuals.The Vaccine Economics Research for Sustainability and Equity (VERSE) tool produces a composite equity metric, which accounts for multiple factors influencing inequity in vaccination coverage. We apply the VERSE tool to Nigeria’s 2018 Demographic and Health Survey (DHS) to cross-sectionally evaluate equity in vaccination status for national immunization program (NIP) vaccines over the following contributing covariates: age of child, sex of child, maternal education level, socioeconomic status, health insurance status, state of residence, and urban or rural designation. We also assess equity for zero-dose, fully immunized for age, and completion of NIP.Results show that socioeconomic status contributes substantially to variation vaccination coverage, but it is not the most substantial factor. For all vaccination statuses, except for NIP completion, maternal education level is the greatest contributor towards a child’s immunization status among model variables. We highlight the outputs for zero-dose, fully immunized at infancy, MCV1 and PENTA1. The percentage point gap in vaccination status between the top and bottom quintiles of disadvantage, as ranked by the composite indicator is 31.1 (29.5–32.7) for zero-dose status, 53.1 (51.3–54.9) for full immunization status, 48.9 (46.9–50.9) for MCV1, and 67.6 (66.0–69.2) for PENTA1. Though concentration indices indicate inequity for all statuses, full immunization coverage is very low at 31.5% suggesting significant gaps in reaching children after initial doses for routine immunizations. Applying the VERSE tool to future Nigeria DHS surveys can allow decisionmakers to track changes in vaccination coverage equity, in a standardized manner, over time.

Identifying Predictors of Homelessness Among Adults in a Large Integrated Health System in Northern California.

Introduction Homelessness contributes to worsening health and increased health care costs. There is little published research that leverages rich electronic health record (EHR) data to predict future homelessness risk and inform interventions to address it. The authors' objective was to develop a model for predicting future homelessness using individual EHR and geographic data covariates. Methods This retrospective cohort study included 2,543,504 adult members (≥ 18 years old) from Kaiser Permanente Northern California and evaluated which covariates predicted a composite outcome of homelessness status (hospital discharge documentation of a homeless patient, medical diagnosis of homelessness, approved medical financial assistance application for homelessness, and/or "homeless/shelter" in address name). The predictors were measured in 2018-2019 and included prior diagnoses and demographic and geographic data. The outcome was measured in 2020. The cohort was split (70:30) into a derivation and validation set, and logistic regression was used to model the outcome. Results Homelessness prevalence was 0.35% in the overall sample. The final logistic regression model included 26 prior diagnoses, demographic, and geographic-level predictors. The regression model using the validation set had moderate sensitivity (80.4%) and specificity (83.2%) for predicting future cases of homelessness and achieved excellent classification properties (area under the curve of 0.891 [95% confidence interval = 0.884-0.897]). Discussion This prediction model can be used as an initial triage step to enhance screening and referral tools for identifying and addressing homelessness, which can improve health and reduce health care costs. Conclusions EHR data can be used to predict chance of homelessness at a population health level.

The heterogeneity of the homebound: A latent class analysis of a national sample of homebound older adults.

Homebound status is a final common pathway for people with a variety of diseases and conditions. There are 7 million homebound older adults in the United States. Despite concerns regarding their high healthcare costs and utilization and limited access to care, the unique subsets within the homebound population are understudied. Better understanding of distinct homebound groups may enable more targeted and tailored approaches to care delivery. Therefore, in a nationally representative sample of homebound older adults we used latent class analysis (LCA) to examine distinct homebound subgroups based on clinical and sociodemographic characteristics. Using data from the National Health and Aging Trends Study (NHATS) 2011-2019, we identified 901 newly homebound persons (defined as never/rarely leaving home or leaving home only with assistance and/or difficulty). Sociodemographic, caregiving context, health and function, and geographic covariates were derived from NHATS via self-report. LCA was used to identify the existence of distinct subgroups within the homebound population. Indices of model fit were compared for models testing 1-5 latent classes. Association between latent class membership and 1 year mortality was examined using a logistic regression. We identified four classes of homebound individuals differentiated by their health, function, sociodemographic characteristics, and caregiving context: (i) Resource constrained (n=264); (ii) Multimorbid/high symptom burden (n=216); (iii) Dementia/functionally impaired (n=307); (iv) Older/assisted living (n=114). One year mortality was highest among the older/assisted living subgroup (32.4%) and lowest among the resource constrained (8.2%). This study identifies subgroups of homebound older adults characterized by distinct sociodemographic and clinical characteristics. These findings will support policymakers, payers, and providers in targeting and tailoring care to the needs of this growing population.