Random Effect Term Research Articles

Impact of Areal Factors on Students’ Travel Mode Choices: A Bayesian Spatial Analysis

A preliminary analysis of the 2018/2019 Austin Travel Survey indicated that most off-campus students in Travis County, TX, tend to use cars rather than more sustainable transportation modes, significantly contributing to traffic congestion and environmental impact. This study aims to analyze the impacts of areal factors, including environmental and transportation factors, on students’ choices of travel mode in order to promote more sustainable transport behaviors. Additionally, we investigate the presence of spatial correlation and unobserved heterogeneity in travel data and their effects on students’ travel mode choices. We have proposed two Bayesian models—a basic model and a spatial model—with structured and unstructured random-effect terms to perform the analysis. The results indicate that the inclusion of spatial random effects considerably improves model performance, suggesting that students’ choices of mode are likely influenced by areal factors often ‘unobserved’ in many individual travel mode choice surveys. Furthermore, we found that the average slope, sidewalk density, and bus-stop density significantly affect students’ travel mode choices. These findings provide insights into promoting sustainable transport systems by addressing environmental and infrastructural factors in an effort to reduce car dependency among students, thereby supporting sustainable urban development.

Estimating the Impact of Hospital-Level Variation on the Use of Inpatient Rehabilitation Facilities Versus Skilled Nursing Facilities on Individual Patients With Stroke.

There is substantial hospital-level variation in the use of Inpatient Rehabilitation Facilities (IRFs) versus Skilled Nursing Facilities (SNFs) among patients with stroke, which is poorly understood. Our objective was to quantify the net effect of the admitting hospital on the probability of receiving IRF or SNF care for individual patients with stroke. Using Medicare claims data (2011-2013), a cohort of patients with acute stroke discharged to an IRF or SNF was identified. We generated 2 multivariable logistic regression models. Model 1 predicted IRF admission (versus SNF) using only patient-level factors, whereas model 2 added a hospital random effect term to quantify the hospital effect. The statistical significance and direction of the random effect terms were used to categorize hospitals as being either IRF-favoring, SNF-favoring, or neutral with respect to their discharge patterns. The hospital's impact on individual patient's probability of IRF discharge was estimated by taking the change in individual predicted probabilities (change in individual predicted probability) between the 2 models. Hospital-level effects were categorized as small (<10%), moderate (10%-19%), or large (≥20%) depending on change in individual predicted probability. The cohort included 135 415 patients (average age, 81.5 [SD=8.0] years, 61% female, 91% ischemic stroke) who were discharged from 1816 acute care hospitals to IRFs (n=66 548) or SNFs (n=68 867). Half of hospitals were classified as being either IRF-favoring (n=461, 25.4%) or SNF-favoring (n=485, 26.7%) with the remainder (n=870, 47.9%) considered neutral. Overall, just over half (n=73 428) of patients were treated at hospitals that had moderate or large independent effects on discharge settings. Hospital effects for neutral hospitals were small (ie, change in individual predicted probability <10%) for most patients (72.5%). However, hospital effects were moderate or large for 78.8% and 84.6% of patients treated at IRF- or SNF-favoring hospitals, respectively. For most patients with stroke, the admitting hospital meaningfully changed the type of rehabilitation care that they received.

Bayesian Shared Component Spatial Modeling for Assessing the Shared and Age Group-Specific Mental Health Disorder Risk of Young and Old Age Groups: A Case Study of Toronto Neighborhoods, Canada

Mental health disorder risks of young and old age groups hold considerable importance for understanding present and future risk burdens. However, assessing mental health risks is significantly constrained by the influence of shared and age group-specific spatial processes and risk factors. Therefore, this study employed Bayesian shared component spatial modeling (BSCSM) to analyze mental health disorder data obtained from young (20–44 years) and old (65+ years) age groups in Toronto. BSCSM was employed to model the shared and age group-specific disorder risk and to identify hotspot areas. The unmeasured covariates, overdispersion, and latent spatial processes were adjusted using spatial and non-spatial random effect terms. The findings from BSCSM were finally compared with non-shared component modeling approaches. The results suggest that over 60% of variations in mental health disorder risk for both age groups could be explained by the shared component. The high-risk neighborhoods were mainly localized in southern and north-central Toronto for the young and old age groups. Deviance information criterion values suggested that models from BSCSM outperformed non-BSCSM models. BSCSM risk maps were also better at identifying high-risk areas. This work demonstrated that both shared and age group-specific risks are essential for assessing mental health disorder risk and devising targeted interventions.

International Variation in Severe Exacerbation Rates in Patients With Severe Asthma

BackgroundExacerbation frequency strongly influences treatment choices in patients with severe asthma. Research QuestionWhat is the extent of the variability of exacerbations rate across countries and its implications in disease management? Study Design and MethodsWe retrieved data from the International Severe Asthma Registry, an international observational cohort of patients with a clinical diagnosis of severe asthma. We identified patients aged ≥18 years who did not initiate any biologics prior to baseline visit. A severe exacerbation was defined as the use of oral corticosteroids for ≥3 days or asthma-related hospitalization/emergency room visit. A series of negative binomial models were applied to estimate country-specific severe exacerbation rates during 365 days of follow-up, starting from a naïve model with country as the only variable, to an adjusted model with country as a random-effect term and patient and disease characteristics as independent variables. ResultsThe final sample included 7,510 patients from 17 countries (56% from the United States), contributing to 1,939 severe exacerbations (0.27/person-year). There was large between-country variation in observed severe exacerbation rate (min: 0.04 [Argentina], max:0.88 [Saudi Arabia], interquartile range [IQR]: 0.13–0.54), which remained substantial after adjusting for patient characteristics and sampling variability (IQR: 0.16–0.39). InterpretationIndividuals with similar patient characteristics but coming from different jurisdictions have varied severe exacerbation risks, even after controlling for patient and disease characteristics. This suggests unknown patient factors or system-level variations at play. Disease management guidelines should recognize such between-country variability. Risk prediction models that are calibrated for each jurisdiction will be needed to optimize treatment strategies. Clinical Trial Registration NumberN/A

Digital pathology for reporting histopathology samples, including cancer screening samples - definitive evidence from a multisite study.

To conduct a definitive multicentre comparison of digital pathology (DP) with light microscopy (LM) for reporting histopathology slides including breast and bowel cancer screening samples. A total of 2024 cases (608 breast, 607 GI, 609 skin, 200 renal) were studied, including 207 breast and 250 bowel cancer screening samples. Cases were examined by four pathologists (16 study pathologists across the four speciality groups), using both LM and DP, with the order randomly assigned and 6 weeks between viewings. Reports were compared for clinical management concordance (CMC), meaning identical diagnoses plus differences which do not affect patient management. Percentage CMCs were computed using logistic regression models with crossed random-effects terms for case and pathologist. The obtained percentage CMCs were referenced to 98.3% calculated from previous studies. For all cases LM versus DP comparisons showed the CMC rates were 99.95% [95% confidence interval (CI) = 99.90-99.97] and 98.96 (95% CI = 98.42-99.32) for cancer screening samples. In speciality groups CMC for LM versus DP showed: breast 99.40% (99.06-99.62) overall and 96.27% (94.63-97.43) for cancer screening samples; [gastrointestinal (GI) = 99.96% (99.89-99.99)] overall and 99.93% (99.68-99.98) for bowel cancer screening samples; skin 99.99% (99.92-100.0); renal 99.99% (99.57-100.0). Analysis of clinically significant differences revealed discrepancies in areas where interobserver variability is known to be high, in reads performed with both modalities and without apparent trends to either. Comparing LM and DP CMC, overall rates exceed the reference 98.3%, providing compelling evidence that pathologists provide equivalent results for both routine and cancer screening samples irrespective of the modality used.

Clustering algorithm for the connected model of repeated measurements and survival data: Application to HIV study

The cluster analysis is dividing the individuals into clusters or groups. This job is valuable and helpful to provide facts and information about individuals. Here in this paper we proposed the cluster analysis for special model. It is the connected model for repeated measures called longitudinal and survival for individuals or patients. The statistical analysis of the connected model from the longitudinal and survival datasets becoming popular recently because it comes together in many medical applications. Here in this study, we utilized two statistical methods, cluster analysis and, connect the two models from longitudinal and survival data. It is beneficial since it gather the information from the repeated measurements, and survival responses. The shared random effect term is used joint multivariate Gaussian distribution (longitudinal) and Cox proportional hazard model (survival) for the same patients. Then, the pseudo-likelihood algorithm (clustering methodology) is performed for the joint model to distinguish the clusters or groups of patents. The application is HIV patient’s dataset with CD4+ counts responses and time to death (survival data) with some independent variables as gender and drug treatments. We conducted the clustering for S=2, and estimate the parameters from the longitudinal and survival models with and without clustering, and compare the estimations. Our results showed the generated clusters are different from each other, the estimation parameters be located around the original estimations (without clustering). It is helpful methodology to identify distinct groups or clusters from population. Finally, there is a big need for this type of application in medical elds.

Comparison of new computational methods for spatial modelling of malaria

BackgroundGeostatistical analysis of health data is increasingly used to model spatial variation in malaria prevalence, burden, and other metrics. Traditional inference methods for geostatistical modelling are notoriously computationally intensive, motivating the development of newer, approximate methods for geostatistical analysis or, more broadly, computational modelling of spatial processes. The appeal of faster methods is particularly great as the size of the region and number of spatial locations being modelled increases.MethodsThis work presents an applied comparison of four proposed ‘fast’ computational methods for spatial modelling and the software provided to implement them—Integrated Nested Laplace Approximation (INLA), tree boosting with Gaussian processes and mixed effect models (GPBoost), Fixed Rank Kriging (FRK) and Spatial Random Forests (SpRF). The four methods are illustrated by estimating malaria prevalence on two different spatial scales—country and continent. The performance of the four methods is compared on these data in terms of accuracy, computation time, and ease of implementation.ResultsTwo of these methods—SpRF and GPBoost—do not scale well as the data size increases, and so are likely to be infeasible for larger-scale analysis problems. The two remaining methods—INLA and FRK—do scale well computationally, however the resulting model fits are very sensitive to the user’s modelling assumptions and parameter choices. The binomial observation distribution commonly used for disease prevalence mapping with INLA fails to account for small-scale overdispersion present in the malaria prevalence data, which can lead to poor predictions. Selection of an appropriate alternative such as the Beta-binomial distribution is required to produce a reliable model fit. The small-scale random effect term in FRK overcomes this pitfall, but FRK model estimates are very reliant on providing a sufficient number and appropriate configuration of basis functions. Unfortunately the computation time for FRK increases rapidly with increasing basis resolution.ConclusionsINLA and FRK both enable scalable geostatistical modelling of malaria prevalence data. However care must be taken when using both methods to assess the fit of the model to data and plausibility of predictions, in order to select appropriate model assumptions and parameters.

PSIX-17 Preliminary Assessment of Preweaning Growth Performance of Dexter and Mashona X Dexter Calves

Abstract Non-traditional cattle breeds are often considered by managers for niche beef production and marketing systems. Some of these alternative breeds are small-framed and thus desired by small-farm operators but have received little research attention because they do not fit traditional commodity beef production systems. Very few of these small breeds also offer heat tolerance. This project was designed to evaluate preweaning performance of straightbred (n = 76) and F1 crossbred calves (n = 81) produced by mature Dexter cows across three years and two Tennessee State University research farms. Spring-born calves were sired by Dexter (n = 7; DD) and Mashona bulls (n = 6; MD) in single-sire, natural service breeding groups following 60-d breeding seasons. Breed of sire was balanced across cow age and farm locations. Calves were weighed within 24-h of birth, at 4 months (125 d of age) of age, and at weaning (227 d of age). Calves were not creep-fed and bull calves were not castrated until after weaning. No cases of dystocia were observed as a result of crossbreeding. Calf performance traits were evaluated using MIXED models of SAS. Fixed effects included sire breed, sex, farm location, and the interaction of sire breed and sex. Random effect terms were production year and sire nested within sire breed. For all the traits evaluated, models were consistent in that the sire breed x calf sex interaction was not significant (P &amp;gt; 0.1), whereas sire breed and calf sex were consistently important (P &amp;lt; 0.01) sources of variation. Bull calves had greater weight and ADG values than heifer calves from birth to weaning. Mean body weight was heavier for MD than DD calves at birth (28.4 ± 0.8 vs. 23.9 ± 0.7 kg), at 4 months of age (120.6 ± 9.7 vs. 96.7 ± 9.6 kg), and at weaning (154 ± 8.6 vs. 124.4 ± 8.6 kg). The MD calves had greater ADG than DD calves from birth to 4 months of age (715.8 ± 36.1 vs. 599.1 ± 36 g/d), from 4 months of age to weaning (334 ± 44.3 vs. 256.8 ± 44 g/d), and from birth to weaning (547 ± 35.7 vs. 447.1 ± 35.7 g/d). Mean 205-d adjusted body weight was heavier for MD than DD (140.2 ± 7.6 vs. 114.7 ± 7.6 kg). Preliminary results indicated that crossbreeding with Mashona bulls significantly increased preweaning growth of calves produced by Dexter cows, while the F1 calves maintained small statute. The spring-born MD F1 calves likely benefited from heat tolerance provided by the Mashona influence and heterosis generated by crossbreeding.

Effectiveness of COVID-19 vaccines against hospitalisation in Latin America during three pandemic waves, 2021–2022: a test-negative case-control design

Vaccine effectiveness (VE) is essential to monitor the performance of vaccines and generate strategic information to guide decision making. We pooled data from six Latin American countries to estimate the effectiveness of COVID-19 vaccines in preventing laboratory-confirmed SARS-CoV-2 hospitalisation during three different pandemic waves from February 2021 to September2022. We used a test-negative case-control design in hospitalised adults in Chile, Costa Rica, Ecuador, Guatemala, Paraguay, and Uruguay. We estimated adjusted VE by age group (18-64 and≥65 years), vaccine type and product for primary series vaccination and booster vaccination and by time since last dose during the Omicron variant dominant period. We used mixed effects logistic regression models adjusting for sex, age, week of onset of symptom onset and pre-existing conditions with country fit as a random effect term. We included 15,241 severe acute respiratory infection (SARI) patients in the analysis. Among adults 18-64 years, VE estimates for primary series vaccination during pre-Delta and Delta periods ranged by product from 66.5% to 95.1% and from 33.5% to 88.2% for older adults. During the Omicron period, VE estimates for primary series were lower and decreased by time since last vaccination, but VE increased to between 26.4% and 57.4% when a booster was administered. mRNA and viral vector vaccines presented higher VE for both primary series and booster. While VE decreased over time, protection against severe COVID-19-associated hospitalisation increased when booster doses were administered. Vaccination with additional doses should be recommended, particularly for persons at increased risk of developing severe COVID-19. This work was supported by a grant from the U.S. Centers for Disease Control and Prevention (CDC) through cooperative agreements with the Pan American Health Organization/World Health Organization.

Understanding the role of environmental and socioeconomic factors in the geographic variation of breast cancer risk in the US-wide Sister Study

ObjectiveTo describe the geographic pattern of breast cancer incidence in a nationwide prospective cohort and investigate whether environmental exposures and/or neighborhood socioeconomic status explain observed geographic disparities. MethodsUsing accelerated failure time models with a spatial random effect term, we mapped the health region-level association between residential location and breast cancer incidence for 44,707 participants in the Sister Study after controlling for established individual-level breast cancer risk factors. We performed a variable selection process to select environmental exposures [i.e., ambient nitrogen dioxide (NO2) and fine particulate matter (PM2.5), PM2.5 chemical composition, outdoor light at night (LAN), ambient noise, ultraviolet radiation, and greenspace] and neighborhood-level factors [i.e., population density and area deprivation index (ADI)] that predicted breast cancer incidence and quantified the spatial variation explained by the selected factors. We also considered whether the geographic pattern and predictors were similar when restricting to estrogen receptor-positive (ER+) tumors. ResultsWe observed a spatial patterning in the incidence of overall breast cancer (Moran's I = 16.7, p < 0.05) and ER+ breast cancer (Moran's I = 13.2, p < 0.05), with a lower risk observed in the South and Southeast and a greater risk in the Northwest and certain areas of the Midwest and Northeast. NO2, LAN, and ADI explained 21.4% of the spatial variation in overall breast cancer incidence whereas NO2, PM2.5 chemical composition, LAN, greenspace, and ADI together explained 63.3% of the spatial variation in ER+ breast cancer incidence. ConclusionsOur findings provide additional evidence for a role of environmental exposures in breast cancer incidence and suggest that geographic-based risk factors may vary according to breast cancer subtype. Our findings support the need for additional research to quantify the relative contributions of geographic-based risk factors for breast cancer.

Benefits and harms of perioperative high fraction inspired oxygen for surgical site infection prevention: a protocol for a systematic review and meta-analysis of individual patient data of randomised controlled trials

IntroductionThe use of high fraction of inspired oxygen (FiO2) intraoperatively for the prevention of surgical site infection (SSI) remains controversial. Promising results of early randomised controlled trials (RCT) have been...

Abstract 5 Umbilical Cord Blood Treatment to Improve Gross Motor Function in Individuals with Cerebral Palsy: Results from an Individual Participant Data Meta-Analysis

Abstract Introduction Umbilical cord blood (UCB) is a novel treatment for cerebral palsy (CP) with clinical trials indicating UCB is safe and can improve gross motor function. However, heterogeneity across research design, participant selection and treatment protocols has limited the ability to interpret overall treatment effect, identify best responders and dosing thresholds. Objectives We sought to conduct a systematic review and individual participant data meta-analysis (IPDMA) to 1) determine the effect of UCB treatment measured by Gross Motor Function Measure (GMFM) in individuals with CP, and 2) determine the effect of UCB cell dose on clinical improvement. Methods CENTRAL, Embase and Medline were systematically searched for clinical studies administering UCB to individuals with CP with or without a control group/comparator, collecting GMFM scores at baseline and 1, 3, 6 and/or 12 months. Individual de-identified participant data were obtained before being cleaned and coded. A one-stage IPDMA was conducted in R to obtain the pooled effect of UCB and the pooled cell dose effect via linear mixed models, including random effects terms to investigate between-study heterogeneity. Results Individual participant data from nine published and two unpublished studies (seven RCTs and four single-arm studies) were obtained, comprising 285 UCB (+/- erythropoietin, EPO)- treated participants and 171 controls for analysis. Median baseline age was 43 months (range 8-227). Most UCB infusions were allogeneic (80%), and 10/11 studies applied UCB intravenously. The median pre-thaw dose of intravenous-infused UCB was 46.7x10^6 total nucleated cells/kg (range 7.7-210.3). The pooled effect at 6 months post-intravenous UCB was an average change from baseline GMFM-66 of 0.93 (95% CI: 0.12, 1.75; p=0.027) points greater among UCB-treated participants (+/- EPO) compared to controls. There was also a statistically significant positive pooled cell dose effect at 6 months post-intervention (p=0.005), indicating that higher cell doses were associated with greater GMFM-66 increases. There was negligible between-study heterogeneity in both the treatment effect and the cell dose effect at 6 months. Discussion UCB improves gross motor function, and higher dose is beneficial. Future exploration of additional timepoints in addition to co-variates including age at infusion, severity, and etiology is underway and may elucidate responder subgroups and reveal increased efficacy.

Quantifying physician’s bias to terminate resuscitation. The TERMINATOR study

ContextDeciding on “termination of resuscitation” (TOR) is a dilemma for any physician facing cardiac arrest. Due to the lack of evidence-based criteria and scarcity of the existing guidelines, crucial arbitration to interrupt resuscitation remains at the practitioner’s discretion. AimEvaluate with a quantitative method the existence of a physician internal bias to terminate resuscitation. MethodWe extracted data concerning OHCAs managed between January 2013 and September 2021 from the RéAC registry. We conducted a statistical analysis using generalized linear mixed models to model the binary TOR decision. Utstein data were used as fixed effect terms and a random effect term to model physicians personal bias towards TOR. Results5,144 OHCAs involving 173 physicians were included. The cohort’s average age was 69 (SD 18) and was composed of 62% of women. Median no-flow and low-flow times were respectively 6 (IQR [0,12]) and 18 (IQR [10,26]) minutes. Our analysis showed a significant (p < 0.001) physician effect on TOR decision.Odds ratio for the “doctor effect” was 2.48 [2.13–2.94] for a doctor one SD above the mean, lower than that of dependency for activities of daily living (41.18 [24.69–65.50]), an age of more than 85 years (38.60 [28.67–51.08]), but higher than that of oncologic, cardiovascular, respiratory disease or no-flow duration between 10 to 20 minutes (1.60 [1.26–2.00]). ConclusionsWe demonstrate the existence of individual physician biases in their decision about TOR. The impact of this bias is greater than that of a no-flow duration lasting ten to twenty minutes. Our results plead in favor developing tools and guidelines to guide physicians in their decision.

A controlled evaluation of filter paper use during staining of sputum smears for tuberculosis microscopy.

Background: Some sputum smear microscopy protocols recommend placing filter paper over sputum smears during staining for Mycobacterium tuberculosis (TB) . We found no published evidence assessing whether this is beneficial. We aimed to evaluate the effect of filter paper on sputum smear microscopy results. Methods: Sputum samples were collected from 30 patients with confirmed pulmonary TB and 4 healthy control participants. From each sputum sample, six smears (204 smears in total) were prepared for staining with Ziehl-Neelsen (ZN), auramine or viability staining with fluorescein diacetate (FDA). Half of the slides subjected to each staining protocol were randomly selected to have Whatman grade 3 filter paper placed over the dried smears prior to stain application and removed prior to stain washing. The counts of acid-fast bacilli (AFB) and precipitates per 100 high-power microscopy fields of view, and the proportion of smear that appeared to have been washed away were recorded. Statistical analysis used a linear regression model adjusted by staining technique with a random effects term to correct for between-sample variability. Results: The inclusion of filter paper in the staining protocol significantly decreased microscopy positivity independent of staining with ZN, auramine or FDA (p=0.01). Consistent with this finding, there were lower smear grades in slides stained using filter paper versus without (p=0.04), and filter paper use reduced AFB counts by 0.28 logarithms (95% confidence intervals, CI=0.018, 0.54, p=0.04) independent of staining technique. In all analyses, auramine was consistently more sensitive with higher AFB counts versus ZN (p=0.001), whereas FDA had lower sensitivity and lower AFB counts (p<0.0001). Filter paper use was not associated with the presence of any precipitate (p=0.5) or the probability of any smear washing away (p=0.6) during the staining process. Conclusions: Filter paper reduced the sensitivity of AFB microscopy and had no detectable beneficial effects so is not recommended.

A deep mixed-effects modeling approach for real-time monitoring of metal additive manufacturing process

In-situ process monitoring of additive manufacturing (AM) is critically important and has been attracting significant efforts to achieve desirable process reliability and quality consistency. Despite enormous progress in embedding various sensors into the AM system, effectively fusing these sensor data for anomaly detection is very challenging, due to complex relationship among sensing signals, process condition and environment. In this article we propose a deep mixed-effects modeling approach to monitor the melt pool temperature for anomaly detection. It consists of a deep neural network (DNN) capturing the relationship between the temperature and other sensing data, a random-effect term accounting for the random variation of mean temperature, and a residual term modeling potential autocorrelations. To train the model (Phase I), an efficient hard expectation-maximization (EM) algorithm is developed, which iteratively optimizes the DNN parameters and the ones for the random-effect and residual terms. In online monitoring (Phase II), a T 2 -based and a generalized likelihood ratio (GLR) test-based control charts are developed to timely detect the process anomalies. The asymptotic behaviors of both the T 2 and GLR statistics are further established. The effectiveness of the proposed approach is demonstrated through thorough simulation study and real case study of an AM process.

Substantial decline in energy storage and stomach fullness in Antarctic minke whales (Balaenoptera bonaerensis) during the 1990s

A substantial decline in energy storage in the Antarctic minke whale during the 18-year Japanese JARPA research programme (austral summers 1987/88–2004/05) was reported in 2008 (Konishi et al., 2008). The statistical method used in the study was simple multiple linear regression. The results have since been thoroughly discussed by the Scientific Committee of the International Whaling Commission because of the potential importance of the findings. Some had suggested that the sampling heterogeneity in the JARPA data was so substantial that generalised linear models (GLMs) with interaction terms and random-effects terms should be explored. For the present article, five response variables related to energy storage and the variable ‘stomach content weight’ are systematically analysed using GLMs. For all five energy storage variables, the results show declines in the interval 3% to 9% over the JARPA period, all significantly different from zero at the 5% level, but no later decline. The weight of sieved stomach contents declined by 25% over the same period. The coefficients of the decline and the coefficients for most other independent variables were similar to values obtained by simple linear regression, but in some cases the standard errors were larger. The results indicate that important changes took place in the Antarctic ecosystem during the 1990s. It is hypothesised that the most important cause of the changes was the simultaneous increase in numbers of other krill feeders, especially humpback whales.

Abstract 76: Estimating The Impact Of Hospital-level Variation In The Use Of Inpatient Rehabilitation Facilities Versus Skilled Nursing Facilities On Individual Stroke Patients

Introduction: Observational studies indicate that stroke rehabilitation at Inpatient Rehabilitation Facilities (IRFs) is superior to Skilled Nursing Facilities (SNFs). Nationally, IRF vs. SNF use varies widely at the hospital level, but the impact on individual patients is unclear. Our objective was to quantify the effect of the admitting hospital on the probability of receiving IRF or SNF care for individual stroke patients. Methods: Medicare claims data (2011-2014) was used to generate a cohort of acute ischemic and hemorrhagic stroke patients who were discharged to IRFs or SNFs. We generated 2 multivariable effects logistic regression models. Model 1 predicted IRF admission (vs. SNF) using only patient-level factors as fixed effects, whereas Model 2 added a hospital random effect. The impact of the admitting hospital on an individual patients’ probability of IRF care was estimated by taking the difference in predicted probabilities (p^) between the 2 models. Hospital effects were categorized as moderate (10-19%) or large (&gt;20%) based on the change in size of p^. The magnitude and direction of the random effect terms in Model 2 was used to categorize individual hospitals as being either neutral, SNF-, or IRF- favoring. Results: The cohort included 1,816 acute care hospitals which discharged 135,415 patients to IRFs (n=66,548) or SNFs (n= 68,867). Half of the hospitals were categorized as neutral (n=870, 47.9%) with the remaining being SNF- favoring (n=485, 26.7%) or IRF- favoring (n=461, 25.4%). For half of all patients, acute care hospital had a moderate influence on the type of rehabilitation that they received. For SNF and IRF-favoring hospitals, there were large (55% and 37% of patients) or moderate (30% and 42% of patients) hospital effects on discharge setting for the majority of patients. Conclusion: For the majority of stroke patients, which acute care hospital they happen to be admitted to meaningfully impacts the type of rehabilitation care they receive

Effect of Temperature and Humidity on Milk Urea Nitrogen Concentration.

This study investigated the effect of ambient temperature and humidity on milk urea nitrogen (MUN) concentration in Holstein cows. Meteorological data corresponding to the dates of milk sampling were collected over six years. A linear mixed-effects model including a random effect term for cow identification was used to assess whether temperature and humidity were predictive of MUN concentration. Age, days in milk, temperature humidity index (THI), ration, milk yield, parity and somatic cell count were also evaluated as main effects in the model. A general linear model including all variables as random effects was then fitted to assess the contribution of each variable towards the variability in MUN concentration. Maximum daily temperature and humidity on the sampling day were positively associated with MUN concentration, but their interaction term was negatively associated, indicating that their effects were not independent and additive. Variables that contributed the most to the variability of MUN concentration were dietary crude protein (21%), temperature (18%) and other factors (24%) that were not assessed in the model (error term). Temperature has a significant influence on urea nitrogen concentration and should therefore always be considered when urea nitrogen concentration data are used to make inferences about the dietary management of dairy cows.

A statistical algorithm for outbreak detection in multisite settings: an application to sick leave monitoring.

Public health authorities monitor cases of health-related problems over time using surveillance algorithms that detect unusually high increases in the number of cases, namely aberrations. Statistical aberrations signal outbreaks when further investigation reveals epidemiological significance. The increasing availability and diversity of epidemiological data and the most recent epidemic threats call for more accurate surveillance algorithms that not just detect aberration times but also detect locations. Sick leave data, for instance, can be monitored across companies to identify companies-related aberrations. In this context, we develop an extension to multisite surveillance of a routinely used aberration detection algorithm, the quasi-Poisson regression Farrington Flexible algorithm. The new algorithm consists of a negative-binomial mixed effects regression model with a random effects term for sites and a new reweighting procedure reducing the effect of past aberrations. A wide range of simulations shows that, compared with Farrington Flexible, the new algorithm produces better false positive rates and similar probabilities of detecting genuine outbreaks, for case counts that exceed historical baselines by 3SD. As expected, higher surges lead to lower false positive rates and higher probabilities of detecting true outbreaks. The new algorithm provides better detection of true outbreaks, reaching 100%, when cases exceed eight baseline standard deviations. We apply our algorithm to sick leave rates in the context of COVID-19 and find that it detects the pandemic effect. The new algorithm is easily implementable over a range of contrasting data scenarios, providing good overall performance and new perspectives for multisite surveillance. All the analyses are performed in the R statistical software using the package glmmTMB. The code for performing the analyses and for generating the simulations can be found online at the following link: https://github.com/TomDuchemin/mixed_surveillance. a.noufaily@warwick.ac.uk.

Assessing climate risk using ensembles: A novel framework for applying and extending open-source climate risk assessment platforms

Climate change adaptation decisions often require the consideration of risk rather than the environmental hazard alone. One approach for quantifying risk is to use a risk assessment framework which combines information about hazard, exposure and vulnerability to estimate risk in a spatially consistent way. In recent years, publicly available, open-source risk assessment frameworks have been made available, including the CLIMADA platform. Such tools are increasingly being used in combination with ensembles of climate model projections to quantify risk on climate time-scales, presenting the ensemble spread as a measure of climate model uncertainty. As climate models are computationally expensive to run, this quantification of uncertainty derived from the ensemble of projections is often limited by the number of members available. We present a novel framework involving the application and extension of the CLIMADA open-source climate risk assessment platform, demonstrating an approach for overcoming this limitation. We first show how the CLIMADA platform can be applied to an ensemble of UKCP18 regional climate projections to assess climate risk coherently across space in an idealised example for the UK. We then show how a Generalised Additive Model, involving an ‘ensemble member’ random effect term, can be used to statistically represent the climate model ensemble summary of risk and be used to simulate many more realisations of risk, representative of a larger collection of plausible ensemble members. Specifically, we apply the framework to an idealised example related to heat-stress and the associated risk of reduced outdoor physical working capacity in the UK, based on three global warming levels (recent past, 2 °C and 4 °C warmer than pre-industrial). We show how, in this idealised example, in a 2 °C warmer world (relative to pre-industrial), the UK could lose on average 15 million (or 2.5% of) days of outdoor physical work in a working year (225 days) as a result of heat-stress, which could equate to more than £1.5 billion of economic loss (roughly 0.07% of UK annual GDP). The uncertainty quantification provided by the framework allows for an upper range estimate which better quantifies climate model uncertainty. In a 4 °C warmer world this indicates the plausibility of38 million (or 6.2% of) working days lost in a year, possibly equating to more than £3.8 billion of economic loss (roughly 0.17% of UK annual GDP). Finally, we discuss limitations of the approach and recommend a number of extensions and areas of future work.