Variance Estimation Research Articles

Application of Sampling Variance Smoothing Methods for Small Area Proportion Estimation

Abstract Sampling variance smoothing is an important topic in small area estimation. In this article, we propose sampling variance smoothing methods for small area proportion estimation. In particular, we consider the generalized variance function and design effect methods for sampling variance smoothing. We evaluate and compare the smoothed sampling variances and small area estimates based on the smoothed variance estimates through analysis of survey data from Statistics Canada. The results from real data analysis and simulation study indicate that the proposed sampling variance smoothing methods perform very well for small area estimation.

Statistical innovations in the API ADAD Colombia trial: a randomized, double‐blind, placebo‐controlled, parallel‐group study in cognitively unimpaired PSEN1 E280A mutation carriers evaluating efficacy and safety of crenezumab

AbstractBackgroundA number of extended families in Colombia with early‐onset ADAD have a PSEN1 mutation at codon 280 (E280A), comprise the largest ADAD kindred in the world. Trial participants have been treated with crenezumab (an experimental anti‐amyloid monoclonal antibody) or placebo. Topline results were presented in August, 2022 at AAIC. Although the trial did not demonstrate a statistically significant clinical benefit in either of its co‐primary endpoints, multiple statistical innovations were incorporated into this trial.MethodThe API ADAD Colombia trial is a randomized, double‐blind, placebo‐controlled, parallel‐group phase 2b study evaluating the efficacy, safety, and tolerability of crenezumab in cognitively unimpaired ADAD PSEN1 E280A mutation carriers in Colombia (NCT01998841). Due to under enrollment and concerns with the sensitivity of the primary endpoint, the protocol was amended mid‐trial to change from 260 weeks of treatment to a common‐close design whereby treatment continued in the double‐blind portion until the last enrolled participant received 260 weeks of treatment. Because of the change in trial design and power considerations, the primary analysis was adapted from a traditional mixed model with repeated measures (MMRM) based analysis to a random coefficient regression model (RCRM) based analysis. To quantitatively examine the adequacy of the RCRM assumptions, a robust variance estimator (Huber 1967; White 1982) was prespecified as a sensitivity analysis to calculate the standard errors and p‐values of the co‐primary endpoints. This study used a type‐I‐error‐controlled statistical procedure to identify and elevate the Free and Cued Selective Reminding Test (FCSRT) cueing index to a co‐primary endpoint. To comprehensively assess the crenezumab effect across primary and secondary endpoints, the prespecified relative reduction point estimates along with the bootstrapped 95% confidence intervals were generated.ResultInnovative statistical methods in this trial will be presented.ConclusionAlthough the API Colombia study did not meet its co‐primary endpoints, statistical learnings from this trial can benefit development of future AD trials. To further maximize the value of this unique trial dataset, participant‐level data will be made publicly available, including clinical, imaging, and biomarker data.Funding: NIA grants RF1‐AG041705, R01‐AG055444, Genentech, Roche, philanthropic contributions to Banner Alzheimer’s Foundation. FlorbetapirPET was partly supported by Avid and Lilly.

Frailty and long-term survival among patients in Australian intensive care units with metastatic cancer (FRAIL-CANCER study): a retrospective registry-based cohort study

Recent advances in cancer therapeutics have improved outcomes, resulting in increasing candidacy of patients with metastatic cancer being admitted to intensive care units (ICUs). A large proportion of patients also have frailty, predisposing them to poor outcomes, yet the literature reporting on this is scarce. We aimed to assess the impact of frailty on survival in patients with metastatic cancer admitted to the ICU. In this retrospective registry-based cohort study, we used data from the Australia and New Zealand Intensive Care Society Adult Patient (age ≥16 years) database to identify patients with advanced (solid and haematological cancer) and a documented Clinical Frailty scale (CFS) admitted to 166 Australian ICUs. Patients without metastatic cancer were excluded. We analysed the effect of frailty (CFS 5-8) on long-term survival, and how this effect changed in specific subgroups (cancer subtypes, age [<65 years or ≥65 years], and those who survived hospitalisation). Because estimates tend to cluster within centres and vary between them, we used Cox proportional hazards regression models with robust sandwich variance estimators to assess the effect of frailty on survival time up to 4 years after ICU admission between groups. Between Jan 1, 2018, and March 31, 2022, 30 026 patients were eligible, and after exclusions 21 174 patients were included in the analysis; of these, 6806 (32·1%) had frailty, and 11 662 (55·1%) were male, 9489 (44·8%) were female, and 23 (0·1%) were intersex or self-reported indeterminate sex. The overall survival was lower for patients with frailty at 4 years compared with patients without frailty (29·5% vs 10·9%; p<0·0001). Frailty was associated with shorter 4-year survival times (adjusted hazard ratio 1·52 [95% CI 1·43-1·60]), and this effect was seen across all cancer subtypes. Frailty was associated with shorter survival times in patients younger than 65 years (1·66 [1·51-1·83]) and aged 65 years or older (1·40 [1·38-1·56]), but its effects were larger in patients younger than 65 years (pinteraction<0·0001). Frailty was also associated with shorter survival times in patients who survived hospitalisation (1·49 [1·40-1·59]). In patients with metastatic cancer admitted to the ICU, frailty was associated with poorer long-term survival. Patients with frailty might benefit from a goal-concordant time-limited trial in the ICU and will need suitable post-intensive care supportive management. None.

AutoSourceID-FeatureExtractor

Aims. In astronomy, machine learning has been successful in various tasks such as source localisation, classification, anomaly detection, and segmentation. However, feature regression remains an area with room for improvement. We aim to design a network that can accurately estimate sources’ features and their uncertainties from single-band image cutouts, given the approximated locations of the sources provided by the previously developed code AutoSourceID-Light (ASID-L) or other external catalogues. This work serves as a proof of concept, showing the potential of machine learning in estimating astronomical features when trained on meticulously crafted synthetic images and subsequently applied to real astronomical data. Methods. The algorithm presented here, AutoSourceID-FeatureExtractor (ASID-FE), uses single-band cutouts of 32x32 pixels around the localised sources to estimate flux, sub-pixel centre coordinates, and their uncertainties. ASID-FE employs a two-step mean variance estimation (TS-MVE) approach to first estimate the features and then their uncertainties without the need for additional information, for example the point spread function (PSF). For this proof of concept, we generated a synthetic dataset comprising only point sources directly derived from real images, ensuring a controlled yet authentic testing environment. Results. We show that ASID-FE, trained on synthetic images derived from the MeerLICHT telescope, can predict more accurate features with respect to similar codes such as SourceExtractor and that the two-step method can estimate well-calibrated uncertainties that are better behaved compared to similar methods that use deep ensembles of simple MVE networks. Finally, we evaluate the model on real images from the MeerLICHT telescope and the Zwicky Transient Facility (ZTF) to test its transfer learning abilities.

Aerobic exercise and volumetric change in the human hippocampus: A meta‐analysis of randomized controlled trials in healthy older individuals

AbstractBackgroundThere is promising, albeit equivocal, evidence that aerobic exercise training (AET) increases human hippocampal volume. The evidence is unclear, as recent randomized controlled trials (RCTs) have failed to replicate AET’s previously reported favorable effects on hippocampal volume in cognitively unimpaired older individuals. We thus conducted a meta‐analysis of RCTs investigating the effects of AET on hippocampal volume and cardiorespiratory fitness (CRF) in cognitively unimpaired, healthy older individuals.MethodA meta‐analysis of RCTs that delivered AET interventions to older individuals. Inclusion criteria were: 1) cognitively unimpaired, healthy individuals aged ≥ 50; 2) AET intervention delivered for at least four weeks; 3) at least one intervention group and one control group; 4) reported data of the effects of AET on hippocampal volume. Two authors independently screened titles, abstracts, and full texts, performed data extraction and methodological quality assessment of included studies. We performed meta‐analyses with random‐effects robust variance estimation models and meta‐regression analyses to test associations between changes in CRF and hippocampal volume. The primary outcome was changes in hippocampal volume, and the secondary outcome was changes in CRF.ResultEight parent trials and one secondary outcome article were eligible. The meta‐analytic summary of 17 effect sizes indicated no significant effects of AET on hippocampal volume (SMD = 0.10, 95%CI ‐0.01 to 0.21, p = 0.073). Meta‐analysis of nine studies and effect sizes showed that AET moderately improved cardiorespiratory fitness (SMD = 0.30, 95%CI 0.12 to 0.48, p = 0.005). Improvement in CRF was not statistically associated with changes in hippocampal volume (bSE = 0.05, SE = 0.51, p = 0.923). Methodological quality was poor in one, fair in three, good in four, and excellent in one trial.ConclusionAET did not show robust effects on hippocampal volume in cognitively unimpaired, healthy older individuals. Since AET significantly improved CRF, a lack of training fidelity is unlikely to explain our findings. The biologically still plausible favorable effects of AET on hippocampal volume in cognitively unimpaired older individuals might be masked by methodological weaknesses and the inclusion of only nine studies. Future studies need to address: the lack of harmonization of hippocampus quantification methods and insufficient power to detect meaningful changes in hippocampal volume across studies.

Simultaneous Confidence Intervals for All Pairwise Differences between Means of Weibull Distributions

The Weibull distribution is a continuous probability distribution that finds wide application in various fields for analyzing real-world data. Specifically, wind speed data often adhere to the Weibull distribution. In our study, our aim is to compare the mean wind speed datasets from different areas in Thailand. To achieve this, we proposed simultaneous confidence intervals for all pairwise differences between the means of Weibull distributions. The generalized confidence interval (GCI), method of variance estimates recovery (MOVER), and a Bayesian approach, utilizing both gamma and uniform prior distributions, are proposed to construct simultaneous confidence intervals. Through simulations, we find that the Bayesian highest posterior density (HPD) interval using a gamma prior distribution demonstrates satisfactory performance, while the GCI proves to be a viable alternative. Finally, we applied these proposed approaches to real wind speed data in northeastern and southern Thailand to illustrate their effectiveness and practicality.

Association of Neighborhood-Level Marginalization With Health Care Use and Clinical Outcomes Following Hospital Discharge in Patients Who Underwent Coronary Catheterization for Acute Myocardial Infarction in a Single-Payer Health Care System.

Canadian data suggest that patients of lower socioeconomic status with acute myocardial infarction receive less beneficial therapy and have worse clinical outcomes, raising questions regarding care disparities even in universal health care systems. We assessed the contemporary association of marginalization with clinical outcomes and health services use. Using clinical and administrative databases in Ontario, Canada, we conducted a population-based study of patients aged ≥65 years hospitalized for their first acute myocardial infarction between April 1, 2010 and March 1, 2019. Patients receiving cardiac catheterization and surviving 7 days postdischarge were included. Our primary exposure was neighborhood-level marginalization, a multidimensional socioeconomic status metric. Neighborhoods were categorized by quintile from Q1 (least marginalized) to Q5 (most marginalized). Our primary outcome was all-cause mortality. A proportional hazards regression model with a robust variance estimator was used to quantify the association of marginalization with outcomes, adjusting for risk factors, comorbidities, disease severity, and regional cardiologist supply. Among 53 841 patients (median age, 75 years; 39.1% female) from 20 640 neighborhoods, crude 1- and 3-year mortality rates were 7.7% and 17.2%, respectively. Patients in Q5 had no significant difference in 1-year mortality (hazard ratio [HR], 1.08 [95% CI, 0.95-1.22]), but greater mortality over 3 years (HR, 1.13 [95% CI, 1.03-1.22]) compared with Q1. Over 1 year, we observed differences between Q1 and Q5 in visits to primary care physicians (Q1, 96.7%; Q5, 93.7%) and cardiologists (Q1, 82.6%; Q5, 72.6%), as well as diagnostic testing. There were no differences in secondary prevention medications dispensed or medication adherence at 1 year. In older patients with acute myocardial infarction who survived to hospital discharge, those residing in the most marginalized neighborhoods had a greater long-term risk of mortality, less specialist care, and fewer diagnostic tests. Yet, there were no differences across socioeconomic status in prescription medication use and adherence.

Misleading overestimation bias in methods to estimate wolf abundance that use spatial models.

Abstract 1. Population abundance is the main criterion used by agencies to manage and conserve species and it allows adaptive decision-making in response to impacts. Its estimation is particularly important for large mammals, especially carnivores that are notoriously difficult to monitor yet have high ecological and economic importance to humans. Reversal of their historic population decline is also vital to restoring ecosystem health through the ecosystem service of trophic interactions. 2. Because bias and precision (variance) are the independent yardsticks of the quality and reliability of population estimates, we preliminarily assessed new abundance estimators for wolves that non-traditionally use spatial models to estimate the area that an observation represents. Among many biases and problems identified by recent assessments, we identified suspected biases due to obvious violation of the closure assumption in occupancy modeling used to determine the area occupied by territorial pack members. Thus, we chose to simulate the effect of spatial resolution (grid size) on the direction and magnitude of that inherent bias in a recent method called iPOM used in Montana. We also examined the potential bias in their estimate of variance and confidence intervals. 3. We found that even the use of small grid cells (relative to wolf territory size), biased the total area occupied and the number of packs used to calculate abundance. The bias rapidly increases with increasing grid cell size. At the grid cell size used in iPOM for Montana (600 km 2 ) there was a severe overestimation bias of 150% that proliferated through the iPOM submodel structure and resulted in estimated wolf abundance 2.5 times larger than true abundance. 4. This bias alone when combined with a misapplication and underreporting of iPOM's estimate of variance (biased low) results in a precariously misleading situation for decision-makers that threatens wolf populations. Other identified biases that inflate abundance likely make this situation worse but they should be further examined and tested. 5. Due to these biases and the high sensitivity of iPOM's spatial models to estimate area, we suggest that such spatial models should not be used in population estimation methods or such methods, iPOM for example, should be improved and/or restructured with submodels robust to assumption violation thereby reducing bias. Given iPOM's current design, there is no ability to detect change let alone determine wolf population size. 6. We provide a comparative framework for testing and improvement and strongly suggest proper model-assisted sampling designs and hierarchical modeling such as is used in capture-recapture models, especially those that use non-invasive procedures that avoid costly capture and marking. We also recommend collaborative activities that lead to using the best available scientific methods to determine carnivore abundance.

Uncertainty propagation in absolute metabolite quantification for in vivo MRS of the human brain.

Absolute spectral quantification is the standard method for deriving estimates of the concentration from metabolite signals measured using in vivo proton MRS (1 H-MRS). This method is often reported with minimum variance estimators, specifically the Cramér-Rao lower bound (CRLB) of the metabolite signal amplitude's scaling factor from linear combination modeling. This value serves as a proxy for SD and is commonly reported in MRS experiments. Characterizing the uncertainty of absolute quantification, however, depends on more than simply the CRLB. The uncertainties of metabolite-specific (T1m , T2m ), reference-specific (T1ref , T2ref ), and sequence-specific (TR , TE ) parameters are generally ignored, potentially leading to an overestimation of precision. In this study, the propagation of uncertainty is used to derive a comprehensive estimate of the overall precision of concentrations from an internal reference. The propagated uncertainty is calculated using analytical derivations and Monte Carlo simulations and subsequently analyzed across a set of commonly measured metabolites and macromolecules. The effect of measurement error from experimentally obtained quantification parameters is estimated using published uncertainties and CRLBs from in vivo 1 H-MRS literature. The additive effect of propagated measurement uncertainty from applied quantification correction factors can result in up to a fourfold increase in the concentration estimate's coefficient of variation compared to the CRLB alone. A case study analysis reveals similar multifold increases across both metabolites and macromolecules. The precision of absolute metabolite concentrations derived from 1 H-MRS experiments is systematically overestimated if the uncertainties of commonly applied corrections are neglected as sources of error.

Trade-off between efficiency and variance estimation of spatially balanced augmented samples

Trade-off between efficiency and variance estimation of spatially balanced augmented samples

Deep and infratentorial cerebral microbleeds are related to wake-up stroke by large-artery atherosclerosis and cardioembolism

Background and aimsCircadian variability of blood pressure (BP) and hypercoagulation in the morning have been proposed as underlying mechanisms of wake-up stroke (WUS). The aim was to determine the impact of cerebral microbleeds (CMBs), showing BP fluctuation and background hypercoagulability, on WUS. MethodsConsecutive patients with acute ischemic stroke onset-to-door time within one week were screened. WUS was defined as an ischemic stroke that occurred during sleep at night. CMBs were categorized into three: “strictly Lobar”, “strictly Deep (D) and/or Infratentorial (I)”, and “Mixed”. Moderate to severe CMBs were defined as having more than three in total. First, whether CMBs are associated with WUS was examined. Second, the same analysis was performed according to the stroke subtype classified as large-artery atherosclerosis (LAA), cardioembolism (CE), and small-vessel occlusion (SVO). ResultsA total of 1477 patients (1059 [72%] male, median age 69 years) were included, and WUS was observed in 363 (25%) patients. On Poisson regression analysis with a robust variance estimator in the total cohort, moderate to severe strictly D and/or I CMBs (PR 1.505, 95% CI 1.154–1.962, p = 0.003) were associated with WUS. From the perspective of stroke subtype, the same result was confirmed in LAA (PR 2.223, 95% CI 1.036–4.768, p = 0.040) and CE (PR 1.668, 95% CI 1.027–2.709, p = 0.039), not SVO. ConclusionsThe presence of moderate to severe strictly D and/or I CMBs might be associated with the development of WUS. By stroke subtype, the same result was confirmed in LAA and CE.

Early introduction of ultra-processed foods is associated with overweight and anaemia in socially vulnerable Brazilian children.

This study aimed to evaluate the early introduction of ultra-processed foods (UPF) and identify its association with overweight and anaemia in Brazilian children living in a situation of social vulnerability. A population-based cross-sectional study was conducted in a Brazilian capital. Children aged 12-59 months were included. The presence of overweight and anaemia was evaluated, as well as the introduction of twelve different UPF in children's first year of life. Association analysis was performed using Poisson regression, with robust estimates of variances. A total of 561 children were studied; 85·5 % had consumed at least one UPF evaluated in the first year of life; 19·1 % were overweight and 52·0 % were anaemic. Adjusted multivariate analyses identified that the early introduction of soft drinks (Prevalence Ratio (PR) = 1·18, 95 % CI (1·02, 1·38)), packaged snacks (PR = 1·17, 95 % CI (1·05, 1·30)) and powdered soft drinks (PR = 1·36, 95 % CI (1·16, 1·60)) increased the likelihood of children being overweight, and the early introduction of chocolate drink (PR = 1·25, 95 % CI (1·02, 1·53)) increased the likelihood of them being anaemic, when comparing children who consumed these UPF before reaching 1 year of age with those who consumed these foods at 12 months of age or older. From the results found, one can see the existing relationship between the early introduction of UPF with overweight and anaemia, being necessary to intensify public health policies to combat malnutrition, focusing on the promotion of proper and healthy eating, especially during the phase of food introduction, focusing on the population living in socially vulnerable situations.

No Survival Disparity between Black and White Patient's with Stage I/II Diffuse Large B-Cell Lymphoma and It's Significance

Background Diffuse large B Cell Lymphoma (DLBCL) is the most common subtype of non-Hodgkin Lymphoma. Limited stage I/II DLBCL has an excellent prognosis with five-year overall survival (OS) of at least 70%. Black patients with DLBCL stage I-IV collectively are reported to have worse outcomes on multivariate analysis. There is scant data to evaluate the disparity between Black and White patients in limited-stage DLBCL. We aimed to determine the characteristics and outcomes for stage I/II DLBCL between Black and White patients using the National Cancer Database (NCDB) from 2004 to 2017. Methods Chi-square and Wilcoxon rank sum tests were performed to assess the differences in characteristic variables between Black and White patients with limited-stage DLBCL. Multivariate Cox regression analysis was conducted to identify the independent determinants affecting outcomes in stage I/II DLBCL. Fifteen variables in the analysis include demographics (age, sex, race, ethnicity), social (distance traveled, facility type, insurance status, median household income, education level), clinical features (B symptoms, HIV status, Charlson-Deyo score), management (year of diagnosis, stage, treatment type). Kaplan-Meier (KM) survival curves were used to evaluate survival between Black and White patients with stage I / II DLBCL. Additionally, exact matching for all characteristic variables between Black and White patients were performed to compare survival using the Cox model with robust variance estimators. Results: For stage I/II DLBCL, KM survival rate estimates at 1-year, 3-year, and 5-year were 78.8%, 68.8%, and 62.1%, respectively. Diagnosis after 2004 measured in 4 intervals did not impact outcome after rituximab approval in 1997. There is no statistical survival difference (p = 0.5263) in stage I/II DLBCL between Black and White patients. Age &amp;gt; 60 years, males, higher Charlson score, more B symptoms, and positive HIV status were independent variables that affected survival (above not shown). In limited-stage DLBCL, Black patients were younger, had higher Charlson scores, more B symptoms, and higher percent were HIV positive(Table 1). Exact matching of paired samples between Black and White patients for all variables using the Cox model with robust variance estimators similarly showed no statistical difference in survival (p=0.1818). Interestingly, Hispanic patients (HR=0.84) have better survival than non-Hispanic patients (Table 2). Conclusion: Black patients have reported worse survival for stages I-IV DLBCL collectively. However, there is no statistically significant survival difference in stages I-II. Our study's independent characteristics associated with worse outcome are similar to DLBCL stage I-IV studies except for race. Yet the observed disparities between Black and White patients with stage I/II DLBCL do not appear to impact survival as an outcome. The exact matching of variables, including demographic and social determinants, confirmed survival remained similar, not better. Therefore, in early-stage DLBCL, the biology of DLBCL may be more predictive of outcome for race. Determinants of outcome should be stratified by stage. Advanced DLBCL remains an independent prognostic feature. Achieving health equity in DLBCL may be more impactful on advanced stage than for limited stage, as there is no survival disparity between Black and White patients in stage I/II DLBCL despite the disparity in demographic and social determinants.

Hospital Associated Disability Among Older Adults with Plasma Cell Disorders Receiving Autologous Stem Cell Transplant

Background: An increasing number of older adults with Plasma Cell Disorders (PCDs) are receiving autologous stem cell transplant (ASCT) in the US. Hospital associated disability (HAD) is a loss in independence in one or more activities of daily living associated with acute care hospitalization and is a common complication among older adults. The prevalence and prognostic significance of HAD among older adults with PCDs undergoing ASCT is unknown. Methods: This retrospective cohort study used consecutive adults receiving ASCT at a large U.S. comprehensive medical center between 1/2013 and 5/2023. Trained nursing staff assessed Katz Activities of Daily Living (ADL) at admission and every 3 days thereafter under our Virtual Acute Care for Elders program. The primary outcome was development of HAD defined as ≥1 point decline on the Katz ADL scale from hospital admission to discharge. We also examined the association between development of HAD and various predictors including age, Karnofsky Performance Status (KPS), and Hematopoietic Cell Transplantation-Specific Comorbidity Index (HCT-CI) using modified Poisson regression models with robust variance estimators. Using similar models, we studied the impact of HAD on unplanned 30-day readmission rates. Results: Of 870 patients undergoing ASCT during the study period, 778 had at least two Katz ADL assessments and were included. The study population was comprised of 56% Caucasian/white ethnicity and 56% male with a median age of 62y (IQR 55-68y). In the overall population, 112 (14.4%) developed HAD. The incidence of HAD was much higher among adults ≥65y compared to their younger counterparts (22% vs. 9%; p=&amp;lt;0.01). In multivariate analysis, age ≥65y (RR 2.22; 95% CI 1.56-3.15; p=&amp;lt;0.001) and baseline KPS &amp;lt;=70 (RR 1.83; 95% CI 1.15-2.92; p=0.011) was associated with increased risk of developing HAD; HCT-CI scores did not have a significant association with HAD. Those with HAD had a two-fold higher risk of 30-day readmission as compared to those without HAD (RR 2.18; 95% CI 1.20-3.66; p=0.003). Conclusions: Nearly 1 in 4 adults ≥65y developed HAD while undergoing ASCT, which was associated with a two-fold increased risk of 30-day readmission. Interventions to prevent HAD and its downstream consequences are critically needed.

Practical Considerations for Sandwich Variance Estimation in 2-Stage Regression Settings.

In this paper, we present a practical approach for computing the sandwich variance estimator in 2-stage regression model settings. As a motivating example for 2-stage regression, we consider regression calibration, a popular approach for addressing covariate measurement error. The sandwich variance approach has rarely been applied in regression calibration, despite its requiring less computation time than popular resampling approaches for variance estimation, specifically the bootstrap. This is probably because it requires specialized statistical coding. Here we first outline the steps needed to compute the sandwich variance estimator. We then develop a convenient method of computation in R for sandwich variance estimation, which leverages standard regression model outputs and existing R functions and can be applied in the case of a simple random sample or complex survey design. We use a simulation study to compare the sandwich estimator to a resampling variance approach for both settings. Finally, we further compare these 2 variance estimation approaches in data examples from the Women's Health Initiative (1993-2005) and the Hispanic Community Health Study/Study of Latinos (2008-2011). In our simulations, the sandwich variance estimator typically had good numerical performance, but simple Wald bootstrap confidence intervals were unstable or overcovered in certain settings, particularly when there was high correlation between covariates or large measurement error.

2260. Differences in the Prevalence of Inappropriate Antibiotic Prescribing by Payer Type

Abstract Background Understanding the relationship between inappropriate prescribing and patient insurance type could inform the targeting of antibiotic stewardship initiatives. However, few national data on this topic exist. Methods We analyzed the National Ambulatory Medical Care Survey, a nationally representative survey of office-based physician visits. We included all years from 2016 to 2019 except 2017, a year during which the survey was not fielded. We identified visits resulting in an oral antibiotic prescription among children and non-elderly adults who were privately insured, publicly insured, or uninsured. The exposure was safety-net status (publicly insured/uninsured versus privately insured). The outcome was inappropriate antibiotic prescribing, defined as an antibiotic prescription in a visit for which none of the diagnosis codes justified antibiotic use (following the approach of Chua et al, BMJ 2019). We assessed the association between safety-net status and inappropriate antibiotic prescribing using linear regression, controlling for patient characteristics, calendar quarter, and specialty. Models accounted for the complex design of the survey and employed designed-based variance estimators. Results Of 413,830,032 weighted visits for children, 16.9% resulted in an antibiotic prescription. Among these visits, the unadjusted prevalence of inappropriate antibiotic prescriptions was 29.1% overall, 23.4% for safety-net patients, and 32.4% for privately insured patients (adjusted difference, safety-net minus private: -11.9%, 95% CI: -20.6%, -3.2%). Of 1,240,610,825 weighted visits for non-elderly adults, 11.0% resulted in an antibiotic prescription. Among these visits, the unadjusted prevalence of inappropriate antibiotic prescriptions was 50.1% overall, 60.7% for safety-net patients, and 47.9% for privately insured patients (adjusted difference, safety-net minus private: 11.3%, 95% CI: 2.4%, 20.3%). Conclusion Safety-net status was associated with a lower rate of inappropriate antibiotic prescribing among children but a higher rate among non-elderly adults. Antibiotic stewardship initiatives targeting clinicians who care for privately insured children and adults with safety-net insurance may be warranted. Disclosures Jeffrey A. Linder, MD, MPH, FACP, Amgen, Biogen, Eli Lily: Stocks/Bonds

Designing individually randomized group treatment trials with repeated outcome measurements using generalized estimating equations.

Individually randomized group treatment (IRGT) trials, in which the clustering of outcome is induced by group-based treatment delivery, are increasingly popular in public health research. IRGT trials frequently incorporate longitudinal measurements, of which the proper sample size calculations should account for correlation structures reflecting both the treatment-induced clustering and repeated outcome measurements. Given the relatively sparse literature on designing longitudinal IRGT trials, we propose sample size procedures for continuous and binary outcomes based on the generalized estimating equations approach, employing the block exchangeable correlation structures with different correlation parameters for the treatment arm and for the control arm, and surveying five marginal mean models with different assumptions of time effect: no-time constant treatment effect, linear-time constant treatment effect, categorical-time constant treatment effect, linear time by treatment interaction, and categorical time by treatment interaction. Closed-form sample size formulas are derived for continuous outcomes, which depends on the eigenvalues of the correlation matrices; detailed numerical sample size procedures are proposed for binary outcomes. Through simulations, we demonstrate that the empirical power agrees well with the predicted power, for as few as eight groups formed in the treatment arm, when data are analyzed using the matrix-adjusted estimating equations for the correlation parameters with a bias-corrected sandwich variance estimator.

Estimation of plant density based on presence/absence data using hybrid inference

Monitoring of plant populations has become more and more important, especially in the current context of environmental change. In this paper, we propose methods to estimate plant density from presence/absence surveys, wherein the presence or absence of each species is recorded on sample plots. Presence/absence sampling is a useful and relatively simple method for monitoring state and change of plant communities. Moreover, it has advantages compared to traditional plant cover assessment, the latter being more prone to observer bias. We present a hybrid estimation framework, that combines model- and design-based inference features, in which a generalised linear model (for binary presence/absence data) and an inhomogeneous Poisson model (for plant locations) are used to estimate plant density in a region of interest. We look at two different cases, the first one with a known area and the second one where the area is unknown and must be estimated. Our methods are applied to real data on Vaccinium vitis-idaea from the Swedish National Forest Inventory as well as simulated data to assess the performance of our estimators of plant density and corresponding variance estimators. The results obtained are promising and indicate that this method has a potential to add considerable analytic strength to monitoring programmes that collect presence/absence data.

Unravelling the genetic variability of host resilience to endo- and ectoparasites in Nellore commercial herds

BackgroundHost resilience (HR) to parasites can affect the performance of animals. Therefore, the aim of this study was to present a detailed investigation of the genetic mechanisms of HR to ticks (TICK), gastrointestinal nematodes (GIN), and Eimeria spp. (EIM) in Nellore cattle that were raised under natural infestation and a prophylactic parasite control strategy. In our study, HR was defined as the slope coefficient of body weight (BW) when TICK, GIN, and EIM burdens were used as environmental gradients in random regression models. In total, 1712 animals were evaluated at five measurement events (ME) at an average age of 331, 385, 443, 498, and 555 days, which generated 7307 body weight (BW) records. Of the 1712 animals, 1075 genotyped animals were used in genome-wide association studies to identify genomic regions associated with HR.ResultsPosterior means of the heritability estimates for BW ranged from 0.09 to 0.54 across parasites and ME. The single nucleotide polymorphism (SNP)-derived heritability for BW at each ME ranged from a low (0.09 at ME.331) to a moderate value (0.23 at ME.555). Those estimates show that genetic progress can be achieved for BW through selection. Both genetic and genomic associations between BW and HR to TICK, GIN, and EIM confirmed that parasite infestation impacted the performance of animals. Selection for BW under an environment with a controlled parasite burden is an alternative to improve both, BW and HR. There was no impact of age of measurement on the estimates of genetic variance for HR. Five quantitative trait loci (QTL) were associated with HR to EIM but none with HR to TICK and to GIN. These QTL contain genes that were previously shown to be associated with the production of antibody modulators and chemokines that are released in the intestinal epithelium.ConclusionsSelection for BW under natural infestation and controlled parasite burden, via prophylactic parasite control, contributes to the identification of animals that are resilient to nematodes and Eimeria ssp. Although we verified that sufficient genetic variation existed for HR, we did not find any genes associated with mechanisms that could justify the expression of HR to TICK and GIN.

Cox regression with linked data.

Record linkage is increasingly used, especially in medical studies, to combine data from different databases that refer to the same entities. The linked data can bring analysts novel and valuable knowledge that is impossible to obtain from a single database. However, linkage errors are usually unavoidable, regardless of record linkage methods, and ignoring these errors may lead to biased estimates. While different methods have been developed to deal with the linkage errors in the generalized linear model, there is not much interest on Cox regression model, although this is one of the most important statistical models in clinical and epidemiological research. In this work, we propose an adjusted estimating equation for secondary Cox regression analysis, where linked data have been prepared by a third-party operator, and no information on matching variables is available to the analyst. Through a Monte Carlo simulation study, the proposed method is shown to lead to substantial bias reductions in the estimation of the parameters of the Cox model caused by false links. An asymptotically unbiased variance estimator for the adjusted estimators of Cox regression coefficients is also proposed. Finally, the proposed method is applied to a linked database from the Brest stroke registry in France.