Imprecise Estimates Research Articles

Using camera traps and N‐mixture models to estimate population abundance: Model selection really matters

Abstract Estimating the abundance or density of wildlife populations is a critical part of species conservation and management, but estimates can vary greatly in precision and accuracy according to the sampling and statistical methods, sampling and ecological variation, and sample size. We used images of moose (Alces americanus) from camera traps to parameterize N‐mixture models and tested the effect of ecological conditions, the spatial scale of measurement, and the criteria used to define independent detections on estimates of population abundance. We compared the model estimates to those generated empirically with aerial survey data, the standard method for many species of ungulate. We explored the sensitivity of estimates to model choice based on the common statistical criterion of parsimony. The two most parsimonious N‐mixture models (i.e. AICc) were considerably biased, producing implausibly large and considerably imprecise estimates of abundance. Most of the other models produced estimates of moose abundance that were ecologically realistic and relatively accurate. The accuracy of population estimates produced by N‐mixture models was not overly sensitive to the formulation of models, the scale at which ecological conditions were measured, or the criteria used to define independent detection and by extension sample size. Our results suggested that parsimony was a poor measure of the predictive accuracy of the population estimates produced with the N‐mixture model. We recommend using a suite of models to generate predictions of abundance instead of the single top‐ranked model. Collecting and processing data from the aerial survey was less expensive and took less time, but data from camera traps provided a broader set of insights into the behaviour of moose and the co‐occurrence of competitors and predators.

Handling missing data in research.

Missing data are an inevitable part of research and lead to a decrease in the size of the analyzable population, and biased and imprecise estimates. In this article, we discuss the types of missing data, methods to handle missing data and suggest ways in which missing data can be minimized.

Review of methodologies for detecting an observer effect in commercial fisheries data

Observers are deployed on commercial fishing trips to collect representative samples of discard rates. However, fishers may change their fishing habits when an observer is onboard (“observer effect”) or observer programs may over- or under-sample portions of the fleet (“deployment effect”). If the extent of these effects are substantial, observer data will not be representative of unobserved trips, potentially biasing the estimation of discards. This sampling bias can impact catch monitoring, stock assessments, and fishery management. The purpose of this study was to examine the power and error rate of several published methods for detecting an observer effect using a simulation of observer and deployment effects at varying sampling ratios (i.e., observer coverage) for several sample statistics. The simplest methods (t-test and F-test for difference of means and variances) provided an accurate although imprecise estimate of the observer effect size, but only when there were no deployment effects. A generalized linear mixed effects model (GLMM) was also not reliable for detecting small bias, but was not confounded by deployment effects and was relatively robust to changing coverage rates. The most complicated tests involved comparing differences in trip characteristics between subsequent trips for observed-unobserved and unobserved-unobserved pairs. These tests were able to detect smaller observer effects and were not confounded by deployment effects, but were unreliable at high coverage rates (>60%), producing both high false positive and false negative rates. Sensitivity tests also showed differing detection accuracy as the distribution of the metric of interest changed. Thus, the optimal test for detecting an observer effect will depend on the metric of interest, the coverage rate, and whether a deployment effect exists. An example from the New England groundfish fishery is provided to illustrate how conflicting results may be explained. Results should always be considered carefully when declaring that an observer effect is or is not occurring because of the sensitivity of the tests.

Data-driven RANS closures for improving mean field calculation of separated flows

Reynolds-averaged Navier-Stokes (RANS) simulations have found widespread use in engineering applications, yet their accuracy is compromised, especially in complex flows, due to imprecise closure term estimations. Machine learning advancements have opened new avenues for turbulence modeling by extracting features from high-fidelity data to correct RANS closure terms. This method entails establishing a mapping relationship between the mean flow field and the closure term through a designated algorithm. In this study, the k-ω SST model serves as the correction template. Leveraging a neural network algorithm, we enhance the predictive precision in separated flows by forecasting the desired learning target. We formulate linear terms by approximating the high-fidelity closure (from Direct Numerical Simulation) based on the Boussinesq assumption, while residual errors (referred to as nonlinear terms) are introduced into the momentum equation via an appropriate scaling factor. Utilizing data from periodic hills flows encompassing diverse geometries, we train two neural networks, each possessing comparable structures, to predict the linear and nonlinear terms. These networks incorporate features from the minimal integrity basis and mean flow. Through generalization performance tests, the proposed data-driven model demonstrates effective closure term predictions, mitigating significant overfitting concerns. Furthermore, the propagation of the predicted closure term to the mean velocity field exhibits remarkable alignment with the high-fidelity data, thus affirming the validity of the current framework. In contrast to prior studies, we notably trim down the total count of input features to 12, thereby simplifying the task for neural networks and broadening its applications to more intricate scenarios involving separated flows.

Community based complex interventions to sustain independence in older people: systematic review and network meta-analysis

ObjectiveTo synthesise evidence of the effectiveness of community based complex interventions, grouped according to their intervention components, to sustain independence for older people.DesignSystematic review and network meta-analysis.Data sourcesMedline, Embase, CINAHL,...

Estimating causality between smoking and abdominal obesity by Mendelian randomization.

Smokers tend to have a lower body weight than non-smokers, but also more abdominal fat. It remains unclear whether or not the relationship between smoking and abdominal obesity is causal. Previous Mendelian randomization (MR) studies have investigated this relationship by relying upon a single genetic variant for smoking heaviness. This approach is sensitive to pleiotropic effects and may produce imprecise causal estimates. We aimed to estimate causality between smoking and abdominal obesity using multiple genetic instruments. MR study using causal analysis using summary effect estimates (CAUSE) and latent heritable confounder MR (LHC-MR) methods that instrument smoking using genome-wide data, and also two-sample MR (2SMR) methods. Genome-wide association studies (GWAS) summary statistics from participants of European ancestry, obtainedfrom the GWAS and Sequencing Consortium of Alcohol and Nicotine use (GSCAN), Genetic Investigation of Anthropometric Traits (GIANT) Consortium and the UK Biobank. We used GWAS results for smoking initiation (n = 1 232 091), life-time smoking (n = 462 690) and smoking heaviness (n = 337 334) as exposure traits, and waist-hip ratio (WHR) and waist and hip circumferences (WC and HC) (n up to 697 734), with and without adjustment for body mass index (adjBMI), as outcome traits. Smoking initiation, life-time smoking, smoking heaviness, WHR, WC, HC, WHRadjBMI, WCadjBMI and HCadjBMI. Both CAUSE and LHC-MR indicated a positive causal effect of smoking initiation on WHR (0.13 [95% confidence interval (CI) = 0.10, 0.16 and 0.49 (0.41, 0.57), respectively] and WHRadjBMI (0.07 (0.03, 0.10) and 0.31 (0.26, 0.37). Similarly, they indicated a positive causal effect of life-time smoking on WHR [0.35 (0.29, 0.41) and 0.44 (0.38, 0.51)] and WHRadjBMI [0.18 (0.13, 0.24) and 0.26 (0.20, 0.31)]. In follow-up analyses, smoking particularly increased visceral fat. There was no evidence of a mediating role by cortisol or sex hormones. Smoking initiation and higher life-time smoking may lead to increased abdominal fat distribution. The increase in abdominal fat due to smoking is characterized by an increase in visceral fat. Thus, efforts to prevent and cease smoking can have the added benefit of reducing abdominal fat.

Performance Analysis of the Particle Swarm Optimization Algorithm in a VLC System for Localization in Hospital Environments

Localization in hospitals can be valuable in improving different services in medical environments. In this sense, an accurate location system in this environment requires adequately enabling communication technology. However, widely adopted technologies such as Wireless Fidelity (WiFi), Bluetooth, and Radio Frequency Identification (RFID) are considered poorly suited to enable hospital localization due to their inherent drawbacks, including high implementation costs, poor signal strength, imprecise estimates, and potential interference with medical devices. The increasing expenses associated with the implementation and maintenance of these technologies, along with their limited accuracy in dynamic hospital environments, underscore the pressing need for alternative solutions. In this context, it becomes imperative to explore and present novel approaches that not only avoid these challenges but also offer more cost effective, accurate, and interference-resistant connectivity to achieve precise localization within the complex and sensitive hospital environment. In the quest to achieve adequate localization accuracy, this article strategically focuses on leveraging Visible Light Communication (VLC) as a fundamental technology to address the specific demands of hospital environments to achieve the precise localization and tracking of life-saving equipment. The proposed system leverages existing lighting infrastructure and utilizes three transmitting LEDs with different wavelengths. The Received Signal Strength (RSS) is used at the receiver, and a trilateration algorithm is employed to determine the distances between the receiver and each LED to achieve precise localization. The accuracy of the localization is further enhanced by integrating a trilateration algorithm with the sophisticated Particle Swarm Optimization (PSO) algorithm. The proposed method improves the localization accuracy, for example, at a height of 1 m, from a 11.7 cm error without PSO to 0.5 cm with the PSO algorithm. This enhanced accuracy is very important to meet the need for precise equipment location in dynamic and challenging hospital environments to meet the demand for life-saving equipment. Furthermore, the performance of the proposed localization algorithm is compared with conventional positioning methods, which denotes improvements in terms of the localization error and position estimation.

Nlmixr2 Versus NONMEM: An Evaluation of Maximum A Posteriori Bayesian Estimates Following External Evaluation of Gentamicin and Tobramycin Population Pharmacokinetic Models.

The objective of this project is to compare the results of the same study carried out on NONMEM and nlmixr2. This analysis consists of evaluating previously published population pharmacokinetic models of gentamicin and tobramycin in our population of interest with sparse concentrations. A literature review was performed to determine the gentamicin and tobramycin models in critically ill adult patients. In parallel, gentamicin and tobramycin dosing data, information on the treatment, the patient, and the bacteria were collected retrospectively in 2 Quebec establishments. The external evaluations were previously performed using NONMEM Version 7.5. Model equations were rewritten with R, and external evaluations were performed using nlmixr2. Predictive performance was assessed based on the estimation of bias and imprecision of the prediction error for maximum a posteriori (MAP) Bayesian PK parameter and observed concentrations. Comparison between nlmixr2 and NONMEM was performed on 4 gentamicin and 3 tobramycin population pharmacokinetic models. Compared to NONMEM, for gentamicin and tobramycin clearance and central volume of distribution, nlmixr2 produced individual pharmacokinetic parameters with bias values ranging from -32.5% to 5.67% and imprecision values ranging from 6.33% to 32.5%. Despite these differences, population bias and imprecision for sparse concentrations were low and ranged from 0% to 5.3% and 0.2% to 6.5%, respectively. The external evaluations performed with both software packages resulted in the same interpretation in terms of population predictive performance for all 7 models. Nlmxir2 showed comparable predictive performance with NONMEM with sparse concentrations that are, at most, sampled twice within a single dose administration (peak and trough).

Machine learning to predict continuous protein properties from binary cell sorting data and map unseen sequence space

Proteins are a diverse class of biomolecules responsible for wide-ranging cellular functions, from catalyzing reactions to recognizing pathogens. The ability to evolve proteins rapidly and inexpensively toward improved properties is a common objective for protein engineers. Powerful high-throughput methods like fluorescent activated cell sorting and next-generation sequencing have dramatically improved directed evolution experiments. However, it is unclear how to best leverage these data to characterize protein fitness landscapes more completely and identify lead candidates. In this work, we develop a simple yet powerful framework to improve protein optimization by predicting continuous protein properties from simple directed evolution experiments using interpretable, linear machine learning models. Importantly, we find that these models, which use data from simple but imprecise experimental estimates of protein fitness, have predictive capabilities that approach more precise but expensive data. Evaluated across five diverse protein engineering tasks, continuous properties are consistently predicted from readily available deep sequencing data, demonstrating that protein fitness space can be reasonably well modeled by linear relationships among sequence mutations. To prospectively test the utility of this approach, we generated a library of stapled peptides and applied the framework to predict affinity and specificity from simple cell sorting data. We then coupled integer linear programming, a method to optimize protein fitness from linear weights, with mutation scores from machine learning to identify variants in unseen sequence space that have improved and co-optimal properties. This approach represents a versatile tool for improved analysis and identification of protein variants across many domains of protein engineering.

Robust fault-tolerant control for dynamic positioning of ships with prescribed performance

This study presents a new fault-tolerant control scheme for ship dynamic positioning. The scheme aims to handle unknown disturbances, imprecise fault estimation, and control input constraints. First, an adaptive sliding mode disturbance observer is introduced. This observer effectively reconstructs disturbances in finite time, even without knowledge of their derivative upper bounds. Next, we design a novel prescribed performance function that sets bounds on tracking errors. This function is combined with an auxiliary intermediate control technique to create a high-level controller. Lastly, a robust fault-tolerant control allocation scheme is proposed to redistribute the generalized forces among faulty actuators. It autonomously finds a feasible force vector in the attainable force set. Simulation results are provided to demonstrate the effectiveness of the proposed control scheme.

SimMACT, a Software Demonstrator to Improve Maximum Actuation Joint Torques Simulation for Ergonomics Assessment.

The maximum actuation joint torques that operators can perform at the workplace are essential parameters for biomechanical risk assessment. However, workstation designers generally only have at their disposal the imprecise and sparse estimates of these quantities provided with digital manikin digital human model (DHM) software. For instance, such tools consider only static postures and ignore important specificities of the human musculoskeletal system such as interjoints couplings. To alleviate the weaknesses of existing approaches implemented in digital human modeling tools relying on torque databases, this paper describes a methodology based on a class of polytopes called zonotopes and musculoskeletal simulation to assess maximum actuation torques. It has two main advantages, the ability to estimate maximum joint torques for any posture and taking into account musculoskeletal specificities unlike existing digital human modeling tools. As a case study, it also compares simulated maximum actuation torques to those recorded during an experiment described in the literature, focusing on an isometric task of the upper limb. This simulation has led to similar or smaller errors than DHM software tools. Hence, this methodology may help in interpreting interjoint couplings, choosing appropriate mathematical models or design experimental protocols. It may also be implemented in DHM software to provide designers with more comprehensive and more reliable data.

Waste generation and macroeconomic drivers: a panel study for European countries and regions

PurposeThe vast amounts of waste generated today threaten economies and societies due to high environmental and management costs. The aim is to investigate the short- and long-term patterns of municipal waste generation (MWG) in response to socio-economic and demographic growth variables at national and regional levels.Design/methodology/approachA panel data approach employing ordinary least squares (OLS), fixed effects (FE), random effects (RE), fully modified least squares (FMOLS) and error correction model (ECM) techniques. A sample of 28 European countries (2000–2020) and 44 European Union (EU) regions (2000–2018) were selected.FindingsDuring periods of economic growth and higher employment rates, consumer confidence tends to increase, leading to elevated levels of consumer spending and consumption. Intensification in the production factors, specifically capital and employment, results in an upsurge in MWG, thereby creating a cycle where waste generation becomes deeply entrenched in the economic system in both the short and long terms. Rapid population growth, attributed to higher fertility rates, is associated with increased MWG. At the regional level, a double-aging process and a shift toward an aging population exert less pressure on MWG in both the short and long term. Promoting higher levels of environment-oriented human development yields various benefits, including the generation of greater knowledge spillovers, enhanced environmental literacy, a shift toward circular thinking and the promotion of greener entrepreneurship. Increased R&D expenditures facilitate the development of innovative waste reduction technologies, fostering improvements in waste management techniques, recycling processes and the utilization of sustainable materials.Research limitations/implicationsThe research examines the short- and long-term adjustments of MWG in response to changes in macroeconomic variables from low aggregation (countries) to high aggregation (regions). By analyzing the relationship between economic growth, urbanization, healthcare system quality, labor market functioning, demographic trends, educational level, technological advancement and MWG, the study fills a research gap and enhances understanding of waste management interventions. However, data availability and waste statistics accuracy should be considered. Future research could explore the relationship between macroeconomic variables and waste generation in sectors beyond MWG, such as industrial or construction waste, for a more comprehensive understanding of waste generation as a whole.Practical implicationsThe positive correlation between economic activity levels and waste generation in both the short and long terms, emphasizes the criticality of investing in waste reduction and recycling infrastructure to mitigate landfill waste. The negative correlation between population density and waste generation stresses the importance of strategic waste facility placement in low-density areas. To effectively manage higher MWG, tailored waste collection systems and initiatives promoting healthy lifestyles are of immense importance. The positive relationship between employment rates and waste generation underscores the necessity of waste reduction programs that generate employment opportunities. The positive correlation between fertility rates and waste generation emphasizes the need for the expansion of extended producer responsibility programs to include products and materials specifically associated with families and child-rearing. Education campaigns and governmental support for research and development (R&D) in waste reduction technologies are also integral components of an effective waste management strategy.Originality/valueThe short- and long-term adjustments of MWG reacts to shifts in macroeconomic variables from low aggregation (countries) to high aggregation (regions). Previous research has neglected the long-term information contained in variables by not incorporating the lagged error correction term (ETM). Neglecting this aspect could result in imprecise estimates of the elasticities.

Optimal Reporting Systems in Bank Runs

ABSTRACT We study the role of reporting systems in the context of bank runs. In our model, a bank receives an early but imprecise estimate of its investment performance, and its financial reporting system generates a report. We find that, from a financial-stability standpoint, the optimal reporting system requires full disclosure when the bank’s early estimate is below a certain threshold, but no disclosure otherwise. Importantly, such optimal reporting threshold should be tailored to the bank’s exposure to bank-run risk. In particular, the threshold is nonmonotonic and U-shaped in the bank-run risk. We also relate our results to current accounting standards and discuss their implications for policy-making and empirical research. JEL Classifications: G21; G28; M41; M48.

Welfare implications of information sharing by integrated e-retailers

ABSTRACT An integrated e-retailer competes with third-party sellers on the retailer’s platform. Due to its access to consumer data and its analytic capability, the integrated retailer possesses demand information that is not available to third-party sellers. When the information is not shared by the retailer, the third-party sellers make an imprecise estimate. Modelling the competition as a Cournot game, we show that by withholding information the retailer benefits at the cost of the third-party sellers only if the actual demand is more than that estimated by the third-party sellers and is less than a threshold. However, consumers’ surplus increases. The gain in consumers’ surplus may be greater than the retailer’s profit gain, resulting in an increase in social welfare. Increases in the commission rate, the number of third-party sellers, and the wholesale price of the product encourage the retailer to share demand information with the third-party sellers. The insights help integrated retailers decide their information sharing strategy, and the regulators in formulating antitrust laws.

Small Area with Multiply Imputed Survey Data

Abstract In this article, we propose a framework for small area estimation with multiply imputed survey data. Many statistical surveys suffer from (a) high nonresponse rates due to sensitive questions and response burden and (b) too small sample sizes to allow for reliable estimates on (unplanned) disaggregated levels due to budget constraints. One way to deal with missing values is to replace them by several plausible/imputed values based on a model. Small area estimation, such as the model by Fay and Herriot, is applied to estimate regionally disaggregated indicators when direct estimates are imprecise. The framework presented tackles simultaneously multiply imputed values and imprecise direct estimates. In particular, we extend the general class of transformed Fay-Herriot models to account for the additional uncertainty from multiple imputation. We derive three special cases of the Fay-Herriot model with particular transformations and provide point and mean squared error estimators. Depending on the case, the mean squared error is estimated by analytic solutions or resampling methods. Comprehensive simulations in a controlled environment show that the proposed methodology leads to reliable and precise results in terms of bias and mean squared error. The methodology is illustrated by a real data example using European wealth data.

A case study evaluating the effect of clustering, publication bias, and heterogeneity on the meta-analysis estimates in implant dentistry.

Meta-analyses may provide imprecise estimates when important meta-analysis parameters are not considered during the synthesis. The aim of this case study was to highlight the influence of meta-analysis parameters that can affect reported estimates using as an example pre-existing meta-analyses on the association between implant survival and sinus membrane perforation. PubMed was searched on 7 July 2021 for meta-analyses comparing implant failure in perforated and non-perforated sinus membranes. Primary studies identified in these meta-analyses were combined in a new random-effects model with odds ratios (ORs), confidence intervals (CIs), and prediction intervals reported. Using this new meta-analysis, further meta-analyses were then undertaken considering the clinical, methodological, and statistical heterogeneity of the primary studies, publication bias, and clustering effects. The meta-analyses with the greatest number and more homogeneous studies provided lower odds of implant failure in non-perforated sites (OR 0.49, 95 % CI=[0.26, 0.92]). However, when considering heterogeneity, publication bias, and clustering (number of implants), the confidence in these results was reduced. Interpretation of estimates reported in systematic reviews can vary depending on the assumptions made in the meta-analysis. Users of these analyses need to carefully consider the impact of heterogeneity, publication bias, and clustering, which can affect the size, direction, and interpretation of the reported estimates.

Bias and mean squared error in Mendelian randomization with invalid instrumental variables.

Mendelian randomization (MR) is a statistical method that utilizes genetic variants as instrumental variables (IVs) to investigate causal relationships between risk factors and outcomes. Although MR has gained popularity in recent years due to its ability to analyze summary statistics from genome-wide association studies (GWAS), it requires a substantial number of single nucleotide polymorphisms (SNPs) as IVs to ensure sufficient power for detecting causal effects. Unfortunately, the complex genetic heritability of many traits can lead to the use of invalid IVs that affect both the risk factor and the outcome directly or through an unobserved confounder. This can result in biased and imprecise estimates, as reflected by a larger mean squared error (MSE). In this study, we focus on the widely used two-stage least squares (2SLS) method and derive formulas for its bias and MSE when estimating causal effects using invalid IVs. Using those formulas, we identify conditions under which the 2SLS estimate is unbiased and reveal how the independent or correlated pleiotropic effects influence the accuracy and precision of the 2SLS estimate. We validate these formulas through extensive simulation studies and demonstrate the application of those formulas in an MR study to evaluate the causal effect of the waist-to-hip ratio on various sleeping patterns. Our results can aid in designing future MR studies and serve as benchmarks for assessing more sophisticated MR methods.

Constructing high frequency economic indicators by imputation

Summary Monthly and weekly economic indicators are often taken to be the largest common factor estimated from high and low frequency data, either separately or jointly. To incorporate mixed frequency information without directly modelling them, we target a low frequency diffusion index that is already available, and treat high frequency values as missing. We impute these values using multiple factors estimated from the high frequency data. In the empirical examples considered, static matrix completion that does not account for serial correlation in the idiosyncratic errors yields imprecise estimates of the missing values irrespective of how the factors are estimated. Single equation and systems-based dynamic procedures that account for serial correlation yield imputed values that are closer to the observed low frequency ones. This is the case in the counterfactual exercise that imputes the monthly values of consumer sentiment series before 1978 when the data was released only on a quarterly basis. This is also the case for a weekly version of the Chicago Fed National Activity Index of economic activity that is imputed using seasonally unadjusted data. The imputed series reveals episodes of increased variability of weekly economic information that are masked by the monthly data, notably around the 2014–2015 collapse in oil prices.

Reducing bias in density estimates for unmarked populations that exhibit reactive behaviour towards camera traps

Abstract Density estimates guide wildlife management, and camera traps are commonly used to estimate sizes of unmarked populations. Unfortunately, animals often alter their natural behaviour in the presence of camera traps, which may bias subsequent density estimates. We simulated 100 populations of known density to test several new and existing methods that aimed to reduce bias in density estimates from camera trap distance sampling (CTDS) and the random encounter model (REM). Within our simulated populations, we introduced different behavioural reactions including attraction towards cameras, freezing when near cameras and fleeing from cameras. CTDS and REM provided density estimates with decent coverage of confidence intervals (CTDS = 94%, REM = 87%), mean coefficient of variation (CTDS = 0.121, REM = 0.071) and minimal bias (root‐mean squared error: CTDS = 1.336, REM = 0.913) for simulated populations with no reactive behaviour. However, failure to implement a method to account for reactive behaviour resulted in low coverage, large bias and potentially imprecise density estimates when 30% of the simulated population potentially reacted by attraction to or fleeing from camera traps. We identified a corrective strategy that enhanced confidence interval coverage, increased precision and reduced bias for every behavioural reaction except when individuals potentially fled from cameras. Synthesis and applications. We provide empirically tested methods for reducing bias of density estimates. Wildlife managers requiring population estimates of animals that exhibit reactive behaviour can use our methods to reduce inaccuracy. We encourage future studies to quantify behavioural responses to camera traps and to implement, test and possibly extend our methods to reduce bias through simulation.

Dementia Attributable Healthcare Utilizations in the Caribbean versus United States.

Despite the high burden of Alzheimer's disease and other dementias among the Hispanic population worldwide, little is known about how dementia affects healthcare utilizations among this population outside of the US, in particular among those in the Caribbean region. This study examines healthcare utilization associated with Alzheimer's disease and other dementias among older adults in the Caribbean as compared to the US. We conducted harmonized analyses of two population-based surveys, the 10/66 Dementia Group Research data collected in Dominican Republic, Cuba, and Puerto Rico, and the US-based Health and Retirement Study. We examined changes in hospital nights and physician visits in response to incident and ongoing dementias. Incident dementia significantly increased the risk of hospitalization and number of hospital nights in both populations. Ongoing dementia increased the risk of hospitalization and hospital nights in the US, with imprecise estimates for the Caribbean. The number of physician visits was elevated in the US but not in the Caribbean. The concentration of increased healthcare utilization on hospital care and among patients with incident dementia suggests an opportunity for improved outpatient management of new and existing dementia patients in the Caribbean.