Inverse probability weighting is a fundamental and general methodology for estimating the causal effects of exposures and interventions, but standard approaches to constructing such weights are often suboptimal.

In this paper, we describe a recent approach for constructing such weights that directly balances covariates while optimizing the stability of the resulting weighting estimator.

To illustrate the use of this approach in mental health research, we present an exploratory study of the effects of exposure to violence on the risk of suicide attempt.

The direct balancing approach to weighting should be given strong consideration in empirical research due to its robustness and transparency in building weighting estimators.

To prevent chronic disease from getting worse, it is important to detect and predict it at an early stage. Therefore, the accuracy of the prediction is particularly important. To investigate the accuracy of different methods, this study compares the out-of-sample errors of machine learning algorithms and traditional econometric methods in predicting diabetes. The object of prediction in this study is fasting blood glucose, and the machine learning algorithms used are stepwise selection, bagging, random forests and support vector machine (SVM). In addition, we demonstrate the linear combination of above machine learning algorithms in this study. The findings indicate that the combined model outperforms both traditional econometric models and individual machine learning algorithms. However, the predictive performance of individual machine learning models does not consistently surpass that of traditional econometric approaches. Based on the data characteristics analyzed in this study, a possible explanation for this finding is that traditional econometric methods may exhibit superior performance in linear data prediction. Finally, the analysis of variable importance suggests that medical indicators and physical condition may play a more significant role in determining fasting blood glucose compared to hereditary factors. To further validate our results, we applied the same methodology to predict hypertension using the same dataset. The findings similarly indicated that the predictive ability of individual machine learning algorithms does not always surpass that of traditional econometric models. And a linear combination of the four machine learning algorithms enhances the predictive accuracy for hypertension.

Cholinesterase inhibitors (ChEIs) are recognized as first-line therapies for patients with mild-to-moderate dementia. However, there is limited comparative evidence regarding antipsychotic initiation risk among individual ChEIs to manage behavioral symptoms of dementia.

This study aims to evaluate and compare the risk of antipsychotic initiation among dementia patients prescribed individual ChEIs.

This is a retrospective cohort study using the 2009-2018 TriNetX electronic medical records data. Dementia patients aged over 60 years who were incident users of rivastigmine, donepezil, or galantamine with a 12-month washout period were included. Patients with a history of antipsychotic use during baseline and 30 days post-initiation of ChEIs were excluded. Patients were followed up to 12 months to identify the antipsychotic use. A generalized boosted model-based inverse probability treatment weights-adjusted Cox Proportional Hazard (CPH) model was applied to compare the risk of antipsychotic initiation across the different ChEIs.

Among the 7,878 eligible dementia patients initiating ChEIs, 89.40% (n=7,043) were incident donepezil users, followed by 8.13% of (n=641) rivastigmine users, and 2.46% (n=194) galantamine users. During the 12-month follow-up, 807 patients (10.24%) initiated antipsychotics. The CPH model showed that rivastigmine users were at an increased risk of antipsychotic use compared to donepezil users (adjusted hazard ratio=1.45, 95% confidence interval: 1.11-1.88). No significant difference was observed in the risk of antipsychotic initiation between galantamine and donepezil users.

This study found that rivastigmine users were more likely to initiate antipsychotics compared to donepezil users, while no significant difference between galantamine and donepezil users was observed. These findings emphasize the importance of careful medication monitoring and management to prevent prescribing cascades and reduce related adverse effects in dementia patients.

Targeted maximum likelihood estimation (TMLE) is an increasingly popular framework for the estimation of causal effects. It requires modeling both the exposure and outcome but is doubly robust in the sense that it is valid if at least one of these models is correctly specified. In addition, TMLE allows for flexible modeling of both the exposure and outcome with machine learning methods. This provides better control for measured confounders since the model specification automatically adapts to the data, instead of needing to be specified by the analyst a priori. Despite these methodological advantages, TMLE remains less popular than alternatives in part because of its less accessible theory and implementation. While some tutorials have been proposed, none address the case of a time-to-event outcome. This tutorial provides a detailed step-by-step explanation of the implementation of TMLE for estimating the effect of a point binary or multilevel exposure on a time-to-event outcome, modeled as counterfactual survival curves and causal hazard ratios. The tutorial also provides guidelines on how best to use TMLE in practice, including aspects related to study design, choice of covariates, controlling biases and use of machine learning. R-code is provided to illustrate each step using simulated data ( https://github.com/detal9/SurvTMLE). To facilitate implementation, a general R function implementing TMLE with options to use machine learning is also provided. The method is illustrated in a real-data analysis concerning the effectiveness of statins for the prevention of a first cardiovascular disease among older adults in Québec, Canada, between 2013 and 2018.

Variability exists in the degree of improvement patients experience following exercise-based interventions (EBIs) for knee osteoarthritis (KOA), but understanding of this heterogeneity is limited. Using a machine learning approach, this study leveraged data from two randomized controlled trials (RCTs) to identify patient characteristics contributing to differential treatment effects.

The RCTs enrolled n ​= ​621 patients and evaluated three EBIs (group-based physical therapy (PT), individual PT, and a Stepped Exercise Program) and an education control group. The primary outcome was change in total Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) score from baseline to end of treatment. Predictors included 25 demographic, clinical, and psychosocial characteristics. Three metalearners with three machine learning algorithms each and a simple interpretable model-based regression tree were used to identify subgroups with differential treatment effects. Fit was evaluated with holdout/validation data using root mean square error and mean absolute error.

The regression tree model outperformed all 9 metalearner models. Tree results suggested group-based PT yielded the largest improvement in mean WOMAC score. Only two subgroups were identified: baseline WOMAC score≤44 versus >44. Group-based PT was the optimal treatment regardless of baseline WOMAC score, but results were more ambiguous for patients with higher initial WOMAC score. For all 3 EBIs, patients with higher baseline WOMAC score made greater improvements.

Results suggest individuals with moderate or greater KOA symptoms may benefit more from EBIs than those with less severe symptoms and that group-based PT is a promising approach for KOA.

Propensity score estimation is often used as a preliminary step to estimate the average treatment effect with observational data. Nevertheless, misspecification of propensity score models undermines the validity of effect estimates in subsequent analyses. Prediction-based machine learning algorithms are increasingly used to estimate propensity scores to allow for more complex relationships between covariates. However, these approaches may not necessarily achieve covariates balancing. We propose a calibration-based method to better incorporate covariate balance properties in a general modeling framework. Specifically, we calibrate the loss function by adding a covariate imbalance penalty to standard parametric (e.g. logistic regressions) or machine learning models (e.g. neural networks). Our approach may mitigate the impact of model misspecification by explicitly taking into account the covariate balance in the propensity score estimation process. The empirical results show that the proposed method is robust to propensity score model misspecification. The integration of loss function calibration improves the balance of covariates and reduces the root-mean-square error of causal effect estimates. When the propensity score model is misspecified, the neural-network-based model yields the best estimator with less bias and smaller variance as compared to other methods considered.

One primary goal of precision medicine is to estimate the individualized treatment rules that optimize patients' health outcomes based on individual characteristics. Health studies with multiple treatments are commonly seen in practice. However, most existing individualized treatment rule estimation methods were developed for the studies with binary treatments. Many require that the outcomes are fully observed. In this article, we propose a matching-based machine learning method to estimate the optimal individualized treatment rules in observational studies with multiple treatments when the outcomes are fully observed or right-censored. We establish theoretical property for the proposed method. It is compared with the existing competitive methods in simulation studies and a hepatocellular carcinoma study.

Counterfactual inference at the distributional level presents new challenges with censored targets, especially in modern healthcare problems. To mitigate selection bias in this context, we exploit the intrinsic structure of reproducing kernel Hilbert spaces (RKHS) harnessing the notion of kernel mean embedding. This enables the development of a non-parametric estimator of counterfactual survival functions. We provide rigorous theoretical guarantees regarding consistency and convergence rates of our new estimator under general hypotheses related to smoothness of the underlying RKHS. We illustrate the practical viability of our methodology through extensive simulations and a relevant case study: The SPRINT trial. Our estimatort presents a distinct perspective compared to existing methods within the literature, which often rely on semi-parametric approaches and confront limitations in causal interpretations of model parameters.

Propensity scores are commonly used to reduce the confounding bias in non-randomized observational studies for estimating the average treatment effect. An important assumption underlying this approach is that all confounders that are associated with both the treatment and the outcome of interest are measured and included in the propensity score model. In the absence of strong prior knowledge about potential confounders, researchers may agnostically want to adjust for a high-dimensional set of pre-treatment variables. As such, variable selection procedure is needed for propensity score estimation. In addition, studies show that including variables related to treatment only in the propensity score model may inflate the variance of the treatment effect estimators, while including variables that are predictive of only the outcome can improve efficiency. In this article, we propose to incorporate outcome-covariate relationship in the propensity score model by including the predicted binary outcome probability as a covariate. Our approach can be easily adapted to an ensemble of variable selection methods, including regularization methods and modern machine-learning tools based on classification and regression trees. We evaluate our method to estimate the treatment effects on a binary outcome, which is possibly censored, across multiple treatment groups. Simulation studies indicate that incorporating outcome probability for estimating the propensity scores can improve statistical efficiency and protect against model misspecification. The proposed methods are applied to a cohort of advanced-stage prostate cancer patients identified from a private insurance claims database for comparing the adverse effects of four commonly used drugs for treating castration-resistant prostate cancer.

With the proliferation of anti-transgender policies in some U.S. jurisdictions, this study examines the general, mental, and physical health of transgender and cisgender populations.

Data from the 2020-2023 Behavioral Risk Factor Surveillance System were analyzed to examine associations between gender identity and health outcomes. Propensity score weighting was used to address potential imbalances among group characteristics. We conducted logistic regression for the binary outcome of self-rated health and quasi-Poisson regression for the number of days reporting poor mental and physical health.

Results reveal significant disparities in health outcomes, with transgender individuals reporting lower proportions of good general health and more days of poor mental and physical health compared to cisgender individuals. In the adjusted analyses, transgender individuals were significantly less likely to report good general health compared to cisgender peers (OR = 0.60, 95% CI = 0.52-0.69). Gender nonconforming (GNC), male-to-female (MTF), and female-to-male (FTM) individuals had lower odds of reporting good general health compared to cisgender individuals (GNC, OR = 0.46, 95% CI = 0.35-0.61; MTF, OR = 0.67, 95% CI = 0.53-0.85; FTM, OR = 0.71, 95% CI = 0.57-0.87). GNC individuals had an 86% higher frequency of poor mental health days (IRR = 1.86, 95% CI = 1.57-2.21) and a 37% higher frequency of poor physical health days (IRR = 1.37, 95% CI = 1.15-1.63) compared to cisgender counterparts. Similarly, MTF and FTM individuals had significantly higher frequencies of poor mental and physical health days.

The study highlights significant health disparities faced by transgender individuals, who report poorer general, mental, and physical health. These findings underscore the need to address the unique challenges and improve health outcomes within the transgender community.

Meta-analysis of randomized controlled trials is commonly used to evaluate treatments and inform policy decisions because it provides comprehensive summaries of all available evidence. However, meta-analyses are limited to draw population inference of treatment effects because they usually do not define target populations of interest specifically, and results of the individual randomized controlled trials in those meta-analyses may not generalize to the target populations. To leverage evidence from multiple randomized controlled trials in the generalizability context, we bridge the ideas from meta-analysis and causal inference. We integrate meta-analysis with causal inference approaches estimating target population average treatment effect. We evaluate the performance of the methods via simulation studies and apply the methods to generalize meta-analysis results from randomized controlled trials of treatments on schizophrenia to adults with schizophrenia who present to usual care settings in the United States. Our simulation results show that all methods perform comparably and well across different settings. The data analysis results show that the treatment effect in the target population is meaningful, although the effect size is smaller than the sample average treatment effect. We recommend applying multiple methods and comparing the results to ensure robustness, rather than relying on a single method.

Isolated anterior cerebral artery occlusions (ACAo) in patients with acute ischemic stroke present significant challenges due to their rarity. The efficacy and safety of endovascular therapy (EVT) in comparison with best medical therapy (BMT) for ACAo remains unclear. This study aimed to assess the outcomes of these treatments.

This multinational, multicenter study analyzed data from the MAD-MT registry. Data were collected retrospectively from 37 sites across North America, Asia, and Europe. Inverse probability of treatment weighting (IPTW) was applied to balance confounding variables. The primary outcome was functional independence (modified Rankin Scale (mRS) scores of 0-2) at 90 days. Secondary outcomes included excellent outcomes (mRS 0-1), mortality at 90 days, and NIH Stroke Scale (NIHSS) score on day 1 post treatment.

Of the 108 patients, 36 received BMT and 72 underwent EVT. The median age was 75 years, and 56% were male. At 90 days, 40% of patients achieved mRS 0-2, with no significant difference between EVT and BMT (38% vs 45%, p=0.46). Procedural success (mTICI 2b-3) was 91% in the EVT group, with a sICH rate of 2.9%. IPTW-adjusted analysis showed no significant difference between EVT and BMT for functional independence (OR 1.17, 95% CI 0.23 to 6.02, p=0.85), mortality (25% vs 21%, p=0.71) or day 1 NIHSS scores (Beta 2.2, 95% CI -0.51 to 4.8, p=0.11).

EVT showed high procedural success but did not significantly improve functional outcomes or mortality compared with BMT in patients with ACAo. Further randomized trials are needed to clarify EVT's role in ACAo.

This response to a Commentary by Delgado-Ron, Jeyabalan, Watt, and Salway (2024) on Cimpian, Timmer, and Kim's (2023) paper discusses and clarifies some key issues in applying the new data-validity sensitivity analysis proposed by Cimpian et al. (2023). The differences in the applications of the method by Delgado-Ron et al. (2024) and Cimpian et al. (2023) present an opportunity to recognize the possibilities of this method, while also noting some challenges and limitations. This response-commentary focuses on five key areas: (1) surmising a set of likely motivations of the survey respondents, (2) selecting the screener items, (3) considering the outcomes examined, (4) reflecting on null results in a sensitivity analysis, and (5) recognizing the advantages and disadvantages of the ease of this new method.

Ruxolitinib is a JAK1/JAK2 inhibitor approved for the treatment of myelofibrosis (MF)-related splenomegaly or symptoms. The recommended starting dose depends on platelet count, regardless of haemoglobin level at baseline. In the recent years, an overall survival (OS) advantage was reported in patients treated with ruxolitinib compared with best available therapy. We analysed a large Italian cohort of 3494 patients identified by Agenzia Italiana del Farmaco (AIFA) monitoring registries. Of them, 2337 (66.9%) started at reduced dose: these patients were older (median age 70 vs. 67), with increased incidence of large splenomegaly (longitudinal diameter 20 vs. 19.1 cm, median volume 1064 cm3 vs. 1016 cm3), with higher IPSS risk (30.9% vs. 26.1%), and worse ECOG score (more than 1 in 14.3% vs. 9.8%). After balancing for baseline characteristics, Kaplan-Meier analysis showed a median OS of 78.2 months (95% CI 65.9-89) for patients who started at full dose and 52.6 (95% CI 49-56.6) months for patients who started with reduced dose (p < 0.001). Group analysis also showed a substantial difference in patients with intermediate-2 and high IPSS risk. The majority of MF patients in real-world analysis started with a reduced dose of ruxolitinib, which is associated with less favourable outcomes.

In the framework of causal inference, average treatment effect (ATE) is one of crucial concerns. To estimate it, the propensity score based estimation method and its variants have been widely adopted. However, most existing methods were developed by assuming that binary treatments are precisely measured. In addition, propensity scores are usually formulated as parametric models with respect to confounders. However, in the presence of measurement error in binary treatments and nonlinear relationship between treatments and confounders, existing methods are no longer valid and may yield biased inference results if these features are ignored. In this paper, we first analytically examine the impact of estimation of ATE and derive biases for the estimator of ATE when treatments are contaminated with measurement error. After that, we develop a valid method to address binary treatments with misclassification. Given the corrected treatments, we adopt the random forest method to estimate the propensity score with nonlinear confounders accommodated and then derive the estimator of ATE. Asymptotic properties of the error-eliminated estimator are established. Numerical studies are also conducted to assess the finite sample performance of the proposed estimator, and numerical results verify the importance of correcting for measurement error effects.

External data (e.g., real-world data (RWD) and historical data) have become more readily available. This has led to rapidly increasing interest in exploring and evaluating ways of utilizing external data to facilitate traditional clinical trials (TCT), especially for rare diseases with high unmet medical needs where a TCT would be impractical and/or unethical. In this article, we focus on hybrid studies that incorporate external data into randomized clinical trials to augment the control arm and explore a complex innovative design. A sequential adaptive design conducts multiple interim assessments to improve the accuracy of estimates of agreement between external data and current data. At each interim assessment, we apply the inverse probability weighted power prior (IPW-PP) method to adaptively borrow information from external data to account for confounding and heterogeneity. The randomization ratio is dynamically adjusted during the interim assessment based on accumulatively augmented information to reduce the sample size of the current trial. Additionally, the proposed design can be extended to allow interim analyses for early efficacy/futility stopping, that is, early assessment of trial success or failure based on accumulated data, potentially reducing ineffective treatment exposure and unnecessary time and resources. The performance of the proposed method and design is evaluated via extensive simulation studies. The sequential adaptive design and IPW-PP approach having desirable properties are implemented.

In the worldwide, real-life setting, some candidates for right colectomy still receive no bowel preparation, some receive oral antibiotics alone, some receive mechanical bowel preparation alone, and some receive mechanical bowel preparation with oral antibiotics, with varying degrees of compliance to preoperative intravenous antibiotic prophylaxis. Previous studies mainly focused on left-sided colorectal anastomoses while less attention has been devoted to right-sided ileocolic anastomoses. When high-level evidence from randomized clinical trials is lacking, multiple-treatment propensity score weighting analysis of prospective data on the basis of generalized boosted model is superior to a simple propensity score-matching analysis and to an inverse probability weighting in terms of external validity and bias reduction.

This is an analysis on the basis of machine-learning procedures of 2,617 patients who underwent elective right colectomies.

The risk of surgical-site infections (5.0% after no bowel preparation) was significantly lower after mechanical bowel preparation with oral antibiotics (4.0%, P = .017), significantly greater after mechanical bowel preparation alone (8.6%, P = .019), and comparable after oral antibiotics alone (3.9%). The risk of anastomotic leakage (3.2% after no bowel preparation) was significantly greater after oral antibiotics alone (4.8%, P = .013). Concerning secondary outcomes, no significant differences were recorded for the risk of overall morbidity and reoperation. The risk of readmission (3.0% after no bowel preparation) was significantly reduced after mechanical bowel preparation with oral antibiotics (1.5%, P = .046), and the risk of major morbidity (5.1% after no bowel preparation) was significantly greater after oral antibiotics alone (6.7%, P = .007).

This multitreatment machine-learning analysis, despite some limitations, showed that mechanical bowel preparation with oral antibiotics is associated with a decrease in surgical-site infections after elective right colectomy compared with no bowel preparation.

Mutations in KRAS and NRAS are predictive of poor response to cetuximab and panitumumab, two anti-epidermal growth factor receptor (EGFR) monoclonal antibodies used in metastatic colorectal cancer (mCRC). Our objective was to explore the value of using KRAS and NRAS mutation status to inform third-line anti-EGFR therapy for mCRC using the value of heterogeneity (VOH) framework.

We used administrative data to identify mCRC patients who were potentially eligible for third-line therapy in 2006-2019 in British Columbia (BC), Canada. We compared three alternative stratification policies in place during the study period: the unstratified policy where anti-EGFR therapy was not offered (2006-2009), stratification by KRAS mutation (2009-2016), and stratification by KRAS+NRAS mutation (2016-2019). We used inverse-probability-of-treatment weighting to balance covariates across the three groups. Cost and survival time were calculated using a 3-year time horizon and adjusted for censoring, with bootstrapping to characterize uncertainty. Mean net monetary benefit (NMB) was calculated at a range of threshold values. The VOH of using KRAS and NRAS mutation status to inform treatment selection was calculated as the change in NMB with increasing stratification, under current (static VOH) or perfect (dynamic VOH) information.

We included 2664 patients in the analysis. At a willingness-to-pay of CA$100,000/ life-year gained (LYG), stratification on KRAS mutation status provided a static VOH of CA$1565 per patient; further stratification on KRAS+NRAS provided additional static VOH of CA$594. The static VOH exceeded the marginal cost of genomic testing under both policies.

Stratification of anti-EGFR therapy by KRAS and NRAS mutation status can provide additional value at a threshold of CA$100,000/LYG. There is diminishing marginal value and increasing marginal costs as the policy becomes more stratified. The VOH framework can illustrate the value of subgroup-specific decisions in a comprehensive way, to better inform targeted treatment policies.

(1) Background: Quasi-experimental design has been widely used in causal inference for health policy impact evaluation. However, due to the non-randomized treatment used, there is great potential for bias in the assessment of the results, which can be reduced by using propensity score (PS) methods. In this context, this article aims to map the literature concerning the use of machine learning (ML) algorithms for propensity score estimation. (2) Methods: A scoping review was carried out in the PubMed, EMBASE, ACM Digital Library, IEEE Explore, LILACS, Web of Science, Scopus, Compendex, and gray literature (ProQuest and Google Scholar) databases, based on the PRISMA-ScR guidelines. This scoping review aims to identify ML models and their accuracy and the characteristics of studies on causal inference for health policy impacts, with a specific focus on PS estimation using ML. (3) Results: Seven studies were included in the review from 3018 references searched. In general, tree-based ML models were used for PS estimation. Most of the studies did not show or mention the performance metrics of the selected models, focusing instead on discussing the treatment effects under analysis. (4) Conclusions: Despite important aspects of model development and evaluation being under-reported, this scoping review provides insights into the recent use of ML algorithms in health policy impact evaluation.

We tested the association between early childcare attendance in the first three years of life and child development at age 3.5 years in the French context, where early childcare is subsidized. In the ELFE (Étude Longitudinale Français depuis l'Enfance) birth cohort study set in metropolitan France, children's development was reported by parents at age 3.5 years (n = 11,033) via the Child Development Inventory (CDI) questionnaire. CDI scores were transformed into a development quotient (DQ), with a DQ < 90 corresponding to possible and a DQ < 85 corresponding to a probable developmental delay. Inverse probability weighted multivariable regression models were used to analyse whether early childcare in the first three years of life (centre-based, childminder, informal or parental care) was associated to development delay. Compared to children in exclusive parental care, those in centre-based childcare (CBC) or with a childminder prior to school entry were significantly less likely to experience possible (OR = 0.56, [95% CI = 0.51-0.61] for CBC and OR = 0.77, [95% CI = 0.72-0.83] for childminder attendance) and probable developmental delay (OR = 0.62, [0.58-0.67] for CBC and OR = 0.80 [0.76-0.83] for childminder). Informal childcare attendance was not significantly associated with children's possible nor probable developmental delay ((OR = 0.97, [0.84-1.12]) and (OR = 0.97, [0.82-1.15]), respectively). Conclusions: Overall, our findings add to the existing scientific literature, showing that in the French context, where childcare can start as early as 3 months of age, early childcare attendance can contribute to child's development. What's Known on This Subject: • Studies on early childcare attendance and child development have shown mixed results, associations with better psychomotor development mainly being observed in Nordic countries, while some studies in other countries such as the USA showed no or negative associations. What This Study Adds: • In a country with broad and subsidized access to childcare such as France, access to early childhood education can positively contribute to children's psychomotor development. However, we found that access to childcare does not appear to reduce social inequalities in children's psychomotor development.

Machine learning (ML) methods are promising and scalable alternatives for propensity score (PS) estimation, but their comparative performance in disease risk score (DRS) estimation remains unexplored.

We used real-world data comparing antihypertensive users to non-users with 69 negative control outcomes, and plasmode simulations to study the performance of ML methods in PS and DRS estimation. We conducted a cohort study using UK primary care records. Further, we conducted a plasmode simulation with synthetic treatment and outcome mimicking empirical data distributions. We compared four PS and DRS estimation methods: 1. Reference: Logistic regression including clinically chosen confounders. 2. Logistic regression with L1 regularisation (LASSO). 3. Multi-layer perceptron (MLP). 4. Extreme Gradient Boosting (XgBoost). Covariate balance, coverage of the null effect of negative control outcomes (real-world data) and bias based on the absolute difference between observed and true effects (for plasmode) were estimated. 632,201 antihypertensive users and nonusers were included.

ML methods outperformed the reference method for PS estimation in some scenarios, both in terms of covariate balance and coverage/bias. Specifically, XgBoost achieved the best performance. DRS-based methods performed worse than PS in all tested scenarios.

We found that ML methods could be reliable alternatives for PS estimation. ML-based DRS methods performed worse than PS ones, likely given the rarity of outcomes.

Despite the proven effectiveness of endovascular therapy (EVT) in acute ischemic strokes (AIS) involving anterior circulation large vessel occlusions, isolated posterior cerebral artery (PCA) occlusions (iPCAo) remain underexplored in clinical trials. This study investigates the comparative effectiveness and safety of EVT against medical management (MM) in patients with iPCAo.

This multinational, multicenter propensity score-weighted study analyzed data from the Multicenter Analysis of primary Distal medium vessel occlusions: effect of Mechanical Thrombectomy (MAD-MT) registry, involving 37 centers across North America, Asia, and Europe. We included iPCAo patients treated with either EVT or MM. The primary outcome was the modified Rankin Scale (mRS) at 90 days, with secondary outcomes including functional independence, mortality, and safety profiles such as hemorrhagic complications.

A total of 177 patients were analyzed (88 MM and 89 EVT). EVT showed a statistically significant improvement in 90-day mRS scores (OR = 0.55, 95% CI = 0.30-1.00, p = 0.048), functional independence (OR = 2.52, 95% CI = 1.02-6.20, p = 0.045), and a reduction in 90-day mortality (OR = 0.12, 95% CI = 0.03-0.54, p = 0.006) compared to MM. Hemorrhagic complications were not significantly different between the groups.

EVT for iPCAo is associated with better neurological outcomes and lower mortality compared to MM, without an increased risk of hemorrhagic complications. Nevertheless, these results should be interpreted with caution due to the study's observational design. The findings are hypothesis-generating and highlight the need for future randomized controlled trials to confirm these observations and establish definitive treatment guidelines for this patient population.

Analyses of cluster randomized trials (CRTs) can be complicated by informative missing outcome data. Methods such as inverse probability weighted generalized estimating equations have been proposed to account for informative missingness by weighing the observed individual outcome data in each cluster. These existing methods have focused on settings where missingness occurs at the individual level and each cluster has partially or fully observed individual outcomes. In the presence of missing clusters, for example, all outcomes from a cluster are missing due to drop-out of the cluster, these approaches ignore this cluster-level missingness and can lead to biased inference if the cluster-level missingness is informative. Informative missingness at multiple levels can also occur in CRTs with a multi-level structure where study participants are nested in subclusters such as healthcare providers, and the subclusters are nested in clusters such as clinics. In this paper, we propose new estimators for estimating the marginal treatment effect in CRTs accounting for missing outcome data at multiple levels based on weighted generalized estimating equations. We show that the proposed multi-level multiply robust estimator is consistent and asymptotically normally distributed provided that one of the multiple propensity score models postulated at each clustering level is correctly specified. We evaluate the performance of the proposed method through extensive simulations and illustrate its use with a CRT evaluating a Malaria risk-reduction intervention in rural Madagascar.

Propensity scores (PS) are typically evaluated using balance metrics that focus on covariate balance, often without considering their predictive power for the outcome. This approach may not always result in optimal bias reduction in the treatment effect estimate. To address this issue, evaluating covariate balance through prognostic scores, which account for the relationship between covariates and the outcome, has been proposed. Similarly, using a typical model averaging approach for PS estimation that minimizes prediction error for treatment status and covariate imbalance does not necessarily optimize PS-based confounding adjustment. As an alternative approach, using the averaged PS model that minimizes inter-group differences in the prognostic score may further reduce bias in the treatment effect estimate. Moreover, since the prognostic score is also an estimated quantity, model averaging in the prognostic scores can help identify a better prognostic score model. Utilizing the model-averaged prognostic scores as the balance metric for constructing the averaged PS model can contribute to further decreasing bias in treatment effect estimates. This paper demonstrates the effectiveness of the PS model averaging approach based on prognostic score balance and proposes a method that uses the model-averaged prognostic score as a balance metric, evaluating its performance through simulations and empirical analysis.

We conduct a series of simulations alongside an analysis of empirical observational data to compare the performances of weighted treatment effect estimates using the proposed and existing approaches. In our examination, we separately provid four candidate estimates for the PS and prognostic score models using traditional regression and machine learning methods. The model averaging of PS based on these candidate estimators is performed to either maximize the prediction accuracy of the treatment or to minimize intergroup differences in covariate distributions or prognostic scores. We also utilize not only the prognostic scores from each candidate model but also an averaged score that best predicted the outcome, for the balance assessment.

The simulation and empirical data analysis reveal that our proposed model-averaging approaches for PS estimation consistently yield lower bias and less variability in treatment effect estimates across various scenarios compared to existing methods. Specifically, using the optimally averaged prognostic scores as a balance metric significantly improves the robustness of the weighted treatment effect estimates.

The prognostic score-based model averaging approach for estimating PS can outperform existing model averaging methods. In particular, the estimator using the model averaging prognostic score as a balance metric can produce more robust estimates. Since our results are obtained under relatively simple conditions, applying them to real data analysis requires adjustments to obtain accurate estimates according to the complexity and dimensionality of the data.

Using the prognostic score as the balance metric for the PS model averaging enhances the performance of the treatment effect estimator, which can be recommended for a wide variety of situations. When applying the proposed method to real-world data, it is important to use it in conjunction with techniques that mitigate issues arising from the complexity and high dimensionality of the data.

The incorporation of real-world data (RWD) into medical product development and evaluation has exhibited consistent growth. However, there is no universally adopted method of how much information to borrow from external data. This paper proposes a study design methodology called Tree-based Monte Carlo (TMC) that dynamically integrates patients from various RWD sources to calculate the treatment effect based on the similarity between clinical trial and RWD. Initially, a propensity score is developed to gauge the resemblance between clinical trial data and each real-world dataset. Utilizing this similarity metric, we construct a hierarchical clustering tree that delineates varying degrees of similarity between each RWD source and the clinical trial data. Ultimately, a Gaussian process methodology is employed across this hierarchical clustering framework to synthesize the projected treatment effects of the external group. Simulation result shows that our clustering tree could successfully identify similarity. Data sources exhibiting greater similarity with clinical trial are accorded higher weights in treatment estimation process, while less congruent sources receive comparatively lower emphasis. Compared with another Bayesian method, meta-analytic predictive prior (MAP), our proposed method's estimator is closer to the true value and has smaller bias.

The use of propensity score methods has become ubiquitous in causal inference. At the heart of these methods is the positivity assumption. Violation of the positivity assumption leads to the presence of extreme propensity score weights when estimating average causal effects, which affects statistical inference. To circumvent this issue, trimming or truncating methods have been widely used. Unfortunately, these methods require that we pre-specify a threshold. There are a number of alternative methods to deal with the lack of positivity when we estimate the average treatment effect (ATE). However, no other methods exist beyond trimming and truncation to deal with the same issue when the goal is to estimate the average treatment effect on the treated (ATT). In this article, we propose a propensity score weight-based alternative for the ATT, called overlap weighted average treatment effect on the treated. The appeal of our proposed method lies in its ability to obtain similar or even better results than trimming and truncation while relaxing the constraint to choose an a priori threshold (or related measures). The performance of the proposed method is illustrated via a series of Monte Carlo simulations and a data analysis on racial disparities in health care expenditures.

Among clinical trialists, there has been a growing interest in using external data to improve decision-making and accelerate drug development in randomized clinical trials (RCTs). Here we propose a novel approach that combines the propensity score weighting (PW) and the multi-source exchangeability modelling (MEM) approaches to augment the control arm of a RCT in the rare disease setting. First, propensity score weighting is used to construct weighted external controls that have similar observed pre-treatment characteristics as the current trial population. Next, the MEM approach evaluates the similarity in outcome distributions between the weighted external controls and the concurrent control arm. The amount of external data we borrow is determined by the similarities in pretreatment characteristics and outcome distributions. The proposed approach can be applied to binary, continuous and count data. We evaluate the performance of the proposed PW-MEM method and several competing approaches based on simulation and re-sampling studies. Our results show that the PW-MEM approach improves the precision of treatment effect estimates while reducing the biases associated with borrowing data from external sources.

Overlap weighting, using weights defined as the probability of receiving the opposite treatment, is a relatively new, alternative propensity score-based weighting technique used to adjust for confounding when estimating causal treatment effects. It has preferable properties compared to inverse probability of treatment weighting, such as exact covariate balance, safeguards against extreme weights and emphasis on medical equipoise, where treatment decisions are most uncertain. In this article, we introduce the overlap weighting methodology, compare it to inverse probability of treatment weighting and provide some strategies for assessing weighting impact, through an applied example of hospital mortality. When the propensity score distributions have large separation, inverse probability of treatment weighting has been shown to produce biased and less efficient estimates of the treatment effect, making overlap weighting a preferred method in such cases.

Motor vehicle crashes (MVCs) are a leading cause of death for law enforcement officers (LEOs) in the U.S. LEOs and more specifically novice LEOs (nLEOs) are susceptible to high cognitive workload while driving which can lead to fatal MVCs. The objective of this study was to develop a machine learning algorithm (MLA) that can estimate cognitive workload of LEOs while performing secondary tasks in a patrol vehicle. A ride-along study was conducted with 24 nLEOs. Participants performed their normal patrol operations while their physiological responses such as heartrate, eye movement, and galvanic skin response were recorded using unobtrusive devices. Findings suggested that the random forest algorithm could predict cognitive workload with relatively high accuracy (>70%) given that it was entirely reliant on physiological signals. The developed MLA can be used to develop adaptive in-vehicle technology based on real-time estimation of cognitive workload, which can reduce the risk of MVCs in police operations.

Pilot tends to have a high prevalence of dyslipidemia. The present study aimed to identify key factors of pilot hyperlipidemia through thorough analysis of physical examination data, and to provide pilot-targeted health guidance to manage hyperlipidemia risks. The physical examination data of 1,253 pilot inpatients from January 2019 to June 2022, were evaluated and divided into two groups based on whether or not the pilot had hyperlipidemia. A total of three multivariate analysis models including logistic model, multilevel model and boosting propensity score were applied to find the risk factors of pilot hyperlipidemia. In the group of pilots with hyperlipidemia, four risk factors, including thrombin time, carbohydrate antigen 199, lymphocyte count and rheumatoid factor, were significantly different from pilots without hyperlipidemia, which might be positively associated with the incidence of hyperlipidemia. In future studies regarding pilots, whether hyperlipidemia is connected to abnormalities in thrombin time, carbohydrate antigen 199 and rheumatoid factor should be further explored. Based on the findings of the present study, pilot health management should be more refined and personalized, and attention should be paid to the risk factors of hyperlipidemia including diet and lifestyle.

Current evidence about intraoperative anastomotic testing after left-sided colorectal resections is still controversial. The aim of this study was to analyze the impact of Indocyanine Green fluorescent angiography (ICG-FA) and air-leak test (ALT) over standard assessment on anastomotic leakage (AL) rates according to surgeon's perception of anastomosis perfusion and/or integrity in clinical practice.

A database of 2061 patients who underwent left-sided colorectal resections was selected from patients enrolled in a prospective multicenter study. It was retrospectively analyzed through a multi-treatment machine-learning model considering standard visual assessment (NW; No. = 899; 43.6%) as the reference treatment arm, compared to ICG-FA alone (WP; No. = 409; 19.8%), ALT alone (WI; No. = 420; 20.4%) or both (WPI; No. = 333; 16.2%). Twenty-four covariates potentially affecting the outcomes were included and balanced into the model within the subgroups. The primary endpoint was AL, the secondary endpoints were overall morbidity (OM), major morbidity (MM), reoperation for AL, and mortality. All the results were reported as odds ratio (OR) with 95% confidence intervals (95%CI).

The WPI subgroup showed significantly higher AL risk (OR 1.91; 95% CI 1.02-3.59; p 0.043), MM risk (OR 2.35; 95% CI 1.39-3.97; p 0.001), and reoperation for AL risk (OR 2.44; 95% CI 1.12-5.31; p 0.025). No other significant differences were recorded.

This study showed that the surgeons' perception of both anastomotic perfusion and integrity (WPI subgroup) was associated to a significantly higher risk of AL and related morbidity, notwithstanding the extensive use of both ICG-FA and ALT testing.

Propensity score matching is vital in epidemiological studies using observational data, yet its estimates relies on correct model-specification. This study assesses supervised deep learning models and unsupervised autoencoders for propensity score estimation, comparing them with traditional methods for bias and variance accuracy in treatment effect estimations.

Utilizing a plasmode simulation based on the Right Heart Catheterization dataset, under a variety of settings, we evaluated (1) a supervised deep learning architecture and (2) an unsupervised autoencoder, alongside two traditional methods: logistic regression and a spline-based method in estimating propensity scores for matching. Performance metrics included bias, standard errors, and coverage probability. The analysis was also extended to real-world data, with estimates compared to those obtained via a double robust approach.

The analysis revealed that supervised deep learning models outperformed unsupervised autoencoders in variance estimation while maintaining comparable levels of bias. These results were supported by analyses of real-world data, where the supervised model's estimates closely matched those derived from conventional methods. Additionally, deep learning models performed well compared to traditional methods in settings where exposure was rare.

Supervised deep learning models hold promise in refining propensity score estimations in epidemiological research, offering nuanced confounder adjustment, especially in complex datasets. We endorse integrating supervised deep learning into epidemiological research and share reproducible codes for widespread use and methodological transparency.

In the healthcare landscape, data science (DS) methods have emerged as indispensable tools to harness real-world data (RWD) from various data sources such as electronic health records, claim and registry data, and data gathered from digital health technologies. Real-world evidence (RWE) generated from RWD empowers researchers, clinicians, and policymakers with a more comprehensive understanding of real-world patient outcomes. Nevertheless, persistent challenges in RWD (e.g., messiness, voluminousness, heterogeneity, multimodality) and a growing awareness of the need for trustworthy and reliable RWE demand innovative, robust, and valid DS methods for analyzing RWD. In this article, I review some common current DS methods for extracting RWE and valuable insights from complex and diverse RWD. This article encompasses the entire RWE-generation pipeline, from study design with RWD to data preprocessing, exploratory analysis, methods for analyzing RWD, and trustworthiness and reliability guarantees, along with data ethics considerations and open-source tools. This review, tailored for an audience that may not be experts in DS, aspires to offer a systematic review of DS methods and assists readers in selecting suitable DS methods and enhancing the process of RWE generation for addressing their specific challenges.

Observational studies are frequently used in clinical research to estimate the effects of treatments or exposures on outcomes. To reduce the effects of confounding when estimating treatment effects, covariate balancing methods are frequently implemented. This study evaluated, using extensive Monte Carlo simulation, several methods of covariate balancing, and two methods for propensity score estimation, for estimating the average treatment effect on the treated using a hazard ratio from a Cox proportional hazards model. With respect to minimizing bias and maximizing accuracy (as measured by the mean square error) of the treatment effect, the average treatment effect on the treated weighting, fine stratification, and optimal full matching with a conventional logistic regression model for the propensity score performed best across all simulated conditions. Other methods performed well in specific circumstances, such as pair matching when sample sizes were large (n = 5000) and the proportion treated was < 0.25. Statistical power was generally higher for weighting methods than matching methods, and Type I error rates were at or below the nominal level for balancing methods with unbiased treatment effect estimates. There was also a decreasing effective sample size with an increasing number of strata, therefore for stratification-based weighting methods, it may be important to consider fewer strata. Generally, we recommend methods that performed well in our simulations, although the identification of methods that performed well is necessarily limited by the specific features of our simulation. The methods are illustrated using a real-world example comparing beta blockers and angiotensin-converting enzyme inhibitors among hypertensive patients at risk for incident stroke.

Inverse probability weighting (IPW) is frequently used to reduce or minimize the observed confounding in observational studies. IPW creates a pseudo-sample by weighting each individual by the inverse of the conditional probability of receiving the treatment level that he/she has actually received. In the pseudo-sample there is no variation among the multiple individuals generated by weighting the same individual in the original sample. This would reduce the variability of the data and therefore bias the variance estimate in the target population. Conventional variance estimation methods for IPW estimators generally ignore this underestimation and tend to produce biased estimates of variance. We here propose a more reasonable method that incorporates this source of variability by using parametric bootstrapping based on intra-stratum variability estimates. This approach firstly uses propensity score stratification and intra-stratum standard deviation to approximate the variability among multiple individuals generated based on a single individual whose propensity score falls within the corresponding stratum. The parametric bootstrapping is then used to incorporate the target variability by re-generating outcomes after adding a random error term to the original data. The performance of the proposed method is compared with three existing methods including the naïve model-based variance estimator, the nonparametric bootstrap variance estimator, and the robust variance estimator in the simulation section. An example of patients with sarcopenia is used to illustrate the implementation of the proposed approach. According to the results, the proposed approach has desirable statistical properties and can be easily implemented using the provided R code.

During the 2017-18 influenza season in the USA, there was a high incidence of influenza illness and mortality. However, no apparent antigenic change was identified in the dominant H3N2 viruses, and the severity of the season could not be solely attributed to a vaccine mismatch. We aimed to investigate whether the altered virus properties resulting from gene reassortment were underlying causes of the increased case number and disease severity associated with the 2017-18 influenza season.

Samples included were collected from patients with influenza who were prospectively recruited during the 2016-17 and 2017-18 influenza seasons at the Johns Hopkins Hospital Emergency Departments in Baltimore, MD, USA, as well as from archived samples from Johns Hopkins Health System sites. Among 647 recruited patients with influenza A virus infection, 411 patients with whole-genome sequences were available in the Johns Hopkins Center of Excellence for Influenza Research and Surveillance network during the 2016-17 and 2017-18 seasons. Phylogenetic trees were constructed based on viral whole-genome sequences. Representative viral isolates of the two seasons were characterised in immortalised cell lines and human nasal epithelial cell cultures, and patients' demographic data and clinical outcomes were analysed.

Unique H3N2 reassortment events were observed, resulting in two predominant strains in the 2017-18 season: HA clade 3C.2a2 and clade 3C.3a, which had novel gene segment constellations containing gene segments from HA clade 3C.2a1 viruses. The reassortant re3C.2a2 viruses replicated with faster kinetics and to a higher peak titre compared with the parental 3C.2a2 and 3C.2a1 viruses (48 h vs 72 h). Furthermore, patients infected with reassortant 3C.2a2 viruses had higher Influenza Severity Scores than patients infected with the parental 3C.2a2 viruses (median 3·00 [IQR 1·00-4·00] vs 1·50 [1·00-2·00]; p=0·018).

Our findings suggest that the increased severity of the 2017-18 influenza season was due in part to two intrasubtypes, cocirculating H3N2 reassortant viruses with fitness advantages over the parental viruses. This information could help inform future vaccine development and public health policies.

The Center of Excellence for Influenza Research and Response in the US, National Science and Technology Council, and Chang Gung Memorial Hospital in Taiwan.

In a prospective observational study (POS) designed to assess the average causal effect of a treatment (e.g. Drug A) compared to a comparator (e.g. Drug B) in the treatment population, enrolling all patients who are assigned to the treatments of interest for follow-up has a potentially large negative impact on the statistical efficiency and bias of the analysis of the outcomes and on the cost of the study. "Up-front matching" is an innovative enrollment method for selecting patients for long-term follow-up among those who have already been assigned to treatment or comparator which uses frequency matching and hence avoids the restrictions of individual matching that other methods have used. To achieve potential statistical and logistical efficiencies in the POS, in up-front matching, a target population is defined based on a retrospective database which then enables selecting populations of patients for follow-up that have desirable statistical properties. In particular, the resulting populations of patients who are enrolled look like the population of treatment patients were randomized to treatment or comparator for the baseline covariates that are used to select patients for follow-up. The method is illustrated in detail for a study designed to assess the effect of injectable antipsychotics versus oral antipsychotics.

Estimating causal effects from large experimental and observational data has become increasingly prevalent in both industry and research. The bootstrap is an intuitive and powerful technique used to construct standard errors and confidence intervals of estimators. Its application however can be prohibitively demanding in settings involving large data. In addition, modern causal inference estimators based on machine learning and optimization techniques exacerbate the computational burden of the bootstrap. The bag of little bootstraps has been proposed in non-causal settings for large data but has not yet been applied to evaluate the properties of estimators of causal effects. In this article, we introduce a new bootstrap algorithm called causal bag of little bootstraps for causal inference with large data. The new algorithm significantly improves the computational efficiency of the traditional bootstrap while providing consistent estimates and desirable confidence interval coverage. We describe its properties, provide practical considerations, and evaluate the performance of the proposed algorithm in terms of bias, coverage of the true 95% confidence intervals, and computational time in a simulation study. We apply it in the evaluation of the effect of hormone therapy on the average time to coronary heart disease using a large observational data set from the Women's Health Initiative.

The United States Centers for Disease Control and Prevention (CDC) advises testing individuals for COVID-19 after exposure or if they display symptoms. However, a deeper understanding of demographic factors associated with testing hesitancy is necessary.

A US nationwide cross-sectional survey of adults with risk factors for developing severe COVID-19 ("high-risk" individuals) was conducted from August 18-September 5, 2023. Objectives included characterizing demographics and attitudes associated with COVID-19 testing. Inverse propensity weighting was used to weight the data to accurately reflect the high-risk adult US population as reflected in IQVIA medical claims data. We describe here the weighted results modeled to characterize demographic factors driving hesitancy.

In the weighted sample of 5019 respondents at high risk for severe COVID-19, 58.2% were female, 37.8% were ≥ 65 years old, 77.1% were White, and 13.9% had a postgraduate degree. Overall, 67% were Non-testers (who indicated that they were unlikely or unsure of their likelihood of being tested within the next 6 months); these respondents were significantly more likely than Testers (who indicated a higher probability of testing within 6 months) to be female (60.2 vs. 54.1%; odds ratio [OR] [95% confidence interval (CI)], 1.3 [1.1‒1.4]), aged ≥ 65 years old (41.5 vs. 30.3%; OR [95% CI] compared with ages 18‒34 years, 0.6 [0.5‒0.7]), White (82.1 vs. 66.8%; OR [95% CI], 1.4 [1.1‒1.8]), and to identify as politically conservative (40.9 vs. 18.1%; OR [95% CI], 2.6 [2.3‒2.9]). In contrast, Testers were significantly more likely than Non-testers to have previous experience with COVID-19 testing, infection, or vaccination; greater knowledge regarding COVID-19 and testing; greater healthcare engagement; and concerns about COVID-19.

Older, female, White, rural-dwelling, and politically conservative high-risk adults are the most likely individuals to experience COVID-19 testing hesitancy. Understanding these demographic factors will help guide strategies to improve US testing rates.

Propensity score methods, such as inverse probability-of-treatment weighting (IPTW), have been increasingly used for covariate balancing in both observational studies and randomized trials, allowing the control of both systematic and chance imbalances. Approaches using IPTW are based on two steps: (i) estimation of the individual propensity scores (PS), and (ii) estimation of the treatment effect by applying PS weights. Thus, a variance estimator that accounts for both steps is crucial for correct inference. Using a variance estimator which ignores the first step leads to overestimated variance when the estimand is the average treatment effect (ATE), and to under or overestimated estimates when targeting the average treatment effect on the treated (ATT). In this article, we emphasize the importance of using an IPTW variance estimator that correctly considers the uncertainty in PS estimation. We present a comprehensive tutorial to obtain unbiased variance estimates, by proposing and applying a unifying formula for different types of PS weights (ATE, ATT, matching and overlap weights). This can be derived either via the linearization approach or M-estimation. Extensive R code is provided along with the corresponding large-sample theory. We perform simulation studies to illustrate the behavior of the estimators under different treatment and outcome prevalences and demonstrate appropriate behavior of the analytical variance estimator. We also use a reproducible analysis of observational lung cancer data as an illustrative example, estimating the effect of receiving a PET-CT scan on the receipt of surgery.