The Right Tool for the Job: Choosing Between Covariate-balancing and Generalized Boosted Model Propensity Scores

Contrasting Health Outcomes following a Severe Smoke Episode and Ambient Air Pollution in Early Life: Findings from an Australian Data Linkage Cohort Study of Hospital Utilization

BACKGROUND

Episodic spikes in air pollution due to landscape fires are increasing, and their potential for longer term health impacts is uncertain.

OBJECTIVE

Our objective is to evaluate associations between exposure in utero and in infancy to severe pollution from a mine fire, background ambient air pollution, and subsequent hospital care.

METHODS

We linked health records of births, emergency department (ED) visits, and hospitalizations of children born in the Latrobe Valley, Australia, 2012-2015, which included a severe pollution episode from a mine fire (9 February 2014 to 25 March 2014). We assigned modeled exposure estimates for fire-related and ambient particulate matter with an aerodynamic diameter of 2.5 μ m (PM 2.5 ) to residential address. We used logistic regression to estimate associations with hospital visits for any cause and groupings of infectious, allergic, and respiratory conditions. Outcomes were assessed for the first year of life in the in utero cohort and the year following the fire in the infant cohort. We estimated exposure-response for both fire-related and ambient PM 2.5 and also employed inverse probability weighting using the propensity score to compare exposed and not/minimally exposed children.

RESULTS

Prenatal exposure to fire-related PM 2.5 was associated with ED presentations for allergies/skin rash [odds ratio ( OR ) = 1.34 , 95% confidence interval (CI): 1.01, 1.76 per 240   μ g / m 3 increase]. Exposure in utero to ambient PM 2.5 was associated with overall presentations (OR = 1.18 , 95% CI: 1.05, 1.33 per 1.4   μ g / m 3 ) and visits for infections (ED: OR = 1.13 , 95% CI: 0.98, 1.29; hospitalizations: OR = 1.23 , 95% CI: 1.00, 1.52). Exposure in infancy to fire-related PM 2.5 compared to no/minimal exposure, was associated with ED presentations for respiratory (OR = 1.37 , 95% CI: 1.05, 1.80) and infectious conditions (any: OR = 1.21 , 95% CI: 0.98, 1.49; respiratory-related: OR = 1.39 , 95% CI: 1.05, 1.83). Early life exposure to ambient PM 2.5 was associated with overall ED visits (OR = 1.17 , 95% CI: 1.05, 1.30 per 1.4   μ g / m 3 increase).

DISCUSSION

Higher episodic and lower ambient concentrations of PM 2.5 in early life were associated with visits for allergic, respiratory, and infectious conditions. Our findings also indicated differences in associations at the two developmental stages. https://doi.org/10.1289/EHP12238.

Reweighting estimators to extend the external validity of clinical trials: methodological considerations

Methods to extend the strong internal validity of randomized controlled trials to reliably estimate treatment effects in target populations are gaining attention. This paper enumerates steps recommended for undertaking such extended inference, discusses currently viable choices for each one, and provides recommendations. We demonstrate a complete extended inference from a clinical trial studying a pharmaceutical treatment for Alzheimer's disease (AD) to a realistic target population of European residents diagnosed with AD. This case study highlights approaches to overcoming practical difficulties and demonstrates limitations of reliably extending inference from a trial to a real-world population.

Using Balancing Weights to Target the Treatment Effect on the Treated when Overlap is Poor

Inverse probability weights are commonly used in epidemiology to estimate causal effects in observational studies. Researchers often focus on either the average treatment effect or the average treatment effect on the treated with inverse probability weighting estimators. However, poor overlap in the baseline covariates between the treated and control groups can produce extreme weights that can result in biased treatment effect estimates. One alternative to inverse probability weights are overlap weights, which target the population with the most overlap on observed covariates. Although estimates based on overlap weights produce less bias in such contexts, the causal estimand can be difficult to interpret. An alternative to model-based inverse probability weights are balancing weights, which directly target imbalances during the estimation process, rather than model fit. Here, we explore whether balancing weights allow analysts to target the average treatment effect on the treated in cases where inverse probability weights lead to biased estimates due to poor overlap. We conduct three simulation studies and an empirical application. We find that balancing weights often allow the analyst to still target the average treatment effect on the treated even when overlap is poor. We show that although overlap weights remain a key tool, more familiar estimands can sometimes be targeted by using balancing weights.

Association between Fluoxetine Use and Overall Survival among Patients with Cancer Treated with PD-1/L1 Immunotherapy

Checkpoint inhibitors can be a highly effective antitumor therapy but only to a subset of patients, presumably due to immunotherapy resistance. Fluoxetine was recently revealed to inhibit the NLRP3 inflammasome, and NLRP3 inhibition could serve as a target for immunotherapy resistance. Therefore, we evaluated the overall survival (OS) in patients with cancer receiving checkpoint inhibitors combined with fluoxetine. A cohort study was conducted among patients diagnosed with lung, throat (pharynx or larynx), skin, or kidney/urinary cancer treated with checkpoint inhibitor therapy. Utilizing the Veterans Affairs Informatics and Computing Infrastructure, patients were retrospectively evaluated during the period from October 2015 to June 2021. The primary outcome was overall survival (OS). Patients were followed until death or the end of the study period. There were 2316 patients evaluated, including 34 patients who were exposed to checkpoint inhibitors and fluoxetine. Propensity score weighted Cox proportional hazards demonstrated a better OS in fluoxetine-exposed patients than unexposed (HR: 0.59, 95% CI 0.371-0.936). This cohort study among cancer patients treated with checkpoint inhibitor therapy showed a significant improvement in the OS when fluoxetine was used. Because of this study's potential for selection bias, randomized trials are needed to assess the efficacy of the association of fluoxetine or another anti-NLRP3 drug to checkpoint inhibitor therapy.

Evaluating the influence of alcohol intoxication on the pre-hospital identification of severe head injury: a multi-center, cohort study

OBJECTIVE

To determine the influence of intoxication on the pre-hospital recognition of severely head-injured patients by Emergency Medical Services (EMS) professionals and to investigate the relationship between suspected alcohol intoxication and severe head injury.

METHODS

This multi-center, retrospective, cohort study included trauma patients, aged ≥ 16 years, transported by an ambulance of the Regional Ambulance Facility Utrecht to any emergency department in the participating trauma regions.

RESULTS

Between January 1, 2015 and December 31, 2017, 19,206 patients were included, of whom 1167 (6.0%) were suspected to have a severe head injury in the field, and 623 (3.2%) were diagnosed with such an injury at the hospital. These injuries were less frequently recognized in patients with a GCS ≥ 13 than in patients with a GCS < 13 (25.0% vs. 76.2%). Patients suspected to be intoxicated had a higher chance to suffer from severe head injury (OR 1.42, 95%-CI 1.22-1.65) and were recognized slightly more often (45.3% vs. 40.2%).

CONCLUSION

Severe head injuries are difficult to recognize in the field, especially in patients without a decreased GCS. Suspicion of alcohol intoxication did not seem to influence pre-hospital injury recognition, as it possibly makes a severe head injury harder to recognize and simultaneously raises caution for a severe injury.

Dealing with confounding in observational studies: A scoping review of methods evaluated in simulation studies with single-point exposure

The aim of this article was to perform a scoping review of methods available for dealing with confounding when analyzing the effect of health care treatments with single-point exposure in observational data. We aim to provide an overview of methods and their performance assessed by simulation studies indexed in PubMed. We searched PubMed for simulation studies published until January 2021. Our search was restricted to studies evaluating binary treatments and binary and/or continuous outcomes. Information was extracted on the methods' assumptions, performance, and technical properties. Of 28,548 identified references, 127 studies were eligible for inclusion. Of them, 84 assessed 14 different methods (ie, groups of estimators that share assumptions and implementation) for dealing with measured confounding, and 43 assessed 10 different methods for dealing with unmeasured confounding. Results suggest that there are large differences in performance between methods and that the performance of a specific method is highly dependent on the estimator. Furthermore, the methods' assumptions regarding the specific data features also substantially influence the methods' performance. Finally, the methods result in different estimands (ie, target of inference), which can even vary within methods. In conclusion, when choosing a method to adjust for measured or unmeasured confounding it is important to choose the most appropriate estimand, while considering the population of interest, data structure, and whether the plausibility of the methods' required assumptions hold.

High maternal glucose exacerbates the association between prenatal per- and polyfluoroalkyl substance exposure and reduced birth weight

BACKGROUND

Per- and polyfluoroalkyl substances (PFAS) exposure has been associated with reduced birth weight. However, the association may be complicated by glucose status due to PFAS impact on fetal growth and placental transport.

OBJECTIVES

To examine whether maternal glucose status modifies the association between prenatal PFAS exposure and birth weight z-score.

METHODS

We analyzed data of 1405 mother-child pairs from the prospective Shanghai Birth Cohort. Plasma concentrations of six PFAS were quantified in the first trimester. Fasting plasma glucose (FPG) was collected at 24-28 gestation weeks. A range of FPG cutoffs (4.9-5.4 mmol/L) covering current recommendations for gestational diabetes mellitus were used to define high and low FPG groups. Association between PFAS concentration and birth weight z-score was evaluated using multivariate linear regression in two FPG groups respectively, and the dose-response relationship was estimated with cutoffs ranging from low to high. We then used propensity score to counterbalance the effects of different PFAS concentrations between the high and low FPG groups, and run the regression again.

RESULTS

A doubling increase in concentrations of several PFAS was inversely associated with birth weight z-score. The association was more evident in high FPG groups and the magnitudes intensified when FPG cutoff increased. The strongest association was observed for PFOA, with the magnitude increased from -0.34 (95 % CI: -0.66, -0.03) for 5.0 mmol/L cutoff, to -0.41 (95 % CI: -0.77, -0.05) for 5.1 mmol/L cutoff, and further to -0.51 (95 % CI: -0.98, -0.03) for 5.3 mmol/L. Propensity score matching yielded similar results.

CONCLUSIONS

High maternal glucose level may increase the risk of reduced birth weight z-score related to prenatal PFAS exposure. Moreover, exploring the effects with different FPG cutoffs may contribute to providing intervention strategies for pregnant women with high PFAS exposure.

Target Trial Emulation: A Design Tool for Cancer Clinical Trials

PURPOSE

To apply target trial emulation to explore the potential impact of eligibility criteria on the primary outcome of a randomized controlled trial.

METHODS

Simulations of a real-world explanatory trial of transarterial radioembolization for advanced unresectable hepatocellular carcinoma with portal vein invasion were performed to examine the effects of cohort specification on survival outcomes and patient sample size. Simulations comprised 24 different permutations of the trial varied on three disease nonspecific eligibility parameters. Treatment and control arms for these emulated trials were drawn from the National Cancer Database and matched by treatment propensity. Target trial emulation served as the causal framework for this analysis, allowing the architecture of a true controlled experiment to address forms of bias routinely encountered in comparative effectiveness studies involving real-world observational data.

RESULTS

Twenty-four propensity score-matched cohorts comprising a wider clinical spectrum of patients than specified by the original target trial were successfully generated using the National Cancer Database. The arms for each of the emulated trials demonstrated exchangeability across all eligibility criteria and other clinical covariates. Significant treatment benefits were associated with only a narrow range of eligibility criteria, indicating that the original target trial was well specified.

CONCLUSION

The impact of patient selection on treatment outcomes can be studied using target trial emulation. This analytical framework can furthermore serve to leverage existing real-world data to inform the task of cohort specification for a randomized controlled trial, facilitating a more data-driven approach for this important step in clinical trial design.

Retirement and elderly health in China: Based on propensity score matching

Background

The relationship between retirement and health is important to the formulation of retirement related policies but is a controversial topic, perhaps because selection bias has not been well-addressed in previous studies through traditional analysis methods. Using data from the China Health and Retirement Longitudinal Study (CHARLS), this study explored the potential impact of retirement on the health of elderly Chinese individuals, adjusting for selection bias.

Methods

We balanced the baseline differences between retirement groups and working groups based on nearest neighbor matching and genetic matching with a generalized boosted model (GBM), and regression analysis was used to evaluate the impact of retirement on the health of elderly individuals.

Results

No significant difference was found in any of the covariates between the two groups after matching. Genetic matching performed better than nearest neighbor matching in balancing the covariates. Compared to the working group, the retirement group had a 0.78 (95% CI: 0.65-0.94, P = 0.026) times higher probability of self-reported physical pain, a 0.76 (95% CI: 0.62-0.93, P = 0.023) times higher probability of depression, and a 0.57-point (95% CI: 0.37-0.78, P < 0.001) improvement in cognitive status score. Among male, the retirement group had a 0.89-point (95% CI: 0.45-1.33, P < 0.001) improvement in cognitive status score for low education, a 0.65 (95% CI: 0.46-0.92, P = 0.042) times higher probability of self-reported physical pain for middle education. For female with low education, the cognitive status of the retirement group was significantly higher by 0.99 points (95% CI: 0.42-1.55, P = 0.004), the probability of depression was 0.56 (95% CI: 0.36-0.87, P = 0.031) times higher in the retirement group than in the working group. There was no difference for the middle and high education.

Conclusion

Retirement can exert a beneficial effect on the health of elderly individuals. Therefore, the government and relevant departments should consider this potential effect when instituting policies that delay retirement.

Higher Moments for Optimal Balance Weighting in Causal Estimation

We expand upon a simulation study that compared three promising methods for estimating weights for assessing the average treatment effect on the treated for binary treatments: generalized boosted models, covariate-balancing propensity scores, and entropy balance. The original study showed that generalized boosted models can outperform covariate-balancing propensity scores, and entropy balance when there are likely to be nonlinear associations in both the treatment assignment and outcome models and when the other two models are fine-tuned to obtain balance only on first-order moments. We explore the potential benefit of using higher-order moments in the balancing conditions for covariate-balancing propensity scores and entropy balance. Our findings showcase that these two models should, by default, include higher-order moments and focusing only on first moments can result in substantial bias in estimated treatment effect estimates from both models that could be avoided using higher moments.

A tutorial comparing different covariate balancing methods with an application evaluating the causal effects of substance use treatment programs for adolescents

Randomized controlled trials are the gold standard for measuring causal effects. However, they are often not always feasible, and causal treatment effects must be estimated from observational data. Observational studies do not allow robust conclusions about causal relationships unless statistical techniques account for the imbalance of pretreatment confounders across groups and key assumptions hold. Propensity score and balance weighting (PSBW) are useful techniques that aim to reduce the observed imbalances between treatment groups by weighting the groups to look alike on the observed confounders. Notably, there are many methods available to estimate PSBW. However, it is unclear a priori which will achieve the best trade-off between covariate balance and effective sample size for a given application. Moreover, it is critical to assess the validity of key assumptions required for robust estimation of the needed treatment effects, including the overlap and no unmeasured confounding assumptions. We present a step-by-step guide to the use of PSBW for estimation of causal treatment effects that includes steps on how to evaluate overlap before the analysis, obtain estimates of PSBW using multiple methods and select the optimal one, check for covariate balance on multiple metrics, and assess sensitivity of findings (both the estimated treatment effect and statistical significance) to unobserved confounding. We illustrate the key steps using a case study examining the relative effectiveness of substance use treatment programs and provide a user-friendly Shiny application that can implement the proposed steps for any application with binary treatments.

Receipt of Early Intervention Services Before Age 3 Years and Performance on Third-Grade Standardized Tests Among Children Exposed to Lead

IMPORTANCE

Research has shown that early intervention programs can improve academic outcomes of children with developmental delays. It has been suggested that similar programs may combat the deleterious effects of lead on children's neurodevelopment. However, to our knowledge, there are no published studies examining this possibility.

OBJECTIVE

The objective of this study was to estimate the association between receipt of early intervention services and third-grade standardized test scores among children exposed to lead before age 3 years.

DESIGN, SETTING, AND PARTICIPANTS

Cohort study including children born in New York City, New York, from 1994 to 1998 within an administrative data linkage of birth, lead monitoring, early intervention, and education data systems. Participants had a blood lead level of 4 μg/dL or greater at any point before age 3 years and later attended public school in New York City.

EXPOSURES

Any use of early intervention services from birth through age 3 years.

MAIN OUTCOMES AND MEASURES

Children who did or did not receive early intervention services were matched using propensity scores. Linear and log-binomial regression were used to estimate the association between receipt of early intervention services before age 3 years and standardized test scores in math and English-language arts in third grade.

RESULTS

There were 2986 children exposed to lead who received early intervention services before age 36 months. Of these children, 2757 were propensity score-matched to 8160 children who did not receive services. Children who received early intervention services did 7% (95% CI, 3%-12%) of an SD better on math and 10% (95% CI, 5%-14%) of an SD better on English-language arts tests than children who did not receive services. In addition, children who received services were 14% (95% CI, 9%-19%) and 16% (95% CI, 9%-23%) more likely to meet test-based standards in math and English-language arts, respectively, than children who did not receive services. These associations became larger in magnitude when analyses were restricted to children with higher blood lead levels.

CONCLUSIONS AND RELEVANCE

By leveraging existing public health data, this study found evidence that receipt of early intervention services may benefit the academic performance of children exposed to lead early in life.

Recommendations for the use of propensity score methods in multiple sclerosis research

BACKGROUND

With many disease-modifying therapies currently approved for the management of multiple sclerosis, there is a growing need to evaluate the comparative effectiveness and safety of those therapies from real-world data sources. Propensity score methods have recently gained popularity in multiple sclerosis research to generate real-world evidence. Recent evidence suggests, however, that the conduct and reporting of propensity score analyses are often suboptimal in multiple sclerosis studies.

OBJECTIVES

To provide practical guidance to clinicians and researchers on the use of propensity score methods within the context of multiple sclerosis research.

METHODS

We summarize recommendations on the use of propensity score matching and weighting based on the current methodological literature, and provide examples of good practice.

RESULTS

Step-by-step recommendations are presented, starting with covariate selection and propensity score estimation, followed by guidance on the assessment of covariate balance and implementation of propensity score matching and weighting. Finally, we focus on treatment effect estimation and sensitivity analyses.

CONCLUSION

This comprehensive set of recommendations highlights key elements that require careful attention when using propensity score methods.

Targeting Rac1 for the prevention of atherosclerosis among U.S. Veterans with inflammatory bowel disease

Evidence suggests that Ras-related C3 botulinum toxin substrate 1 (Rac1) might be a target in atherosclerotic disease (AD). We hypothesize that due to their ability to inhibit Rac1, thiopurines are associated with a lower risk of AD. We fit a time-dependent cox proportional hazards model estimating the hazard of AD among a national cohort of US veterans with inflammatory bowel disease. Patients exposed to thiopurines had a 7.5% lower risk of AD (HR = 0.925; 95% CI = (0.87-0.984)) compared to controls. The propensity score weighted analysis reveals thiopurine exposure reduces the risk of AD by 6.6% (HR = 0.934; 95% CI = (0.896-0.975)), compared to controls. Further exploration and evaluation of Rac1 inhibition as a target for AD is warranted.

Pneumococcal vaccination prevented severe LRTIs in adults: a causal inference framework applied in registry data

OBJECTIVES

We estimated the effect of pneumococcal vaccination (PV) on acute lower respiratory tract infections (LRTIs) in various age and risk groups using different methods within a causal inference methodological framework.

STUDY DESIGN AND SETTING

We used data from a general practitioners' morbidity registry for the year 2019. Both traditional statistical methods (regression-based and propensity score methods) and machine learning techniques were deployed. Multiple imputation was used to account for missing data. Relative risks (RRs) with 95% confidence intervals were estimated. Sensitivity analyses were performed to account for the severity of LRTIs and differences in vaccination registration.

RESULTS

All methods showed a standardized mean difference below 0.1 for each covariate. No method was found to be superior to another. PV (combination of conjugate and polysaccharide vaccine) had an overall protective effect for severe LRTIs. PV was protective in different age and risk groups, especially in people aged 50-84 years with an intermediate risk group.

CONCLUSION

Using several techniques, PV was found to prevent severe LRTIs and confirmed the recommendations of the Belgian Superior Health Council.

Pharmacoepidemiology, Machine Learning, and COVID-19: An Intent-to-Treat Analysis of Hydroxychloroquine, With or Without Azithromycin, and COVID-19 Outcomes Among Hospitalized US Veterans

Hydroxychloroquine (HCQ) was proposed as an early therapy for coronavirus disease 2019 (COVID-19) after in vitro studies indicated possible benefit. Previous in vivo observational studies have presented conflicting results, though recent randomized clinical trials have reported no benefit from HCQ amongst hospitalized COVID-19 patients. We examined the effects of HCQ alone, and in combination with azithromycin, in a hospitalized COVID-19 positive, United States (US) Veteran population using a propensity score adjusted survival analysis with imputation of missing data. From March 1, 2020 through April 30, 2020, 64,055 US Veterans were tested for COVID-19 based on Veteran Affairs Healthcare Administration electronic health record data. Of the 7,193 positive cases, 2,809 were hospitalized, and 657 individuals were prescribed HCQ within the first 48-hours of hospitalization for the treatment of COVID-19. There was no apparent benefit associated with HCQ receipt, alone or in combination with azithromycin, and an increased risk of intubation when used in combination with azithromycin [Hazard Ratio (95% Confidence Interval): 1.55 (1.07, 2.24)]. In conclusion, we assessed the effectiveness of HCQ with or without azithromycin in treating patients hospitalized with COVID-19 using a national sample of the US Veteran population. Using rigorous study design and analytic methods to reduce confounding and bias, we found no evidence of a survival benefit from the administration of HCQ.

Reflection on modern methods: when worlds collide-prediction, machine learning and causal inference

Causal inference requires theory and prior knowledge to structure analyses, and is not usually thought of as an arena for the application of prediction modelling. However, contemporary causal inference methods, premised on counterfactual or potential outcomes approaches, often include processing steps before the final estimation step. The purposes of this paper are: (i) to overview the recent emergence of prediction underpinning steps in contemporary causal inference methods as a useful perspective on contemporary causal inference methods, and (ii) explore the role of machine learning (as one approach to 'best prediction') in causal inference. Causal inference methods covered include propensity scores, inverse probability of treatment weights (IPTWs), G computation and targeted maximum likelihood estimation (TMLE). Machine learning has been used more for propensity scores and TMLE, and there is potential for increased use in G computation and estimation of IPTWs.

Instrumental variable estimation of truncated local average treatment effects

Instrumental variable (IV) analysis is used to address unmeasured confounding when comparing two nonrandomized treatment groups. The local average treatment effect (LATE) is a causal estimand that can be identified by an IV. The LATE approach is appealing because its identification relies on weaker assumptions than those in other IV approaches requiring a homogeneous treatment effect assumption. If the instrument is confounded by some covariates, then one can use a weighting estimator, for which the outcome and treatment are weighted by instrumental propensity scores. The weighting estimator for the LATE has a large variance when the IV is weak and the target population, i.e., the compliers, is relatively small. We propose a truncated LATE that can be estimated more reliably than the regular LATE in the presence of a weak IV. In our approach, subjects who contribute substantially to the weak IV are identified by their probabilities of being compliers, and they are removed based on a pre-specified threshold. We discuss interpretation of the proposed estimand and related inference method. Simulation and real data experiments demonstrate that the proposed truncated LATE can be estimated more precisely than the standard LATE.

Validation of Questionnaire-based Case Definitions for Chronic Obstructive Pulmonary Disease

BACKGROUND

Various questionnaire-based definitions of chronic obstructive pulmonary disease (COPD) have been applied using the US representative National Health and Nutrition Examination Survey (NHANES), but few have been validated against objective lung function data. We validated two prior definitions that incorporated self-reported physician diagnosis, respiratory symptoms, and/or smoking. We also validated a new definition that we developed empirically using gradient boosting, an ensemble machine learning method.

METHODS

Data came from 7,996 individuals 40-79 years who participated in NHANES 2007-2012 and underwent spirometry. We considered participants "true" COPD cases if their ratio of postbronchodilator forced expiratory volume in 1 second to forced vital capacity was below 0.7 or the lower limit of normal. We stratified all analyses by smoking history. We developed a gradient boosting model for smokers only; predictors assessed (25 total) included sociodemographics, inhalant exposures, clinical variables, and respiratory symptoms.

RESULTS

The spirometry-based COPD prevalence was 26% for smokers and 8% for never smokers. Among smokers, using questionnaire-based definitions resulted in a COPD prevalence ranging from 11% to 16%, sensitivity ranging from 18% to 35%, and specificity ranging from 88% to 92%. The new definition classified participants based on age, bronchodilator use, body mass index (BMI), smoking pack-years, and occupational organic dust exposure, and resulted in the highest sensitivity (35%) and specificity (92%) among smokers. Among never smokers, the COPD prevalence ranged from 4% to 5%, and we attained good specificity (96%) at the expense of sensitivity (9-10%).

CONCLUSION

Our results can be used to parametrize misclassification assumptions for quantitative bias analysis when pulmonary function data are unavailable.

Free Access to a Broad Contraceptive Method Mix and Women's Contraceptive Choice: Evidence from Sub-Saharan Africa

Financial barriers may restrict women's ability to use their preferred contraceptive methods, especially long-acting reversible contraceptives (LARC). Providing free access to a broad contraceptive method mix, including both LARC and short-acting reversible contraceptives (SARC), may increase contraceptive use, meet women's various fertility needs, and increase their agency in contraceptive decisions. Linking facility and individual data from eight countries in sub-Saharan Africa, we use a propensity score approach combined with machine learning techniques to examine how free access to a broad contraceptive method mix affects women's contraceptive choice. Free access to both LARC and SARC was associated with an increase of 3.2 percentage points (95 percent confidence interval: 0.006, 0.058) in the likelihood of contraceptive use, driven by greater use of SARC. Among contraceptive users, free access did not prompt women to switch to LARC and had no effect on contraceptive decision-making. The price effects were larger among older and more educated women, but free access was associated with lower contraceptive use among adolescents. While free access to contraceptives is associated with a modest increase in contraceptive use for some women, removing user fees alone does not address all barriers women face, especially for the most vulnerable groups of women.

A machine learning compatible method for ordinal propensity score stratification and matching

Although machine learning techniques that estimate propensity scores for observational studies with multivalued treatments have advanced rapidly in recent years, the development of propensity score adjustment techniques has not kept pace. While machine learning propensity models provide numerous benefits, they do not produce a single variable balancing score that can be used for propensity score stratification and matching. This issue motivates the development of a flexible ordinal propensity scoring methodology that does not require parametric assumptions for the propensity model. The proposed method fits a one-parameter power function to the cumulative distribution function (CDF) of the generalized propensity score (GPS) vector resulting from any machine learning propensity model, and is henceforth called the GPS-CDF method. The estimated parameter from the GPS-CDF method, ã , is a scalar balancing score that can be used to group similar subjects in outcome analyses. Specifically, subjects who received different levels of the treatment are stratified or matched based on their ã value to produce unbiased estimates of the average treatment effect (ATE). Simulation studies presented show remediation of covariate balance, minimal bias in ATEs, and maintain coverage probability. The proposed method is applied to the Mexican-American Tobacco use in Children (MATCh) study to determine whether an ordinal treatment of exposure to smoking imagery in movies causes cigarette experimentation in Mexican-American adolescents.

Adverse Outcomes Associated With Delaying or Withholding β-Blockers After Cardiac Surgery: A Retrospective Single-Center Cohort Study

BACKGROUND

Ideal timing of postoperative β-blockers is unclear. We hypothesized that patients who do not receive β-blockers immediately after cardiac surgery would have increased in-hospital mortality (primary outcome) and postoperative hemodynamic, pulmonary, neurologic, or respiratory complications (secondary outcomes).

METHODS

We performed a retrospective cohort study evaluating patients who underwent cardiac surgery at our institution from January 1, 2013 to September 30, 2017. We compared outcomes between patients who received β-blockers by postoperative day (POD) 5 with outcomes in patients who did not receive β-blockers at any time or received them after POD 5. Inverse probability of treatment weighting was used to minimize confounding. Univariate logistic regression analyses were performed on the weighted sets using absent or delayed β-blockers as the independent variable and each outcome as dependent variables in separate analyses. A secondary analysis was performed in patients prescribed preoperative β-blockers. E-values were calculated for significant outcomes.

RESULTS

All results were confounder adjusted. Among patients presenting for cardiac surgery, not receiving β-blockers by POD 5 or at any time was not associated with the primary outcome in-hospital mortality, estimated odds ratio (OR; 99.5% confidence interval [CI]) of 1.6 (0.49-5.1), P = .28. Not receiving β-blockers by POD 5 or at any time was associated with postoperative atrial fibrillation, estimated OR (99.5% CI) of 1.5 (1.1-2.1), P < .001, and pulmonary complications, estimated OR (99.5% CI) of 3.0 (1.8-5.2), P < .001. E-values were 2.4 for postoperative atrial fibrillation and 5.6 for pulmonary complications. Among patients presenting for cardiac surgery taking preoperative β-blockers, not receiving β-blockers by POD 5 or at any time was not associated with the primary outcome mortality, with estimated OR (99.5% CI) of 1.3 (0.43-4.1), P = .63. In this subset, not receiving β-blockers by POD 5 or at any time was associated with increased adjusted ORs of postoperative atrial fibrillation (OR = 1.6; 99.5% CI, 1.1-2.4; P < .001) and postoperative pulmonary complications (OR = 2.8; 99.5% CI, 1.6-5.2; P < .001). Here, e-values were 2.7 for postoperative atrial fibrillation and 5.1 for pulmonary complications. For the sensitivity analyses for secondary outcomes, exposure and outcome periods overlap. Outcomes may have occurred before or after postoperative β-blocker administration.

CONCLUSIONS

Among patients who undergo cardiac surgery, not receiving postoperative β-blockers within the first 5 days after cardiac surgery or at any time is not associated with in-hospital mortality and is associated with, but may not necessarily cause, postoperative atrial fibrillation and pulmonary complications.

Machine learning outcome regression improves doubly robust estimation of average causal effects

BACKGROUND

Doubly robust estimation produces an unbiased estimator for the average treatment effect unless both propensity score (PS) and outcome models are incorrectly specified. Studies have shown that the doubly robust estimator is subject to more bias than the standard weighting estimator when both PS and outcome models are incorrectly specified.

METHOD

We evaluated whether various machine learning methods can be used for estimating conditional means of the potential outcomes to enhance the robustness of the doubly robust estimator to various degrees of model misspecification in terms of reducing bias and standard error. We considered four types of methods to predict the outcomes: least squares, tree-based methods, generalized additive models and shrinkage methods. We also considered an ensemble method called the Super Learner (SL), which is a linear combination of multiple learners. We conducted simulations considering different scenarios by the complexity of PS and outcome-generating models and some ranges of treatment prevalence.

RESULTS

The shrinkage methods performed well with robust doubly robust estimates in term of bias and mean squared error across the scenarios when the models became rich by including all 2-way interactions of the covariates. The SL performed similarly to the best method in each scenario.

CONCLUSIONS

Our findings indicate that machine learning methods such as the SL or the shrinkage methods using interaction models should be used for more accurate doubly robust estimators.

Quantifying the bias due to observed individual confounders in causal treatment effect estimates

It is often of interest to use observational data to estimate the causal effect of a target exposure or treatment on an outcome. When estimating the treatment effect, it is essential to appropriately adjust for selection bias due to observed confounders using, for example, propensity score weighting. Selection bias due to confounders occurs when individuals who are treated are substantially different from those who are untreated with respect to covariates that are also associated with the outcome. A comparison of the unadjusted, naive treatment effect estimate with the propensity score adjusted treatment effect estimate provides an estimate of the selection bias due to these observed confounders. In this article, we propose methods to identify the observed covariate that explains the largest proportion of the estimated selection bias. Identification of the most influential observed covariate or covariates is important in resource-sensitive settings where the number of covariates obtained from individuals needs to be minimized due to cost and/or patient burden and in settings where this covariate can provide actionable information to healthcare agencies, providers, and stakeholders. We propose straightforward parametric and nonparametric procedures to examine the role of observed covariates and quantify the proportion of the observed selection bias explained by each covariate. We demonstrate good finite sample performance of our proposed estimates using a simulation study and use our procedures to identify the most influential covariates that explain the observed selection bias in estimating the causal effect of alcohol use on progression of Huntington's disease, a rare neurological disease.

Estimation of average treatment effects among multiple treatment groups by using an ensemble approach

In observational studies, generalized propensity score (GPS)-based statistical methods, such as inverse probability weighting (IPW) and doubly robust (DR) method, have been proposed to estimate the average treatment effect (ATE) among multiple treatment groups. In this article, we investigate the GPS-based statistical methods to estimate treatment effects from two aspects. The first aspect of our investigation is to obtain an optimal GPS estimation method among four competing GPS estimation methods by using a rank aggregation approach. We further examine whether the optimal GPS-based IPW and DR methods would improve the performance for estimating ATE. It is well known that the DR method is consistent if either the GPS or the outcome models are correctly specified. The second aspect of our investigation is to examine whether the DR method could be improved if we ensemble outcome models. To that end, bootstrap method and rank aggregation method are used to obtain the ensemble optimal outcome model from several competing outcome models, and the resulting outcome model is incorporated into the DR method, resulting in an ensemble DR (enDR) method. Extensive simulation results indicate that the enDR method provides the best performance in estimating the ATE regardless of the method used for estimating GPS. We illustrate our methods using the MarketScan healthcare insurance claims database to examine the treatment effects among three different bones and substitutes used for spinal fusion surgeries. We draw conclusions based on the estimates from the enDR method coupled with the optimal GPS estimation method.

Mutant p53 amplifies a dynamin-1/APPL1 endosome feedback loop that regulates recycling and migration

Feedback loops arising from crosstalk between early endocytic trafficking and receptor signaling can be co-opted or amplified in cancer cells to enhance their metastatic abilities. Lakoduk et al. reveal that mutant p53 upregulates dynamin-1 expression and recruitment of the APPL1 signaling scaffold to a spatially localized subpopulation of endosomes to increase receptor recycling and cell migration.

Multiple mechanisms contribute to cancer cell progression and metastatic activity, including changes in endocytic trafficking and signaling of cell surface receptors downstream of gain-of-function (GOF) mutant p53. We report that dynamin-1 (Dyn1) is up-regulated at both the mRNA and protein levels in a manner dependent on expression of GOF mutant p53. Dyn1 is required for the recruitment and accumulation of the signaling scaffold, APPL1, to a spatially localized subpopulation of endosomes at the cell perimeter. We developed new tools to quantify peripherally localized early endosomes and measure the rapid recycling of integrins. We report that these perimeter APPL1 endosomes modulate Akt signaling and activate Dyn1 to create a positive feedback loop required for rapid recycling of EGFR and β1 integrins, increased focal adhesion turnover, and cell migration. Thus, Dyn1- and Akt-dependent perimeter APPL1 endosomes function as a nexus that integrates signaling and receptor trafficking, which can be co-opted and amplified in mutant p53–driven cancer cells to increase migration and invasion.

Chasing balance and other recommendations for improving nonparametric propensity score models

In this article, we carefully examine two important implementation issues when estimating propensity scores using generalized boosted models (GBM), a promising machine learning technique. First, we examine which of the following methods for tuning GBM lead to better covariate balance and inferences about causal effects: pursuing covariate balance between the treatment groups or tuning the propensity score model on the basis of a model fit criterion. Second, we examine how well GBM can handle irrelevant covariates that are included in the estimation model. We find that chasing balance rather than model fit when estimating propensity scores yielded better covariate balance and more accurate treatment effect estimates. Additionally, we find that adding irrelevant covariates to GBM increased imbalance and bias in the treatment effects. The findings from this paper have useful implications for other work focused on improving methods for estimating propensity scores.