Optimizing matching and analysis combinations for estimating causal effects

Considerations for master protocols using external controls

There has been an increasing use of master protocols in oncology clinical trials because of its efficiency to accelerate cancer drug development and flexibility to accommodate multiple substudies. Depending on the study objective and design, a master protocol trial can be a basket trial, an umbrella trial, a platform trial, or any other form of trials in which multiple investigational products and/or subpopulations are studied under a single protocol. Master protocols can use external data and evidence (e.g. external controls) for treatment effect estimation, which can further improve efficiency of master protocol trials. This paper provides an overview of different types of external controls and their unique features when used in master protocols. Some key considerations in master protocols with external controls are discussed including construction of estimands, assessment of fit-for-use real-world data, and considerations for different types of master protocols. Similarities and differences between regular randomized controlled trials and master protocols when using external controls are discussed. A targeted learning-based causal roadmap is presented which constitutes three key steps: (1) define a target statistical estimand that aligns with the causal estimand for the study objective, (2) use an efficient estimator to estimate the target statistical estimand and its uncertainty, and (3) evaluate the impact of causal assumptions on the study conclusion by performing sensitivity analyses. Two illustrative examples for master protocols using external controls are discussed for their merits and possible improvement in causal effect estimation.

Unravelling risk selection in Spanish general government employee mutual funds: evidence from cancer hospitalizations in the public health network

Government employees in Spain are covered by public Mutual Funds that purchase a uniform basket of benefits, equal to the ones served to the general population, from private companies. Companies apply as private bidders for a fixed per capita premium hardly adjusted by age. Our hypothesis is that this premium does not cover risks, and companies have incentives for risk selection, which are more visible in high-cost patients. We focus on a particularly costly disease, cancer, whose prevalence is similar among government employees and the general population. We compare hospitalisations in the public hospitals of the government employees that have chosen public provision and the general population. We analysed a database of hospital discharges in the Valencian Community from 2010 to 2015 (3 million episodes). Using exact matching and logistic models, we find significant risk selection; thus, in hospitalised government employees, the likelihood for a solid metastatic carcinoma and non-metastatic cancer to appear in the registry is 31% higher than in the general population. Lymphoma shows the highest odds ratio of 2.64. We found quantitatively important effects. This research provides indirect evidence of risk selection within Spanish Mutual Funds for government employees, prompting action to reduce incentives for such a practice. More research is needed to figure out if what we have observed with cancer patients occurs in other conditions.

Automated generation of comparator patients in the electronic medical record

Background

Well-designed randomized trials provide high-quality clinical evidence but are not always feasible or ethical. In their absence, the electronic medical record (EMR) presents a platform to conduct comparative effectiveness research, central to the emerging academic learning health system (aLHS) model. A barrier to realizing this vision is the lack of a process to efficiently generate a reference comparison group for each patient.

Objective

To test a multi-step process for the selection of comparators in the EMR.

Materials and Methods

We conducted a mixed-methods study within a large aLHS in North Carolina. We (1) created a list of 35 candidate variables; (2) surveyed 270 researchers to assess the importance of candidate variables; and (3) built consensus rankings around survey-identified variables (ie, importance scores >7) across two panels of 7-8 clinical research experts. Prioritized algorithm inputs were collected from the EMR and applied using a greedy matching technique. Feasibility was measured as the percentage of patients with 100 matched comparators and performance was measured via computational time and Euclidean distance.

Results

Nine variables were selected: age, sex, race, ethnicity, body mass index, insurance status, smoking status, Charlson Comorbidity Index, and neighborhood percentage in poverty. The final process successfully generated 100 matched comparators for each of 1.8 million candidate patients, executed in less than 100 min for the majority of strata, and had average Euclidean distance 0.043.

Conclusion

EMR-derived matching is feasible to implement across a diverse patient population and can provide a reproducible, efficient source of comparator data for observational studies, with additional testing in clinical research applications needed.

The effect of carbon fertilization on naturally regenerated and planted US forests

Over the last half century in the United States, the per-hectare volume of wood in trees has increased, but it is not clear whether this increase has been driven by forest management, forest recovery from past land uses, such as agriculture, or other environmental factors such as elevated carbon dioxide, nitrogen deposition, or climate change. This paper uses empirical analysis to estimate the effect of elevated carbon dioxide on aboveground wood volume in temperate forests of the United States. To accomplish this, we employ matching techniques that allow us to disentangle the effects of elevated carbon dioxide from other environmental factors affecting wood volume and to estimate the effects separately for planted and natural stands. We show that elevated carbon dioxide has had a strong and consistently positive effect on wood volume while other environmental factors yielded a mix of both positive and negative effects. This study, by enabling a better understanding of how elevated carbon dioxide and other anthropogenic factors are influencing forest stocks, can help policymakers and other stakeholders better account for the role of forests in Nationally Determined Contributions and global mitigation pathways to achieve a 1.5 degree Celsius target.

The CO2 fertilisation effect in forests remains controversial. Here, the authors disentangle the effect of CO2 on forest wood volume from other environmental factors, showing that elevated CO2 had a positive effect on wood volume in planted and natural US temperate forests.

Recurrent glioblastoma in elderly: Options and decision for the treatment

Background:

Glioblastoma (GBM) is the most common primary malignant brain tumor in adult. Its incidence increases with age and nearly half of the all newly diagnosed GBM cases are older than 65 years. Management of GBM in elderly is challenging and recurrence poses further challenge. This article aims to review the literature, evaluate the various options, and to decide the treatment plan in elderly cases with GBM recurrence.

Methods:

A systemic search was performed with the phrase “recurrent GBM (rGBM) in elderly and management” as a search term in PubMed central, Medline, and Embase databases to identify all the articles published on the subject till February 2022. The review included peer-reviewed original articles, review articles, clinical trials, and keywords in title and abstract.

Results:

Out of 473 articles searched, 15 studies followed our inclusion criteria and were included in this review. In 15 studies, ten were original and five were review articles. The minimum age group included in these studies was ≥65 years. Out of 15 studies, eight studies had described the role of resurgery, four chemotherapy, three resurgery and/or chemotherapy, and only one study on role of reradiotherapy in patients with rGBM. Out of eight studies described the role of resurgery, six have mentioned improved survival and two have no survival advantage of resurgery in cases of rGBM.

Conclusion:

Resurgery is the main treatment option in selected elderly rGBM cases in good performance status. In patients with poor performance status, chemotherapy has better post progression survival than best supportive care.

Frequency and Consequences of Cervical Lymph Node Overstaging in Head and Neck Carcinoma

Clinical lymph node staging in head and neck carcinoma (HNC) is fraught with uncertainties. Established clinical algorithms are available for the problem of occult cervical metastases. Much less is known about clinical lymph node overstaging. We identified HNC patients clinically classified as lymph node positive (cN+), in whom surgical neck dissection (ND) specimens were histopathologically negative (pN0) and in addition the subgroup, in whom an originally planned postoperative radiotherapy (PORT) was omitted. We compared these patients with surgically treated patients with clinically and histopathologically negative neck (cN0/pN0), who had received selective ND. Using a fuzzy matching algorithm, we identified patients with closely similar patient and disease characteristics, who had received primary definitive radiotherapy (RT) with or without systemic therapy (RT ± ST). Of the 980 patients with HNC, 292 received a ND as part of primary treatment. In 128/292 patients with cN0 neck, ND was elective, and in 164 patients with clinically positive neck (cN+), ND was therapeutic. In 43/164 cN+ patients, ND was histopathologically negative (cN+/pN−). In 24 of these, initially planned PORT was omitted. Overall, survival did not differ from the cN0/pN0 and primary RT ± ST control groups. However, more RT ± ST patients had functional problems with nutrition (p = 0.002). Based on these data, it can be estimated that lymph node overstaging is 26% (95% CI: 20% to 34%). In 15% (95% CI: 10% to 21%) of surgically treated cN+ HNC patients, treatment can be de-escalated without the affection of survival.

Heterogeneous treatment effect analysis based on machine-learning methodology

Heterogeneous treatment effect (HTE) analysis focuses on examining varying treatment effects for individuals or subgroups in a population. For example, an HTE‐informed understanding can critically guide physicians to individualize the medical treatment for a certain disease. However, HTE analysis has not been widely recognized and used, even given the explosive increase of data availability attributed to the arrival of the Big Data era. Part of the reason behind its underuse is that data are often of high dimension and high complexity, which pose significant challenges for applying conventional HTE analysis methods. To meet these challenges, a newly developed causal forest HTE method has been derived from the random forest machine‐learning algorithm. We conducted a systematic performance evaluation for the causal forest method against the conventional two‐step method by simulating scenarios with different levels of complexity for the analysis. Our results show that causal forest outperforms the conventional HTE method in assessing treatment effect, especially when data are complex (e.g., nonlinear) and high dimensional, suggesting that causal forest is a promising tool for real‐world applications of HTE analysis.

Does a health crisis change how we value health?

General population health state values are used in healthcare resource allocation, including health technology assessment. We examine whether UK general population health valuations changed during the COVID-19 pandemic. Ratings of EQ-5D-5L health states 11111 (no problems), 55555 (extreme problems), and dead were collected in a UK general population survey during the pandemic (April-May 2020) using the 0 = worst imaginable health, 100 = best imaginable health visual analog scale (EQ-VAS). Ratings for 55555 were transformed to a full health = 1, dead = 0 scale. Responses were compared to similar data collected pre-pandemic (2018). After propensity score matching to minimize sample differences, EQ-VAS responses were analyzed using Tobit regressions. On the 0-100 scale, 11111 was rated on average 8.67 points lower, 55555 rated 9.56 points higher, and dead rated 7.45 points lower post-pandemic onset compared to pre-pandemic. On the full health = 1, dead = 0 scale, 55555 values were 0.09 higher post-pandemic onset. There was evidence of differential impacts of COVID-19 by gender, age, and ethnicity, although only age impacted values on the 1-0 scale. COVID-19 may have affected how people value health. It is unknown whether the effect is large enough to have policy relevance, but caution should be taken in assuming pre-COVID-19 values are unchanged.

Matching Methods for Confounder Adjustment: An Addition to the Epidemiologist's Toolbox

Propensity score weighting and outcome regression are popular ways to adjust for observed confounders in epidemiological research. Here, we provide an introduction to matching methods, which serve the same purpose but can offer advantages in robustness and performance. A key difference between matching and weighting methods is that matching methods do not directly rely on the propensity score and so are less sensitive to its misspecification or to the presence of extreme values. Matching methods offer many options for customization, which allow a researcher to incorporate substantive knowledge and carefully manage bias/variance trade-offs in estimating the effects of nonrandomized exposures. We review these options and their implications, providing guidance for their use, and comparison with weighting methods. Because of their potential advantages over other methods, matching methods should have their place in an epidemiologist's methodological toolbox.

Assessing Heterogeneity of Treatment Effects in Observational Studies

Here we describe methods for assessing heterogeneity of treatment effects over prespecified subgroups in observational studies, using outcome-model-based (g-formula), inverse probability weighting, doubly robust, and matching estimators of subgroup-specific potential outcome means, conditional average treatment effects, and measures of heterogeneity of treatment effects. We compare the finite-sample performance of different estimators in simulation studies where we vary the total sample size, the relative frequency of each subgroup, the magnitude of treatment effect in each subgroup, and the distribution of baseline covariates, for both continuous and binary outcomes. We find that the estimators' bias and variance vary substantially in finite samples, even when there is no unobserved confounding and no model misspecification. As an illustration, we apply the methods to data from the Coronary Artery Surgery Study (August 1975-December 1996) to compare the effect of surgery plus medical therapy with that of medical therapy alone for chronic coronary artery disease in subgroups defined by previous myocardial infarction or left ventricular ejection fraction.

Effect of perinatal depression on risk of adverse infant health outcomes in mother-infant dyads in Gondar town: a causal analysis

BACKGROUND

Approximately one-third of pregnant and postnatal women in Ethiopia experience depression posing a substantial health burden for these women and their families. Although associations between postnatal depression and worse infant health have been observed, there have been no studies to date assessing the causal effects of perinatal depression on infant health in Ethiopia. We applied longitudinal data and recently developed causal inference methods that reduce the risk of bias to estimate associations between perinatal depression and infant diarrhea, Acute Respiratory Infection (ARI), and malnutrition in Gondar Town, Ethiopia.

METHODS

A cohort of 866 mother-infant dyads were followed from infant birth for 6 months and the cumulative incidence of ARI, diarrhea, and malnutrition were assessed. The Edinburgh Postnatal Depression Scale (EPDS) was used to assess the presence of maternal depression, the Integrated Management of Newborn and Childhood Illnesses (IMNCI) guidelines were used to identify infant ARI and diarrhea, and the mid upper arm circumference (MUAC) was used to identify infant malnutrition. The risk difference (RD) due to maternal depression for each outcome was estimated using targeted maximum likelihood estimation (TMLE), a doubly robust causal inference method used to reduce bias in observational studies.

RESULTS

The cumulative incidence of diarrhea, ARI and malnutrition during 6-month follow-up was 17.0% (95%CI: 14.5, 19.6), 21.6% (95%CI: 18.89, 24.49), and 14.4% (95%CI: 12.2, 16.9), respectively. There was no association between antenatal depression and ARI (RD = - 1.3%; 95%CI: - 21.0, 18.5), diarrhea (RD = 0.8%; 95%CI: - 9.2, 10.9), or malnutrition (RD = -7.3%; 95%CI: - 22.0, 21.8). Similarly, postnatal depression was not associated with diarrhea (RD = -2.4%; 95%CI: - 9.6, 4.9), ARI (RD = - 3.2%; 95%CI: - 12.4, 5.9), or malnutrition (RD = 0.9%; 95%CI: - 7.6, 9.5).

CONCLUSION

There was no evidence for an association between perinatal depression and the risk of infant diarrhea, ARI, and malnutrition amongst women in Gondar Town. Previous reports suggesting increased risks resulting from maternal depression may be due to unobserved confounding.

Estimating counterfactuals for evaluation of ecological and conservation impact: an introduction to matching methods

Matching methods encompass non-parametric approaches to estimating counterfactual states through a rigorous selection of control units with similar characteristics to units submitted to an intervention. These methods enable comparisons between treated and control units in a way that facilitates understanding of causal relationships between interventions and outcomes. Matching methods have been used only recently in ecology and conservation biology, where such applications changed the way the field investigates causal questions, for example, in impact-evaluation studies. However, the strengths and limitations of matching methods are not well understood by most ecologists and environmental scientists. Herein, we review state-of-the-art matching methods aiming to help fill this gap in understanding. First, we present relevant theoretical concepts related to matching methods and related subjects such as counterfactual states and causation. Next, we propose guidelines and strategies for the application of matching methods in ecology and conservation biology. Finally, we discuss the possibilities for future applications of matching methods in the environmental sciences.

G-computation and machine learning for estimating the causal effects of binary exposure statuses on binary outcomes

In clinical research, there is a growing interest in the use of propensity score-based methods to estimate causal effects. G-computation is an alternative because of its high statistical power. Machine learning is also increasingly used because of its possible robustness to model misspecification. In this paper, we aimed to propose an approach that combines machine learning and G-computation when both the outcome and the exposure status are binary and is able to deal with small samples. We evaluated the performances of several methods, including penalized logistic regressions, a neural network, a support vector machine, boosted classification and regression trees, and a super learner through simulations. We proposed six different scenarios characterised by various sample sizes, numbers of covariates and relationships between covariates, exposure statuses, and outcomes. We have also illustrated the application of these methods, in which they were used to estimate the efficacy of barbiturates prescribed during the first 24 h of an episode of intracranial hypertension. In the context of GC, for estimating the individual outcome probabilities in two counterfactual worlds, we reported that the super learner tended to outperform the other approaches in terms of both bias and variance, especially for small sample sizes. The support vector machine performed well, but its mean bias was slightly higher than that of the super learner. In the investigated scenarios, G-computation associated with the super learner was a performant method for drawing causal inferences, even from small sample sizes.

Redlines and Greenspace: The Relationship between Historical Redlining and 2010 Greenspace across the United States

INTRODUCTION

Redlining, a racist mortgage appraisal practice of the 1930s, established and exacerbated racial residential segregation boundaries in the United States. Investment risk grades assigned >80y ago through security maps from the Home Owners' Loan Corporation (HOLC) are associated with current sociodemographics and adverse health outcomes. We assessed whether historical HOLC investment grades are associated with 2010 greenspace, a health-promoting neighborhood resource.

OBJECTIVES

We compared 2010 normalized difference vegetation index (NDVI) across previous HOLC neighborhood grades using propensity score restriction and matching.

METHODS

Security map shapefiles were downloaded from the Mapping Inequality Project. Neighborhood investment risk grades included A (best, green), B (blue), C (yellow), and D (hazardous, red, i.e., redlined). We used 2010 satellite imagery to calculate the average NDVI for each HOLC neighborhood. Our main outcomes were 2010 annual average NDVI and summer NDVI. We assigned areal-apportioned 1940 census measures to each HOLC neighborhood. We used propensity score restriction, matching, and targeted maximum likelihood estimation to limit model extrapolation, reduce confounding, and estimate the association between HOLC grade and NDVI for the following comparisons: Grades B vs. A, C vs. B, and D vs. C.

RESULTS

Across 102 urban areas (4,141 HOLC polygons), annual average ±standard deviation (SD) 2010 NDVI was 0.47 (±0.09), 0.43 (±0.09), 0.39 (±0.09), and 0.36 (±0.10) in Grades A-D, respectively. In analyses adjusted for current ecoregion and census region, 1940s census measures, and 1940s population density, annual average NDVI values in 2010 were estimated at -0.039 (95% CI: -0.045, -0.034), -0.024 (95% CI: -0.030, -0.018), and -0.026 (95% CI: -0.037, -0.015) for Grades B vs. A, C vs. B, and D vs. C, respectively, in the 1930s.

DISCUSSION

Estimates adjusted for historical characteristics indicate that neighborhoods assigned worse HOLC grades in the 1930s are associated with reduced present-day greenspace. https://doi.org/10.1289/EHP7495.

Using propensity scores to estimate effects of treatment initiation decisions: State of the science

Confounding can cause substantial bias in nonexperimental studies that aim to estimate causal effects. Propensity score methods allow researchers to reduce bias from measured confounding by summarizing the distributions of many measured confounders in a single score based on the probability of receiving treatment. This score can then be used to mitigate imbalances in the distributions of these measured confounders between those who received the treatment of interest and those in the comparator population, resulting in less biased treatment effect estimates. This methodology was formalized by Rosenbaum and Rubin in 1983 and, since then, has been used increasingly often across a wide variety of scientific disciplines. In this review article, we provide an overview of propensity scores in the context of real-world evidence generation with a focus on their use in the setting of single treatment decisions, that is, choosing between two therapeutic options. We describe five aspects of propensity score analysis: alignment with the potential outcomes framework, implications for study design, estimation procedures, implementation options, and reporting. We add context to these concepts by highlighting how the types of comparator used, the implementation method, and balance assessment techniques have changed over time. Finally, we discuss evolving applications of propensity scores.

Optimal subset selection for causal inference using machine learning ensembles and particle swarm optimization

We suggest and evaluate a method for optimal construction of synthetic treatment and control samples for the purpose of drawing causal inference. The balance optimization subset selection problem, which formulates minimization of aggregate imbalance in covariate distributions to reduce bias in data, is a new area of study in operations research. We investigate a novel metric, cross-validated area under the receiver operating characteristic curve (AUC) as a measure of balance between treatment and control groups. The proposed approach provides direct and automatic balancing of covariate distributions. In addition, the AUC-based approach is able to detect subtler distributional differences than existing measures, such as simple empirical mean/variance and count-based metrics. Thus, optimizing AUCs achieves a greater balance than the existing methods. Using 5 widely used real data sets and 7 synthetic data sets, we show that optimization of samples using existing methods (Chi-square, mean variance differences, Kolmogorov–Smirnov, and Mahalanobis) results in samples containing imbalance that is detectable using machine learning ensembles. We minimize covariate imbalance by minimizing the absolute value of the distance of the maximum cross-validated AUC on $$ M $$ M folds from 0.50, using evolutionary optimization. We demonstrate that particle swarm optimization (PSO) outperforms modified cuckoo swarm (MCS) for a gradient-free, non-linear noisy cost function. To compute AUCs, we use supervised binary classification approaches from the machine learning and credit scoring literature. Using superscore ensembles adds to the classifier-based two-sample testing literature. If the mean cross-validated AUC based on machine learning is 0.50, the two groups are indistinguishable and suitable for causal inference.

G-computation, propensity score-based methods, and targeted maximum likelihood estimator for causal inference with different covariates sets: a comparative simulation study

Controlling for confounding bias is crucial in causal inference. Distinct methods are currently employed to mitigate the effects of confounding bias. Each requires the introduction of a set of covariates, which remains difficult to choose, especially regarding the different methods. We conduct a simulation study to compare the relative performance results obtained by using four different sets of covariates (those causing the outcome, those causing the treatment allocation, those causing both the outcome and the treatment allocation, and all the covariates) and four methods: g-computation, inverse probability of treatment weighting, full matching and targeted maximum likelihood estimator. Our simulations are in the context of a binary treatment, a binary outcome and baseline confounders. The simulations suggest that considering all the covariates causing the outcome led to the lowest bias and variance, particularly for g-computation. The consideration of all the covariates did not decrease the bias but significantly reduced the power. We apply these methods to two real-world examples that have clinical relevance, thereby illustrating the real-world importance of using these methods. We propose an R package RISCA to encourage the use of g-computation in causal inference.

Evaluation of opioid discontinuation after non-orthopaedic surgery among chronic opioid users: a population-based cohort study

BACKGROUND

Many patients use opioids chronically before surgery; it is unclear if surgery alters the likelihood of ongoing opioid consumption in these patients.

METHODS

We performed a population-based matched cohort study of adults in Ontario, Canada undergoing one of 16 non-orthopaedic surgical procedures and who were chronically using opioids, defined as (1) an opioid prescription that overlapped the index date and (2) either a total of 120 or more cumulative calendar days of filled opioid prescriptions, or 10 or more prescriptions filled in the prior year. Each surgical patient was matched based on age, sex, Charlson comorbidity index, and daily preoperative opioid dose to three non-surgical patients who were also chronic opioid users. The primary outcome was time to opioid discontinuation.

RESULTS

The cohort included 4755 surgical and 14 265 matched non-surgical patients. After adjustment for sociodemographic characteristics and comorbidities, surgery was associated with an increased likelihood of opioid discontinuation (adjusted hazard ratio: 1.34, 95% confidence interval [CI]: 1.27, 1.42). Among surgical patients, factors associated with a reduced odds of discontinuation included a mean preoperative opioid dose above 90 morphine milligram equivalents (adjusted odds ratio [aOR]: 0.39; 95% CI: 0.32, 0.49) or filling a prescription for oxycodone (aOR: 0.73; 95% CI: 0.56, 0.98). Receipt of an in-patient Acute Pain Service consultation (aOR: 1.34; 95% CI: 1.06, 1.69) or residing in the highest neighbourhood income quintile (aOR: 1.35; 95% CI: 1.04, 1.79) were associated with a greater odds of opioid discontinuation.

CONCLUSIONS

For chronic opioid users, surgery was associated with an increased likelihood of discontinuation of opioids in the following year compared with non-surgical chronic opioid users.

Nearest Neighbour Propensity Score Matching and Bootstrapping for Estimating Binary Patient Response in Oncology: A Monte Carlo Simulation

Nearest Neighbour (NN) propensity score (PS) matching methods are commonly used in pharmacoepidemiology to estimate treatment response using observational data. Unfortunately, there is limited evidence on the optimal approach for accurately estimating binary treatment response and, more so, to estimate its variance. Bootstrapping, although commonly used to accurately estimate variance, is rarely used together with PS matching. In this Monte Carlo simulation-based study, we examined the performance of bootstrapping used in conjunction with PS matching, as opposed to different NN matching techniques, on a simulated dataset exhibiting varying levels of real world complexity. Thus, an experimental design was set up that independently varied the proportion of patients treated, the proportion of outcomes censored and the amount of PS matches used. Simulation results were externally validated on a real observational dataset obtained from the Belgian Cancer Registry. We found all investigated PS methods to be stable and concordant, with k-NN matching to be optimally dealing with the censoring problem, typically present in chronic cancer-related datasets, whilst being the least computationally expensive. In contrast, bootstrapping used in conjunction with PS matching, being the most computationally expensive, only showed superior results in small patient populations with long-term largely unobserved treatment effects.

Resection of recurrent glioblastoma multiforme in elderly patients: a pseudo-randomized analysis revealed clinical benefit

INTRODUCTION

Elderly patients constitute an expanding part of our society. Due to a continuously increasing life expectancy, an optimal quality of life is expected even into advanced age. Glioblastoma (GBM) is more common in older patients, but they are still often withheld from efficient treatment due to worry of worse tolerance and have a significantly worse prognosis compared to younger patients. Our retrospective observational study aimed to investigate the therapeutic benefit from a second resection in recurrent glioblastoma of elderly patients.

MATERIALS AND METHODS

We included a cohort of 39 elderly patients (> 65 years) with a second resection as treatment option in the case of a tumor recurrence. A causal inference model was built by multiple non- and semiparametric models, which was used to identify matched patients from our elderly GBM database which comprises 538 patients. The matched cohorts were analyzed by a Cox-regression model adjusted by time-dependent covariates.

RESULTS

The Cox-regression analysis showed a significant survival benefit (Hazard Ratio: 0.6, 95% CI 0.36-0.9, p-value = 0.0427) for the re-resected group (18.0 months, 95% CI 13.97-23.2 months) compared to the group without re-resection (10.1 months, 95% CI 8.09-20.9 months). No differences in the co-morbidities or hemato-oncological side effects during chemotherapy could be detected. Anesthetic- and surgical complications were rare and comparable to the complication rate of patients undergoing the first-line resection.

CONCLUSION

Taken together, in elderly patients, re-resection is an acceptable treatment option in the recurrent state of a glioblastoma. The individual evaluation of the patients' medical status as well as the chances of withstanding general anesthesia needs to be done in close interdisciplinary consultation. If these requirements are met, elderly patients benefit from a re-resection.

Environmental noise and sleep and mental health outcomes in a nationally representative sample of urban US adolescents

BACKGROUND

Environmental noise has been linked to negative health outcomes, like poor sleep, poor mental health, and cardiovascular disease, and likely accounts for more than 1 million disability-adjusted life years annually in Western Europe. Adolescence may be a particularly sensitive period for noise exposure due to an increased need for sleep, failure to meet sleep guidelines, and increased risk for first onset of some mental health disorders. However, the potential health effects of living in high-noise environments have not been studied in US adolescents, rarely in European adolescents, and mental health outcomes studied have not corresponded to diagnoses from the Diagnostic and Statistical Manual of Mental Disorders (DSM).

METHODS

Using a US-based nationally representative survey of urban adolescents (N = 4,508), we estimated associations of day-night average sound levels exceeding the US Environmental Protection Agency's 55 decibel limit with sleep outcomes and lifetime mental health DSM diagnoses. We implemented doubly robust targeted minimum loss-based estimation coupled with propensity score matching to account for numerous potential adolescent, household, and environmental confounders.

RESULTS

Living in a high- versus low-noise Census block group was associated with later bedtimes on weeknights (0.48 hours, 95% confidence interval [CI] = -0.15, 1.12) and weekend nights (0.65 hours, 95% CI = 0.37, 0.93), but not with total hours slept. Associations between living in a high- versus low-noise Census block group and mental disorders were mixed, with wide CIs, and not robust to sensitivity analyses.

CONCLUSIONS

We find evidence for an association between residence in a high-noise area and later bedtimes among urban adolescents but no consistent evidence of such an association with mental health disorders.

Performance of Matching Methods as Compared With Unmatched Ordinary Least Squares Regression Under Constant Effects

Matching methods are assumed to reduce the likelihood of a biased inference compared with ordinary least squares (OLS) regression. Using simulations, we compared inferences from propensity score matching, coarsened exact matching, and unmatched covariate-adjusted OLS regression to identify which methods, in which scenarios, produced unbiased inferences at the expected type I error rate of 5%. We simulated multiple data sets and systematically varied common support, discontinuities in the exposure and/or outcome, exposure prevalence, and analytical model misspecification. Matching inferences were often biased in comparison with OLS, particularly when common support was poor; when analysis models were correctly specified and common support was poor, the type I error rate was 1.6% for propensity score matching (statistically inefficient), 18.2% for coarsened exact matching (high), and 4.8% for OLS (expected). Our results suggest that when estimates from matching and OLS are similar (i.e., confidence intervals overlap), OLS inferences are unbiased more often than matching inferences; however, when estimates from matching and OLS are dissimilar (i.e., confidence intervals do not overlap), matching inferences are unbiased more often than OLS inferences. This empirical "rule of thumb" may help applied researchers identify situations in which OLS inferences may be unbiased as compared with matching inferences.

An educational intervention to improve knowledge about prevention against occupational asthma and allergies using targeted maximum likelihood estimation

PURPOSE

Occupational asthma and allergies are potentially preventable diseases affecting 5-15% of the working population. However, the use of preventive measures is often insufficient. The aim of this study was to estimate the average treatment effect of an educational intervention designed to improve the knowledge of preventive measures against asthma and allergies in farm apprentices from Bavaria (Southern Germany).

METHODS

Farm apprentices at Bavarian farm schools were asked to complete a questionnaire evaluating their knowledge about preventive measures against occupational asthma and allergies (use of personal protective equipment, personal and workplace hygiene measures). Eligible apprentices were randomized by school site to either a control or an intervention group. The intervention consisted of a short educational video about use of preventive measures. Six months after the intervention, subjects were asked to complete a post-intervention questionnaire. Of the 116 apprentices (70 intervention group, 46 control group) who answered the baseline questionnaire, only 47 subjects (41%; 17 intervention group, 30 control group) also completed the follow-up questionnaire. We, therefore, estimated the causal effect of the intervention using targeted maximum likelihood estimation. Models were controlled for potential confounders.

RESULTS

Based on the targeted maximum likelihood estimation, the intervention would have increased the proportion of correct answers on all six preventive measures by 18.4% (95% confidence interval 7.3-29.6%) had all participants received the intervention vs. had they all been in the control group.

CONCLUSIONS

These findings indicate the improvement of knowledge by the educational intervention.

Evaluation of Complementary and Integrative Health Approaches Among US Veterans with Musculoskeletal Pain Using Propensity Score Methods

Objectives

To examine the treatment effectiveness of complementary and integrative health approaches (CIH) on chronic pain using Propensity Score (PS) methods.

Design, Settings, and Participants

A retrospective cohort of 309,277 veterans with chronic musculoskeletal pain assessed over three years after initial diagnosis.

Methods

CIH exposure was defined as one or more clinical visits for massage, acupuncture, or chiropractic care. The treatment effect of CIH on self-rated pain intensity was examined using a longitudinal model. PS-matching and inverse probability of treatment weighting (IPTW) were used to account for potential selection and confounding biases.

Results

At baseline, veterans with (7,621) and without (301,656) CIH exposure differed significantly in 21 out of 35 covariates. During the follow-up period, on average CIH recipients had 0.83 (95% confidence interval [CI] = 0.77 to 0.89) points higher pain intensity ratings (range = 0-10) than nonrecipients. This apparent unfavorable effect size was reduced to 0.37 (95% CI = 0.28 to 0.45) after PS matching, 0.36 (95% CI = 0.29 to 0.44) with IPTW on the treated (IPTW-T) weighting, and diminished to null when integrating IPTW-T with PS matching (0.004, 95% CI = -0.09 to 0.10). An alternative IPTW model and conventional covariate adjustment appeared least powerful in terms of potential bias reduction. Sensitivity analyses restricting the follow-up period to one year after CIH initiation derived consistent results.

Conclusions

PS-based causal methods successfully eliminated baseline difference between exposure groups in all measured covariates, yet they did not detect a significant difference in the self-rated pain intensity outcome between veterans who received CIHs and those who did not during the follow-up period.

Performing an Informatics Consult: Methods and Challenges

Our health care system is plagued by missed opportunities, waste, and harm. Data generated in the course of care are often underutilized, scientific insight goes untranslated, and evidence is overlooked. To address these problems, we envisioned a system where aggregate patient data can be used at the bedside to provide practice-based evidence. To create that system, we directly connect practicing physicians to clinical researchers and data scientists through an informatics consult. Our team processes and classifies questions posed by clinicians, identifies the appropriate patient data to use, runs the appropriate analyses, and returns an answer, ideally in a 48-hour time window. Here, we discuss the methods that are used for data extraction, processing, and analysis in our consult. We continue to refine our informatics consult service, moving closer to a learning health care system.

Predicting the Population Health Impacts of Community Interventions: The Case of Alcohol Outlets and Binge Drinking

A substitution estimator can be used to predict how shifts in population exposures might change health. We illustrated this method by estimating how an upper limit on alcohol outlet density might alter binge drinking in the New York Social Environment Study (n = 4000), and provided statistical code and sample data. The largest differences in binge drinking were for an upper limit of 70 outlets per square mile; there was a -0.7% difference in binge drinking prevalence for New York City overall (95% confidence interval [CI] = -0.2%, -1.3%) and a -2.4% difference in binge drinking prevalence for the subset of communities the intervention modified (95% CI = -0.5%, -4.0%). A substitution estimator is a flexible tool for estimating population intervention parameters and improving the translation of public health research results to practitioners.