Sensitivity analysis for principal ignorability violation in estimating complier and noncomplier average causal effects.

An important strategy for identifying principal causal effects (popular estimands in settings with noncompliance) is to invoke the principal ignorability (PI) assumption. As PI is untestable, it is important to gauge how sensitive effect estimates are to its violation. We focus on this task for the common one-sided noncompliance setting where there are two principal strata, compliers and noncompliers. Under PI, compliers and noncompliers share the same outcome-mean-given-covariates function under the control condition. For sensitivity analysis, we allow this function to differ between compliers and noncompliers in several ways, indexed by an odds ratio, a generalized odds ratio, a mean ratio, or a standardized mean difference sensitivity parameter. We tailor sensitivity analysis techniques (with any sensitivity parameter choice) to several types of PI-based main analysis methods, including outcome regression, influence function (IF) based and weighting methods. We discuss range selection for the sensitivity parameter. We illustrate the sensitivity analyses with several outcome types from the JOBS II study. This application estimates nuisance functions parametrically - for simplicity and accessibility. In addition, we establish rate conditions on nonparametric nuisance estimation for IF-based estimators to be asymptotically normal - with a view to inform nonparametric inference.

Causal Inference for Social Network Data.

We describe semiparametric estimation and inference for causal effects using observational data from a single social network. Our asymptotic results are the first to allow for dependence of each observation on a growing number of other units as sample size increases. In addition, while previous methods have implicitly permitted only one of two possible sources of dependence among social network observations, we allow for both dependence due to transmission of information across network ties and for dependence due to latent similarities among nodes sharing ties. We propose new causal effects that are specifically of interest in social network settings, such as interventions on network ties and network structure. We use our methods to reanalyze an influential and controversial study that estimated causal peer effects of obesity using social network data from the Framingham Heart Study; after accounting for network structure we find no evidence for causal peer effects.

The Pregnancy, Arsenic, and Immune Response (PAIR) Study in rural northern Bangladesh.

BACKGROUND: Arsenic exposure and micronutrient deficiencies may alter immune reactivity to influenza vaccination in pregnant women, transplacental transfer of maternal antibodies to the foetus, and maternal and infant acute morbidity. OBJECTIVES: The Pregnancy, Arsenic, and Immune Response (PAIR) Study was designed to assess whether arsenic exposure and micronutrient deficiencies alter maternal and newborn immunity and acute morbidity following maternal seasonal influenza vaccination during pregnancy. POPULATION: The PAIR Study recruited pregnant women across a large rural study area in Gaibandha District, northern Bangladesh, 2018-2019. DESIGN: Prospective, longitudinal pregnancy and birth cohort. METHODS: We conducted home visits to enrol pregnant women in the late first or early second trimester (11-17 weeks of gestational age). Women received a quadrivalent seasonal inactivated influenza vaccine at enrolment. Follow-up included up to 13 visits between enrolment and 3 months postpartum. Arsenic was measured in drinking water and maternal urine. Micronutrient deficiencies were assessed using plasma biomarkers. Vaccine-specific antibody titres were measured in maternal and infant serum. Weekly telephone surveillance ascertained acute morbidity symptoms in women and infants. PRELIMINARY RESULTS: We enrolled 784 pregnant women between October 2018 and March 2019. Of 784 women who enrolled, 736 (93.9%) delivered live births and 551 (70.3%) completed follow-up visits to 3 months postpartum. Arsenic was detected (≥0.02 µg/L) in 99.7% of water specimens collected from participants at enrolment. The medians (interquartile ranges) of water and urinary arsenic at enrolment were 5.1 (0.5, 25.1) µg/L and 33.1 (19.6, 56.5) µg/L, respectively. Water and urinary arsenic were strongly correlated (Spearman's ⍴ = 0.72) among women with water arsenic ≥ median but weakly correlated (⍴ = 0.17) among women with water arsenic < median. CONCLUSIONS: The PAIR Study is well positioned to examine the effects of low-moderate arsenic exposure and micronutrient deficiencies on immune outcomes in women and infants. REGISTRATION: NCT03930017.

Finding influential subjects in a network using a causal framework.

Researchers across a wide array of disciplines are interested in finding the most influential subjects in a network. In a network setting, intervention effects and health outcomes can spill over from one node to another through network ties, and influential subjects are expected to have a greater impact than others. For this reason, network research in public health has attempted to maximize health and behavioral changes by intervening on a subset of influential subjects. Although influence is often defined only implicitly in most of the literature, the operative notion of influence is inherently causal in many cases: influential subjects are those we should intervene on to achieve the greatest overall effect across the entire network. In this work, we define a causal notion of influence using potential outcomes. We review existing influence measures, such as node centrality, that largely rely on the particular features of the network structure and/or on certain diffusion models that predict the pattern of information or diseases spreads through network ties. We provide simulation studies to demonstrate when popular centrality measures can agree with our causal measure of influence. As an illustrative example, we apply several popular centrality measures to the HIV risk network in the Transmission Reduction Intervention Project and demonstrate the assumptions under which each centrality can represent the causal influence of each participant in the study.

Causal mediation analysis: From simple to more robust strategies for estimation of marginal natural (in)direct effects.

This paper aims to provide practitioners of causal mediation analysis with a better understanding of estimation options. We take as inputs two familiar strategies (weighting and model-based prediction) and a simple way of combining them (weighted models), and show how a range of estimators can be generated, with different modeling requirements and robustness properties. The primary goal is to help build intuitive appreciation for robust estimation that is conducive to sound practice. We do this by visualizing the target estimand and the estimation strategies. A second goal is to provide a "menu" of estimators that practitioners can choose from for the estimation of marginal natural (in)direct effects. The estimators generated from this exercise include some that coincide or are similar to existing estimators and others that have not previously appeared in the literature. We note several different ways to estimate the weights for cross-world weighting based on three expressions of the weighting function, including one that is novel; and show how to check the resulting covariate and mediator balance. We use a random continuous weights bootstrap to obtain confidence intervals, and also derive general asymptotic variance formulas for the estimators. The estimators are illustrated using data from an adolescent alcohol use prevention study. R-code is provided.

Hydroxychloroquine/chloroquine for the treatment of hospitalized patients with COVID-19: An individual participant data meta-analysis.

BACKGROUND: Results from observational studies and randomized clinical trials (RCTs) have led to the consensus that hydroxychloroquine (HCQ) and chloroquine (CQ) are not effective for COVID-19 prevention or treatment. Pooling individual participant data, including unanalyzed data from trials terminated early, enables more detailed investigation of the efficacy and safety of HCQ/CQ among subgroups of hospitalized patients. METHODS: We searched ClinicalTrials.gov in May and June 2020 for US-based RCTs evaluating HCQ/CQ in hospitalized COVID-19 patients in which the outcomes defined in this study were recorded or could be extrapolated. The primary outcome was a 7-point ordinal scale measured between day 28 and 35 post enrollment; comparisons used proportional odds ratios. Harmonized de-identified data were collected via a common template spreadsheet sent to each principal investigator. The data were analyzed by fitting a prespecified Bayesian ordinal regression model and standardizing the resulting predictions. RESULTS: Eight of 19 trials met eligibility criteria and agreed to participate. Patient-level data were available from 770 participants (412 HCQ/CQ vs 358 control). Baseline characteristics were similar between groups. We did not find evidence of a difference in COVID-19 ordinal scores between days 28 and 35 post-enrollment in the pooled patient population (odds ratio, 0.97; 95% credible interval, 0.76-1.24; higher favors HCQ/CQ), and found no convincing evidence of meaningful treatment effect heterogeneity among prespecified subgroups. Adverse event and serious adverse event rates were numerically higher with HCQ/CQ vs control (0.39 vs 0.29 and 0.13 vs 0.09 per patient, respectively). CONCLUSIONS: The findings of this individual participant data meta-analysis reinforce those of individual RCTs that HCQ/CQ is not efficacious for treatment of COVID-19 in hospitalized patients.

Long-Term Fine Particulate Matter Concentrations and Prevalence of Severe Acute Respiratory Syndrome Coronavirus 2: Differential Relationships by Socioeconomic Status Among Pregnant Individuals in New York City.

We aimed to determine whether long-term ambient concentrations of fine particulate matter (particulate matter with an aerodynamic diameter less than or equal to 2.5 µm (PM2.5)) were associated with increased risk of testing positive for coronavirus disease 2019 (COVID-19) among pregnant individuals who were universally screened at delivery and whether socioeconomic status (SES) modified this relationship. We used obstetrical data collected from New-York Presbyterian Hospital/Columbia University Irving Medical Center in New York, New York, between March and December 2020, including data on Medicaid use (a proxy for low SES) and COVID-19 test results. We linked estimated 2018-2019 PM2.5 concentrations (300-m resolution) with census-tract-level population density, household size, income, and mobility (as measured by mobile-device use) on the basis of residential address. Analyses included 3,318 individuals; 5% tested positive for COVID-19 at delivery, 8% tested positive during pregnancy, and 48% used Medicaid. Average long-term PM2.5 concentrations were 7.4 (standard deviation, 0.8) µg/m3. In adjusted multilevel logistic regression models, we saw no association between PM2.5 and ever testing positive for COVID-19; however, odds were elevated among those using Medicaid (per 1-µg/m3 increase, odds ratio = 1.6, 95% confidence interval: 1.0, 2.5). Further, while only 22% of those testing positive showed symptoms, 69% of symptomatic individuals used Medicaid. SES, including unmeasured occupational exposures or increased susceptibility to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) due to concurrent social and environmental exposures, may explain the increased odds of testing positive for COVID-19 being confined to vulnerable pregnant individuals using Medicaid.

Neighborhood Socioeconomic Environment and Risk of Type 2 Diabetes: Associations and Mediation Through Food Environment Pathways in Three Independent Study Samples.

OBJECTIVE: We examined whether relative availability of fast-food restaurants and supermarkets mediates the association between worse neighborhood socioeconomic conditions and risk of developing type 2 diabetes (T2D). RESEARCH DESIGN AND METHODS: As part of the Diabetes Location, Environmental Attributes, and Disparities Network, three academic institutions used harmonized environmental data sources and analytic methods in three distinct study samples: 1) the Veterans Administration Diabetes Risk (VADR) cohort, a national administrative cohort of 4.1 million diabetes-free veterans developed using electronic health records (EHRs); 2) Reasons for Geographic and Racial Differences in Stroke (REGARDS), a longitudinal, epidemiologic cohort with Stroke Belt region oversampling (N = 11,208); and 3) Geisinger/Johns Hopkins University (G/JHU), an EHR-based, nested case-control study of 15,888 patients with new-onset T2D and of matched control participants in Pennsylvania. A census tract-level measure of neighborhood socioeconomic environment (NSEE) was developed as a community type-specific z-score sum. Baseline food-environment mediators included percentages of 1) fast-food restaurants and 2) food retail establishments that are supermarkets. Natural direct and indirect mediating effects were modeled; results were stratified across four community types: higher-density urban, lower-density urban, suburban/small town, and rural. RESULTS: Across studies, worse NSEE was associated with higher T2D risk. In VADR, relative availability of fast-food restaurants and supermarkets was positively and negatively associated with T2D, respectively, whereas associations in REGARDS and G/JHU geographies were mixed. Mediation results suggested that little to none of the NSEE-diabetes associations were mediated through food-environment pathways. CONCLUSIONS: Worse neighborhood socioeconomic conditions were associated with higher T2D risk, yet associations are likely not mediated through food-environment pathways.

Hydroxychloroquine/chloroquine for the treatment of hospitalized patients with COVID-19: An individual participant data meta-analysis.

Background: Results from observational studies and randomized clinical trials (RCTs) have led to the consensus that hydroxychloroquine (HCQ) and chloroquine (CQ) are not effective for COVID-19 prevention or treatment. Pooling individual participant data, including unanalyzed data from trials terminated early, enables more detailed investigation of the efficacy and safety of HCQ/CQ among subgroups of hospitalized patients. Methods: We searched ClinicalTrials.gov in May and June 2020 for US-based RCTs evaluating HCQ/CQ in hospitalized COVID-19 patients in which the outcomes defined in this study were recorded or could be extrapolated. The primary outcome was a 7-point ordinal scale measured between day 28 and 35 post enrollment; comparisons used proportional odds ratios. Harmonized de-identified data were collected via a common template spreadsheet sent to each principal investigator. The data were analyzed by fitting a prespecified Bayesian ordinal regression model and standardizing the resulting predictions. Results: Eight of 19 trials met eligibility criteria and agreed to participate. Patient-level data were available from 770 participants (412 HCQ/CQ vs 358 control). Baseline characteristics were similar between groups. We did not find evidence of a difference in COVID-19 ordinal scores between days 28 and 35 post-enrollment in the pooled patient population (odds ratio, 0.97; 95% credible interval, 0.76-1.24; higher favors HCQ/CQ), and found no convincing evidence of meaningful treatment effect heterogeneity among prespecified subgroups. Adverse event and serious adverse event rates were numerically higher with HCQ/CQ vs control (0.39 vs 0.29 and 0.13 vs 0.09 per patient, respectively). Conclusions: The findings of this individual participant data meta-analysis reinforce those of individual RCTs that HCQ/CQ is not efficacious for treatment of COVID-19 in hospitalized patients.

Clarifying causal mediation analysis: Effect identification via three assumptions and five potential outcomes.

Causal mediation analysis is complicated with multiple effect definitions that require different sets of assumptions for identification. This article provides a systematic explanation of such assumptions. We define five potential outcome types whose means are involved in various effect definitions. We tackle their mean/distribution's identification, starting with the one that requires the weakest assumptions and gradually building up to the one that requires the strongest assumptions. This presentation shows clearly why an assumption is required for one estimand and not another, and provides a succinct table from which an applied researcher could pick out the assumptions required for identifying the causal effects they target. Using a running example, the article illustrates the assembling and consideration of identifying assumptions for a range of causal contrasts. For several that are commonly encountered in the literature, this exercise clarifies that identification requires weaker assumptions than those often stated in the literature. This attention to the details also draws attention to the differences in the positivity assumption for different estimands, with practical implications. Clarity on the identifying assumptions of these various estimands will help researchers conduct appropriate mediation analyses and interpret the results with appropriate caution given the plausibility of the assumptions.

A Warning About Using Predicted Values From Regression Models for Epidemiologic Inquiry.

In many settings, researchers may not have direct access to data on 1 or more variables needed for an analysis and instead may use regression-based estimates of those variables. Using such estimates in place of original data, however, introduces complications and can result in uninterpretable analyses. In simulations and observational data, we illustrate the issues that arise when an average treatment effect is estimated from data where the outcome of interest is predicted from an auxiliary model. We show that bias in any direction can result, under both the null and alternative hypotheses.

Causal inference, social networks and chain graphs.

Traditionally, statistical inference and causal inference on human subjects rely on the assumption that individuals are independently affected by treatments or exposures. However, recently there has been increasing interest in settings, such as social networks, where individuals may interact with one another such that treatments may spill over from the treated individual to their social contacts and outcomes may be contagious. Existing models proposed for causal inference using observational data from networks of interacting individuals have two major shortcomings. First, they often require a level of granularity in the data that is infeasible in practice to collect in most settings and, second, the models are high dimensional and often too big to fit to the available data. We illustrate and justify a parsimonious parameterization for network data with interference and contagion. Our parameterization corresponds to a particular family of graphical models known as chain graphs. We argue that, in some settings, chain graph models approximate the marginal distribution of a snapshot of a longitudinal data-generating process on interacting units. We illustrate the use of chain graphs for causal inference about collective decision making in social networks by using data from US Supreme Court decisions between 1994 and 2004 and in simulations.

Integrated HIV testing, prevention, and treatment intervention for key populations in India: a cluster-randomised trial.

BACKGROUND: To achieve reductions in HIV incidence, we need strategies to engage key population at risk for HIV in low-income and middle-income countries. We evaluated the effectiveness of integrated care centres in India that provided single-venue HIV testing, prevention, and treatment services for people who inject drugs (PWID) and men who have sex with men (MSM). METHODS: We did baseline respondent-driven sampling surveys in 27 sites across India, and selected 22 of these (12 PWID and ten MSM) for a cluster randomised trial on the basis of high HIV prevalence and logistical considerations. We used stratified (by PWID and MSM), restricted randomisation to allocate sites to either the integrated care intervention or usual care (11 sites per group). We implemented integrated care centres in 11 cities (six PWID integrated care centres embedded within opioid agonist treatment centres and five MSM centres within government-sponsored health services), with a single integrated care centre per city in all but one city. After a 2-year intervention phase, we did respondent-driven sampling evaluation surveys of target population members who were aged 18 years or older at all sites. The primary outcome was self-reported HIV testing in the previous 12 months (recent testing), determined via the evaluation survey. We used a biometric identification system to estimate integrated care centre exposure (visited an integrated care centre at least once) among evaluation survey participants at intervention sites. This trial is registered with ClinicalTrials.gov, number NCT01686750. FINDINGS: Between Oct 1, 2012, and Dec 19, 2013, we recruited 11 993 PWID and 9997 MSM in the baseline survey and, between Aug, 1 2016, and May 27, 2017, surveyed 11 721 PWID and 10 005 MSM in the evaluation survey using respondent-driven sampling, across the 22 trial sites. During the intervention phase, integrated care centres provided HIV testing for 14 698 unique clients (7630 PWID and 7068 MSM. In the primary population-level analysis, recent HIV testing was 31% higher at integrated care centres than at usual care sites (adjusted prevalence ratio [PR] 1·31, 95% CI 0·95-1·81, p=0·09). Among survey participants at intervention sites, integrated care centre exposure was lower than expected (median exposure 40% at PWID sites and 24% at MSM sites). In intervention sites, survey participants who visited an integrated care centre were more likely to report recent HIV testing than were participants who had not (adjusted PR 3·46, 2·94-4·06). INTERPRETATION: Although integrated care centres increased HIV testing among visitors, our low exposure findings suggest that the scale-up of a single integrated care centre in most cities was insufficient to serve the large PWID and MSM populations. Future work should address the use of population size estimates to guide the dose of combination HIV interventions targeting key populations. FUNDING: US National Institutes of Health and the Elton John AIDS Foundation.

Analysis of covariance in randomized trials: More precision and valid confidence intervals, without model assumptions.

"Covariate adjustment" in the randomized trial context refers to an estimator of the average treatment effect that adjusts for chance imbalances between study arms in baseline variables (called "covariates"). The baseline variables could include, for example, age, sex, disease severity, and biomarkers. According to two surveys of clinical trial reports, there is confusion about the statistical properties of covariate adjustment. We focus on the analysis of covariance (ANCOVA) estimator, which involves fitting a linear model for the outcome given the treatment arm and baseline variables, and trials that use simple randomization with equal probability of assignment to treatment and control. We prove the following new (to the best of our knowledge) robustness property of ANCOVA to arbitrary model misspecification: Not only is the ANCOVA point estimate consistent (as proved by Yang and Tsiatis, 2001) but so is its standard error. This implies that confidence intervals and hypothesis tests conducted as if the linear model were correct are still asymptotically valid even when the linear model is arbitrarily misspecified, for example, when the baseline variables are nonlinearly related to the outcome or there is treatment effect heterogeneity. We also give a simple, robust formula for the variance reduction (equivalently, sample size reduction) from using ANCOVA. By reanalyzing completed randomized trials for mild cognitive impairment, schizophrenia, and depression, we demonstrate how ANCOVA can achieve variance reductions of 4 to 32%.

Retirements of Coal and Oil Power Plants in California: Association With Reduced Preterm Birth Among Populations Nearby.

Coal and oil power plant retirements reduce air pollution nearby, but few studies have leveraged these natural experiments for public health research. We used California Department of Public Health birth records and US Energy Information Administration data from 2001-2011 to evaluate the relationship between the retirements of 8 coal and oil power plants and nearby preterm (gestational age of <37 weeks) birth. We conducted a difference-in-differences analysis using adjusted linear mixed models that included 57,005 births-6.3% of which were preterm-to compare the probability of preterm birth before and after power plant retirement among mothers residing within 0-5 km and 5-10 km of the 8 power plants. We found that power plant retirements were associated with a decrease in the proportion of preterm birth within 5 km (-0.019, 95% CI: -0.031, -0.008) and 5-10 km (-0.015, 95% CI: -0.024, -0.007), controlling for secular trends with mothers living 10-20 km away. For the 0-5-km area, this corresponds to a reduction in preterm birth from 7.0% to 5.1%. Subgroup analyses indicated a potentially larger association among non-Hispanic black and Asian mothers than among non-Hispanic white and Hispanic mothers and no differences in educational attainment. Future coal and oil power plant retirements may reduce preterm birth among nearby populations.