Toward Causal Inference With Interference

Maternal and child health intervention to promote behaviour change: a population-level cluster-randomised controlled trial in Honduras

OBJECTIVES

To assess the efficacy of a sustained educational intervention to affect diverse outcomes across the pregnancy and infancy timeline.

SETTING

A multi-arm cluster-randomised controlled trial in 99 villages in Honduras' Copán region, involving 16 301 people in 5633 households from October 2015 to December 2019.

PARTICIPANTS

Residents aged 12 and older were eligible. A photographic census involved 93% of the population, with 13 881 and 10 263 individuals completing baseline and endline surveys, respectively.

INTERVENTION

22-month household-based counselling intervention aiming to improve practices, knowledge and attitudes related to maternal, neonatal and child health.

PRIMARY AND SECONDARY OUTCOME MEASURES

Primary outcomes were prenatal/postnatal care behaviours, facility births, exclusive breast feeding, parental involvement, treatment of diarrhoea and respiratory illness, reproductive health, and gender/reproductive norms. Secondary outcomes were knowledge and attitudes related to the primary outcomes.

RESULTS

Parents targeted for the intervention were 16.4% (95% CI 3.1%-29.8%, p=0.016) more likely to have their newborn's health checked in a health facility within 3 days of birth; 19.6% (95% CI 4.2%-35.1%, p=0.013) more likely to not wrap a fajero around the umbilical cord in the first week after birth; and 8.9% (95% CI 0.3%-17.5%, p=0.043) more likely to report that the mother breast fed immediately after birth. Changes in knowledge and attitudes related to these primary outcomes were also observed. We found no significant effect on various other practices.

CONCLUSION

A sustained counselling intervention delivered in the home setting by community health workers can meaningfully change practices, knowledge and attitudes related to proper newborn care following birth, including professional care-seeking, umbilical cord care and breast feeding.

TRIAL REGISTRATION NUMBER

NCT02694679.

Causal Inference About the Effects of Interventions From Observational Studies in Medical Journals

Importance

Many medical journals, including JAMA, restrict the use of causal language to the reporting of randomized clinical trials. Although well-conducted randomized clinical trials remain the preferred approach for answering causal questions, methods for observational studies have advanced such that causal interpretations of the results of well-conducted observational studies may be possible when strong assumptions hold. Furthermore, observational studies may be the only practical source of information for answering some questions about the causal effects of medical or policy interventions, can support the study of interventions in populations and settings that reflect practice, and can help identify interventions for further experimental investigation. Identifying opportunities for the appropriate use of causal language when describing observational studies is important for communication in medical journals.

Observations

A structured approach to whether and how causal language may be used when describing observational studies would enhance the communication of research goals, support the assessment of assumptions and design and analytic choices, and allow for more clear and accurate interpretation of results. Building on the extensive literature on causal inference across diverse disciplines, we suggest a framework for observational studies that aim to provide evidence about the causal effects of interventions based on 6 core questions: what is the causal question; what quantity would, if known, answer the causal question; what is the study design; what causal assumptions are being made; how can the observed data be used to answer the causal question in principle and in practice; and is a causal interpretation of the analyses tenable?

Conclusions and Relevance

Adoption of the proposed framework to identify when causal interpretation is appropriate in observational studies promises to facilitate better communication between authors, reviewers, editors, and readers. Practical implementation will require cooperation between editors, authors, and reviewers to operationalize the framework and evaluate its effect on the reporting of empirical research.

Revisiting the effects of maternal education on adolescents' academic performance: Doubly robust estimation in a network-based observational study

In many contexts, particularly when study subjects are adolescents, peer effects can invalidate typical statistical requirements in the data. For instance, it is plausible that a student's academic performance is influenced both by their own mother's educational level as well as that of their peers. Since the underlying social network is measured, the Add Health study provides a unique opportunity to examine the impact of maternal college education on adolescent school performance, both direct and indirect. However, causal inference on populations embedded in social networks poses technical challenges, since the typical no interference assumption no longer holds. While inverse probability-of-treatment weighted (IPW) estimators have been developed for this setting, they are often highly unstable. Motivated by the question of maternal education, we propose doubly robust (DR) estimators combining models for treatment and outcome that are consistent and asymptotically normal if either model is correctly specified. We present empirical results that illustrate the DR property and the efficiency gain of DR over IPW estimators even when the treatment model is misspecified. Contrary to previous studies, our robust analysis does not provide evidence of an indirect effect of maternal education on academic performance within adolescents' social circles in Add Health.

Approximate balancing weights for clustered observational study designs

In a clustered observational study, a treatment is assigned to groups and all units within the group are exposed to the treatment. We develop a new method for statistical adjustment in clustered observational studies using approximate balancing weights, a generalization of inverse propensity score weights that solve a convex optimization problem to find a set of weights that directly minimize a measure of covariate imbalance, subject to an additional penalty on the variance of the weights. We tailor the approximate balancing weights optimization problem to the clustered observational study setting by deriving an upper bound on the mean square error and finding weights that minimize this upper bound, linking the level of covariate balance to a bound on the bias. We implement the procedure by specializing the bound to a random cluster-level effects model, leading to a variance penalty that incorporates the signal-to-noise ratio and penalizes the weight on individuals and the total weight on groups differently according to the the intra-class correlation.

G-formula for observational studies under stratified interference, with application to bed net use on malaria

Assessing population-level effects of vaccines and other infectious disease prevention measures is important to the field of public health. In infectious disease studies, one person's treatment may affect another individual's outcome, that is, there may be interference between units. For example, the use of bed nets to prevent malaria by one individual may have an indirect effect on other individuals living in close proximity. In some settings, individuals may form groups or clusters where interference only occurs within groups, that is, there is partial interference. Inverse probability weighted estimators have previously been developed for observational studies with partial interference. Unfortunately, these estimators are not well suited for studies with large clusters. Therefore, in this paper, the parametric g-formula is extended to allow for partial interference. G-formula estimators are proposed for overall effects, effects when treated, and effects when untreated. The proposed estimators can accommodate large clusters and do not suffer from the g-null paradox that may occur in the absence of interference. The large sample properties of the proposed estimators are derived assuming no unmeasured confounders and that the partial interference takes a particular form (referred to as 'weak stratified interference'). Simulation studies are presented demonstrating the finite-sample performance of the proposed estimators. The Demographic and Health Survey from the Democratic Republic of the Congo is then analyzed using the proposed g-formula estimators to assess the effects of bed net use on malaria.

Target Trial Emulation: Does surgical versus non-surgical management of cranial cruciate ligament rupture in dogs cause different outcomes?

Target trial emulation applies design principles from randomised controlled trials to the analysis of observational data for causal inference and is increasingly used within human epidemiology. Using anonymised veterinary clinical data from the VetCompass Programme, this study applied the target trial emulation framework to determine whether surgical (compared to non-surgical) management for cranial cruciate ligament (CCL) rupture in dogs causes improved short- and long-term lameness and analgesia outcomes. The emulated target trial included dogs diagnosed with CCL rupture between January 1, 2019 and December 31, 2019 within the VetCompass database. Inclusion in the emulated trial required dogs aged ≥ 1.5 and < 12 years, first diagnosed with unilateral CCL rupture during 2019 and with no prior history of contralateral ligament rupture or stifle surgery. Dogs were retrospectively observed to have surgical or non-surgical management. Informed from a directed acyclic graph derived from expert opinion, data on the following variables were collected: age, breed, bodyweight, neuter status, insurance status, non-orthopaedic comorbidities, orthopaedic comorbidities and veterinary group. Inverse probability of treatment weighting (IPTW) was used to adjust for confounding, with weights calculated based on a binary logistic regression exposure model. Censored dogs were accounted for in the IPTW analysis using inverse probability of censoring weighting (IPCW). The IPCWs were combined with IPTWs and used to weight each dog's contribution to binary logistic regression outcome models. Standardized mean differences (SMD) examined the balance of covariate distribution between treatment groups. The emulated trial included 615 surgical CCL rupture cases and 200 non-surgical cases. The risk difference for short-term lameness in surgically managed cases (compared with non-surgically managed cases) was -25.7% (95% confidence interval (CI) -36.7% to -15.9%) and the risk difference for long-term lameness -31.7% (95% CI -37.9% to -18.1%). The study demonstrated the application of the target trial framework to veterinary observational data. The findings show that surgical management causes a reduction in short- and long-term lameness compared with non-surgical management in dogs.

Methods for the estimation of direct and indirect vaccination effects by combining data from individual- and cluster-randomized trials

Both individually and cluster randomized study designs have been used for vaccine trials to assess the effects of vaccine on reducing the risk of disease or infection. The choice between individually and cluster randomized designs is often driven by the target estimand of interest (eg, direct versus total), statistical power, and, importantly, logistic feasibility. To combat emerging infectious disease threats, especially when the number of events from one single trial may not be adequate to obtain vaccine effect estimates with a desired level of precision, it may be necessary to combine information across multiple trials. In this article, we propose a model formulation to estimate the direct, indirect, total, and overall vaccine effects combining data from trials with two types of study designs: individual-randomization and cluster-randomization, based on a Cox proportional hazards model, where the hazard of infection depends on both vaccine status of the individual as well as the vaccine status of the other individuals in the same cluster. We illustrate the use of the proposed model and assess the potential efficiency gain from combining data from multiple trials, compared to using data from each individual trial alone, through two simulation studies, one of which is designed based on a cholera vaccine trial previously carried out in Matlab, Bangladesh.

Cluster Randomized Trials Designed to Support Generalizable Inferences

When planning a cluster randomized trial, evaluators often have access to an enumerated cohort representing the target population of clusters. Practicalities of conducting the trial, such as the need to oversample clusters with certain characteristics in order to improve trial economy or support inferences about subgroups of clusters, may preclude simple random sampling from the cohort into the trial, and thus interfere with the goal of producing generalizable inferences about the target population. We describe a nested trial design where the randomized clusters are embedded within a cohort of trial-eligible clusters from the target population and where clusters are selected for inclusion in the trial with known sampling probabilities that may depend on cluster characteristics (e.g., allowing clusters to be chosen to facilitate trial conduct or to examine hypotheses related to their characteristics). We develop and evaluate methods for analyzing data from this design to generalize causal inferences to the target population underlying the cohort. We present identification and estimation results for the expectation of the average potential outcome and for the average treatment effect, in the entire target population of clusters and in its non-randomized subset. In simulation studies, we show that all the estimators have low bias but markedly different precision. Cluster randomized trials where clusters are selected for inclusion with known sampling probabilities that depend on cluster characteristics, combined with efficient estimation methods, can precisely quantify treatment effects in the target population, while addressing objectives of trial conduct that require oversampling clusters on the basis of their characteristics.

Causal Effects of Stochastic PrEP Interventions on HIV Incidence Among Men Who Have Sex With Men

Antiretroviral preexposure prophylaxis (PrEP) is highly effective in preventing human immunodeficiency virus (HIV) infection, but uptake has been limited and inequitable. Although interventions to increase PrEP uptake are being evaluated in clinical trials among men who have sex with men (MSM), those trials cannot evaluate effects on HIV incidence. Estimates from observational studies of the causal effects of PrEP-uptake interventions on HIV incidence can inform decisions about intervention scale-up. We used longitudinal electronic health record data from HIV-negative MSM accessing care at Fenway Health, a community health center in Boston, Massachusetts, from January 2012 through February 2018, with 2 years of follow-up. We considered stochastic interventions that increased the chance of initiating PrEP in several high-priority subgroups. We estimated the effects of these interventions on population-level HIV incidence using a novel inverse-probability weighted estimator of the generalized g-formula, adjusting for baseline and time-varying confounders. Our results suggest that even modest increases in PrEP initiation in high-priority subgroups of MSM could meaningfully reduce HIV incidence in the overall population of MSM. Interventions tailored to Black and Latino MSM should be prioritized to maximize equity and impact.

Investigating environmental effects on phonology using diachronic models

Previous work has proposed various mechanisms by which the environment may affect the emergence of linguistic features. For example, dry air may cause careful control of pitch to be more effortful, and so affect the emergence of linguistic distinctions that rely on pitch such as lexical tone or vowel inventories. Criticisms of these proposals point out that there are both historical and geographic confounds that need to be controlled for. We take a causal inference approach to this problem to design the most detailed test of the theory to date. We analyse languages from the Bantu language family, using a prior geographic-phylogenetic tree of relationships to establish where and when languages were spoken. This is combined with estimates of humidity for those times and places, taken from historical climate models. We then estimate the strength of causal relationships in a causal path model, controlling for various influences of inheritance and borrowing. We find no evidence to support the previous claims that humidity affects the emergence of lexical tone. This study shows how using causal inference approaches lets us test complex causal claims about the cultural evolution of language.

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.

Instrumental variable analysis for cost outcome: Application to the effect of primary care visit on medical cost among low-income adults

Medical cost data often consist of zero values as well as extremely right-skewed positive values. A two-part model is a popular choice for analyzing medical cost data, where the first part models the probability of a positive cost using logistic regression and the second part models the positive cost using a lognormal or Gamma distribution. To address the unmeasured confounding in studies on cost outcome under two-part models, two instrumental variable (IV) methods, two-stage residual inclusion (2SRI) and two-stage prediction substitution (2SPS) are widely applied. However, previous literature demonstrated that both the 2SRI and the 2SPS could fail to consistently estimate the causal effect among compliers under standard IV assumptions for binary and survival outcomes. Our simulation studies confirmed that it continued to be the case for a two-part model, which is another nonlinear model. In this article, we develop a model-based IV approach, Instrumental Variable with Two-Part model (IV2P), to obtain a consistent estimate of the causal effect among compliers for cost outcome under standard IV assumptions. In addition, we develop sensitivity analysis approaches to allow the evaluation of the sensitivity of the causal conclusions to potential quantified violations of the exclusion restriction assumption and the randomization of IV assumption. We apply our method to a randomized cash incentive study to evaluate the effect of a primary care visit on medical cost among low-income adults newly covered by a primary care program.

Causal models and causal modelling in obesity: foundations, methods and evidence

Discussing causes in science, if we are to do so in a way that is sensible, begins at the root. All too often, we jump to discussing specific postulated causes but do not first consider what we mean by, for example, causes of obesity or how we discern whether something is a cause. In this paper, we address what we mean by a cause, discuss what might and might not constitute a reasonable causal model in the abstract, speculate about what the causal structure of obesity might be like overall and the types of things we should be looking for, and finally, delve into methods for evaluating postulated causes and estimating causal effects. We offer the view that different meanings of the concept of causal factors in obesity research are regularly being conflated, leading to confusion, unclear thinking and sometimes nonsense. We emphasize the idea of different kinds of studies for evaluating various aspects of causal effects and discuss experimental methods, assumptions and evaluations. We use analogies from other areas of research to express the plausibility that only inelegant solutions will be truly informative. Finally, we offer comments on some specific postulated causal factors. This article is part of a discussion meeting issue 'Causes of obesity: theories, conjectures and evidence (Part II)'.

Target trial emulation: Do antimicrobials or gastrointestinal nutraceuticals prescribed at first presentation for acute diarrhoea cause a better clinical outcome in dogs under primary veterinary care in the UK?

Target trial emulation applies design principles from randomised controlled trials to the analysis of observational data for causal inference and is increasingly used within human epidemiology. Veterinary electronic clinical records represent a potentially valuable source of information to estimate real-world causal effects for companion animal species. This study employed the target trial framework to evaluate the usefulness on veterinary observational data. Acute diarrhoea in dogs was used as a clinical exemplar. Inclusion required dogs aged ≥ 3 months and < 10 years, presenting for veterinary primary care with acute diarrhoea during 2019. Treatment strategies were: 1. antimicrobial prescription compared to no antimicrobial prescription and 2. gastrointestinal nutraceutical prescription compared to no gastrointestinal nutraceutical prescription. The primary outcome was clinical resolution (defined as no revisit with ongoing diarrhoea within 30 days from the date of first presentation). Informed from a directed acyclic graph, data on the following covariates were collected: age, breed, bodyweight, insurance status, comorbidities, vomiting, reduced appetite, haematochezia, pyrexia, duration, additional treatment prescription and veterinary group. Inverse probability of treatment weighting was used to balance covariates between the treatment groups for each of the two target trials. The risk difference (RD) of 0.4% (95% CI -4.5% to 5.3%) was non-significant for clinical resolution in dogs treated with antimicrobials compared with dogs not treated with antimicrobials. The risk difference (RD) of 0.3% (95% CI -4.5% to 5.0%) was non-significant for clinical resolution in dogs treated with gastrointestinal nutraceuticals compared with dogs not treated with gastrointestinal nutraceuticals. This study successfully applied the target trial framework to veterinary observational data. The findings show that antimicrobial or gastrointestinal prescription at first presentation of acute diarrhoea in dogs causes no difference in clinical resolution. The findings support the recommendation for veterinary professionals to limit antimicrobial use for acute diarrhoea in dogs.

Joint evaluation of placebo and treatment effects in cluster randomized trials by causal inference models

The term placebo effect refers to the psychobiological effect of a patient's knowledge or belief of being treated. A placebo effect is patient-driven, which makes it fundamentally different from the usual treatment effect resulting from external actions. In modern clinical research, the presence of a placebo effect is often treated as a nuisance issue, something to be "adjusted away" in estimating a treatment effect of primary interest. However, from a patient-centered perspective, we believe that a possible placebo produces substantial improvements in patient-centered outcomes. Understanding placebo effects is therefore an important part of patient-centered outcomes research. The available methods for estimating placebo effects are designed for individually randomized trials and are not directly applicable to cluster randomized trials (CRTs). There are several challenges in estimating placebo effects in CRTs. A major challenge is the possible presence of interference within clusters, in the sense that a subject's outcome may depend on the beliefs subjects in the same cluster about treatment assignment (mentality) and therefore possible correlation in outcome and mentality among subjects exists in the same cluster. In this article, we extend the previously developed causal inference framework to also encompass CRTs, using the G-Computation and inverse probability weighting (IPW) approaches. We also develop methodologies and further extend the G-Computation and IPW approaches to handle missingness for jointly evaluating placebo effect and treatment-specific effect, specifically in the context of CRTs. The proposed methods are demonstrated in simulation studies and a cluster randomized trial on effect of fermented dairy drink.

Statistical inference and power analysis for direct and spillover effects in two-stage randomized experiments

Two-stage randomized experiments become an increasingly popular experimental design for causal inference when the outcome of one unit may be affected by the treatment assignments of other units in the same cluster. In this paper, we provide a methodological framework for general tools of statistical inference and power analysis for two-stage randomized experiments. Under the randomization-based framework, we consider the estimation of a new direct effect of interest as well as the average direct and spillover effects studied in the literature. We provide unbiased estimators of these causal quantities and their conservative variance estimators in a general setting. Using these results, we then develop hypothesis testing procedures and derive sample size formulas. We theoretically compare the two-stage randomized design with the completely randomized and cluster randomized designs, which represent two limiting designs. Finally, we conduct simulation studies to evaluate the empirical performance of our sample size formulas. For empirical illustration, the proposed methodology is applied to the randomized evaluation of the Indian National Health Insurance Program. An open-source software package is available for implementing the proposed methodology.

Generalized propensity score approach to causal inference with spatial interference

Many spatial phenomena exhibit interference, where exposures at one location may affect the response at other locations. Because interference violates the stable unit treatment value assumption, standard methods for causal inference do not apply. We propose a new causal framework to recover direct and spill-over effects in the presence of spatial interference, taking into account that exposures at nearby locations are more influential than exposures at locations further apart. Under the no unmeasured confounding assumption, we show that a generalized propensity score is sufficient to remove all measured confounding. To reduce dimensionality issues, we propose a Bayesian spline-based regression model accounting for a sufficient set of variables for the generalized propensity score. A simulation study demonstrates the accuracy and coverage properties. We apply the method to estimate the causal effect of wildland fires on air pollution in the Western United States over 2005-2018.

Association Between Stimulant Treatment and Substance Use Through Adolescence Into Early Adulthood

Importance

Possible associations between stimulant treatment of attention-deficit/hyperactivity disorder (ADHD) and subsequent substance use remain debated and clinically relevant.

Objective

To assess the association of stimulant treatment of ADHD with subsequent substance use using the Multimodal Treatment Study of ADHD (MTA), which provides a unique opportunity to test this association while addressing methodologic complexities (principally, multiple dynamic confounding variables).

Design, Setting, and Participants

MTA was a multisite study initiated at 6 sites in the US and 1 in Canada as a 14-month randomized clinical trial of medication and behavior therapy for ADHD but transitioned to a longitudinal observational study. Participants were recruited between 1994 and 1996. Multi-informant assessments included comprehensively assessed demographic, clinical (including substance use), and treatment (including stimulant treatment) variables. Children aged 7 to 9 years with rigorously diagnosed DSM-IV combined-type ADHD were repeatedly assessed until a mean age of 25 years. Analysis took place between April 2018 and February 2023.

Exposure

Stimulant treatment of ADHD was measured prospectively from baseline for 16 years (10 assessments) initially using parent report followed by young adult report.

Main Outcomes and Measures

Frequency of heavy drinking, marijuana use, daily cigarette smoking, and other substance use were confidentially self-reported with a standardized substance use questionnaire.

Results

A total of 579 children (mean [SD] age at baseline, 8.5 [0.8] years; 465 [80%] male) were analyzed. Generalized multilevel linear models showed no evidence that current (B [SE] range, -0.62 [0.55] to 0.34 [0.47]) or prior stimulant treatment (B [SE] range, -0.06 [0.26] to 0.70 [0.37]) or their interaction (B [SE] range, -0.49 [0.70] to 0.86 [0.68]) were associated with substance use after adjusting for developmental trends in substance use and age. Marginal structural models adjusting for dynamic confounding by demographic, clinical, and familial factors revealed no evidence that more years of stimulant treatment (B [SE] range, -0.003 [0.01] to 0.04 [0.02]) or continuous, uninterrupted stimulant treatment (B [SE] range, -0.25 [0.33] to -0.03 [0.10]) were associated with adulthood substance use. Findings were the same for substance use disorder as outcome.

Conclusions and Relevance

This study found no evidence that stimulant treatment was associated with increased or decreased risk for later frequent use of alcohol, marijuana, cigarette smoking, or other substances used for adolescents and young adults with childhood ADHD. These findings do not appear to result from other factors that might drive treatment over time and findings held even after considering opposing age-related trends in stimulant treatment and substance use.

Estimating Causal Effects of HIV Prevention Interventions with Interference in Network-based Studies among People Who Inject Drugs

Evaluating causal effects in the presence of interference is challenging in network-based studies of hard-to-reach populations. Like many such populations, people who inject drugs (PWID) are embedded in social networks and often exert influence on others in their network. In our setting, the study design is observational with a non-randomized network-based HIV prevention intervention. Information is available on each participant and their connections that confer possible HIV risk through injection and sexual behaviors. We considered two inverse probability weighted (IPW) estimators to quantify the population-level spillover effects of non-randomized interventions on subsequent health outcomes. We demonstrated that these two IPW estimators are consistent, asymptotically normal, and derived a closed-form estimator for the asymptotic variance, while allowing for overlapping interference sets (groups of individuals in which the interference is assumed possible). A simulation study was conducted to evaluate the finite-sample performance of the estimators. We analyzed data from the Transmission Reduction Intervention Project, which ascertained a network of PWID and their contacts in Athens, Greece, from 2013 to 2015. We evaluated the effects of community alerts on subsequent HIV risk behavior in this observed network, where the connections or links between participants were defined by using substances or having unprotected sex together. In the study, community alerts were distributed to inform people of recent HIV infections among individuals in close proximity in the observed network. The estimates of the risk differences for spillover using either IPW estimator demonstrated a protective effect. The results suggest that HIV risk behavior could be mitigated by exposure to a community alert when an increased risk of HIV is detected in the network.

Impact of family networks on uptake of health interventions: evidence from a community-randomized control trial aimed at increasing HIV testing in South Africa

INTRODUCTION

While it is widely acknowledged that family relationships can influence health outcomes, their impact on the uptake of individual health interventions is unclear. In this study, we quantified how the efficacy of a randomized health intervention is shaped by its pattern of distribution in the family network.

METHODS

The "Home-Based Intervention to Test and Start" (HITS) was a 2×2 factorial community-randomized controlled trial in Umkhanyakude, KwaZulu-Natal, South Africa, embedded in the Africa Health Research Institute's population-based demographic and HIV surveillance platform (ClinicalTrials.gov # NCT03757104). The study investigated the impact of two interventions: a financial micro-incentive and a male-targeted HIV-specific decision support programme. The surveillance area was divided into 45 community clusters. Individuals aged ≥15 years in 16 randomly selected communities were offered a micro-incentive (R50 [$3] food voucher) for rapid HIV testing (intervention arm). Those living in the remaining 29 communities were offered testing only (control arm). Study data were collected between February and November 2018. Using routinely collected data on parents, conjugal partners, and co-residents, a socio-centric family network was constructed among HITS-eligible individuals. Nodes in this network represent individuals and ties represent family relationships. We estimated the effect of offering the incentive to people with and without family members who also received the offer on the uptake of HIV testing. We fitted a linear probability model with robust standard errors, accounting for clustering at the community level.

RESULTS

Overall, 15,675 people participated in the HITS trial. Among those with no family members who received the offer, the incentive's efficacy was a 6.5 percentage point increase (95% CI: 5.3-7.7). The efficacy was higher among those with at least one family member who received the offer (21.1 percentage point increase (95% CI: 19.9-22.3). The difference in efficacy was statistically significant (21.1-6.5 = 14.6%; 95% CI: 9.3-19.9).

CONCLUSIONS

Micro-incentives appear to have synergistic effects when distributed within family networks. These effects support family network-based approaches for the design of health interventions.

Examining the Causal Mediating Role of Cardiovascular Disease on the Effect of Subclinical Cardiovascular Disease on Cognitive Impairment via Separable Effects

BACKGROUND

An important epidemiological question is understanding how vascular risk factors contribute to cognitive impairment. Using data from the Cardiovascular Health Cognition Study, we investigated how subclinical cardiovascular disease (sCVD) relates to cognitive impairment risk and the extent to which the hypothesized risk is mediated by the incidence of clinically manifested cardiovascular disease (CVD), both overall and within apolipoprotein E-4 (APOE-4) subgroups.

METHODS

We adopted a novel "separable effects" causal mediation framework that assumes that sCVD has separably intervenable atherosclerosis-related components. We then ran several mediation models, adjusting for key covariates.

RESULTS

We found that sCVD increased overall risk of cognitive impairment (risk ratio [RR] = 1.21, 95% confidence interval [CI]: 1.03, 1.44); however, there was little or no mediation by incident clinically manifested CVD (indirect effect RR = 1.02, 95% CI: 1.00, 1.03). We also found attenuated effects among APOE-4 carriers (total effect RR = 1.09, 95% CI: 0.81, 1.47; indirect effect RR = 0.99, 95% CI: 0.96, 1.01) and stronger findings among noncarriers (total effect RR = 1.29, 95% CI: 1.05, 1.60; indirect effect RR = 1.02, 95% CI: 1.00, 1.05). In secondary analyses restricting cognitive impairment to only incident dementia cases, we found similar effect patterns.

CONCLUSIONS

We found that the effect of sCVD on cognitive impairment does not seem to be mediated by CVD, both overall and within APOE-4 subgroups. Our results were critically assessed via sensitivity analyses, and they were found to be robust. Future work is needed to fully understand the relationship between sCVD, CVD, and cognitive impairment.

Assessing intervention effects in a randomized trial within a social network

Studies of social networks provide unique opportunities to assess the causal effects of interventions that may impact more of the population than just those intervened on directly. Such effects are sometimes called peer or spillover effects, and may exist in the presence of interference, that is, when one individual's treatment affects another individual's outcome. Randomization-based inference (RI) methods provide a theoretical basis for causal inference in randomized studies, even in the presence of interference. In this article, we consider RI of the intervention effect in the eX-FLU trial, a randomized study designed to assess the effect of a social distancing intervention on influenza-like-illness transmission in a connected network of college students. The approach considered enables inference about the effect of the social distancing intervention on the per-contact probability of influenza-like-illness transmission in the observed network. The methods allow for interference between connected individuals and for heterogeneous treatment effects. The proposed methods are evaluated empirically via simulation studies, and then applied to data from the eX-FLU trial.

Double Negative Control Inference in Test-Negative Design Studies of Vaccine Effectiveness

The test-negative design (TND) has become a standard approach to evaluate vaccine effectiveness against the risk of acquiring infectious diseases in real-world settings, such as Influenza, Rotavirus, Dengue fever, and more recently COVID-19. In a TND study, individuals who experience symptoms and seek care are recruited and tested for the infectious disease which defines cases and controls. Despite TND's potential to reduce unobserved differences in healthcare seeking behavior (HSB) between vaccinated and unvaccinated subjects, it remains subject to various potential biases. First, residual confounding bias may remain due to unobserved HSB, occupation as healthcare worker, or previous infection history. Second, because selection into the TND sample is a common consequence of infection and HSB, collider stratification bias may exist when conditioning the analysis on testing, which further induces confounding by latent HSB. In this paper, we present a novel approach to identify and estimate vaccine effectiveness in the target population by carefully leveraging a pair of negative control exposure and outcome variables to account for potential hidden bias in TND studies. We illustrate our proposed method with extensive simulation and an application to study COVID-19 vaccine effectiveness using data from the University of Michigan Health System.

Robust weights that optimally balance confounders for estimating marginal hazard ratios

Covariate balance is crucial in obtaining unbiased estimates of treatment effects in observational studies. Methods that target covariate balance have been successfully proposed and largely applied to estimate treatment effects on continuous outcomes. However, in many medical and epidemiological applications, the interest lies in estimating treatment effects on time-to-event outcomes. With this type of data, one of the most common estimands of interest is the marginal hazard ratio of the Cox proportional hazards model. In this article, we start by presenting robust orthogonality weights, a set of weights obtained by solving a quadratic constrained optimization problem that maximizes precision while constraining covariate balance defined as the correlation between confounders and treatment. By doing so, robust orthogonality weights optimally deal with both binary and continuous treatments. We then evaluate the performance of the proposed weights in estimating marginal hazard ratios of binary and continuous treatments with time-to-event outcomes in a simulation study. We finally apply robust orthogonality weights in the evaluation of the effect of hormone therapy on time to coronary heart disease and on the effect of red meat consumption on time to colon cancer among 24,069 postmenopausal women enrolled in the Women's Health Initiative observational study.

Disentangling causality: assumptions in causal discovery and inference

Causality has been a burgeoning field of research leading to the point where the literature abounds with different components addressing distinct parts of causality. For researchers, it has been increasingly difficult to discern the assumptions they have to abide by in order to glean sound conclusions from causal concepts or methods. This paper aims to disambiguate the different causal concepts that have emerged in causal inference and causal discovery from observational data by attributing them to different levels of Pearl’s Causal Hierarchy. We will provide the reader with a comprehensive arrangement of assumptions necessary to engage in causal reasoning at the desired level of the hierarchy. Therefore, the assumptions underlying each of these causal concepts will be emphasized and their concomitant graphical components will be examined. We show which assumptions are necessary to bridge the gaps between causal discovery, causal identification and causal inference from a parametric and a non-parametric perspective. Finally, this paper points to further research areas related to the strong assumptions that researchers have glibly adopted to take part in causal discovery, causal identification and causal inference.

Methods for Assessing Spillover in Network-Based Studies of HIV/AIDS Prevention among People Who Use Drugs

Human Immunodeficiency Virus (HIV) interventions among people who use drugs (PWUD) often have spillover, also known as interference or dissemination, which occurs when one participant's exposure affects another participant's outcome. PWUD are often members of networks defined by social, sexual, and drug-use partnerships and their receipt of interventions can affect other members in their network. For example, HIV interventions with possible spillover include educational training about HIV risk reduction, pre-exposure prophylaxis, or treatment as prevention. In turn, intervention effects frequently depend on the network structure, and intervention coverage levels and spillover can occur even if not measured in a study, possibly resulting in an underestimation of intervention effects. Recent methodological approaches were developed to assess spillover in the context of network-based studies. This tutorial provides an overview of different study designs for network-based studies and related methodological approaches for assessing spillover in each design. We also provide an overview of other important methodological issues in network studies, including causal influence in networks and missing data. Finally, we highlight applications of different designs and methods from studies of PWUD and conclude with an illustrative example from the Transmission Reduction Intervention Project (TRIP) in Athens, Greece.

The mobility effects hypothesis: Methods and applications

We discuss hypotheses researchers have put forth to explain how outcomes of socially mobile and immobile individuals might differ and/or how mobility experiences are related to outcomes of interest. Next, we examine the methodological literature on this topic, culminating in the development of the diagonal mobility model (DMM, also called the diagonal reference model in some studies), the primary tool of use since the 1980's. We then discuss some of the many applications of the DMM. Although the model was proposed to examine the effects of social mobility on outcomes of interest, the estimated relationships between mobility and outcomes that researchers have called mobility effects are more appropriately regarded as partial associations. When mobility is not associated with outcomes, as is often found in empirical work, the outcomes of movers from origin o to destination d are a weighted average of the outcomes of individuals who remained in states o and d respectively, and the weights capture the relative salience of origins and destinations in the acculturation process. In light of this attractive feature of the model, we briefly develop several generalizations of the current DMM that future researchers should also find useful. Finally, we propose new estimands of mobility effects, based on the explicit notion that a unit effect of mobility is a comparison of an individual with herself under two conditions, one in which she is mobile, the other in which she is immobile, and we discuss some of the challenges in identifying such effects.

How causal machine learning can leverage marketing strategies: Assessing and improving the performance of a coupon campaign

We apply causal machine learning algorithms to assess the causal effect of a marketing intervention, namely a coupon campaign, on the sales of a retailer. Besides assessing the average impacts of different types of coupons, we also investigate the heterogeneity of causal effects across different subgroups of customers, e.g., between clients with relatively high vs. low prior purchases. Finally, we use optimal policy learning to determine (in a data-driven way) which customer groups should be targeted by the coupon campaign in order to maximize the marketing intervention's effectiveness in terms of sales. We find that only two out of the five coupon categories examined, namely coupons applicable to the product categories of drugstore items and other food, have a statistically significant positive effect on retailer sales. The assessment of group average treatment effects reveals substantial differences in the impact of coupon provision across customer groups, particularly across customer groups as defined by prior purchases at the store, with drugstore coupons being particularly effective among customers with high prior purchases and other food coupons among customers with low prior purchases. Our study provides a use case for the application of causal machine learning in business analytics to evaluate the causal impact of specific firm policies (like marketing campaigns) for decision support.

Propensity Score in the Face of Interference: Discussion of

Rosenbaum and Rubin's (1983) propensity score revolutionized the field of causal inference and has emerged as a standard tool when researchers reason about cause-and-effect relationship across many disciplines. This discussion centers around the key "no interference" assumption in Rosenbaum and Rubin's original development of the propensity score and reviews some recent advances in extending the propensity score to studies involving dependent happenings.

Network experiment designs for inferring causal effects under interference

Current approaches to A/B testing in networks focus on limiting interference, the concern that treatment effects can "spill over" from treatment nodes to control nodes and lead to biased causal effect estimation. In the presence of interference, two main types of causal effects are direct treatment effects and total treatment effects. In this paper, we propose two network experiment designs that increase the accuracy of direct and total effect estimations in network experiments through minimizing interference between treatment and control units. For direct treatment effect estimation, we present a framework that takes advantage of independent sets and assigns treatment and control only to a set of non-adjacent nodes in a graph, in order to disentangle peer effects from direct treatment effect estimation. For total treatment effect estimation, our framework combines weighted graph clustering and cluster matching approaches to jointly minimize interference and selection bias. Through a series of simulated experiments on synthetic and real-world network datasets, we show that our designs significantly increase the accuracy of direct and total treatment effect estimation in network experiments.

Randomization, design and analysis for interdependency in aging research: no person or mouse is an island

Investigators traditionally use randomized designs and corresponding analysis procedures to make causal inferences about the effects of interventions, assuming independence between an individual's outcome and treatment assignment and the outcomes of other individuals in the study. Often, such independence may not hold. We provide examples of interdependency in model organism studies and human trials and group effects in aging research and then discuss methodologic issues and solutions. We group methodologic issues as they pertain to (1) single-stage individually randomized trials; (2) cluster-randomized controlled trials; (3) pseudo-cluster-randomized trials; (4) individually randomized group treatment; and (5) two-stage randomized designs. Although we present possible strategies for design and analysis to improve the rigor, accuracy and reproducibility of the science, we also acknowledge real-world constraints. Consequences of nonadherence, differential attrition or missing data, unintended exposure to multiple treatments and other practical realities can be reduced with careful planning, proper study designs and best practices.

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.

A spatial causal analysis of wildland fire-contributed PM2.5 using numerical model output

Wildland fire smoke contains hazardous levels of fine particulate matter (PM_2.5), a pollutant shown to adversely effect health. Estimating fire attributable PM_2.5 concentrations is key to quantifying the impact on air quality and subsequent health burden. This is a challenging problem since only total PM_2.5 is measured at monitoring stations and both fire-attributable PM_2.5 and PM_2.5 from all other sources are correlated in space and time. We propose a framework for estimating fire-contributed PM_2.5 and PM_2.5 from all other sources using a novel causal inference framework and bias-adjusted chemical model representations of PM_2.5 under counterfactual scenarios. The chemical model representation of PM_2.5 for this analysis is simulated using Community Multiscale Air Quality Modeling System (CMAQ), run with and without fire emissions across the contiguous U.S. for the 2008-2012 wildfire seasons. The CMAQ output is calibrated with observations from monitoring sites for the same spatial domain and time period. We use a Bayesian model that accounts for spatial variation to estimate the effect of wildland fires on PM_2.5 and state assumptions under which the estimate has a valid causal interpretation. Our results include estimates of the contributions of wildfire smoke to PM_2.5 for the contiguous U.S. Additionally, we compute the health burden associated with the PM_2.5 attributable to wildfire smoke.

Estimating the total treatment effect in randomized experiments with unknown network structure

Randomized experiments are widely used to estimate the causal effects of a proposed treatment in many areas of science, from medicine and healthcare to the physical and biological sciences, from the social sciences to engineering, and from public policy to the technology industry. Here we consider situations where classical methods for estimating the total treatment effect on a target population are considerably biased due to confounding network effects, i.e., the fact that the treatment of an individual may impact its neighbors’ outcomes, an issue referred to as network interference or as nonindividualized treatment response. A key challenge in these situations is that the network is often unknown and difficult or costly to measure. We assume a potential outcomes model with heterogeneous additive network effects, encompassing a broad class of network interference sources, including spillover, peer effects, and contagion. First, we characterize the limitations in estimating the total treatment effect without knowledge of the network that drives interference. By contrast, we subsequently develop a simple estimator and efficient randomized design that outputs an unbiased estimate with low variance in situations where one is given access to average historical baseline measurements prior to the experiment. Our solution does not require knowledge of the underlying network structure, and it comes with statistical guarantees for a broad class of models. Due to their ease of interpretation and implementation, and their theoretical guarantees, we believe our results will have significant impact on the design of randomized experiments.

Accounting for Context in Randomized Trials after Assignment

Many preventive trials randomize individuals to intervention condition which is then delivered in a group setting. Other trials randomize higher levels, say organizations, and then use learning collaboratives comprised of multiple organizations to support improved implementation or sustainment. Other trials randomize or expand existing social networks and use key opinion leaders to deliver interventions through these networks. We use the term contextually driven to refer generally to such trials (traditionally referred to as clustering, where groups are formed either pre-randomization or post-randomization - i.e., a cluster-randomized trial), as these groupings or networks provide fixed or time-varying contexts that matter both theoretically and practically in the delivery of interventions. While such contextually driven trials can provide efficient and effective ways to deliver and evaluate prevention programs, they all require analytical procedures that take appropriate account of non-independence, something not always appreciated. Published analyses of many prevention trials have failed to take this into account. We discuss different types of contextually driven designs and then show that even small amounts of non-independence can inflate actual Type I error rates. This inflation leads to rejecting the null hypotheses too often, and erroneously leading us to conclude that there are significant differences between interventions when they do not exist. We describe a procedure to account for non-independence in the important case of a two-arm trial that randomizes units of individuals or organizations in both arms and then provides the active treatment in one arm through groups formed after assignment. We provide sample code in multiple programming languages to guide the analyst, distinguish diverse contextually driven designs, and summarize implications for multiple audiences.

Spillover benefit of pre-exposure prophylaxis for HIV prevention: evaluating the importance of effect modification using an agent-based model

We developed an agent-based model using a trial emulation approach to quantify effect measure modification of spillover effects of pre-exposure prophylaxis (PrEP) for HIV among men who have sex with men (MSM) in the Atlanta-Sandy Springs-Roswell metropolitan area, Georgia. PrEP may impact not only the individual prescribed, but also their partners and beyond, known as spillover. We simulated a two-stage randomised trial with eligible components (≥3 agents with ≥1 HIV+ agent) first randomised to intervention or control (no PrEP). Within intervention components, agents were randomised to PrEP with coverage of 70%, providing insight into a high PrEP coverage strategy. We evaluated effect modification by component-level characteristics and estimated spillover effects on HIV incidence using an extension of randomisation-based estimators. We observed an attenuation of the spillover effect when agents were in components with a higher prevalence of either drug use or bridging potential (if an agent acts as a mediator between ≥2 connected groups of agents). The estimated spillover effects were larger in magnitude among components with either higher HIV prevalence or greater density (number of existing partnerships compared to all possible partnerships). Consideration of effect modification is important when evaluating the spillover of PrEP among MSM.

Targeted maximum likelihood estimation of causal effects with interference: A simulation study

Interference, the dependency of an individual's potential outcome on the exposure of other individuals, is a common occurrence in medicine and public health. Recently, targeted maximum likelihood estimation (TMLE) has been extended to settings of interference, including in the context of estimation of the mean of an outcome under a specified distribution of exposure, referred to as a policy. This paper summarizes how TMLE for independent data is extended to general interference (network-TMLE). An extensive simulation study is presented of network-TMLE, consisting of four data generating mechanisms (unit-treatment effect only, spillover effects only, unit-treatment and spillover effects, infection transmission) in networks of varying structures. Simulations show that network-TMLE performs well across scenarios with interference, but issues manifest when policies are not well-supported by the observed data, potentially leading to poor confidence interval coverage. Guidance for practical application, freely available software, and areas of future work are provided.

Randomization for the susceptibility effect of an infectious disease intervention

Randomized trials of infectious disease interventions, such as vaccines, often focus on groups of connected or potentially interacting individuals. When the pathogen of interest is transmissible between study subjects, interference may occur: individual infection outcomes may depend on treatments received by others. Epidemiologists have defined the primary parameter of interest-called the "susceptibility effect"-as a contrast in infection risk under treatment versus no treatment, while holding exposure to infectiousness constant. A related quantity-the "direct effect"-is defined as an unconditional contrast between the infection risk under treatment versus no treatment. The purpose of this paper is to show that under a widely recommended randomization design, the direct effect may fail to recover the sign of the true susceptibility effect of the intervention in a randomized trial when outcomes are contagious. The analytical approach uses structural features of infectious disease transmission to define the susceptibility effect. A new probabilistic coupling argument reveals stochastic dominance relations between potential infection outcomes under different treatment allocations. The results suggest that estimating the direct effect under randomization may provide misleading conclusions about the effect of an intervention-such as a vaccine-when outcomes are contagious. Investigators who estimate the direct effect may wrongly conclude an intervention that protects treated individuals from infection is harmful, or that a harmful treatment is beneficial.

Toward a Clearer Definition of Selection Bias When Estimating Causal Effects

Selection bias remains a subject of controversy. Existing definitions of selection bias are ambiguous. To improve communication and the conduct of epidemiologic research focused on estimating causal effects, we propose to unify the various existing definitions of selection bias in the literature by considering any bias away from the true causal effect in the referent population (the population before the selection process), due to selecting the sample from the referent population, as selection bias. Given this unified definition, selection bias can be further categorized into two broad types: type 1 selection bias owing to restricting to one or more level(s) of a collider (or a descendant of a collider) and type 2 selection bias owing to restricting to one or more level(s) of an effect measure modifier. To aid in explaining these two types-which can co-occur-we start by reviewing the concepts of the target population, the study sample, and the analytic sample. Then, we illustrate both types of selection bias using causal diagrams. In addition, we explore the differences between these two types of selection bias, and describe methods to minimize selection bias. Finally, we use an example of "M-bias" to demonstrate the advantage of classifying selection bias into these two types.

Toward evaluation of disseminated effects of medications for opioid use disorder within provider-based clusters using routinely-collected health data

Routinely-collected health data can be employed to emulate a target trial when randomized trial data are not available. Patients within provider-based clusters likely exert and share influence on each other's treatment preferences and subsequent health outcomes and this is known as dissemination or spillover. Extending a framework to replicate an idealized two-stage randomized trial using routinely-collected health data, an evaluation of disseminated effects within provider-based clusters is possible. In this article, we propose a novel application of causal inference methods for dissemination to retrospective cohort studies in administrative claims data and evaluate the impact of the normality of the random effects distribution for the cluster-level propensity score on estimation of the causal parameters. An extensive simulation study was conducted to study the robustness of the methods under different distributions of the random effects. We applied these methods to evaluate baseline prescription for medications for opioid use disorder among a cohort of patients diagnosed with opioid use disorder and adjust for baseline confounders using information obtained from an administrative claims database. We discuss future research directions in this setting to better address unmeasured confounding in the presence of disseminated effects.