Assessing the causal effects of a stochastic intervention in time series data: are heat alerts effective in preventing deaths and hospitalizations?

The methodological development of this article is motivated by the need to address the following scientific question: does the issuance of heat alerts prevent adverse health effects? Our goal is to address this question within a causal inference framework in the context of time series data. A key challenge is that causal inference methods require the overlap assumption to hold: each unit (i.e., a day) must have a positive probability of receiving the treatment (i.e., issuing a heat alert on that day). In our motivating example, the overlap assumption is often violated: the probability of issuing a heat alert on a cooler day is near zero. To overcome this challenge, we propose a stochastic intervention for time series data which is implemented via an incremental time-varying propensity score (ItvPS). The ItvPS intervention is executed by multiplying the probability of issuing a heat alert on day $t$-conditional on past information up to day $t$-by an odds ratio $\delta_t$. First, we introduce a new class of causal estimands, which relies on the ItvPS intervention. We provide theoretical results to show that these causal estimands can be identified and estimated under a weaker version of the overlap assumption. Second, we propose nonparametric estimators based on the ItvPS and derive an upper bound for the variances of these estimators. Third, we extend this framework to multisite time series using a spatial meta-analysis approach. Fourth, we show that the proposed estimators perform well in terms of bias and root mean squared error via simulations. Finally, we apply our proposed approach to estimate the causal effects of increasing the probability of issuing heat alerts on each warm-season day in reducing deaths and hospitalizations among Medicare enrollees in 2837 US counties.

Parametric g-formula for Testing Time-Varying Causal Effects: What It Is, Why It Matters, and How to Implement It in Lavaan.

Psychologists leverage longitudinal designs to examine the causal effects of a focal predictor (i.e., treatment or exposure) over time. But causal inference of naturally observed time-varying treatments is complicated by treatment-dependent confounding in which earlier treatments affect confounders of later treatments. In this tutorial article, we introduce psychologists to an established solution to this problem from the causal inference literature: the parametric g-computation formula. We explain why the g-formula is effective at handling treatment-dependent confounding. We demonstrate that the parametric g-formula is conceptually intuitive, easy to implement, and well-suited for psychological research. We first clarify that the parametric g-formula essentially utilizes a series of statistical models to estimate the joint distribution of all post-treatment variables. These statistical models can be readily specified as standard multiple linear regression functions. We leverage this insight to implement the parametric g-formula using lavaan, a widely adopted R package for structural equation modeling. Moreover, we describe how the parametric g-formula may be used to estimate a marginal structural model whose causal parameters parsimoniously encode time-varying treatment effects. We hope this accessible introduction to the parametric g-formula will equip psychologists with an analytic tool to address their causal inquiries using longitudinal data.

Current Challenges With the Use of Test-Negative Designs for Modeling COVID-19 Vaccination and Outcomes.

The widespread testing for severe acute respiratory syndrome coronavirus 2 infection has facilitated the use of test-negative designs (TNDs) for modeling coronavirus disease 2019 (COVID-19) vaccination and outcomes. Despite the comprehensive literature on TND, the use of TND in COVID-19 studies is relatively new and calls for robust design and analysis to adapt to a rapidly changing and dynamically evolving pandemic and to account for changes in testing and reporting practices. In this commentary, we aim to draw the attention of researchers to COVID-specific challenges in using TND as we are analyzing data amassed over more than two years of the pandemic. We first review when and why TND works and general challenges in TND studies presented in the literature. We then discuss COVID-specific challenges which have not received adequate acknowledgment but may add to the risk of invalid conclusions in TND studies of COVID-19.

Addressing missing data in the estimation of time-varying treatments in comparative effectiveness research.

Comparative effectiveness research is often concerned with evaluating treatment strategies sustained over time, that is, time-varying treatments. Inverse probability weighting (IPW) is often used to address the time-varying confounding by re-weighting the sample according to the probability of treatment receipt at each time point. IPW can also be used to address any missing data by re-weighting individuals according to the probability of observing the data. The combination of these two distinct sets of weights may lead to inefficient estimates of treatment effects due to potentially highly variable total weights. Alternatively, multiple imputation (MI) can be used to address the missing data by replacing each missing observation with a set of plausible values drawn from the posterior predictive distribution of the missing data given the observed data. Recent studies have compared IPW and MI for addressing the missing data in the evaluation of time-varying treatments, but they focused on missing confounders and monotone missing data patterns. This article assesses the relative advantages of MI and IPW to address missing data in both outcomes and confounders measured over time, and across monotone and non-monotone missing data settings. Through a comprehensive simulation study, we find that MI consistently provided low bias and more precise estimates compared to IPW across a wide range of scenarios. We illustrate the implications of method choice in an evaluation of biologic drugs for patients with severe rheumatoid arthritis, using the US National Databank for Rheumatic Diseases, in which 25% of participants had missing health outcomes or time-varying confounders.

Prospectively assessed perceived stress associated with early pregnancy losses among women with history of pregnancy loss.

STUDY QUESTION: What is the association between perceived stress during peri-conception and early pregnancy and pregnancy loss among women who have experienced a prior pregnancy loss? SUMMARY ANSWER: Daily perceived stress above the median is associated with over a 2-fold risk of early pregnancy loss among women who have experienced a prior loss. WHAT IS KNOWN ALREADY?: Women who have experienced a pregnancy loss may be more vulnerable to stress while trying to become pregnant again. While prior research has indicated a link between psychological stress and clinically confirmed miscarriages, research is lacking among a pre-conceptional cohort followed prospectively for the effects of perceived stress during early critical windows of pregnancy establishment on risk of both hCG-detected pregnancy losses and confirmed losses, while considering important time-varying confounders. STUDY DESIGN, SIZE, DURATION: Secondary data analysis of the EAGeR trial (2007-2011) among women with an hCG-detected pregnancy (n = 797 women). PARTICIPANTS/MATERIALS, SETTING, METHODS: Women from four US clinical centers enrolled pre-conceptionally and were followed ≤6 cycles while attempting pregnancy and, as applicable, throughout pregnancy. Perceived stress was captured via daily diaries and end-of-month questionnaires. Main outcome measures include hCG-detected and clinically recognized pregnancy losses. MAIN RESULTS AND THE ROLE OF CHANCE: Among women who had an hCG-confirmed pregnancy, 188 pregnancies (23.6%) ended in loss. Women with high (>50th percentile) versus low (≤50th percentile) peri-implantation or early pregnancy weekly perceived stress had an elevated risk of experiencing any pregnancy loss (hazard ratio (HR): 1.69, 95% CI: 1.13, 2.54) or clinical loss (HR: 1.58, 95% CI: 0.96, 2.60), with higher risks observed for women experiencing an hCG-detected loss (HR: 2.16, 95% CI: 1.04, 4.46). Models accounted for women's age, BMI, employment, marital status, income, education, race, parity, prior losses, exercise and time-varying nausea/vomiting, caffeine, alcohol and smoking. LIMITATIONS, REASONS FOR CAUTION: We were limited in our ability to clearly identify the mechanisms of stress on pregnancy loss due to our sole reliance on self-reported perceived stress, and the lack of biomarkers of different pathways of stress. WIDER IMPLICATIONS OF THE FINDINGS: This study provides new insight on early pregnancy perceived stress and risk of pregnancy loss, most notably hCG-detected losses, among women with a history of a prior loss. Our study is an improvement over past studies in its ability to account for time-varying early pregnancy symptoms, such as nausea/vomiting, and lifestyle factors, such as caffeine, alcohol and smoking, which are also risk factors for psychological stress and pregnancy loss. STUDY FUNDING/COMPETING INTEREST(S): This work was supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, Maryland (Contract numbers: HHSN267200603423, HHSN267200603424, HHSN267200603426, HHSN275201300023I). Additionally, K.C.S. was supported by the National Institute on Aging of the National Institutes of Health under Award Number K01AG058781. The authors have no conflicts of interest to disclose. TRIAL REGISTRATION NUMBER: #NCT00467363.

Bias of time-varying exposure effects due to time-varying covariate measurement strategies.

PURPOSE: In studies of effects of time-varying drug exposures, adequate adjustment for time-varying covariates is often necessary to properly control for confounding. However, the granularity of the available covariate data may not be sufficiently fine, for example when covariates are measured for participants only when their exposure levels change. METHODS: To illustrate the impact of choices regarding the frequency of measuring time-varying covariates, we simulated data for a large target trial and for large observational studies, varying in covariate measurement design. Covariates were measured never, on a fixed-interval basis, or each time the exposure level switched. For the analysis, it was assumed that covariates remain constant in periods of no measurement. Cumulative survival probabilities for continuous exposure and non-exposure were estimated using inverse probability weighting to adjust for time-varying confounding, with special emphasis on the difference between 5-year event risks. RESULTS: With monthly covariate measurements, estimates based on observational data coincided with trial-based estimates, with 5-year risk differences being zero. Without measurement of baseline or post-baseline covariates, this risk difference was estimated to be 49% based on the available observational data. With measurements on a fixed-interval basis only, 5-year risk differences deviated from the null, to 29% for 6-monthly measurements, and with magnitude increasing up to 35% as the interval length increased. Risk difference estimates diverged from the null to as low as -18% when covariates were measured depending on exposure level switching. CONCLUSION: Our simulations highlight the need for careful consideration of time-varying covariates in designing studies on time-varying exposures. We caution against implementing designs with long intervals between measurements. The maximum length required will depend on the rates at which treatments and covariates change, with higher rates requiring shorter measurement intervals.

Longitudinal Causal Effects of Normalized Protein Catabolic Rate on All-Cause Mortality in Patients With End-Stage Renal Disease: Adjusting for Time-Varying Confounders Using the G-Estimation Method.

In this study, we aimed to estimate the causal effect of normalized protein catabolic rate (nPCR) on mortality among end-stage renal disease (ESRD) patients in the presence of time-varying confounding affected by prior exposure using g-estimation. Information about 553 ESRD patients was retrospectively collected over an 8-year period (2011-2019) from hemodialysis facilities in Kerman, Iran. nPCR was dichotomized as <1.2 g/kg/day versus ≥1.2 g/kg/day. Then a standard time-varying accelerated failure time (AFT) Weibull model was built, and results were compared with those generated by g-estimation. After appropriate adjustment for time-varying confounders, weighted g-estimation yielded 78% shorter survival time (95% confidence interval (95% CI): -81, -73) among patients with a continuous nPCR <1.2 g/kg/day than among those who had nPCR ≥1.2 g/kg/day during follow-up, though it was 18% (95% CI: -57, 54) in the Weibull model. Moreover, hazard ratio estimates of 4.56 (95% CI: 3.69, 5.37) and 1.20 (95% CI: 0.66, 2.17) were obtained via weighted g-estimation and the Weibull model, respectively. G-estimation indicated that inadequate dietary protein intake characterized by nPCR increases all-cause mortality among ESRD patients, but the Weibull model provided an effect estimate that was substantially biased toward the null.

Common Methods for Handling Missing Data in Marginal Structural Models: What Works and Why.

Marginal structural models (MSMs) are commonly used to estimate causal intervention effects in longitudinal nonrandomized studies. A common challenge when using MSMs to analyze observational studies is incomplete confounder data, where a poorly informed analysis method will lead to biased estimates of intervention effects. Despite a number of approaches described in the literature for handling missing data in MSMs, there is little guidance on what works in practice and why. We reviewed existing missing-data methods for MSMs and discussed the plausibility of their underlying assumptions. We also performed realistic simulations to quantify the bias of 5 methods used in practice: complete-case analysis, last observation carried forward, the missingness pattern approach, multiple imputation, and inverse-probability-of-missingness weighting. We considered 3 mechanisms for nonmonotone missing data encountered in research based on electronic health record data. Further illustration of the strengths and limitations of these analysis methods is provided through an application using a cohort of persons with sleep apnea: the research database of the French Observatoire Sommeil de la Fédération de Pneumologie. We recommend careful consideration of 1) the reasons for missingness, 2) whether missingness modifies the existing relationships among observed data, and 3) the scientific context and data source, to inform the choice of the appropriate method(s) for handling partially observed confounders in MSMs.

Using generalized linear models to implement g-estimation for survival data with time-varying confounding.

Using data from observational studies to estimate the causal effect of a time-varying exposure, repeatedly measured over time, on an outcome of interest requires careful adjustment for confounding. Standard regression adjustment for observed time-varying confounders is unsuitable, as it can eliminate part of the causal effect and induce bias. Inverse probability weighting, g-computation, and g-estimation have been proposed as being more suitable methods. G-estimation has some advantages over the other two methods, but until recently there has been a lack of flexible g-estimation methods for a survival time outcome. The recently proposed Structural Nested Cumulative Survival Time Model (SNCSTM) is such a method. Efficient estimation of the parameters of this model required bespoke software. In this article we show how the SNCSTM can be fitted efficiently via g-estimation using standard software for fitting generalised linear models. The ability to implement g-estimation for a survival outcome using standard statistical software greatly increases the potential uptake of this method. We illustrate the use of this method of fitting the SNCSTM by reanalyzing data from the UK Cystic Fibrosis Registry, and provide example R code to facilitate the use of this approach by other researchers.

Short interpregnancy interval and pregnancy outcomes: How important is the timing of confounding variable ascertainment?

BACKGROUND: Estimation of causal effects of short interpregnancy interval on pregnancy outcomes may be confounded by time-varying factors. These confounders should be ascertained at or before delivery of the first ("index") pregnancy, but are often only measured at the subsequent pregnancy. OBJECTIVES: To quantify bias induced by adjusting for time-varying confounders ascertained at the subsequent (rather than the index) pregnancy in estimated effects of short interpregnancy interval on pregnancy outcomes. METHODS: We analysed linked records for births in British Columbia, Canada, 2004-2014, to women with ≥2 singleton pregnancies (n = 121 151). We used log binomial regression to compare short (<6, 6-11, 12-17 months) to 18-23-month reference intervals for 5 outcomes: perinatal mortality (stillbirth and neonatal death); small for gestational age (SGA) birth and preterm delivery (all, early, spontaneous). We calculated per cent differences between adjusted risk ratios (aRR) from two models with maternal age, low socio-economic status, body mass index, and smoking ascertained in the index pregnancy and the subsequent pregnancy. We considered relative per cent differences <5% minimal, 5%-9% modest, and ≥10% substantial. RESULTS: Adjustment for confounders measured at the subsequent pregnancy introduced modest bias towards the null for perinatal mortality aRRs for <6-month interpregnancy intervals [-9.7%, 95% confidence interval [CI] -15.3, -6.2). SGA aRRs were minimally biased towards the null (-1.1%, 95% CI -2.6, 0.8) for <6-month intervals. While early preterm delivery aRRs were substantially biased towards the null (-10.4%, 95% CI -14.0, -6.6) for <6-month interpregnancy intervals, bias was minimal for <6-month intervals for all preterm deliveries (-0.6%, 95% CI -2.0, 0.8) and spontaneous preterm deliveries (-1.3%, 95% CI -3.1, 0.1). For all outcomes, bias was attenuated and minimal for 6-11-month and 12-17-month interpregnancy intervals. CONCLUSION: These findings suggest that maternally linked pregnancy data may not be needed for appropriate confounder adjustment when studying the effects of short interpregnancy interval on pregnancy outcomes.

Fitting Marginal Structural and G-Estimation Models Under Complex Treatment Patterns: Investigating the Association Between De Novo Vitamin D Supplement Use After Breast Cancer Diagnosis and All-Cause Mortality Using Linked Pharmacy Claim and Registry Data.

Studies have shown that accounting for time-varying confounding through time-dependent Cox proportional hazards models may provide biased estimates of the causal effect of treatment when the confounder is also a mediator. We explore 2 alternative approaches to addressing this problem while examining the association between vitamin D supplementation initiated after breast cancer diagnosis and all-cause mortality. Women aged 50-80 years were identified in the National Cancer Registry Ireland (n = 5,417) between 2001 and 2011. Vitamin D use was identified from linked prescription data (n = 2,570). We sought to account for the time-varying nature of vitamin D use and time-varying confounding by bisphosphonate use using 1) marginal structural models (MSMs) and 2) G-estimation of structural nested accelerated failure-time models (SNAFTMs). Using standard adjusted Cox proportional hazards models, we found a reduction in all-cause mortality in de novo vitamin D users compared with nonusers (hazard ratio (HR) = 0.84, 95% confidence interval (CI): 0.73, 0.99). Additional adjustment for vitamin D and bisphosphonate use in the previous month reduced the hazard ratio (HR = 0.45, 95% CI: 0.33, 0.63). Results derived from MSMs (HR = 0.44, 95% CI: 0.32, 0.61) and SNAFTMs (HR = 0.45, 95% CI: 0.34, 0.52) were similar. Utilizing MSMs and SNAFTMs to account for time-varying bisphosphonate use did not alter conclusions in this example.

Sustained adherence to a Mediterranean diet and physical activity on all-cause mortality in the Melbourne Collaborative Cohort Study: application of the g-formula.

BACKGROUND: Adherence to a traditional Mediterranean diet has been associated with lower mortality and cardiovascular disease risk. The relative importance of diet compared to other lifestyle factors and effects of dietary patterns over time remains unknown. METHODS: We used the parametric G-formula to account for time-dependent confounding, in order to assess the relative importance of diet compared to other lifestyle factors and effects of dietary patterns over time. We included healthy Melbourne Collaborative Cohort Study participants attending a visit during 1995-1999. Questionnaires assessed diet and physical activity at each of three study waves. Deaths were identified by linkage to national registries. We estimated mortality risk over approximately 14 years (1995-2011). RESULTS: Of 22,213 participants, 2163 (9.7%) died during 13.6 years median follow-up. Sustained high physical activity and adherence to a Mediterranean-style diet resulted in an estimated reduction in all-cause mortality of 1.82 per 100 people (95% confidence interval (CI): 0.03, 3.6). The population attributable fraction was 13% (95% CI: 4, 23%) for sustained high physical activity, 7% (95% CI: - 3, 17%) for sustained adherence to a Mediterranean-style diet and 18% (95% CI: 0, 36%) for their combination. CONCLUSIONS: A small reduction in mortality may be achieved by sustained elevated physical activity levels in healthy middle-aged adults, but there may be comparatively little gain from increasing adherence to a Mediterranean-style diet.

Modeling survival distribution as a function of time to treatment discontinuation: A dynamic treatment regime approach.

We consider estimating the effect that discontinuing a beneficial treatment will have on the distribution of a time to event clinical outcome, and in particular assessing whether there is a period of time over which the beneficial effect may continue after discontinuation. There are two major challenges. The first is to make a distinction between mandatory discontinuation, where by necessity treatment has to be terminated and optional discontinuation which is decided by the preference of the patient or physician. The innovation in this article is to cast the intervention in the form of a dynamic regime "terminate treatment optionally at time v unless a mandatory treatment-terminating event occurs prior to v" and consider estimating the distribution of time to event as a function of treatment regime v. The second challenge arises from biases associated with the nonrandom assignment of treatment regimes, because, naturally, optional treatment discontinuation is left to the patient and physician, and so time to discontinuation may depend on the patient's disease status. To address this issue, we develop dynamic-regime Marginal Structural Models and use inverse probability of treatment weighting to estimate the impact of time to treatment discontinuation on a time to event outcome, compared to the effect of not discontinuing treatment. We illustrate our methods using the IMPROVE-IT data on cardiovascular disease.

A comparison of estimators from self-controlled case series, case-crossover design, and sequence symmetry analysis for pharmacoepidemiological studies.

BACKGROUND: Despite the frequent use of self-controlled methods in pharmacoepidemiological studies, the factors that may bias the estimates from these methods have not been adequately compared in real-world settings. Here, we comparatively examined the impact of a time-varying confounder and its interactions with time-invariant confounders, time trends in exposures and events, restrictions, and misspecification of risk period durations on the estimators from three self-controlled methods. This study analyzed self-controlled case series (SCCS), case-crossover (CCO) design, and sequence symmetry analysis (SSA) using simulated and actual electronic medical records datasets. METHODS: We evaluated the performance of the three self-controlled methods in simulated cohorts for the following scenarios: 1) time-invariant confounding with interactions between the confounders, 2) time-invariant and time-varying confounding without interactions, 3) time-invariant and time-varying confounding with interactions among the confounders, 4) time trends in exposures and events, 5) restricted follow-up time based on event occurrence, and 6) patient restriction based on event history. The sensitivity of the estimators to misspecified risk period durations was also evaluated. As a case study, we applied these methods to evaluate the risk of macrolides on liver injury using electronic medical records. RESULTS: In the simulation analysis, time-varying confounding produced bias in the SCCS and CCO design estimates, which aggravated in the presence of interactions between the time-invariant and time-varying confounders. The SCCS estimates were biased by time trends in both exposures and events. Erroneously short risk periods introduced bias to the CCO design estimate, whereas erroneously long risk periods introduced bias to the estimates of all three methods. Restricting the follow-up time led to severe bias in the SSA estimates. The SCCS estimates were sensitive to patient restriction. The case study showed that although macrolide use was significantly associated with increased liver injury occurrence in all methods, the value of the estimates varied. CONCLUSIONS: The estimations of the three self-controlled methods depended on various underlying assumptions, and the violation of these assumptions may cause non-negligible bias in the resulting estimates. Pharmacoepidemiologists should select the appropriate self-controlled method based on how well the relevant key assumptions are satisfied with respect to the available data.

Statin use and risk of multiple myeloma: An analysis from the cancer research network.

Animal and human data suggest statins may be protective against developing multiple myeloma; however, findings may be biased by the interrelationship with lipid levels. We investigated the association between statin use and risk of multiple myeloma in a large US population, with an emphasis on accounting for this potential bias. We conducted a case-control study nested within 6 US integrated healthcare systems participating in the National Cancer Institute-funded Cancer Research Network. Adults aged ≥40 years who were diagnosed with multiple myeloma from 1998-2008 were identified through cancer registries (N = 2,532). For each case, five controls were matched on age, sex, health plan, and membership duration prior to diagnosis/index date. Statin prescriptions were ascertained from electronic pharmacy records. To address potential biases related to lipid levels and medication prescribing practices, multivariable marginal structural models were used to model statin use (≥6 cumulative months) and risk of multiple myeloma, with examination of multiple latency periods. Statin use 48-72 months prior to diagnosis/index date was associated with a suggestive 20-28% reduced risk of developing multiple myeloma, compared to non-users. Recent initiation of statins was not associated with myeloma risk (risk ratio range 0.90-0.99 with 0-36 months latency). Older patients had more consistent protective associations across all latency periods (risk ratio range 0.67-0.87). Our results suggest that the association between statin use and multiple myeloma risk may vary by exposure window and age. Future research is warranted to investigate the timing of statin use in relation to myeloma diagnosis.

Confounding in longitudinal studies in addiction treatment research.

Background: The effectiveness of treatment for people with substance use disorders is usually examined using longitudinal cohorts. In these studies, treatment is often considered as a time-varying exposure. The aim of this commentary is to examine confounding in this context, when the confounding variable is time-invariant and when it is time-varying. Method: Types of confounding are described with examples and illustrated using path diagrams. Simulations are used to demonstrate the direction of confounding bias and the extent that it is accounted for using standard regression adjustment techniques. Results: When the confounding variable is time invariant or time varying and not influenced by prior treatment, then standard adjustment techniques are adequate to control for confounding bias, provided that in the latter scenario the time-varying form of the variable is used. When the confounder is time varying and affected by prior treatment status (i.e. it is a mediator of treatment), then standard methods of adjustment result in inconsistency. Conclusions: In longitudinal cohorts where treatment exposure is time varying, confounding is an issue which should be considered, even if treatment exposure is initially randomized. In these studies, standard methods of adjustment may result be inadequate, even when all confounders have been identified. This occurs when the confounder is also a mediator of treatment. This is a likely scenario in many studies in addiction.

Longitudinal Study of the Effects of Bacteremia and Sepsis on 5-year Risk of Cardiovascular Events.

BACKGROUND: The long-term and cumulative effect of multiple episodes of bacteremia and sepsis across multiple hospitalizations on the development of cardiovascular (CV) events is uncertain. METHODS: We conducted a longitudinal study of 156 380 hospitalizations in 47 009 patients (≥18 years old) who had at least 2 inpatient admissions at an academic tertiary care center in St Louis, Missouri, from 1 January 2008 through 31 December 2012. We used marginal structural models, estimated by inverse probability weighting (IPW) of bacteremia or sepsis and IPW of censoring, to estimate the marginal causal effects of bacteremia and sepsis on developing the first observed incident CV event, including stroke, transient ischemic attack, and myocardial infarction (MI), during the study period. RESULTS: Bacteremia and sepsis occurred during 4923 (3.1%) and 5544 (3.5%) hospitalizations among 3932 (8.4%) and 4474 (9.5%) patients, respectively. CV events occurred in 414 (10.5%) and 538 (12.0%) patients with prior episodes of bacteremia or sepsis, respectively, vs 3087 (7.2%) and 2963 (7.0%) patients without prior episodes of bacteremia or sepsis. The causal odds of experiencing a CV event was 1.52-fold (95% confidence interval [CI], 1.21- to 1.90-fold) and 2.39-fold (95% CI, 1.88- to 3.03-fold) higher in patients with prior instances of bacteremia or sepsis, respectively, compared to those without. Prior instances of septic shock resulted in a 6.91-fold (95% CI, 5.34- to 8.93-fold) increase in the odds of MI. CONCLUSIONS: Prior instances of bacteremia and sepsis substantially increase the 5-year risk of CV events.

Antidepressant treatment and suicide attempts and self-inflicted injury in children and adolescents.

PURPOSE: In the 2004, FDA placed a black box warning on antidepressants for risk of suicidal thoughts and behavior in children and adolescents. The purpose of this paper is to examine the risk of suicide attempt and self-inflicted injury in depressed children ages 5-17 treated with antidepressants in two large observational datasets taking account time-varying confounding. METHODS: We analyzed two large US medical claims databases (MarketScan and LifeLink) containing 221,028 youth (ages 5-17) with new episodes of depression, with and without antidepressant treatment during the period of 2004-2009. Subjects were followed for up to 180 days. Marginal structural models were used to adjust for time-dependent confounding. RESULTS: For both datasets, significantly increased risk of suicide attempts and self-inflicted injury were seen during antidepressant treatment episodes in the unadjusted and simple covariate adjusted analyses. Marginal structural models revealed that the majority of the association is produced by dynamic confounding in the treatment selection process; estimated odds ratios were close to 1.0 consistent with the unadjusted and simple covariate adjusted association being a product of chance alone. CONCLUSIONS: Our analysis suggests antidepressant treatment selection is a product of both static and dynamic patient characteristics. Lack of adjustment for treatment selection based on dynamic patient characteristics can lead to the appearance of an association between antidepressant treatment and suicide attempts and self-inflicted injury among youths in unadjusted and simple covariate adjusted analyses. Marginal structural models can be used to adjust for static and dynamic treatment selection processes such as that likely encountered in observational studies of associations between antidepressant treatment selection, suicide and related behaviors in youth.

The relationship between lifetime health trajectories and socioeconomic attainment in middle age.

A large literature demonstrates the direct and indirect influence of health on socioeconomic attainment, and reveals the ways in which health and socioeconomic background simultaneously and dynamically affect opportunities for attainment and mobility. Despite an increasing understanding of the effects of health on social processes, research to date remains limited in its conceptualization and measurement of the temporal dimensions of health, especially in the presence of socioeconomic circumstances that covary with health over time. Guided by life course theory, we use data from the British National Child Development Study, an ongoing panel study of a cohort born in 1958, to examine the association between lifetime health trajectories and socioeconomic attainment in middle age. We apply finite mixture modeling to identify distinct trajectories of health that simultaneously account for timing, duration and stability. Moreover, we employ propensity score weighting models to account for the presence of time-varying socioeconomic factors in estimating the impact of health trajectories. We find that, when poor health is limited to the childhood years, the disadvantage in socioeconomic attainment relative to being continuously healthy is either insignificant or largely explained by time-varying socioeconomic confounders. The socioeconomic impact of continuously deteriorating health over the life course is more persistent, however. Our results suggest that accounting for the timing, duration and stability of poor health throughout both childhood and adulthood is important for understanding how health works to produce social stratification. In addition, the findings highlight the importance of distinguishing between confounding and mediating effects of time-varying socioeconomic circumstances.

Optimal restricted estimation for more efficient longitudinal causal inference.

Efficient semiparametric estimation of longitudinal causal effects is often analytically or computationally intractable. We propose a novel restricted estimation approach for increasing efficiency, which can be used with other techniques, is straightforward to implement, and requires no additional modeling assumptions.