Variable selection for causal mediation analysis using LASSO-based methods

Causal mediation effect estimates can be obtained from marginal structural models using inverse probability weighting with appropriate weights. In order to compute weights, treatment and mediator propensity score models need to be fitted first. If the covariates are high-dimensional, parsimonious propensity score models can be developed by regularization methods including LASSO and its variants. Furthermore, in a mediation setup, more efficient direct or indirect effect estimators can be obtained by using outcome-adaptive LASSO to select variables for propensity score models by incorporating the outcome information. A simulation study is conducted to assess how different regularization methods can affect the performance of estimated natural direct and indirect effect odds ratios. Our simulation results show that regularizing propensity score models by outcome-adaptive LASSO can improve the efficiency of the natural effect estimators and by optimizing balance in the covariates, bias can be reduced in most cases. The regularization methods are then applied to MIMIC-III database, an ICU database developed by MIT.

Download Full-text

Causal mediation analysis with multiple causally non-ordered mediators

Statistical Methods in Medical Research ◽

10.1177/0962280215615899 ◽

2015 ◽

Vol 27 (1) ◽

pp. 3-19 ◽

Cited By ~ 15

Author(s):

Masataka Taguri ◽

John Featherstone ◽

Jing Cheng

Keyword(s):

Indirect Effect ◽

Interactive Effect ◽

Mediation Effect ◽

Mediation Effects ◽

Mediation Analyses ◽

Intermediate Variable ◽

Causal Mediation ◽

Health Studies ◽

Using Data ◽

The Individual

In many health studies, researchers are interested in estimating the treatment effects on the outcome around and through an intermediate variable. Such causal mediation analyses aim to understand the mechanisms that explain the treatment effect. Although multiple mediators are often involved in real studies, most of the literature considered mediation analyses with one mediator at a time. In this article, we consider mediation analyses when there are causally non-ordered multiple mediators. Even if the mediators do not affect each other, the sum of two indirect effects through the two mediators considered separately may diverge from the joint natural indirect effect when there are additive interactions between the effects of the two mediators on the outcome. Therefore, we derive an equation for the joint natural indirect effect based on the individual mediation effects and their interactive effect, which helps us understand how the mediation effect works through the two mediators and relative contributions of the mediators and their interaction. We also discuss an extension for three mediators. The proposed method is illustrated using data from a randomized trial on the prevention of dental caries.

Download Full-text

Effect of Timely Lactate Measurement on In-hospital Mortality among Adults with Hypotension and Hyperlactatemia: An Observational Study on Two Cohorts

10.21203/rs.3.rs-60524/v1 ◽

2020 ◽

Author(s):

Hui Chen ◽

Xiebing Bao ◽

Ying Xu ◽

Yanxia Guo ◽

Mingqin Zhou ◽

...

Keyword(s):

Observational Study ◽

Hospital Mortality ◽

Lactate Level ◽

Lower Risk ◽

Mediation Effect ◽

Research Database ◽

Icu Admission ◽

Causal Mediation ◽

Lactate Measurement ◽

Mimic Iii

Abstract Background: Whether patients presented with hypotension and hyperlactatemia can benefit from timely lactate measurement and further lactate-guide resuscitation were not fully understood.Methods: This was a retrospective observational study based on the data from the Medical Information Mart for Intensive Care (MIMIC)-III Database and the eICU Collaborative Research Database (eICU). Patients with hypotension (defined as a minimal systolic blood pressure ≤90 mm Hg or minimal mean arterial pressure ≤65 mm Hg or requiring any vasopressors support during the first 24 h after ICU admission) and hyperlactatemia (defined as an initial lactate level > 2.0 mmol/L after ICU admission) were eligible.The primary exposure was the timely lactate measurement, which was defined as an initial lactate level measured within 1 h after ICU admission. The primary outcome was in-hospital mortality. The statistical approaches included multivariate regression, propensity score matching (PSM) and an inverse probability of treatment weighing (IPTW) and causal mediation analysis (CMA) were utilized to elucidate the relationship between timely lactate measurement and in-hospital mortality. Results: A total of 9978 patients were identified, of which 4257 in MIMIC-III and 5721 in eICU. Timely lactate measurement was associated with lower risk-adjusted in-hospital mortality both in MIMIC (OR 0.70 (95%CI 0.58-0.85; p<0.001)) and eICU (OR 0.75 (95%CI 0.64-0.88; p<0.001)). Time to initial intravenous fluid (IVF) in MIMIC mediated 6.7% (95%CI 1.4%-38%; p<0.001) of the beneficial effect of timely lactate measurement (p<0.001 for average causal mediation effect (ACME)) in terms of in-hospital mortality. Finally, delayed initial lactate measurements are also associated an increased in-hospital mortality in MIMIC and eICU.Conclusions: Timely lactate measurement is associated with a lower risk-adjusted in-hospital mortality among patients with hypotension and hyperlactatemia, which was proportional mediated through shortening the time to IVF. Delay in initial lactate measurement showed a positive association with in-hospital mortality.

Download Full-text

Measurement of central venous pressure is associated with better outcomes in septic patients: a retrospective observational study

10.21203/rs.2.24067/v1 ◽

2020 ◽

Author(s):

Hui Chen ◽

Zhu Zhu ◽

Chenyan Zhao ◽

Yanxia Guo ◽

Dongyu Chen ◽

...

Keyword(s):

Central Venous Pressure ◽

Medical Information ◽

Venous Pressure ◽

Serum Lactate ◽

Mediation Effect ◽

Central Venous ◽

Causal Mediation ◽

Causal Mediation Analysis ◽

Vasopressor Use ◽

The Relationship

Abstract Purpose: With the proper insights, measurement of central venous pressure (CVP) can be a useful clinical aid. However, the formal utility of CVP measurement on mortality in septic patients has never been proved.Methods: The Medical Information Mart for Intensive Care III (MIMIC-III) was applied to identify septic patients who had and did not have CVP measured. The primary outcome was 28-day mortality. The statistical approaches including multivariate regression, propensity score matching (PSM) and an inverse probability of treatment weighing (IPTW) and causal mediation analysis (CMA) were utilized to elucidate the relationship between CVP measurement and 28-day mortality.Results: A total of 10275 patients were included in our study, of which 4516 patients (44%) had CVP measured within 24 h after ICU admission. A significant beneficial effect of CVP measurement in terms of 28-day mortality was observed (OR 0.60 (95% CI 0.51–0.70; p<0.001)). Patients in CVP group received more fluid on day 1, had a shorter duration of mechanical ventilation and vasopressor use, and the reduction of serum lactate was higher than that in the no CVP group. The mediation effect of serum lactate reduction was significant for the whole cohort (p=0.04 for average causal mediation effect (ACME)) and patients in the CVP group with an initial CVP level below 8mmHg (p=0.04 for ACME).Conclusion: CVP measurement is associated with a lower risk-adjusted 28-day mortality among patients with sepsis, which is proportionally mediated through serum lactate reduction.

Download Full-text

On Partial Identification of the Natural Indirect Effect

Journal of Causal Inference ◽

10.1515/jci-2016-0004 ◽

2017 ◽

Vol 5 (2) ◽

Cited By ~ 1

Author(s):

Caleb Miles ◽

Phyllis Kanki ◽

Seema Meloni ◽

Eric Tchetgen Tchetgen

Keyword(s):

Antiretroviral Therapy ◽

Virological Failure ◽

Indirect Effect ◽

Mediation Analysis ◽

Partial Identification ◽

Nonparametric Identification ◽

Causal Mediation ◽

Causal Mediation Analysis

AbstractIn causal mediation analysis, nonparametric identification of the natural indirect effect typically relies on, in addition to no unobserved pre-exposure confounding, fundamental assumptions of (i) so-called “cross-world-counterfactuals” independence and (ii) no exposure-induced confounding. When the mediator is binary, bounds for partial identification have been given when neither assumption is made, or alternatively when assuming only (ii). We extend existing bounds to the case of a polytomous mediator, and provide bounds for the case assuming only (i). We apply these bounds to data from the Harvard PEPFAR program in Nigeria, where we evaluate the extent to which the effects of antiretroviral therapy on virological failure are mediated by a patient’s adherence, and show that inference on this effect is somewhat sensitive to model assumptions.

Download Full-text

Identifying Mediators Underlying the Intergenerational Cycle of Obesity: A Causal Mediation Analysis

Current Developments in Nutrition ◽

10.1093/cdn/nzaa063_053 ◽

2020 ◽

Vol 4 (Supplement_2) ◽

pp. 1655-1655

Author(s):

Tiange Liu ◽

Noel Mueller ◽

Sara Benjamin-Neelon

Keyword(s):

Birth Weight ◽

Gestational Age ◽

Mediation Analysis ◽

Mediation Effect ◽

Delivery Mode ◽

Z Score ◽

Causal Mediation ◽

Causal Mediation Analysis ◽

Intergenerational Cycle ◽

Infant Birth

Abstract Objectives To understand the mechanisms of the intergenerational cycle of obesity between women and offspring. Methods We recruited pregnant women into the Nurture study (North Carolina, US) and prospectively followed up their offspring until 1 year of age from 2013–2017. The exposure of this analysis was self-reported maternal pre-pregnancy body mass index (BMI) calculated using weight and height. The outcome was researcher-measured infant weight-for-length z-score (WFLZ) at 1 year, calculated based on the WHO Child Growth Standards. We conducted a causal mediation analysis to estimate the average mediation effect of each mediator, including gestational weight gain (GWG), delivery mode, infant birth weight-for-gestational age z-score, and duration of breastfeeding. We adjusted for maternal age, race, parity, smoking status prior to pregnancy, education, household income, food security, and gestational age (when not examining birth weight-for-gestational age z-score). Results We included 380 dyads. Among mothers, there were 65.5% black, 22.6% white, and 11.8% other/multiple race. Prior to pregnancy, 19.5% were overweight and 45.3% were obese. A 10 kg/m2 increment of pre-pregnancy BMI was associated with 0.16 (95% CI: 0.06, 0.27) higher infant WFLZ at 1 year. When examining mediators individually, birth weight-for-gestational age z-score had a statistically significant mediation effect (0.05, 95% CI: 0.02, 0.08), corresponding to 30.2% (95% CI: 20.0%, 62.9%) of the total effect of pre-pregnancy BMI on infant WFLZ. The average mediation effect by GWG was −0.04 (95% CI: −0.08, 0.00), by cesarean delivery was 0.01 (95% CI: −0.01, 0.04), and by breastfeeding duration was 0.02 (95% CI: −0.01, 0.06). Treating mediators as potential confounders for one another did not alter the results. Conclusions Infant birth weight-for-gestational age z-score mediated, in part (∼30%), the relation between maternal pre-pregnancy BMI and infant WFLZ at 1 year. In contrast, GWG, delivery mode, and breastfeeding were not mediators in our sample. This highlights the importance of primordial prevention of maternal obesity, ideally prior to conception, to mitigate the intergeneration cycle of obesity. Research exploring the potential mediating role of factors such as the gut microbiome is needed. Funding Sources The National Institutes of Health.

Download Full-text