-
Bias in studies of prenatal exposures using real-world data due to pregnancy identification method
Authors:
Chase D. Latour,
Jessie K. Edwards,
Michele Jonsson Funk,
Elizabeth A. Suarez,
Kim Boggess,
Mollie E. Wood
Abstract:
Background: Researchers typically identify pregnancies in healthcare data based on observed outcomes (e.g., delivery). This outcome-based approach misses pregnancies that received prenatal care but whose outcomes were not recorded (e.g., at-home miscarriage), potentially inducing selection bias in effect estimates for prenatal exposures. Alternatively, prenatal encounters can be used to identify p…
▽ More
Background: Researchers typically identify pregnancies in healthcare data based on observed outcomes (e.g., delivery). This outcome-based approach misses pregnancies that received prenatal care but whose outcomes were not recorded (e.g., at-home miscarriage), potentially inducing selection bias in effect estimates for prenatal exposures. Alternatively, prenatal encounters can be used to identify pregnancies, including those with unobserved outcomes. However, this prenatal approach requires methods to address missing data. Methods: We simulated 10,000,000 pregnancies and estimated the total effect of initiating treatment on the risk of preeclampsia. We generated data for 36 scenarios in which we varied the effect of treatment on miscarriage and/or preeclampsia; the percentage with missing outcomes (5% or 20%); and the cause of missingness: (1) measured covariates, (2) unobserved miscarriage, and (3) a mix of both. We then created three analytic samples to address missing pregnancy outcomes: observed deliveries, observed deliveries and miscarriages, and all pregnancies. Treatment effects were estimated using non-parametric direct standardization. Results: Risk differences (RDs) and risk ratios (RRs) from the three analytic samples were similarly biased when all missingness was due to unobserved miscarriage (log-transformed RR bias range: -0.12-0.33 among observed deliveries; -0.11-0.32 among observed deliveries and miscarriages; and -0.11-0.32 among all pregnancies). When predictors of missingness were measured, only the all pregnancies approach was unbiased (-0.27-0.33; -0.29-0.03; and -0.02-0.01, respectively). Conclusions: When all missingness was due to miscarriage, the analytic samples returned similar effect estimates. Only among all pregnancies did bias decrease as the proportion of missingness due to measured variables increased.
△ Less
Submitted 16 April, 2025;
originally announced April 2025.
-
Using Statistical Precision Medicine to Identify Optimal Treatments in a Heart Failure Setting
Authors:
Arti Virkud,
Jessie K. Edwards,
Michele Jonsson Funk,
Patricia Chang,
Abhijit V. Kshirsagar,
Emily W. Gower,
Michael R. Kosorok
Abstract:
Identifying optimal medical treatments to improve survival has long been a critical goal of pharmacoepidemiology. Traditionally, we use an average treatment effect measure to compare outcomes between treatment plans. However, new methods leveraging advantages of machine learning combined with the foundational tenets of causal inference are offering an alternative to the average treatment effect. H…
▽ More
Identifying optimal medical treatments to improve survival has long been a critical goal of pharmacoepidemiology. Traditionally, we use an average treatment effect measure to compare outcomes between treatment plans. However, new methods leveraging advantages of machine learning combined with the foundational tenets of causal inference are offering an alternative to the average treatment effect. Here, we use three unique, precision medicine algorithms (random forests, residual weighted learning, efficient augmentation relaxed learning) to identify optimal treatment rules where patients receive the optimal treatment as indicated by their clinical history. First, we present a simple hypothetical example and a real-world application among heart failure patients using Medicare claims data. We next demonstrate how the optimal treatment rule improves the absolute risk in a hypothetical, three-modifier setting. Finally, we identify an optimal treatment rule that optimizes the time to outcome in a real-world heart failure setting. In both examples, we compare the average time to death under the optimized, tailored treatment rule with the average time to death under a universal treatment rule to show the benefit of precision medicine methods. The improvement under the optimal treatment rule in the real-world setting is greatest (additional ~9 days under the tailored rule) for survival time free of heart failure readmission.
△ Less
Submitted 13 January, 2025;
originally announced January 2025.
-
Healthy Live Births Should be Considered as Competing Events when Estimating the Total Effect of Prenatal Medication Use on Pregnancy Outcomes
Authors:
Chase D. Latour,
Mark Klose,
Jessie K. Edwards,
Zoey Song,
Michele Jonsson Funk,
Mollie E. Wood
Abstract:
Pregnancy loss is recognized as an important competing event in studies of prenatal medication use. However, a healthy live birth also precludes subsequent adverse pregnancy outcomes, yet these events are often censored. Using Monte Carlo simulation, we examine bias that results from failure to account for healthy live birth as a competing event in estimates of the total effect of prenatal medicat…
▽ More
Pregnancy loss is recognized as an important competing event in studies of prenatal medication use. However, a healthy live birth also precludes subsequent adverse pregnancy outcomes, yet these events are often censored. Using Monte Carlo simulation, we examine bias that results from failure to account for healthy live birth as a competing event in estimates of the total effect of prenatal medication use on pregnancy outcomes. We simulated data for 12 trials estimating the effect of antihypertensive initiation versus non-initiation on two outcomes: (1) composite fetal death or severe prenatal preeclampsia and (2) small-for-gestational-age (SGA) live birth. We used time-to-event methods to estimate absolute risks, risk differences and risk ratios. For the composite outcome, we conducted two analyses where non-preeclamptic live birth was (1) a censoring event and (2) a competing event. For SGA live birth, we conducted three analyses where fetal death and non-SGA live birth were (1) censoring events, (2) a competing event and censoring event, respectively; and (3) competing events. In all analyses, censoring healthy live births led to inflated absolute risk estimates as well as bias and imprecise treatment effect estimates. Studies of prenatal exposures on pregnancy outcomes should analyze healthy live births as competing risks to estimate unbiased total treatment effects.
△ Less
Submitted 30 October, 2024;
originally announced October 2024.
-
A Causal Inference Framework for Leveraging External Controls in Hybrid Trials
Authors:
Michael Valancius,
Herb Pang,
Jiawen Zhu,
Stephen R Cole,
Michele Jonsson Funk,
Michael R Kosorok
Abstract:
We consider the challenges associated with causal inference in settings where data from a randomized trial is augmented with control data from an external source to improve efficiency in estimating the average treatment effect (ATE). Through the development of a formal causal inference framework, we outline sufficient causal assumptions about the exchangeability between the internal and external c…
▽ More
We consider the challenges associated with causal inference in settings where data from a randomized trial is augmented with control data from an external source to improve efficiency in estimating the average treatment effect (ATE). Through the development of a formal causal inference framework, we outline sufficient causal assumptions about the exchangeability between the internal and external controls to identify the ATE and establish the connection to a novel graphical criteria. We propose estimators, review efficiency bounds, develop an approach for efficient doubly-robust estimation even when unknown nuisance models are estimated with flexible machine learning methods, and demonstrate finite-sample performance through a simulation study. To illustrate the ideas and methods, we apply the framework to a trial investigating the effect of risdisplam on motor function in patients with spinal muscular atrophy for which there exists an external set of control patients from a previous trial.
△ Less
Submitted 15 May, 2023;
originally announced May 2023.
