3 - (IOP 39) Causal Inference in Clinical Psychology: Using Target Trial Emulation to Improve Observational Research

Clinical researchers often use observational data (e.g., social media posts, electronic health records) to study mental health patterns and treatment outcomes among individuals with or at risk for psychopathology. These datasets offer several benefits, such as large sample sizes, inclusion of diverse and underserved populations, and detailed longitudinal information about real-world clinical decisions. Furthermore, observational data allow researchers to answer clinical questions that cannot feasibly be addressed in an RCT. To assess whether poverty increases risk for depression, for example, it would be unethical to randomize individuals to endure poverty.

Despite these benefits, observational studies present challenges that impede our ability to draw causal conclusions about treatments (e.g., lack of randomization, selection bias). When an RCT—the gold standard for causal inference—cannot be conducted, target trial emulation (TTE) offers a framework for mimicking an RCT using observational data (Hernán et al., 2025), thereby reducing methodological bias and improving causal estimates. The TTE approach is used widely in epidemiology, but rarely in clinical psychology. Accordingly, clinical psychological research is prone to biases in observational studies, and remains limited in its ability to determine whether, when, and for whom mental health treatments work.

This oral paper describes the TTE approach for psychological research, and illustrates its application with an empirical example. First, this talk outlines the two-step process for emulating an RCT using observational data: (1) designing the ideal RCT that would answer the causal question of interest, and (2) mimicking the RCT by using causal inference techniques (e.g., standardization, clone-censor-weight) to emulate randomization, adjust for confounding, and bypass design-induced biases commonly found in observational research.

Next, this talk presents a study using TTE to assess whether lithium discontinuation (vs. maintenance) increases suicide risk in adults with mood disorders. In this study, we created a pseudo-observational dataset by pooling data across three RCTs conducted in the United States from 2009 to 2019 (Katz et al., 2022; Nierenberg et al., 2013; Nierenberg et al., 2016). Participants in the TTE include 636 adults (ages 18+) diagnosed with bipolar disorder or major depressive disorder and receiving lithium treatment. To highlight the benefits of TTE, this talk will show how suicide risk estimates differ when using TTE vs. traditional approaches (e.g., demonstrating that confidence intervals are wider when ignoring confounds that vary throughout treatment, showing how selection bias emerges when a baseline timepoint is not specified).

Overall, this talk exemplifies how clinical psychologists can leverage rigorous causal inference approaches from epidemiology to improve mental health treatment research. When RCTs are impractical, the TTE approach helps researchers define a causal question, obtain more precise treatment estimates, and identify which participants do (or do not) benefit from treatment. In turn, TTE can inform clinical decisions, facilitate personalized treatment, and advance equitable care for mental health.