Accurate choices for modeling the probability distribution of variables are crucial in statistical analysis. However, in real-world applications, especially for discrete outcomes, these variables are often treated as continuous data, neglecting their inherent discrete nature. In this paper, we introduce a novel approach leveraging the Simultaneous Transforming and Rounding Process for Bayesian Additive Regression Trees (STAR BART) which is tailored to the causal inference framework. In contrast to well-known methods in causal inference, such as Causal BART and Bayesian Causal Forest (BCF), which are designed for continuous outcomes, an accurate adjusted STAR BART offers an alternative that explicitly addresses the challenges of discrete outcomes. Through a simulation study, we show the performance of our proposed approach in capturing heterogeneous treatment effects in datasets with discrete outcomes and compare it to Causal BART and BCF.