Phenomenon: Compassion, and related constructs such as empathy, are core values in healthcare, with known benefits for both patients and staff. Yet research on the factors that affect compassion and compassion-related constructs remains scattered. This review systematizes and synthesizes studies investigating the predictors of compassion and related constructs among medical students, allowing for a better understanding of the factors that both positively and negatively contribute to the development of compassionate future physicians. Approach: A systematic review of 12 databases for studies from database inception up until April 2020 was conducted. Non–peer-reviewed literature and studies in which >50% of the sample were non-medical students were excluded. Intervention studies were also out of the scope of this review. We assessed risk of bias and confidence in the findings using standardized tools. Data were categorized within the Transactional Model of Physician Compassion, a framework in which compassion is influenced by personal (student), environmental, patient/family, and clinical factors. Findings: Of 14,060 retrieved articles, 222 studies were included. Of these, 95% studied student factors, but only 25% studied environmental, 9% studied patient, and 6% studied clinical factors. Predictors of greater compassion included maturity; work and life experiences; personality traits of openness to experience and agreeableness; skills such as perspective taking, reflection, and mindfulness; and positive role modeling. Conversely, negative attitudes/emotions, burnout, stress, detachment, operating in cultures prioritizing knowledge and efficiency over humanistic care, negative role models, time constraints, and heavy workloads predicted lower compassion. Patient-related factors included “difficult” and “noncompliant” patients or those perceived as responsible for their illness. Overall, 60% of studies had a serious risk of bias, particularly confounding and participant selection biases. Insights: Medical student compassion is predicted by a wide range of factors relating to the student, their training environment, their patients, and the clinical situation. However, existing research has largely focused on student factors (e.g., sociodemographic and dispositional traits), many of which are not amenable to intervention. Skills such as perspective taking, reflection, and mindfulness are associated with higher compassion and may present opportunities for intervention. There is also strong evidence that environmental factors shape students’ compassion. Researchers and educators should continue to explore the impact of patient and clinical factors on students’ compassion. Studies remain at high risk of bias.