To investigate the risk factors for early-onset psychosis in Parkinson's disease (PD) in a cohort of patients from the Parkinson's Progression Markers Initiative. Longitudinal data on motor and non-motor features, dopamine transporter (DAT) imaging, and cerebrospinal fluid (CSF) measurements were collected. The survival probability of psychotic symptoms, potential risk factors for psychosis development over a 5-year follow-up period, and the performance of the prediction model were evaluated. Among the 338 newly diagnosed patients with PD, 83 developed psychotic symptoms. Gastrointestinal autonomic dysfunction, presence of probable rapid-eye-movement sleep behavior disorder, and the ratio Aβ42: total-tau could independently predict onset of psychosis in PD (hazard ratio (HR) = 1.157, 95% confidence interval (CI) 1.022-1.309, p = 0.021, HR = 2.596, 95% CI 1.287-5.237, p = 0.008, and HR = 0.842, 95% CI 0.723-0.980, p = 0.027, respectively). The combined model integrating baseline clinical predictors, DAT imaging, and CSF measurements achieved better sensitivity than the clinical predictors alone (area under the curve = 0.770 [95% CI 0.672-0.868] vs. 0.714 [95% CI 0.625-0.802], p = 0.098). We identified clinical and CSF predictors of early-onset psychosis in patients with PD. Our study provides evidence and implications for prognostic stratification and therapeutic approaches for PD psychosis.