<p>Diagnosing gastric diseases must address skewed medical data, where positive cases of any specific gastric disease constitute only a small proportion of all cases due to the variety of gastric diseases. Therefore, a new approach is called for to effectively diagnose multiple gastric diseases with high scalability. In this study, a skewed medical data-oriented parallel modular diagnosis approach is proposed for gastric diseases via scalable belief-ruled hierarchy. First, analytical lower-to-middle factors integration functions are constructed. Then, a middle-to-top diagnosis model is constructed using a belief-ruled hierarchy. Third, a skewed data-oriented optimization model is constructed followed by an optimization algorithm. Finally, a parallel modular diagnosis mechanism is proposed for multiple gastric diseases. A practical case with 172 cases and nine types of gastric diseases gathered from a tertiary hospital in Hangzhou, China, during a three-month period is studied for validation. Case study results demonstrate that the proposed approach diagnoses positive cases of nine gastric diseases with an average accuracy of 86.93%. Compared with five other conditions with varied parameter configurations and four machine learning approaches, the proposed approach produces optimal results (e.g., 53.61%, 15.28%, 42.73%, and 40.30% by BPNN, SVM, RF, and GPR, respectively) owing to its penalty coefficient customized to the skewed gastric data to balance all-case diagnosis accuracy and positive-case accuracy.</p>