In this study, we developed an LLM-based model for retest decision in clinical chemistry tests and evaluated its feasibility for manual verification support. The study data consisted of a subset of clinical chemistry test results and manual verification records from the clinical laboratory in Asan Medical Center. The model was designed to predict the retest decision and generate its rationale based on test results and clinical information. A rule-based algorithm was also implemented as a reference. The LLM-based model showed an accuracy of 0.987 on the test set. Its rationale generation performance was BLEU 0.886 and ROUGE-L 0.881. In qualitative assessment by clinician, the generated rationales showed high scores for decision consistency and readability, while factual accuracy was relatively low. The rule-based algorithm achieved an accuracy of 0.991. These results suggest that the LLM-based model has potential as a supportive tool for explanation generation in manual verification of clinical chemistry tests. This study provides baseline evidence for future decision support systems combining rule-based decision structures with LLM-based rationale generation.