Objective: To analyze the research hotspots and development trends of pressure injury risk prediction based on machine learning algorithms by using bibliometric methods. Methods: CNKI, Wanfang, VIP and PubMed databases were retrieved from the inception of databases to October 31, 2022. The bibliographic co-occurrence analysis system software was used to extract the literature data. CiteSpace and gCLUTO software were used to visually cluster the literature data. Results:A total of 1 504 articles were retrieved, including 1 239 Chinese articles and 265 English articles. According to the inclusion and exclusion criteria, 89 articles were included, including 38 Chinese articles(42.70%) and 51 English articles (57.30%). The number of papers published in Chinese and English in this field showed an overall increasing trend. The included articles were published in 63 journals, including 25 Chinese journals and 38 English journals. The co-author rate of the included literature was 91.01 %. There were 45 articles with fund support（50.56 %）. The study mainly involved ICU patients, surgical patients and elderly patients. A total of 98 keywords were extracted from Chinese literature and 59 keywords were extracted from English literature. The research hotspots of Chinese literature were divided into four groups including Braden score, prediction, pressure injury and prediction model. The research hotspots of English literature were divided into four groups including intensive care unit, pressure ulcers, nomogram and machine learning. Conclusion: The researches on the risk prediction of pressure injury based on machine learning algorithm is still in the developmental stage. In the future, the scientific and integrity of the research and design of the prediction model should be strengthened. The following can also be used as future research directions such as the construction and verification of the risk prediction model of pressure injury in different specialties, the training of machine learning algorithms and nursing education research.

目的：应用文献计量学方法分析基于机器学习算法的压力性损伤风险预测的研究热点和发展趋势。 方法：检索中国知网、万方、维普和 PubMed 数据库自建库至 2022 年 10 月 31 日收录的相关文献， 运用书目共现分析系统软件提取文献数据， 运用 CiteSpace 和gCLUTO 对文献数据进行可视化聚类分析。结果：共检索到 1 504 篇文献，其中中文文献共 1 239 篇，英文文献 265 篇。按照纳入、排除标准筛选后共纳入 89 篇文献，包括中文文献 38 篇（42.70%）、英文文献 51 篇（57.30%）。 该领域中英文文献发文量整体呈增长趋势。纳入的文献发表于 63 种期刊， 其中中文期刊 25 种， 英文期刊 38 种。 纳入文献的合著率为 91.01%。 有基金支持的文献共 45 篇（50.56%）。 研究主要涉及的人群包括 ICU 患者、手术患者以及老年患者。 中文文献共提取 98 个主题词，英文文献共提取主题词 59个。 中文文献的研究热点分为 4 组：Braden 评分、预测、压力性损伤、预测模型。 英文文献的研究热点分为 4 组：ICU（重症监护病房）、Pressure ulcers（压力性损伤）、nomogrand（列线图） and machine learning（机器学习）。结论：基于机器学习算法的压力性损伤风险预测研究尚处于发展阶段，今后研究应强化预测模型研究设计的科学性和完整性，未来可以开展不同专科压力性损伤风险预测模型的构建及验证研究、机器学习算法的培训及护理教育研究。