Abstract: Objective: To systematically review the machine learning models for risk prediction of Pressure Injury (PI). Methods: A search was performed on the machine learning models for PI risk prediction in PubMed, Embase, Cochrane Library, CBM, CNKI and Wanfang databases from the establishment of the database to March 1, 2022. Two researchers independently screened literature, extracted data and evaluated the quality of the included studies based on PROBAST. Results: Totally 17 studies were included, including 4 development studies and 13 development and validation studies. The area under the subject operating characteristic curve ranged from 0.790 to 0.897. The overall applicability of the study was good, but there was a certain bias, mainly because blind method was not adopted or reported, the missing data processing method was not reported or improper, sample size was insufficient, independent variable processing was inappropriate, and model performance and fitting situation were not considered. Conclusion: The development of PI risk prediction machine learning models is still in the developing stage, and the extrapolation needs to be further discussed. In the future, attention should be paid to study design and clinical data processing, and the development of models suitable for the Chinese population.

Key words: Pressure Injury; risk prediction; machine learning model; systematic review