Abstract:

Objective To systematically analyze and evaluate the risk prediction model for pneumonia in patients with intracerebral hemorrhage (ICH). Methods Articles reporting risk prediction model for pneumonia in ICH patients published prior to February 2023 were searched in the online databases of Pubmed, Web of Science, Embase, The Cochrane Library, Scopus, Ovid Medline, CNKI (China National Knowledge Infrastructure), WanFang Data, VIP and CBM (Chinese Biomedical Literature Database). Two researchers were independently responsible for screening literature and extracting data. The quality of the literature included in this study was rigorously evaluated, and both the risk of bias and adaptability were assessed in accordance with the Transparent Reporting of a Multivariable Prediction Model for Individal Prognosis or Diagnosis(TRIPOD), and the Prediction Model Risk of Bias Assessment Tool(PROBAST). Results A total of 12 relevant studies were included, involving 7 registered studies, 1 ovarian case-control study, 3 single-center case-control studies, and 1 retrospective cohort study. Logistic regression and machine learning were used for modeling. Eight studies were validated internally, 2 studies were only validated externally, and 2 studies were validated both. The area under the receiver operating characteristic curve of the model was 0.740-0.920. The range of predictors in the 12 studies ranged from 4 to 11, and the common predictors were the age, the National Institutes of Health Stroke Scale score, the Glasgow Coma Scale score, dysphagia, smoking, chronic obstructive pulmonary disease, and nasogastric tube feeding. Model calibration was performed in 9 studies and not in 3 studies. The model was mainly presented in the form of risk score, risk calculation formula and nomogram. The included studies exhibited moderate quality and a high risk of bias. Conclusion The current model for predicting the risk of pneumonia in ICH patients demonstrates good predictive ability, and the predictive factors are relatively easy to obtain. However, there are also significant defects and high bias. In future research, it is recommended that researchers adhere to the TRIPOD guideline and PROBAST statement when conducting prediction model studies. It is important to summarize the advantages and disadvantages of existing models and to conduct external verification, thus developing a risk prediction model for pneumonia in ICH patients with excellent predictive performance and ease of use.

Key words: intracerebral hemorrhage, pneumonia, prediction model, systematic review