Abstract
BACKGROUND: This study aimed to evaluate systematically the risk prediction model for heart failure after myocardial infarction, assess its prediction performance and effect, and provide a reference basis for clinical practice and scientific research. METHODS: Computerized searches of CNKI, WanFang Data, China Science and Technology Journal Database (VIP), China Biology Medicine Database, PubMed, Web of Science, Cochrane Library, Embase, and other databases were performed to retrieve all studies on risk prediction models for heart failure after myocardial infarction from the time of library construction to June 1, 2023. Two researchers independently screened the literature, extracted the information, and evaluated the risk of bias in the included studies. RESULTS: A total of 28 studies were included, of which 25 were retrospective and 3 were prospective, all conducted in China and in patients with acute myocardial infarction. Twenty-two studies were modeled using logistic regression, 5 studies were modeled using Cox regression, and 1 study was modeled using full-subset regression. Twenty-eight studies had an area under the work characteristic curve of the subjects or a C-statistic between 0.6 and 0.9, and only 2 studies were divided into modeling and validation groups for external validation. The most common predictors in the predictive models were age, Gensini score, ultrasensitive C-reactive protein level, multiple diseased vessels, left ventricular ejection fraction, and GRACE score, among others. Heterogeneity and risk of bias were high in the prediction models, with the risk of bias high in 26 models and unclear in 2. Fourteen of the 28 models had high overall applicability. CONCLUSION: The overall predictive efficacy of the heart failure risk prediction model in patients with acute myocardial infarction was good; however, the model heterogeneity and risk of bias were high, and the overall quality needs to be improved. In the future, the TRIPOD reporting specifications should be strictly followed, and the model should be described objectively in 22 aspects, including the title and abstract, introduction, methods, results, statistical analysis, and discussion, to improve the quality of the model and facilitate the prediction model to better serve the clinic.