| pubmed-article:10171190 | pubmed:abstractText | Large databases are being used for outcome prediction analysis with increasing frequency. This review examines four separate databases used to provide risk analysis in the cardiac surgery population. Populations in the databases range in size from 3500 to over 116,000 patients. All of the databases were applied on the clinical, and in one instance, institutional level. Outcome prediction from databases is not without its limitations. Data collection, model bias, and methodologic variation all contribute to weaknesses in the application of databases for outcome prediction. | lld:pubmed |
