Derivation and validation of a preoperative risk model for postoperative mortality (SAMPE model) : an approach to care stratification

• Autor(a) principal: Stefani, Luciana Paula Cadore
• Data de Publicação: 2017
• Outros Autores: Gutierrez, Cláudia de Souza, Castro, Stela Maris de Jezus, Zimmer, Rafael Leal, Diehl, Felipe Polgati, Meyer, Leonardo Elman, Caumo, Wolnei
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UFRGS
• Texto Completo: http://hdl.handle.net/10183/172571
• topic: Bioestatística; HIV

Ascertaining which patients are at highest risk of poor postoperative outcomes could improve care and enhance safety. This study aimed to construct and validate a propensity index for 30-day postoperative mortality. A retrospective cohort study was conducted at Hospital de ClõÂnicas de Porto Alegre, Brazil, over a period of 3 years. A dataset of 13524 patients was used to develop the model and another dataset of 7254 was used to validate it. The primary outcome was 30-day in-hospital mortality. Overall mortality in the development dataset was 2.31% [n = 311; 95% confidence interval: 2.06±2.56%]. Four variables were significantly associated with outcome: age, ASA class, nature of surgery (urgent/emergency vs elective), and surgical severity (major/intermediate/minor). The index with this set of variables to predict mortality in the validation sample (n = 7253) gave an AUROC = 0.9137, 85.2% sensitivity, and 81.7% specificity. This sensitivity cut-off yielded four classes of death probability: class I, <2%; class II, 2±5%; class III, 5±10%; class IV, >10%. Model application showed that, amongst patients in risk class IV, the odds of death were approximately fivefold higher (odds ratio 5.43, 95% confidence interval: 2.82±10.46) in those admitted to intensive care after a period on the regular ward than in those sent to the intensive care unit directly after surgery. The SAMPE (Anaesthesia and Perioperative Medicine Service) model accurately predicted 30-day postoperative mortality. This model allows identification of high-risk patients and could be used as a practical tool for care stratification and rational postoperative allocation of critical care resources.