Developing tools to predict outcomes following cardiovascular surgery Maggi Boult,* Kate Fitzpatrick,* Mary Barnes,† Guy Maddern* and Robert Fitridge* *Department of Surgery, The University of Adelaide, The Queen Elizabeth Hospital, Woodville South †CSIRO Mathematical and Information Sciences – Adelaide, Glen Osmond, South Australia, Australia Key words cardiac surgery, decision tool, model development, predictive modelling, vascular surgery. Correspondence Professor Guy Maddern, Department of Surgery, The University of Adelaide, 28 Woodville Road, Woodville South, SA 5011, Australia. Email: guy.maddern@adelaide.edu.au M. Boult BSc, GDIM; K. Fitzpatrick BSc (Hons), PhD; M. Barnes BAppSci, GradDipMaths; G. Maddern MB BS, PhD, MS, MD, FRACS; R. Fitridge MB BS, MS, FRACS Accepted for publication 9 November 2010. doi: 10.1111/j.1445-2197.2010.05644.x Abstract Background: Surgical decision-making tools may help surgeons achieve better out- comes by providing more personally relevant information to patients. This paper describes approaches to developing statistical tools capable of estimating the prob- ability of morbidity and mortality after cardiovascular surgery. Our aim is to inform surgeons about the important stages that contribute to the development of decision tools. Methods: The key elements described include study design (data quality, cohort size, etc.) and statistical methodology for developing and testing decision tools. Mention is made of the delivery of decision tools, simplicity of use, ease of interpretation of results and accessibility. Information specific to cardiac and vascular surgery is included. Results: Development of useful and effective decision tools is dependent on robust and reliable data, unambiguous outcome requirements and considerable statistical expertise. Decision tools must also be extensively tested for validity and reliability, both internally and with external data. Conclusion: Understanding the development and assumptions that underlie surgical decision tool development will help cardiovascular surgeons appreciate the value of applying such techniques at a clinical level. Introduction Surgical decision tools may help surgeons optimize outcomes by providing more personally relevant information to patients. Many relationships between preoperative variables and outcomes have been reported 1 , but few individualized predictive models have been developed that are simple to use, readily accessible and predict pertinent outcomes based on a limited number of preoperative vari- ables. This paper discusses some of the steps that go into developing such decision tools, including the data requirements and statistical considerations. Some level of risk will always exist for patients undergoing sur- gical procedures and these risks are higher for specific patients. To help the ‘real-world’ patient make decisions, medical science must inform individuals, not populations. 2 Hence, the aim of surgical decision tools is to identify patients at higher risk of adverse out- comes BEFORE they undergo a given procedure. Personalized risk assessment is an effective means of providing better clinical decision making, especially where high-risk subgroups can be identified prior to surgery. Liu et al. 3 coined the term ‘decision tools’ to describe computer- ized decision support systems and defined them as an: ‘active knowledge resource that uses patient data to generate case- specific advice which support decision-making about individual patients by health professionals, the patient themselves or others con- cerned about them.’ Other terms in general use include ‘risk stratification models’, ‘prediction models’, ‘risk scoring’ and ‘risk models’. In this review, we have adopted Liu’s nomenclature ‘decision tool’ to describe methods used to predict outcomes for individual patients. 3 We have put these into the context of cardiac and vascular surgery and discuss their usefulness, accessibility and validity. Decision support is important in all surgical areas, not least those associated with cardiovascular disease, as part of which cardiac REVIEW ARTICLE ANZJSurg.com © 2011 The Authors ANZ Journal of Surgery © 2011 Royal Australasian College of Surgeons ANZ J Surg •• (2011) ••–••