Automated Medical Algorithms: Issues for Medical Errors

Abstract

Introduction The use of medical algorithms can enhance clinical judgment1 and favorably influence patient outcome2. Computerized algorithms can provide timely clinical decision support, improve adherence to evidence-based guidelines, and be a resource for education and research. The Medical Algorithm Project (MEDAL) The Medical Algorithm Project3 (http://www.medal.org) is a free, web-based, compendium of over 2,650 algorithms, divided into 44 categories, that range from simple calculations, such as body mass index, to complicated outcome prediction formulas and clinical decision rules. There are more than 50 anesthesiology algorithms, derived from peer-reviewed literature, including risk prediction models for postoperative nausea and cardiac complications, functional status, sedation and anesthesia recovery scores, predictors of difficult tracheal intubation, equipment and machine checklists, and patient satisfaction measures. Additional algorithms relevant to anesthesiology are found in sections devoted to critical care (severity-of-illness and organ dysfunction scores, outcome prediction models) and other clinical and non-clinical specialties, as well as to specific organ systems (respiratory, cardiovascular, etc.), laboratory calculations, unit conversions and decision analysis. Algorithms are presented in a Microsoft Excel spreadsheet format that is designed for future automation. Each one is represented as: a) overview of documentation, b) unit conversion, c) entry of data, d) intermediate calculations, e) interpretation, and f) data tables. This representation makes explicit the logic, rules and thresholds for variables, and generates explicit and executable instructions. In time-pressured environments like the operating room, medical algorithms should ideally be integrated into "point of care" decision support systems, such as handheld computers and anesthesia workstations. Algorithms could be triggered by data from monitors or electronic record systems. MEDAL provides a source of material for development of such integrated decision support tools. Potential Benefits of Algorithm Automation MEDAL brings together at one site algorithms from diverse sources, thus addressing one of the primary barriers to the practice of evidence-based medicine4. The use of more complex algorithms, such as those involving logistic regression analysis, is facilitated. Automation of medical algorithms may decrease medical errors through reduction of data entry and calculation mistakes, elimination of incorrect recall of algorithm details, and by providing documentation that facilitates proper algorithm selection and application. Algorithms can be used in preoperative risk assessment and triage, automation of routine patient questioning, and decision support for preoperative testing, which may improve the effectiveness and decrease the cost of preoperative evaluation. MEDAL provides a centralized algorithm repository for educators and researchers, especially those interested in anesthesia clinical outcome studies. Outcome measurement algorithms may facilitate departmental performance improvement programs. Risk indices enable comparison between groups ("risk-adjustment") and can help clinicians understand the elements of perioperative risk5. Conclusions and Future Directions MEDAL is open source software, with benefits for a global audience. MEDAL may facilitate the translation of knowledge into available and effective perioperative clinical decision support applications. A larger collaborative effort is needed, so that algorithms are identified, validated, updated, appropriately classified and structured, and transferred into a web-accessible database format. We encourage programmers to incorporate algorithms into perioperative clinical information systems. We hope clinicians will contribute algorithms and make use of them in practice, guided of course by appropriate clinical judgement and experience.