Abstract

To the Editor

Clinicians are subject to many stresses in the midst of healthcare reform. Some question the role of the traditional four tools of clinical management: history taking, physical examination, laboratory investigation, and radiologic evaluation. The US healthcare system is attempting to switch from a treatment model to a health maintenance model that emphasizes prevention. The effective implementation of such preventive measures often overlooks the step of analyzing existing medical records and data. MEDLINE adds approximately 2000 to 4000 references daily,¹ and enhanced electronic storage of data and electronic medical records likely will accelerate this phenomenon. In the quest for analyzing clinical data, there is another tool that is emerging as a worthy addition to the clinician’s repertoire. This method uses a data-mining approach that can capitalize on largely untapped but often readily available sources of health information. Predictive analytics is an approach in which data are analyzed for meaningful patterns that can provide actionable insights, which can be used to enhance preventive actions and provide cost-effective management to improve healthcare outcomes.² Our interest was piqued many years ago by Baxt, who demonstrated that when compared with clinicians, an artificial neural network had a significantly better predictive value for diagnosing myocardial infarction in the emergency department.³ Since then, myriad approaches, including machine intelligence, have been added to standard statistical techniques. The advent of IBM’s natural language artificial intelligence computer system, Watson, provides an example of the capacity to quickly extract salient information from extensive patient records⁴ and analyze them⁵ for the benefit of both provider and patient.