Methods for Evaluating Diagnostic Processes in Outpatient Care

Brown University, Providence RI

Investigators

Abstract

ABSTRACT Diagnostic errors are estimated to affect 12 million patients in the U.S. each year and are an important source of preventable morbidity and mortality. One study concluded most errors involved process breakdowns in the patient-provider clinical encounter (78.9%), based on review of the EHR, but there are no studies linking diagnostic errors with actual observation of clinical encounters. Literature on diagnostic process is largely based on experiments using simulated situations, or experience in medical education. Diagnostic errors are commonly ascribed to cognitive biases or synthesis errors on the part of the physician, but in fact the clinical encounter is a dyadic interaction in which a patient complaint ordinarily initiates the diagnostic process and patient reports are an important source of information. The information patients volunteer, questions physicians ask and how they ask them, how patients express information, patients own opinions, all may be contribute to the diagnostic outcome. Linking recordings of outpatient encounters with subsequent information about diagnostic accuracy is a logistical challenge in that is impossible to know in advance what encounters will result in errors. However, members of our group have been involved in studies in which hundreds of encounters were audio recorded. It is feasible to record even larger numbers of encounters with the stipulation that only those in which diagnostic performance issues are subsequently identified will be analyzed. As preliminary work, we will generate descriptive information about real-world diagnostic processes using 147 transcripts of outpatient encounters we have previously collected and coded for other purposes. The coding systems isolate provider's information gathering processes, and diagnostic and treatment outcomes, with detailed information about elements such as symptoms, lab tests, physical examination, disease labels, and interactions within the diagnostic process ? e.g., types of questions, information giving, expression of goals or affect, concern, reassurance, and instructions. Physical examinations can be deduced from the dialogue, and physicians generally announce their observations. We will build on the existing coding, specifically focusing on diagnostic processes. We will examine the diagnostic process within the clinical encounter and identify points of vulnerability for diagnostic failure such as premature closure, presumptive interpretation of ambiguous representations, or discounting of inconsistent evidence. The long term goal of this research is to develop and test physician, patient, and system-focused interventions to improve diagnostic accuracy in primary care and other routine outpatient care. The objective of this proposal is to analyze diagnosis-related elements of the clinical encounter for vulnerabilities, develop a taxonomy of diagnostic process failures identified and understand diagnostic performance within the patient- provider encounter using existing audiotaped data. Our team includes both social science expertise in analysis of communication, and clinical expertise in diagnosis.