The Institute of Medicine report Improving Diagnosis in Health Care called for tools to monitor physicians' diagnostic process. We addressed this need by developing a tool for clinicians to record and analyze their diagnostic process. The tool was a secure web application in which clinicians used a structured grading system to assess the relative impact of clinical, laboratory, and neuroimaging data for every new diagnosis. Four neurohospitalists used the tool for 6.5 months on a general neurology ward service at a single tertiary-level teaching hospital. Process measures of tool use included number of diagnoses entered, time spent on each data entry, and concordance of diagnoses compared to the medical record. We also aggregated the data across clinicians to examine the average process scores across common inpatient disorders. The 4 clinicians entered 254 new diagnoses that took approximately 3 minutes per patient. In 50 randomly chosen cases, the neurohospitalists' diagnoses entered into the tool agreed with 92% of diagnoses in the medical record, which was better than the agreement between billing code and medical record diagnoses (74%). The diagnostic process varied across disease categories, showing a spectrum of clinical-dominant (eg, headache), laboratory-dominant (eg, encephalitis), and neuroimaging-dominant (eg, stroke) disorders. This study demonstrated the feasibility of a clinician-driven diagnostic process monitoring system, along with preliminary characterization of the process for common disorders. The tracking of diagnostic process has the potential to promote reflection on clinical practice, deconstruct neurologists' clinical decision-making, and improve health-care safety.
Keywords: clinical decision-making; diagnosis; process assessment.