Background: Diagnostic errors are common in medicine. The identification of diagnostic errors is often delayed, typically occurring only after a patient has experienced harm. Care transitions are inherent to the practice of hospital medicine and although care transitions have long been recognized as an area of vulnerability for hospitalized patients, they also represent opportunities for error recognition and improvement as the oncoming clinician reviews and revises the plan established by the prior clinician. We sought to leverage care transitions to provide an opportunity for critical review of diagnoses and to quantify the rate of diagnostic errors in hospital medicine.
Methods: We created a questionnaire based on prior research in this area to assess changes in diagnoses and whether those changes were due to disagreement or other factors (e.g. disease evolution or diagnostic refinement).1 We then created a report in our Electronic Data Warehouse to identify patients admitted from Thursday-Saturday to inpatient hospital medicine services and handed off from one hospitalist to another at end of each week on service. We randomly identify 1-2 patients per hospitalist on service meeting inclusion criteria to assign on the questionnaire and distribute it to the hospitalist assuming care for those patients. Completed questionnaires, in which there was a change in diagnosis, were then reviewed independently by two physician team members using the Revised SaferDx Intrument.2 If there was disagreement between the two independent reviewers on whether the diagnostic change was an error or not, a third reviewer was assigned the chart to review. All reviewers were blinded to each other’s reviews. Using these reviews, we established the frequency of diagnostic changes and errors.
Results: Overall, 167 of 174 questionnaires (96%) were returned over the study period from April 3rd to October 31st, 2019. Questionnaires included data for 281 patients, of which 53 had diagnostic changes (18%). After chart reviews using the Revised SaferDx tool, 19 of these 53 diagnostic changes were deemed to be due to errors, translating into an overall diagnostic error rate of 6.8%. Of the 19 errors, 18 directly reached the patient involved. Of those, 5 required further monitoring to confirm no harm occurred or required intervention to preclude harm. We also found 5 of the errors may have contributed to, at least, temporary harm to the patient. We observed a favorable inter-rater reliability with a Cohen Kappa Statistic of 0.68 suggestive of substantial agreement between reviewers.
Conclusions: Handoffs in patient care can be used to identify changes in diagnoses and diagnostic errors. With this data, we hope to set a benchmark on the baseline rate of diagnostic errors in hospital medicine. We plan to continue this research by evaluating the effect of providing diagnostic feedback to hospital medicine physicians on the rate of errors.