MLMIC DentalLog In

< Back to Publications & Resources

Strategies for Reducing Diagnostic Errors Related to COVID-19

diagnostic process related to COVID-19

Concerns have been raised that the diagnostic process related to COVID-19 may be leading to missed or delayed diagnosis of the virus and subsequent harm. Authored by the University of California Davis School of Medicine’s Patrick Romano, a recent Agency for Healthcare Research and Quality (AHRQ) primer examines why diagnostic error has increasingly become a patient safety issue associated with the pandemic. Romano explains it is essential to correct the problem because accurate and timely communication of test results to patients is critical in ensuring effective treatment and preventing transmission of COVID-19.

To avoid potential diagnostic errors, according to the author, physicians and healthcare organizations must address the following biases that are often present during clinical decision-making and can be amplified by stressful circumstances, such as the pandemic:

  • Availability bias which leads clinicians to over diagnose conditions that are available in their memories, based on reading or clinical experience, or under diagnose conditions based on limited experience or exposure. This bias can influence the consideration a physician gives to the probability of COVID-19.
  • Anchoring bias, writes Romano, causes “physicians to resist altering their initial diagnostic impression, despite subsequent information that contradicts that impression.” Anchoring bias may prompt a physician to make a presumptive diagnosis of COVID-19 or fail to consider the possibility of infection.
  • Framing biases can lead clinicians to different diagnostic impressions based on how information is presented or framed, such as an over diagnosis of COVID-19 in some populations and under diagnosis among other at-risk populations; and
  • Implicit Biases which, explains Romano, “involves associations outside conscious awareness that lead to a negative evaluation of a person based on irrelevant characteristics such as race and ethnicity, nationality, disability, and socio-economic status.” This type of bias may interfere with patient-physician communication about COVID-19 risk factors, potential exposure and symptoms.

He says these scenarios often result from “automatic thinking,” and physicians must shift to “reflective thinking” during the diagnostic process by applying the following approaches:

  • pause to examine the working diagnosis and consider other possibilities;
  • review evidence that may counter the working diagnosis;
  • consider the risk of two concurrent diagnoses;
  • use “artificial intelligence” to support clinical decision-making and provide analytics on the probability of alternative diagnoses;
  • carefully consider the response to initial treatment and the possibility of an alternative diagnosis if the outcome is negative;
  • avoid the consideration of identifying information that may not be relevant to clinical decision-making, such as the patient’s national origin, race or ethnicity;
  • avoid communication that suggests diagnostic uncertainty and frames the patient’s condition in a way that fosters premature closure; and
  • implement education and training for medical professionals to reduce biases in the diagnostic process.

To further reduce the likelihood of a COVID-19-related diagnostic error, Romano says healthcare providers should be diligent in adhering to the Centers for Disease Control and Prevention’s (CDC) SARS-CoV-2 testing protocol, including initial collection of an upper respiratory tract specimen using a nasopharyngeal swab, use of only synthetic fiber swabs with plastic or wire shafts and the immediate placement of swabs into a sterile specimen tube containing 2-3 mL of an appropriate transport medium.

MLMIC encourages and supports a proactive approach to improving diagnostic accuracy and preventing errors. Additionally, clinicians are advised to monitor guidance for the management and treatment of COVID-19  on our resources page and blog: