Missed diagnosis of stroke in emergency medicine settings is an important patient safety problem. In this study, researchers interviewed emergency medicine physicians about their perspectives on diagnostic neurology and use of clinical decision support (CDS) tools. Themes emerged related to challenges in diagnosis, neurological complaints, and challenges in diagnostic decision-making emergency medicine, more generally. Participating physicians were enthusiastic about the possibility of involving CDS tools to improve diagnosis for non-specific neurological complaints.
Clinical decision support (CDS) systems are designed to improve diagnosis. Researchers surveyed emergency department physicians about their evaluation of human factors-based CDS systems to improve diagnosis of pulmonary embolism. Although perceived usability was high, use of the CDS tool in the real clinical environment was low; the authors identified several barriers to use, including lack of workflow integration.
Stark N, Kerrissey M, Grade M, et al. West J Emerg Med. 2020;21:1095-1101.
This article describes the development and implementation of a digital tool to centralize and standardize COVID-19-related resources for use in the emergency department (ED). Clinician feedback suggests confirms that the tool has affected their management of COVID-19 patients. The tool was found to be easily adaptable to accommodate rapidly evolving guidance and enable organizational capacity for improvisation and resiliency.
McCradden MD, Joshi S, Anderson JA, et al. J Am Med Info Asso. 2020;27:2024-2027.
The authors discuss the challenges social inequalities pose to machine learning models, provide several recommendations for adopting ethical principles for the delivery of machine learning-assisted health care.
Carayon P, Hoonakker P, Hundt AS, et al. BMJ Qual Saf. 2020;29:329-340.
This simulation study assessed whether integrating human factors engineering into a clinical decision support system can improve the diagnosis of pulmonary embolism (PE) in the ED. Authors found that this approach can improve the PE diagnostic process by saving time, reducing perceived workload and improving physician satisfaction with the technology.
Rodriguez-Ruiz A, Lång K, Gubern-Merida A, et al. J Natl Cancer Inst. 2019;111:916-922.
Artificial intelligence (AI) may have the capacity to improve diagnosis. Researchers found that an AI system was able to detect breast cancer using mammography with accuracy similar to that of the average of the 101 radiologists whose interpretations were included in the study.
Massalha S, Clarkin O, Thornhill R, et al. Can J Cardiol. 2018;34:827-838.
Decision support tools can help reduce diagnostic uncertainty. Discussing how artificial intelligence can be utilized to inform diagnostic decision making and improve the accuracy of cardiac image interpretation, this review suggests that use of such technology can reduce production pressure and cognitive load for imaging physicians.
Meyer AND, Thompson PJ, Khanna A, et al. J Am Med Inform Assoc. 2018;25:841-847.
Clinical decision support is a widely recommended patient safety strategy. This study examined whether a mobile application created by the Centers for Disease Control and Prevention improved clinician decision-making about anticoagulation test ordering for simulated case vignettes. Each participating physician completed a series of vignettes; half used the application and half did not. When using the application, physicians demonstrated greater diagnostic accuracy and confidence, and they needed less time to complete each vignette. The authors suggest that mobile applications may be useful for providing decision support.
Auerbach AD, Neinstein A, Khanna R. Ann Intern Med. 2018;168:733-734.
Digital tools have the potential to improve diagnosis, patient self-care, and patient–clinician communication. This commentary argues that digital tools that alter diagnosis or treatment require examination to ensure safety. The authors provide recommendations such as involving experts in evaluating the tools, engaging information technologists, and continuous local review and assessment to identify and address risks associated with use of such tools in practice.
Liberman AL, Newman-Toker DE. BMJ Qual Saf. 2018;27:557-566.
Patient safety measurement remains challenging. This article describes a framework to address gaps in measuring diagnostic error. The authors propose utilizing big data to develop diagnostic performance dashboards and benchmarking tools that support proactive learning and improvement strategies.
Bejnordi BE, Veta M, van Diest PJ, et al. JAMA. 2017;318:2199-2210.
Diagnostic error is a growing area of focus within patient safety. Artificial intelligence has the potential to improve the diagnostic process, both in terms of accuracy and efficiency. In this study, investigators compared the use of automated deep learning algorithms for detecting metastatic disease in stained tissue sections of lymph nodes of women with breast cancer to pathologists' diagnoses. The algorithms were developed by researchers as part of a competition and their performance was assessed on a test set of 129 slides, 49 with metastatic disease and 80 without. A panel of 11 pathologists evaluated the same slides with a 2-hour time limit and one pathologist evaluated the slides without any time constraints. The authors conclude that some of the algorithms demonstrated better diagnostic performance than the pathologists did, but they suggest that further testing in a clinical setting is warranted. An accompanying editorial discusses the potential of artificial intelligence in health care.
Brush JE, Brophy JM. JAMA Intern Med. 2017;177:1245-1246.
Improving diagnosis has been recently promoted as a core area in patient safety. This commentary discusses how physicians can utilize cognitive decision making and likelihood ratios to address uncertainty and augment diagnosis. The authors suggest that clinicians learn to assess their diagnostic practices and skills to improve their performance.
Stafos A, Stark S, Barbay K, et al. Am J Nurs. 2017;117:26-31.
This study compared nurses' identification of patients at risk for harm to an electronic predictive model and found that nurses more commonly identified psychological or social risks as relevant to harm. The nurses did not identify some patients whom the predictive model deemed high risk in cases where the risk had been incorporated into the plan of care. The authors suggest that nurse perceptions could inform more accurate predictive models, though neither approach was tested against an actual safety outcome.
Cahan A, Cimino JJ. J Med Internet Res. 2017;19:e54.
Although advanced computing can assist in diagnosis, these systems are not routinely utilized. This commentary suggests a framework to develop diagnostic support technologies that capture physician knowledge to enhance diagnostic safety. The authors encourage drawing from crowdsourced data to guide improvements at a system level to address future practice and educational needs.
Information technology approaches have been advocated as a means of preventing diagnostic error. This study compared the diagnostic accuracy of computerized symptom checkers (software programs that use diagnostic algorithms based on patients' self-reported symptoms to suggest diagnoses) with that of practicing physicians. Physicians consistently arrived at more accurate diagnoses across a variety of simulated cases.
Amland RC, Hahn-Cover KE. Am J Med Qual. 2016;31:103-10.
Sepsis is a clinical condition that can be rapidly fatal, thus prompt recognition and treatment is critical. This multicenter retrospective study describes the performance of a cloud-based computerized decision support system aimed at identifying sepsis in patients before infection was suspected.
Riches N, Panagioti M, Alam R, et al. PLoS One. 2016;11:e0148991.
Despite increasing focus on diagnostic error, it remains a controversial patient safety issue. The Institute of Medicine recently suggested that further research is needed regarding electronic tools to improve diagnosis. Differential diagnosis generators provide a list of possible diagnoses for a problem. The investigators conducted a systematic review and found that differential diagnosis generators have been shown to improve diagnostic accuracy when a clinician has an opportunity to re-review the case using the software in pre-post studies. The degree of improvement varied between studies. The effect on actual clinician behaviors—such as test ordering, clinical outcomes, and cost—is unclear. Clinicians need prospective studies in order to determine whether such tools enhance diagnosis in actual practice. A recent PSNet perspective discussed future research avenues to ensure progress in diagnostic safety.
Jutel A, Lupton D. Diagnosis (Berl). 2015;2:89-96.
This study examined currently available smartphone and software applications (or apps) designed to aid with accurate diagnosis. Although the authors described some of the potential benefits of these apps, they note that their research suggests apps should be used with caution by both clinicians and consumers, due to problems with transparency regarding sources, evidence, and credentials.
Nishikawa RM, Schmidt RA, Linver MN, et al. American Journal of Roentgenology. 2012;198.
A computerized clinical decision support system helped radiologists reduce diagnostic errors in mammogram interpretation. However, radiologists ignored more than two-thirds of the prompts provided by the system.
Murff HJ, FitzHenry F, Matheny ME, et al. JAMA. 2011;306:848-55.
Many adverse event identification methods cannot detect errors until well after the event has occurred, as they rely on screening administrative data or review of the entire chart after discharge. Electronic medical records (EMRs) offer several potential patient safety advantages, such as decision support for averting medication or diagnostic errors. This study, conducted in the Veterans Affairs system, reports on the successful development of algorithms for screening clinicians' notes within EMRs to detect postoperative complications. The algorithms accurately identified a range of postoperative adverse events, with a lower false negative rate than the Patient Safety Indicators. As the accompanying editorial notes, these results extend the patient safety possibilities of EMRs to potentially allow for real time identification of adverse events.
Please select your preferred way to submit a case. Note that even if you have an account, you can still choose to submit a case as a guest. And if you do choose to submit as a logged-in user, your name will not be publicly associated with the case. Learn more information here.