Findings from this review suggest scientific progress in nearly all the diagnostic domains but progress has varied across domains. Certain domains, such as incidence, measurement, and health IT use, have experienced significant scientific progress. However, application of the research to clinical practice and operational safety improvement is lagging. Research is inadequate on cross-cutting systems and processes of care issues related to diagnostic safety, as well as research focused on closing implementation gaps.
We identified several gaps related to both science and practice that need progress. For instance, specific gaps identified in the measurement data and methods domain included low rates of operationalizing reporting tools clinicians can use and limited monitoring of diagnostic error incidence in actual practice. Clinicians have a lot of information about diagnostic errors that can be leveraged by organizations thinking of data-based improvement strategies.
We also found inadequate use of electronic data (such as data available in EHR repositories) to both measure and improve diagnostic safety. Another domain that needs further development was that of cognition, where we found that studies in the field of cognitive processes were mostly experimental and missing real-world context.
In addition, we found gaps in the culture, workflow, and work system issues domain, where studies related to how culture, behavior, and work system influence diagnostic safety were limited. Data on disparities was limited but suggested the need to address diagnostic equity and ensure collection of segmented data to see how various underrepresented groups are affected.
Progress in certain domains, such as patient/family engagement and interventions, was mostly from published perspectives and thought pieces with fewer empirical or evaluation studies. The domains of intervention and implementation are still emerging, with inadequate ways to close the implementation research to practice gaps. Real-world studies that involved health systems as partners for learning and improvement activities are also lacking. Gaps in the health IT domain included limited application or effectiveness studies of computer algorithms and decision support to specifically augment diagnostic accuracy.
The domains were not mutually exclusive, and we found several overlapping themes that cut across the various domains. This overlap underscores the need to develop a more multidisciplinary approach to diagnostic safety that would cut across domains. For example, while a body of literature exists related to cognitive science, we did not find it very well integrated into clinical practice or with other fields. Similarly, while emerging use of health IT is improving diagnosis, we did not find adequate studies that used human factors and cognitive science approaches to optimize cognitive support to clinicians and to inform technology design.
In addition, there has been rapid development of resources and tools while implementation work that focuses on strengthening systems and processes of care is lacking. We also found minimal use of learning health systems approaches and embedded research models that could accelerate practice transformation. Our findings suggest the need to approach diagnostic safety research from multiple lenses and use of multidisciplinary scientific teams to accelerate progress.
To promote broad-scale improvement in diagnostic safety, we recommend research efforts to address the identified gaps within each domain. We also recommend approaches for better dissemination of this work. For example, engaging boards of different specialties and their associated professional organizations to promote diagnostic safety in a more consistent and uniform way could lead to better uptake of emerging research findings.
Another important step would be to engage policymakers who could promote uptake of resources that have been developed and consider policy and payment reforms to incentivize health systems to promote diagnostic safety. As AI becomes more integrated into clinical care and population health, studies are needed to observe how clinicians interact with generative AI to formulate a correct and timely clinical diagnosis, including asking patients the right questions.
This issue brief has several limitations. We used a rapid narrative review methodology to scan diagnostic safety materials that was not systematic, so we likely missed some publications. We addressed this limitation by performing a second search with the domains and subdomains as search terms as well as incorporating input from experts.
Another limitation was including search materials mostly since 2013 onward, so we missed certain key items published just prior to 2013. However, our focus was the current state of diagnostic safety and because many of these publications built on prior work and because we included external experts, this choice had minimal impact on findings. The 10 domains had overlap, so certain papers ideally belonged in more than one category. Finally, the field of diagnostic safety is broad and emerging, so this issue brief may not cover all areas.
We summarized the state of the science of diagnostic safety for the past decade. Despite progress in various domains of diagnostic safety, several research gaps remain. Our findings and recommendations have implications for future investments and research funding for diagnostic safety.
