Diagnostic Error: How Big a Deal Is It?
V. Katherine Gray, PhD, Sage Health
02/05/2019
Most Americans have heard about medication errors that can cause harm or death, but an even bigger issue is diagnostic errors causing harm or death. The original publication from the Institute of Medicine on medical errors (To Err is Human in 1999) provided data on medication errors rate at 11% of the deaths from medical errors, which were estimated at 99,000 deaths per year. These results were impressive, and many healthcare professionals began to focus on solutions to the problem of medication errors. However, the irony is that there were many more deaths from errors of diagnosis than from medications. In fact, diagnostic errors represented 17% of the deathms in the 1999 data. Shockingly, it took another 15 years for the Institute of Medicine to recognize and highlight the problem of deaths from diagnostic errors.
Even if the diagnostic error doesn’t result in death, currently equating to 10% of the medical errors, there are many other negative outcomes in healthcare that are of great concern and significant cost (Improving Diagnosis in Healthcare, 2015). For example, 6-17% of the adverse events in hospitals are due to diagnostic errors. Diagnostic errors are also the leading type of paid malpractice claims, as well as representing twice as many patients who die than other malpractice claims. Finally, the prevalence rate of having a diagnostic error is 5% of all U.S. adults in outpatient care. According to the data from the CDC using all ambulatory visits in 2015 of 841.3 million (including physician office, hospital outpatient and emergency department visits) more than 42 million patients have diagnostic errors each year. Therefore diagnostic errors are a common experience for patients.
What can be done to improve the situation? There are several “big ideas” for how to improve our diagnostic process. First, examine the ordering process and support it with evidence-based clinical decision support. Secondly make the workflow for diagnosing a “learning” system where health care professionals can review whether the diagnosis was correct or not and what might have caused the error. These pieces of information can assist in improving the clinical decision support tools. Finally, use data to help with these quality improvement tasks. In this case, it is by looking at the forest rather than “tree by tree” that will give your organization maximum efficiency in improving the diagnostic process.
Some health care organizations have tried “prior authorization” to change the imaging utilization, but in general it has been a negative experience for the ordering providers due to the administrative burden that it places on the provider to get procedures ordered for their patients. The American Medical Association (AMA) has stated that they believe that prior authorization is overused, and existing processes present significant administrative and clinical concerns (AMA, 2019). In addition, prior authorization does not assist the provider in selecting or learning about the most appropriate procedure given a set of patient criteria; thus, prior authorization never tries to educate the providers, it only tries to stop providers from using services. In fact, our research has shown that since the focus of prior authorization is on stopping an order, that once the patient has many indications, the prior authorization will agree to whatever the provider ordered—even if it is not correct as that is not their focus. Our data show that following imaging recommendations in imaging CDS will result in highly significant improvement in getting to the correct diagnosis, as well as minimize inappropriate resource utilization.
A key innovation (Sage Ox®) developed by our organization improves the imaging Clinical Decision Support recommendations by using Natural Language Processing (NLP) or artificial intelligence (AI) to improve the diagnostic process. Sage Ox analyzes the radiologists’ reports of the imaging procedures to categorize the working diagnosis as being either Positive or Negative compared to the reason for the exam. By having such a single “finding” for each diagnosis, it is much easier to evaluate other aspects of the diagnostic process to understand what variables might be causing errors in ordering and what variables may be recognized as valuable for making a correct diagnosis. An example would be the working diagnosis is pneumonia, but the radiology report finds no pneumonia, but find a lesion consistent with lung cancer. Therefore the finding are positive but for a different diagnosis.
The Center for Medicare & Medicaid Services (CMS) is taking a major step toward requiring the use of Clinical Decision Support in part to reduce unnecessary imaging procedures (PAMA, 2014), but the benefit of this new requirement, which will take effect initially in 2020, will be to produce large amounts of data to focus the health care community on the diagnostic process. For the first time in the United States, we will have a national database of what ordering professionals consider the reason for an exam or a suspected or working diagnosis when they order an advanced imaging procedure (i.e. MRI, CT, PET, etc.) for a Medicare patient. These data will allow the inspection of trends to see where improvements can be made, and ultimately begin to standardize our diagnostic process to improve the quality of care for patients.
An article in JAMA Online (January 7, 2019) recently complained about the waste and possible harm resulting from incidental findings on imaging. The problem posed by such incidental findings it the uncertainty surrounding their predictive value whether additional follow-up testing and treatment is warranted/required. Much more data collection is needed on the incidentals before we know whether or not additional testing/treatment is wasteful or beneficial. At present, we generally respond to everything because we lack reliable data on the predictive value of the incidentals.
As the National Academies (2015) have suggested we must dedicate more funding for diagnostic research, not just treatment. In the future, if the U.S. doesn’t focus on diagnosis improvement, as disease and care increases in complexity the number of diagnostic errors will increase as well. If patients and families understand this, they will lose trust in the quality and expertise of our health systems and our management of care. Right now, patients and caregivers rate “Diagnostic Accuracy” as the number one priority for health plans, and professionals rate it at the second highest priority (Health Services Advisory Group, 2017).
As a health plan, you can look at the current requirement of PAMA imaging CDS as a gift to use for improving diagnostic processes. There are processes in place (as every CMS-qualified CDS Mechanism is standardized) that produce data to be used to review providers and processes. In addition, there are incentives for organizations who participate in the use of the imaging CDS as part of the Quality Payment Program (QPP), so even the finances favor focusing on diagnosis. The outcomes are also labeled by CMS as adherent, non-adherent, or not applicable (i.e. no content available). Therefore, these provide your organization with valuable “outcomes data.”
The outcome data can be combined with information such as the “Findings” codes from a tool like Sage Ox or other analytical tools, to review how the various diagnoses, procedures, workflows, and findings impact the quality of diagnosis and care. Furthermore, predictive items such as patient demographics, provider type, place of service, and specific symptoms can more clearly improve the diagnostic process. The goal is to increase the diagnostic accuracy and lower the costs for inappropriate or unnecessary care, which includes imaging as well as treatment based on the “findings.” As a benefit to your health plan and participating providers, there is also the possibility of using these indicators for incentives as part of your contracting for quality improvements.
As a parent, my son was diagnosed with a blood-clotting disorder and for a year he had weekly treatment with some very negative side-effects. Due to the negative side-effects, I took him off the treatment as he had never exhibited specific symptoms of the disorder. Later I found that the testing that had been done would produce a positive result, even if he did not have the disorder. Therefore, consider your patients and caregivers and implore your organization to initiate a program of improving the diagnostic process in your organization. It will mean a lot to your patients and their families. Treat the PAMA requirements as a gift to improve the diagnostic process!
References
American Medical Association, (2019). Consensus Statement on Improving the Prior Authorization Process. Available at https://www.ama-assn.org/practice-management/sustainability/prior-authorization-reform-initiatives
Balogh, E., Miller, B.T., and Ball, J.R. (eds.) (2015). Improving Diagnosis in Health Care. Washington, DC: National Academies Press.
Health Services Advisory Group, (February 17, 2017). CMS Quality Measure Development Plan Technical Expert Panel Meeting Summary (MACRA Section 102), page 9.
Kohn, L.T., Corrigan, J.M., and Donaldson, M.S. (eds.), (November 1999). To Err Is Human: Building a Safer Health System.
Protecting Access to Medicare Act (PAMA) (2014). Section 1834A of the Act, as established by Section 216.