By Richard R. Watkins, MD, MS, FACP, FIDSA, FISAC
SYNOPSIS: A national retrospective cohort study from the Veterans Administration found a high rate of diagnostic discordance for patients admitted and discharged for community-acquired pneumonia (CAP). Improvement in the diagnosis of CAP is needed.
SOURCE: Jones BE, Chapman AB, Ying J, et al. Diagnostic discordance, uncertainty, and treatment ambiguity in community-acquired pneumonia: A national cohort study of 115 U.S. Veterans Affairs Hospitals. Ann Intern Med 2024;177:1179-1189.
While community-acquired pneumonia (CAP) remains very common in clinical practice, diagnosing it often is difficult for clinicians. Indeed, there is significant overlap between the signs and symptoms of CAP and other frequently encountered conditions, such as exacerbations of congestive heart failure (CHF) and chronic obstructive pulmonary disease (COPD). Furthermore, because of the current fragmented care structure, clinicians who make a diagnosis of CAP rarely learn about the ultimate diagnosis or treatment decisions. Jones and colleagues aimed to better characterize the diagnosis of CAP and examine the factors that lead to diagnostic discordance.
The study was a large retrospective cohort study from 118 Veterans Affairs (VA) medical centers that included adult patients 18 years of age and older who were diagnosed with CAP between January 2015 and January 2022. The investigators previously had developed a natural language approach (NLP) for the diagnosis of pneumonia with an accuracy of 0.90. They employed this NLP, diagnostic codes in the emergency department (ED) and at time of discharge, radiologic findings based on chest radiographs and computed tomography (CT), and the order or administration of antibiotics or respiratory-directed antivirals to evaluate for diagnostic discordance at the time of admission and discharge. Phrases of diagnostic uncertainty (e.g., possible pneumonia) also were identified.
There were 317,437 patients included in the analysis. Discordance between the ED diagnosis of CAP and the discharge diagnosis occurred in 179,834 (56.7%) cases. Among those diagnosed with CAP initially, 36% lacked a discharge diagnosis and 21% lacked positive initial radiology findings. In 33% of patients, there was a discharge diagnosis of CAP but not an ED diagnosis. Patients with diagnostic discordance were more likely to present to high-complexity facilities with high ED and inpatient volumes.
Uncertainty about the diagnosis of CAP was found in 58% of clinical notes from the ED and in 48% at discharge. This resulted in 27% of patients with CAP being treated with diuretics; 36% receiving corticosteroids; and 10% receiving antibiotics, steroids, and diuretics within 24 hours. Patients with diagnostic concordance were more likely to present with fever, an abnormal leukocyte count, an elevated C-reactive protein, and an elevated procalcitonin. Fewer microbiology cultures were obtained in cases with discordance, but the percentage of positive cultures was similar for all tested patients.
The most common sources of clinician error in discordant cases were interpreting chest imaging as positive when it was not, failing to see radiology reports, and failing to appreciate the presence or absence of signs and symptoms of infection. Patients with CAP who lacked an initial diagnosis had a greater 30-day mortality rate (14.4%) compared to those with concordant diagnoses (10.6%). Furthermore, those with concordance had a lower 30-day readmission rate (15.9%) than patients with a negative diagnosis in the ED, positive imaging, and a discharge diagnosis of CAP (18.6%).
Commentary
The ambiguousness of respiratory symptoms such as cough and dyspnea often leads to misdiagnosis and inappropriate care. Indeed, the fact that 10% of patients in the current study received steroids, antibiotics, and diuretics within 24 hours of hospitalization is concerning. Clinicians need to do better. Given that CAP is such a common condition, overprescribing antibiotics for CAP is an important contributor to the ongoing spread of antimicrobial resistance.1 Diagnosing CAP is an example of clinical equipoise in which the tradeoffs from overdiagnosis vs. underdiagnosis are uncertain and need further elucidation. The study by Jones and colleagues provides evidence that the initial diagnosis of CAP is a complicated process with significant uncertainty that may not have been appreciated in previous analyses. Notably, many of these have been used to formulate current clinical guidelines.
There is not much mystery as to why the diagnosis of CAP frequently is wrong. Chest radiographs are discordant with chest CTs in CAP in approximately one-third of cases. Furthermore, the inter-observer agreement on chest radiographs for the presence or absence of pneumonia has been reported to be below 90%, even for radiologists. One potential solution is to increase the use of chest CT. Although initially this seems appealing, CTs are more expensive than chest radiographs, and such a policy likely would increase financial strain on hospitals. Microbiological and rapid antigen testing can be helpful in diagnosing CAP, especially in ambiguous cases. But for tests to be useful and improve outcomes, someone needs to follow up the results, which may not be the clinicians who order them. Thus, there is a significant risk for test results to be missed.
The study had some limitations. First, because it was conducted at the VA, the results might not be generalizable to other patient populations. Second, the retrospective design may have led to unmeasured confounding variables to affect the results. Third, the investigators did not take into account the possibility of patients having more than one diagnosis, such as both pneumonia and an exacerbation of heart failure.
Accurately diagnosing and treating CAP is crucial for both quality and patient safety standards. Providing feedback to clinicians who initially diagnose CAP about the final diagnosis at discharge is a low-cost intervention that potentially could improve outcomes. Additional research on this approach is warranted.
Reference
- Antoñanzas F, Juárez-Castelló CA, Rodríguez-Ibeas R. Does diagnostic testing always decrease antibiotics prescriptions? Eur J Health Econ 2023;24:673-678.
Richard Watkins, MD, is Professor of Medicine, Division of Infectious Diseases, Northeast Ohio Medical University, Rootstown, OH.
