Pneumonia Diagnoses Largely Inaccurate?

Over half of inpatient pneumonia diagnoses changed from the emergency department (ED) to discharge, and uncertainty was more common than not in clinical notes, according to a nationwide retrospective analysis of Veterans Affairs (VA) data.

Among over 317,000 hospitalizations for pneumonia, 57% had discordance between initial and discharge diagnoses, and 58% of patient records examined indicated uncertainty in diagnosis, Barbara Jones, MD, MS, of the Salt Lake City VA Healthcare System and the University of Utah there, and colleagues reported in the Annals of Internal Medicine.

“We should anticipate diagnostic discordances and uncertainty in pneumonia,” the authors wrote. “Clinicians often encounter pressure to provide distinct diagnostic labels to facilitate hospital admission, simplify handoffs, or satisfy patients.”

Of patients discharged with a positive initial chest image and a diagnosis of pneumonia, 33.4% never received an initial diagnosis in the ED. And of those who did have an initial diagnosis of pneumonia, 36.3% did not have it as a discharge diagnosis and 21% lacked positive initial chest imaging.

While the study couldn’t attribute causality, mortality differed between these groups. Pneumonia diagnosed by discharge but not initially had a 14.4% 30-day mortality rate compared with 10.6% when it was recorded on both initial and discharge diagnoses.

“Pneumonia can be a difficult diagnosis,” Mark Metersky, MD, of the University of Connecticut School of Medicine in Farmington, and Grant Waterer, MBBS, PhD, of the University of Western Australia and Curtin University in Perth, commented in an accompanying editorial. “Very often at the time of admission, we do not know if the patient has pneumonia, an exacerbation of chronic obstructive pulmonary disease, heart failure, or a combination of these conditions,” they pointed out.

“We often treat for 2 or even all 3 diagnoses,” the editorialists wrote. “The patient almost always improves and leaves the hospital with ongoing uncertainty about the correct diagnosis.” And, in many cases, discordant diagnoses might not be inaccurate in patients with multiple conditions, they pointed out.

“There is little mystery about why the diagnosis of pneumonia is often inaccurate,” Metersky and Waterer commented, noting chest radiographs frequently lack inter-observer agreement on the presence or absence of pneumonia, even among radiologists.

The most common sources of diagnostic error in the analysis were misinterpretation of chest imaging as positive when it was negative, failure to see chest image reports, and failure to appreciate whether signs of infection were present or absent.

When researchers evaluated 50 patient charts, they found that 58% of ED notes and 49% of discharge notes expressed diagnostic uncertainty. Chart review also revealed that about 90% of patients received antibiotics, 27% received diuretics, and 36% received corticosteroids in the first 24 hours of admission. About 10% received treatment with all three medications. Patients with discordant diagnoses had greater uncertainty documented in clinical notes and received more treatments than those with concordant diagnoses.

“These findings highlight substantial diagnostic uncertainty and treatment ambiguity in pneumonia that warrants recognition by systems, clinicians, and patients,” Jones and colleagues wrote. The editorialists emphasized that “[c]linical pathways should support diagnostic uncertainty as a productive component of clinical reasoning.”

Patients with diagnostic discordance had slightly more renal disease and heart failure, ED stays of longer than 8 hours, and treatment with both corticosteroids and diuretics than those with concordance. Those with concordant diagnoses were more likely to present with fever, abnormal white blood cell count, and elevated C-reactive protein and procalcitonin.

Patients with discordance were also more likely to have less severe illness and lower levels of inflammatory markers. However, they underwent substantially fewer CT scans than patients with concordant diagnoses — 7.4% versus 33.2%, respectively. Patients with discordance also had fewer microbiology tests performed, but the percentage of positive cultures were similar between the two groups.

Researchers also found variations in diagnostic discordance between individual hospitals, noting that it was more common at high-complexity facilities with high patient loads.

The analysis included over 2.4 million patient encounters at EDs that resulted in hospitalizations in the VA healthcare system. In total, 13.3% of encounters involved a diagnosis of pneumonia at the ED visit or by discharge, with 9.1% receiving an initial diagnosis and 10% getting a discharge diagnosis.

Researchers identified pneumonia diagnoses using diagnostic codes assigned in the ED or at hospital discharge and based on notes from the ED, clinician, discharge, and radiology that identified a pneumonia diagnosis, as well as orders for or administration of antibiotics or antiviral medications. Using natural language processing tools, Jones and colleagues also looked for evidence of uncertainty in a subset of 50 randomly selected patient records — for example, terms such as “possible pneumonia.”

The mean age of the pneumonia patients was about 72 years, with about three-quarters reporting white race and most being male, reflecting the VA population.

The analysis had some limitations, the authors pointed out. Since the study relied on retrospective data, some discordance may have been due to coding or documentation errors. Also, the study did not address confounding factors and could not establish causal effects of diagnostic discordance on clinical outcomes.

Please enable JavaScript to view the

comments powered by Disqus.