High Diagnostic Uncertainty and Inaccuracy in Adult Emergency Department Patients With Dyspnea: A National Database Analysis.

Abstract

Dyspnea is the second leading cause for visit among the 20 million annual U.S. emergency department (ED) visits in patients aged ≥65 years old and is most commonly caused by pneumonia, acute exacerbations of chronic obstructive pulmonary disease (COPD), and heart failure (HF) in older adults.1, 2 Despite its frequency and substantial associated morbidity and mortality,3 diagnosing the cause of dyspnea in an older adult poses substantial challenges to the emergency physician.2, 4-8 Approximately 20% of older adults with dyspnea have a diagnosis missed in the ED2 and 31% of adult ED patients with dyspnea are treated for more than one of the common causes of dyspnea (pneumonia, COPD, and HF).8 ED physician diagnostic uncertainty is common,8 manifesting either as treatment for multiple conditions (cotreatment) or as documentation of multiple primary diagnoses (codiagnosis). ED diagnostic uncertainty is associated with increased admission rates, longer hospital lengths of stay, and increases in death or rehospitalization within 1 year.8 We sought to characterize the population of patients presenting to U.S. EDs with dyspnea. We describe the rates of cotreatment, codiagnosis, and diagnostic uncertainty, stratified by age and focusing on those with pneumonia, COPD, and HF. The National Hospital Ambulatory Medical Care Survey (NHAMCS) is conducted annually using previously described methods to describe ambulatory emergency care at U.S. hospitals.9 Data from calendar years 2010 to 2014, the most recent available, were included. This study was exempt from institutional review board review. Visits by individuals aged ≥ 18 years old with a complaint of dyspnea were identified using NHAMCS reason for visit codes (1415.0, 1420.0, 1425.0, 1430.0, 1430.1, and 1403.2). To allow comparison with previous literature4 and exclude patients who had clear etiologies that can also cause dyspnea (e.g., atrial fibrillation with rapid ventricular response), analyses of cotreatment, codiagnosis, and diagnostic uncertainty were limited to the subset of patients with ED diagnoses of pneumonia, COPD, or HF. Treatments for pneumonia, COPD, and HF were determined based on ED medications administered that were distinct for one of these conditions using the drug categories in NHAMCS and following prior work.4 Pneumonia treatment included penicillin, cephalosporin, fluoroquinolone, macrolide, vancomycin, tetracycline, aminoglycoside, or carbapenem antibiotics. Treatment for COPD included glucocorticoids, and for HF, loop diuretics, vasodilators, or positive inotropes.4, 9 ED and hospital diagnoses of pneumonia, COPD, HF, and dyspnea not otherwise specified (NOS) were defined by ICD-9-CM codes.4 The primary outcomes were the proportion of ED visits with codiagnosis, cotreatment (treatment or diagnosis for one or more of pneumonia, COPD, and HF) and diagnostic uncertainty (codiagnosis, cotreatment, or a lone diagnosis of dyspnea NOS). We included treatment as well as diagnosis, as ED documentation of diagnoses can be incomplete and may not represent whether the treating physician felt a condition was present. In admitted patients, we describe the proportion of agreement between ED and hospital diagnoses. Descriptive statistics were calculated, stratified by age ≥ 65 years. Confidence intervals and p-values are not reported as statistical significance would not correlate with clinical significance given large weighted sample sizes. Standard NHAMCS weighting procedures were utilized to obtain nationally representative estimates.9 Data management was conducted using SAS 9.4 (SAS Institute, Inc.) and STATA 15 (StataCorp). From 2010 to 2014, the NHAMCS contained 34,832,195 weighted visits by adults with dyspnea characteristics of the study population are shown in Table 1. Older adults had more comorbidities (including HF) and higher admission rates. Among groups of ICD-9 diagnoses, older adults had greater proportions of heart disease (26% vs. 8%) and pneumonia (15% vs 8%) with less acute respiratory infection (<5% vs. 8%). There were similar proportions of COPD and allied conditions (27 and 31%), but there was a shift from an asthma predominance in younger to COPD in older adults. Using specific ICD-9 codes, of all ED visits by patients with dyspnea, 10% were diagnosed with pneumonia, 13% with COPD, and 10% with HF with all diagnoses more common in older adults. Across all age groups, dyspnea NOS was a common diagnosis (35%) and was the only diagnosis in 28% of visits. Codiagnosis occurred in 3.5% of the entire study population and cotreatment in 10%, but both were more common in older than younger adults (Table 1). Almost half (n = 6,021,403, 47%) of older adults were diagnosed with pneumonia, COPD, and/or HF versus 20% (n = 4,324,921) of younger adults. For patients with one of these three diagnoses, 68% of all adult visits diagnosed with pneumonia were treated, 54% for COPD and 64% for HF. Younger adults were less likely to be treated for pneumonia (61% vs.74%) but more likely for COPD (56% vs. 53%) and HF (67% vs. 44%). In this subset of patients, cotreatment rates (21% across the ages) were similar between older and young adults as was sole diagnosis of dyspnea NOS (24%) but codiagnosis was higher (13% vs. 7.9%). In this subset, diagnostic uncertainty was present in almost half (45%) with similar rates between younger and older adults. Among hospitalized patients with one of these three diagnoses, hospital diagnosis agreement with ED diagnosis of pneumonia, COPD, and HF was low. Overall, 47% of ED pneumonia diagnoses had a hospital discharge diagnosis of pneumonia. Rates were 56% for COPD and 54% for HF. Diagnosis agreement was similar in younger and older adults. Diagnostic uncertainty in older ED patients with dyspnea is associated with increased admission rates, longer hospital lengths of stay, and increased death or rehospitalization within 1 year.8 In addition, ED treatment directed toward an incorrect diagnosis of dyspnea increased mortality from 11% to 25% in older adults with respiratory failure2 with similar results in the inpatient setting.4 Overall, older adults with dyspnea (all diagnoses) had greater rates of codiagnosis (6% vs. 2%) and cotreatment (15% vs. 6%) than younger adults. Older adults are particularly susceptible to adverse effects of these issues due to their much higher rates of diagnosis with pneumonia, COPD, and/or HF than younger adults, particularly given their proven susceptibility and poor outcomes in the setting of uncertainty and inappropriate treatment.2, 4, 8 Although older and younger adults with these three diagnoses had similar rates of codiagnosis, cotreatment, and diagnostic uncertainty, the much larger proportion of older adults with one of the three diagnoses makes the problem particularly important in this population. Therefore, improving diagnostic accuracy and, as a result, appropriate treatment could have a large impact on morbidity and mortality among the half of older ED patients with dyspnea diagnosed with pneumonia, COPD, and HF. The high rate of ED codiagnosis for these three conditions indicates a need to improve diagnostic accuracy. This is further supported by the poor agreement between ED and inpatient diagnoses although the reason for the discrepancy is unclear. With more information or further differentiation of the patient's presentation, the inpatient physicians could be narrowing the diagnosis. Alternatively, the ED diagnosis fields could be inaccurate or incomplete leading to underestimate of codiagnosis. Emergency department cotreatment in the three conditions was common in patients of all ages. This could demonstrate that when diagnostic uncertainty exists, physicians treat for several etiologies in the ED. Our estimates of cotreatment in older adults (pneumonia 29%, COPD 29%, HF 12%) are almost identical for pneumonia, higher for COPD, and much lower for HF when compared with the study by Dharmarajan et al.4 study of older inpatients (pneumonia 32%, COPD 19%, HF 38%). This may reflect differences between an ED and inpatient population where some ED patients are discharged, ED stays are short, or fewer treatments are provided in the ED. Patients could also develop additional diagnoses in the inpatient setting. Data are not available for outcome comparison to previous work, but it is known that inappropriate treatment for the etiology of dyspnea increases mortality in both the ED2 and inpatient settings.4 A prospective study would better study these outcomes in dyspneic ED patients. Emergency department diagnostic uncertainty in dyspnea is likely multifactorial with patient, physician, and diagnostic factors. In older adults, diagnosis is complicated by atypical presentations,10-14 decreased sensation of dyspnea,15, 16 multiple comorbidities,17 failure of clinical prediction rules,18-20 and decreased biomarker accuracy.21 As a result, diagnosing the etiology of dyspnea is challenging for emergency physicians.2, 6-8, 22 These contributing factors could not be examined. Future studies should use a prospective cohort where data can be collected directly from physicians. To improve diagnostic uncertainty and unnecessary cotreatment of ED patients with dyspnea, we need better rapid tests for diagnoses contributing to dyspnea such as biomarkers and algorithms specific to the older adult population. Our approach has limitations. First, the NHAMCS data set has recognized limitations.23-25 Rather than admitting hospital diagnosis, NHAMCS only includes hospital discharge diagnoses. Therefore, we may have overestimated the ED and hospital diagnosis disagreement; however, given our high disagreement rate and previous literature, disagreement is likely to exist.2, 7, 8 Older adults frequently present to the ED with dyspnea and suffer from increased rates of codiagnosis and cotreatment than younger adults overall. Half of older adults with dyspnea are diagnosed with pneumonia, chronic obstructive pulmonary disease, or heart failure (vs. 20% of younger adults) with high rates of codiagnosis, cotreatment, and diagnostic uncertainty. As ED care is known to profoundly impact subsequent care, improved ED diagnostic accuracy is necessary to improve morbidity and mortality.