Introduction

Chronic obstructive pulmonary disease (COPD) is a common, chronic lung disease characterized by progressive and not fully reversible airflow limitation [1]. Persistent worsening in patients’ baseline symptoms of dyspnea, cough and sputum production (quantity and purulence) are required to define an acute exacerbation of COPD (AECOPD). As the disease progresses, exacerbations become more frequent and severe [2]. Exacerbations often result in emergency department (ED) presentations and when severe, require hospitalization. In fact, in patients with COPD, the disease itself is the single most important determinant of overall health outcomes [3], as it is associated with increased morbidity, substantial ED visits, hospital admission and death [4–6].

Exacerbations occur due to a variety of reasons including, but not limited to viral and bacterial infections, environmental pollution, and medication non-adherence. In addition, patients with AECOPD can have complications such as heart failure, pneumonia, arrythmias and pulmonary embolism (PE). The standard ED management of COPD exacerbations consists of complete blood count, electrolyte and renal function panels, simple chest radiography, electrocardiography, and selective cardiorespiratory biomarkers (e.g., D-dimer, troponin, and/or brain natriuretic peptide [BNP]). Debate exists over the frequency of PE in this patient group and how to investigate for the condition in the ED management of AECOPD [7].

A recent systematic review and meta-analysis found that approximately 12% (95% confidence interval [CI]: 9% to 16%) of patients with AECOPD had an underlying PE [7]. Among the included studies, only one study included patients from the ED and found a lower prevalence of PE during AECOPD (3.3%) [8]; however, other ED-based studies have identified over-testing for PE and low PE rates in ED patients [9]. Reviews have showed that the prevalence of PE was significantly higher among patients who required hospitalization for AECOPD; however, the prevalence was also highly variable, ranging from 2% to 33% across studies. This variability raises concerns about potentially missed cases of PE and suggests that clinicians should consider the possibility of PE in patients presenting to the ED with AECOPD.

Most currently available studies may not reflect the actual prevalence of PE during AECOPD in the ED setting [7,10,11]. Additionally, little is known about how often patients experiencing an AECOPD are subjected to PE workups, particularly using pre-test probability scales, appropriate biomarkers, and advanced imaging, including ventilation-perfusion (VQ) lung scans and computerized tomography (CT) to rule-out PE. The goal of this study was to determine the frequency of PE investigation and confirmation, and to explore factors associated with PE investigations for patients presenting to the ED with AECOPD.

Methods

Results

Characteristics of study participants

Of the 25,510 patients with a primary or secondary ED diagnosis of AECOPD, 12,164 (48%) were included over the 6-year study period after applying exclusion criteria; 2,072 (17%) patients underwent further investigations for PE (Fig 1). Characteristics at presentation are reported in Table 1. The median age was 70 years (IQR: 60, 79) and 50.4% were male. Shortness of breath (60%) was the most common presenting complaint, and most patients received chest radiography and CBC testing in the ED. Among those who underwent further investigations documented for PE, 83.7% received D-dimer testing, 44.1% received a chest CT scan and 2.4% received a VQ lung scan. In addition, there was significant variation observed in the proportion of patients with AECOPD receiving at least one test for PE across sites, from a low of 11.8% at a community site to a high of 19.8% at a teaching site (p < 0.0001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of 12,164 patients presenting to emergency departments with an acute exacerbation of chronic obstructive pulmonary disease based on their investigations for pulmonary embolism.

https://doi.org/10.1371/journal.pone.0340308.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram of cohort development.

Abbreviations: AECOPD = acute exacerbation of COPD; CT = computer tomography; ED = Emergency department; PE = pulmonary embolism; V/Q = Ventilation-perfusion lung scan; AMI = acute myocardial infarction; ACS = acute coronary syndrome.

https://doi.org/10.1371/journal.pone.0340308.g001

Primary outcome

Overall, 68 (0.5%) patients received a diagnosis of PE; 41 (0.3%) received a PE co-diagnosis in the ED and 27 (0.2%) patients received a primary PE diagnosis while hospitalized (Table 1). Of 10,429 patients who did not have D-dimer testing, 337 (3.2%) had either one of the advanced images performed (chest CT scan or VQ lung scan) and 0.1% received a diagnosis of PE in the ED (Fig 2). Of the 1,735 patient who received D-dimer testing, 944 (54%) were classified as negative using the AADD compared to 714 (41%) with the conventional cutoffs (13% increase). Among 944 patients who had a negative AADD result, 16% had further imaging investigations and 0.5% were diagnosed with PE. Among 791 patients who were classified as having a positive AADD, 60% had imaging and 3.2% received a diagnosis of PE.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Patients evaluated for pulmonary embolism during their ED visit.

a = if age-adjusted D-dimer cut-off employed.

https://doi.org/10.1371/journal.pone.0340308.g002

Secondary outcomes

As displayed in Table 2, dispositions were similar between those investigated and not investigated for PE. Overall, 46% (5,603) of the patients were admitted. The ED LOS was longer for those investigated than for those not investigated for PE (601 minutes [IQR: 397, 1040] vs. 483 minutes [IQR: 298, 962]; median differences with 95% CI: 117.5 [95.9–139.0]). There were no statistical/clinically relevant differences in LOS among patients who were admitted to the hospital; however, a significant increase in LOS by 113 minutes occurred among patients who were discharged.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Outcomes of 12,164 patients presenting to emergency departments with an acute exacerbation of chronic obstructive pulmonary disease based on their investigations for pulmonary embolism.

https://doi.org/10.1371/journal.pone.0340308.t002

Post ED outcomes included 852 (7%) return visits within 30 days, 2,270 (19%) return visits within 180 days, and 490 (4%) deaths within 30 days with no statistically significant differences between those investigated and not investigated for PE (Log-Rank test p = 0.10). Patients who underwent PE investigations were more likely to have the same testing in their next ED visit compared to those who didn’t undergo PE investigations (25.6% vs. 15.9%; p < 0.0001).

Factors associated with PE investigations

In the multivariable regression model (Table 3), the following factors were significantly associated with PE investigations while in the ED: CTAS scores 1 or 2 (aOR=1.16; 95% CI: 1.04–1.28), presentation with shortness of breath (aOR=1.29; 95% CI: 1.12–1.48) or any chest pain (aOR=2.71; 95% CI: 2.24–3.28), and a history of cancer (aOR = 1.65; 95% CI: 1.40–1.95) or venous thromboembolism (aOR=1.64; 95% CI: 1.30–2.07). Several factors were significantly associated with a lower likelihood of undergoing PE investigations including older age (aOR=0.99; 95% CI: 0.98–0.99), CTAS score 4 or 5 (aOR=0.67; 95% CI: 0.51–0.88), EMS arrival (aOR=0.67; 95% CI: 0.76–0.94), evening presentation (aOR=0.81; 95% CI: 0.73–0.91), presentation with cough/congestion (aOR=0.57; 95% CI: 0.46–0.70), a higher Charlson comorbidity index score (aOR = 0.94; 95% CI: 0.91–0.97) and a history of congestive heart failure (aOR=0.84; 95% CI: 0.73–0.98).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Factors associated with investigations for pulmonary embolism among patients presenting to emergency departments with an acute exacerbation of chronic obstructive pulmonary disease.

https://doi.org/10.1371/journal.pone.0340308.t003

Discussion

Using linked administrative databases, we explored acute COPD presentations to EDs across one province in Canada with special emphasis on the investigative approach ordered during these important encounters within the Canadian health care system. Overall, nearly one in six patients with AECOPD as a primary or secondary ED diagnosis received some form of investigation for PE. By far, the most common test was a blood d-dimer. Despite the frequency of testing, the prevalence of PE in the ED or after hospitalization was lower than reported elsewhere [7]. Among studies included in a systematic review on the topic, only one included patients from the ED and found a similarly low prevalence of PE during AECOPD (3.3%) [8].

It may seem surprising that D-dimers were drawn and, even when positive, subsequent advanced imaging was not performed. Clinicians are not obligated to “chase” D-dimer results and an elevated D-dimer in the setting of AECOPD may reflect systemic inflammation and infection in the chest cavity. Importantly, minimal elevations were often negated by using an age-adjusted cut-off, which may have been used in this setting. Moreover, many patients underwent CTPA without D-dimer testing and even when D-dimer testing was positive (either by conventional or AADD cut-offs), the results were apparently ignored, since further advanced imaging was not ordered.

One of the unique findings in this study was the granular description of ED times. First, the PIA values fail to provide evidence that crowding metrics were factors influencing the delays observed in the ED visit. Second, the proportion of patients who were designated for admission, regardless of investigations for PE, were similar. Third, the investigation of PE in the ED added nearly 2 hours on to an ED visit, appeared to contribute to ED crowding and to delay departure of patients to their place of residence and comfort. Another unique feature of this study was the comprehensive follow-up and outcome assessment. For example, regardless of whether patients received investigations for PE or not, both groups had similar returns to ED at 30- and 180-days, death within 30 days (for all causes) and death within 30 days of discharge.

In addition, the overall prevalence of PE among patients with AECOPD presenting to the ED was higher in patients with a positive D-dimer. Despite significantly higher proportion of patients undergoing advanced imaging for PE, the results suggest that using an AADD cutoff and considering other patient factors might be cost-effective and safe in patients with low probability of PE, preventing unnecessary advanced imaging in approximately 13%.

To understand the clinical practice, we explored factors associated with PE investigations through adjusted analyses. Reassuringly, some factors readily available at the ED presentation such as existing conditions (e.g., cancer), severity indicators (e.g., CTAS score of 1 or 2), and specific presenting complaints (e.g., chest pain and shortness of breath) were significantly associated with investigations for PE. Moreover, there were some factors such as increasing age, existing conditions (e.g., CHF), severity indicators (e.g., CTAS score of 4 or 5), and specific presenting complaints (e.g., cough) that were negatively associated with investigations for PE. Finally, there were perplexing results whereby arrival in evening hours and mode of arrival via EMS were negatively associated with investigations for PE. Overall, these results don’t entirely explain the variation observed.

Concerted efforts have been made through international organizations such as Choosing Wisely to limit the use of unnecessary testing, procedures and/or medications in health care [15]. Many tests are performed by clinicians for fear of missing a diagnosis, medico-legal concerns, and patient expectations. Repeated studies have identified excessive testing for PE in EDs, and calls have been made to adopt a more evidence-based approach [9]. Despite the frequency of testing for venous thromboembolism (17%) as a cause of the exacerbation in this study, the event rates in our and another ED study were lower than expected. Moreover, it appears that testing doesn’t influence important post-ED outcomes such as re-visits and mortality. Since PE is a serious condition, AECOPD and PE share similar, albeit not identical, features (e.g., dyspnea), and misdiagnosis of PE is well known, we conclude that rather than dismiss the diagnosis, clinicians should adopt a more rational approach to investigation. Fortuitously, this approach has recently been endorsed by investigators who developed and validated the use of clinical signs (non-purulent acute exacerbation of COPD, evidence of DVT and likely diagnosis of PE) and D-dimer measurement for the diagnosis of PE in a hospitalized cohort of patients with COPD [16].

There are many reasons why clinicians should be concerned with the issue of extra testing. First, advanced imaging delays departure from the ED by approximately two hours. This adds to the burden of ED crowding and opportunity costs. Second, unnecessary testing increases costs to the health care system. Third, excessive testing can increase the risk of false-positive results, which have their own consequences (e.g., anticoagulation risks, insurance issues for patients). Fourth, while patients receiving CT scans are being subjected to needless radiation exposure, many of these patients present repeatedly as their disease progresses and will also be receiving cumulative radiation exposure.

Overall, the decision to investigate a patient for PE is complex and multi-factorial; more rational approaches to investigating PE appear urgently needed. Recent efforts show promise [16] and advances in artificial intelligence/machine learning have the potential to contribute to ED decision-making in the futur [17].

Limitations

This study has several limitations that require discussion. First, the study relied on administrative data and did not examine individual charts. Consequently, data on other behavioural factors (e.g., smoking status, activity level/exercise, diet, etc.), while often incompletely documented, were not collected. In other studies using ICD coding, however, chart review failed to identify patients with a diagnoses of PE who were missed by ICD-10 coding [9]. If any bias exists, the use of administrative data codes likely overestimates the frequency of PE [18].

Second, we have included all J44.x codes for COPD; however, other ICD-10 codes may have been used (e.g., community acquired pneumonia, heart failure, diabetes) that reduced the number of patients with AECOPD in the cohort. For example, heart failure is a risk factor for PE and up to a third of patients ultimately diagnosed with PE are initially misdiagnosed as having an exacerbation of heart failure [19]. Third, we did not collect some specific PE risk factors (e.g., immobility, recent surgery, travel, etc.) and physical findings (e.g., leg circumference, pitting oedema, varicose veins, alternative diagnoses, etc.), since these are unavailable in the administrative databases employed. This prevented the calculation of a clinical decision rule score and pre-test probability. Moreover, it precluded the assessment of appropriateness of the investigations undertaken.

Fourth, the administrative data does not provide reasons for why clinicians ordered CTPA scans; other reasons to request advanced imaging in AECOPD include pneumonia, pleural effusions, and pneumothorax/pneumomediastinum. In these other conditions, patients generally receive non-contrast CT scans, not CTPA. In a small percentage of cases, clinicians may not have been searching for a PE; however, admission decisions may have been delayed by consulting services requesting PE exclusion/confirmation. Finally, the ED sites in this study were not using the age-adjusted D-dimer threshold at the time of the study, and these results may not apply to other centres that do. Although individual emergency physicians may have applied the age-adjusted D-dimer threshold in their clinical practice, the data collected and reported in the electronic medical system did not account for these adjustments.

Notwithstanding the above concerns, this multi-centered study used robust and valid linked administrative databases to study a large Canadian patient sample that were typical for AECOPD presentations. The linkage to laboratory and imaging databases enhanced our confidence tin the findings.

Conclusion

Although one in six patients with AECOPD were investigated for PE in these Canadian EDs, the overall prevalence of PE was less than 1% in this cohort. Despite a high proportion of patients undergoing advanced imaging for PE, the results suggest that employing pathways using valid CDRs that provide pre-test probabilities, selective investigations (e.g., AADD for PE) and information readily available at presentation might be a more cost-effective and safer strategy to prevent unnecessary imaging, especially in patients with low probability of PE. These study findings will also provide clinicians with more valid and robust evidence upon which to base shared decision-making in the acute setting.

Acknowledgments

This project received in-kind support from the Emergency Medicine Research Group (EMeRG) in the Department of Emergency Medicine, Faculty of Medicine and Dentistry, College of Health Sciences at the University of Alberta.