Reducing PE Protocol CT Ordering Rates in the ED

August 2011

Reducing PE Protocol CT Ordering Rates in the ED

Toolboxstic algorithms in the ED so you conduct a PUBMED search combining the terms “pulmonary embolism”, “diagnostic algorithm”, and “emergency medicine” yielding 56 citations including the manuscripts below. (

Congressional oversight committees are considering a recommendation that emergency physicians who do not appropriately risk stratify patients for pulmonary embolism before ordering a CT scan may not be fully reimbursed for the diagnostic test in the near future. While cost-reduction is certainly a motivating influence, eliminating avoidable radiation exposure is another key factor. However, emergency physicians and radiologists often provide differing perspectives on the balance between acute disease risk and long-term radiation exposure risks.

Accordingly, you have been “empowered” to devise an evidence-based algorithm with input from your hospital’s risk management team, malpractice attorneys, Radiology, and Critical Care services. After reviewing the Division of Emergency Medicine consensus on the clinical accuracy of history/physical exam, the diagnostic accuracy of PE-protocol CT and V/Q scans, the role of unstructured clinical gestalt, Well’s criteria, D-dimer and the PERC rule, this is the straw-man algorithm you put forth to begin the discussion.

PICO Question

Population: ED patients with suspected pulmonary embolism (low, moderate, and high risk)

Intervention: Algorithm-aided diagnostic decision making

Comparison: Routine ED care

Outcome: PE protocol CT order rates, PE-related mortality


First years: Critical issues in the evaluation and management of adult patients presenting to the emergency department with suspected pulmonary embolism, Ann Emerg Med 2011; 57: 628-652. (

Second years: Success of a safe and simple algorithm to reduce use of CT pulmonary angiography in the emergency department, Am J Roentgenol 2010; 194: 392-397. (

Third years: A computerized handheld decision-support system to improve pulmonary embolism diagnosis: a randomized trial, Ann Intern Med 2009; 151: 677-686. (

Fourth years: Effectiveness and acceptability of a computerized decision support system using modified Wells criteria for evaluation of suspected pulmonary embolism, Ann Emerg Med 2011; 57: 613-621 (


Article 1: Critical Issues in the Evaluation and Management of Adult Patients Presenting to the Emergency Department with Suspected Pulmonary Embolism, Ann Emerg Med 2011; 57: 628-652

Article 2: Success of a Safe and Simple Algorithm to Reduce Use of CT Pulmonary Angiography in the Emergency Department, AJR 2010; 194:392-397

Article 3: A Computerized Handheld Decision-Support System to Improve Pulmonary Embolism Diagnosis, Ann Intern Med 2009; 151:677-686

Article 4: Effectiveness and Acceptability of a Computerized Decision Support System Using Modified Wells Criteria for Evaluation of Suspected Pulmonary Embolism, Ann Emerg Med 2011; 57:613-621

Bottom Line

Background and Commentary

As many as 200,000 people in the United States die each year from pulmonary embolism (PE) and non-fatal PE afflicts another 400,000 patients. Survivors of PE can suffer from pulmonary hypertension or post-thrombotic syndrome. The standard of care for initial management of stable PE is heparin, but this therapy is based upon only two trials. In 1959 Barritt and Jordan conducted the only trial thus far However, the trial involved no placebo and no blinding, enrolled 35 subjects, and did not confirm the diagnosis of PE with any standard that would be accepted today (pulmonary angiography, V/Q, autopsy). By the time of enrollment, their subjects were typically hypotensive and nearly 20% died. Fifty years later Kline conducted an observational trial of over 8000 patients with ED testing for PE that included 500 confirmed PE cases and noted a 0.2% PE-related mortality rate – a 100-fold difference than that noted in 1959. A 2006 Cochrane review discovered only 1 controlled trial evaluating anticoagulation for known DVT noting no benefit in clot resolution, prevention of PE, or mortality. Thus, the efficacy of anticoagulation for PE remains in doubt. In fact, Newman and Schriger used conservative estimates of harm and liberal estimates of benefits in contemplating the value of CT pulmonary angiography for all patients with suspected PE and concluded that testing prevented 6 PE-related deaths and 24 major non-fatal PE events, but caused 36 deaths and 37 non-fatal major medical events – testing causes roughly 6 times more deaths than lives saved! When combined with recently increasing reports that 1.5%-2% of all future cancers in the United States may result from medical radiation, and the fact that despite increasing CT ordering rates PE mortality has not changed in decades, one could conclude that CT’s may represent this generation’s asbestosis so emergency physicians should strongly consider evidence-based protocols to limit CT’s to those patients most likely to benefit.

CT utilization rates have not increased without reason. In a prevailing malpractice environment where the courtroom perspective is that available technology is considered fair game for any & all with no margin for errors of omission (not ordering a test), time-pressured emergency physicians have felt compelled to continue lowering the threshold in which PE imaging occurs. And since CT is usually readily available around-the-clock (as compared to V/Q and lower extremity ultrasound), it has become the default test of choice. In addition, CT provides valuable alternative diagnoses that are often being considered in patients who present with a chief complaint (dyspnea, chest pain) rather than a diagnosis which is often paradoxically obvious when the case is later reviewed retrospectively.

Survey Results (StLouisSurvey and Virginia Beach results)

From 57 respondents that included 9 community physicians and 51% residents, 100% use clinical decision rules (CDR) in their decision making, including 30% who do so daily and 91% when evaluating a patient for PE. The Well’s criteria is the favored CDR, but 33% also rely on clinical gestalt and 25% use the PERC rule of which 89% of respondents were familiar. The majority of respondents (70%) felt that the PERC rule had been sufficiently validated to safely use in clinical settings, and 47% do so quite often or very often. All respondents have access to D-dimer testing but 75% were not aware of the type of assay used at their hospital. Most (89%) use D-dimer testing in low-risk patients and 21% also use it to evaluate intermediate risk patients.

Over half (52%) feel that emergency physicians over-order CTs for PE and another 34% thought that EP’s might over-order, though only 23% felt that they personally over-order CTs. The most commonly identified barrier to more restrictive use of CTs for PE were malpractice concerns (85%) and lack of additional diagnostic resources (ultrasound, V/Q – 57%) and time-constraints making diagnostic algorithms undesirable (41%). Published guidelines (83%), medico-legal support (68%), around-the-clock access to alternative diagnostic tests (66%), and administrative support (47%) were identified as mechanisms to efficiently reduce ED CT ordering rates for PE. Almost all respondents (98%) would support a guideline-based diagnostic algorithm.

Analysis of the Evidence

Paper-based guideline dissemination that does not integrate decision support tools into physician work flow at the time and location of decision making have limited effect on moving emergency clinicians towards more Bayesian-appropriate PE diagnostic strategies. Alternatively, computer decision support systems forcing function using validated algorithms to standardize CTPA test ordering in ED patients with suspected PE can safely reduce CTPA utilization rates without increasing the prevalence of undiagnosed PE. However, despite up-front buy-in from EP’s and a 1-2 minute maximum work-delay to input the Well’s Criteria and consider the algorithm recommendations EP’s rejected this intervention. Future PE-CDSS should be designed before implementation with the following considerations: speed, real-time delivery, workflow impediments, usability, physician objections, altering rather than stopping behavior, simplicity, minimization of information requested, result assessment and feedback, and maintaining up-to-date, evidence-based CDSS. (Bates 2003)

Small Group Discussions

Group 1 — Barriers and Implementable Solutions

An acceptable PE miss rate by emergency physicians was felt to be 0-10% amongst malpractice attorneys, 5% amongst Radiologists, and 3-5% amongst emergency physicians. Barriers identified included those internal and external to the ED. Internal barriers were physician personalities resistant to change, but these were felt far less important than external barriers which included admitting services that focus on misses with hypercritical assessments of individual physicians without acknowledging the limitations of the diagnostic technology or emerging evidence of over-testing in the United States. An additional barrier to alternative imaging modalities such as V/Q and Doppler ultrasound studies was that the emergency department is open around-the-clock whereas some hospital services operate on “banker’s hours”. If hospital administrators are committed to reducing CT order rates for PE then there must be a commitment to provide V/Q and Doppler ultrasounds 24/7, too. Some sustainable solutions offered by this group included opinion leader driven departmental educational efforts within the ED and then driven by department heads through the Medical Executive Committee of each hospital. Once a protocol and miss rate are accepted by the MEC, malpractice insurance providers should be encouraged to promote the protocol and individual physician audit-and-feedback efforts initiated. If a protocol was accepted and implemented across a region, the malpractice attorneys unanimously believed that this regional standard of care would protect physicians in a court of law for the occasional miss that would occur. Annual regional report cards demonstrating CT order rates were proposed as another potential strategy to encourage those slow to accept an algorithmic approach, but risk management and hospital attorneys did not feel that this solution would be well accepted by administrators across healthcare networks.

Based upon the evidence analyzed tonight, a paper-based algorithm is much less likely to succeed in reducing CTPA ordering efficiency so electronic medical records will be essential to remind and assist physicians in the use of any diagnostic or therapeutic algorithms that become acceptable and available.

One important point raised by an academic physician is that without a single leader to take charge after this Journal Club, the entire effort is unlikely to be initiated at all. At Journal Club nobody was willing to take on this role, but perhaps such a leader who is driven by the diagnostic and therapeutic issues surrounding the emergency management of PE will step forward.

Group 2 – Defining an Acceptable and Sustainable Diagnostic Algorithm

Based upon the ACEP Guidelines, independent review of the evidence, and group consensus amongst community & academic emergency physicians, radiologists, and risk management, the following algorithm was proposed and is now being implemented at several of the participating hospitals.

When applying this algorithm two important details must not be overlooked.

The ACEP Guidelines, PIOPED II, and our prior Journal Club consensus have all emphasized that discordant results (high pre-test probability, negative CTPA; low pre-test probability, positive CTPA) should be evaluated with a further test since neither result changes the post-test probability enough to justify safe exclusion of (negative CTPA) or ruled in (positive CTPA) PE in these situations. This statement is a bit contentious and should not result in reflexive double-testing for these small subsets of patients with suspected PE. Why is the statement contentious? First, the ACEP Guidelines are based largely on PIOPED II which used 4-slice CT scanners. In 2011 we use 64-slice CT scanners (or greater). Although some literature suggests that the higher slice scanners only identify more sub-segmental PE’s (which may be clinically inconsequential), Radiology feels that recommending additional imaging for low probability patients with a positive CT is illogical in most situations since no other test will be as accurate as CTPA. Second, although the guidelines and PIOPED II treat CTPA as a dichotomous test, it really is not since different findings that all fall under the “CT positive for PE” domain are not equally reliable, accurate, or prognostically significant. For example, a saddle embolism is more likely to be identified by different Radiologists and is associated with a worse prognosis than sub-segmental peripheral emboli and is probably definitive enough to merit anticoagulation without further imaging even in the low pretest probability patient. Therefore, when faced with a low risk patient and a CT that is positive for PE, consider the following

  • Was the patient truly low-risk?
  • Could the interpretation represent a false-positive if your clinical impression is reviewed with Radiology?
  • What is your local institution’s PE protocol CT diagnostic accuracy (sensitivity, specificity, LR+, LR-) and how are these figures derived (i.e. what gold standard is used)?
  • What is the level of confidence associated with the Radiology interpretation?
  • What is the prognostic significance of the PE identified (sub-segmental versus saddle embolism)?
  • What is the risk of anticoagulation in this patient and is the patient agreeable to the risk if 60% do not have a PE as suggested by the PIOPED II data?

For the opposite scenario where a high probability patient has a negative CT, there is agreement that a Doppler exam of the lower extremities is indicated before excluding PE.

The PERC rule was not designed to apply to undifferentiated patients with suspected PE. Patients should first be risk-stratified (using gestalt or another non-PERC CDR) and if after that they are low-risk then the PERC rule can be applied in lieu of a D-dimer to identify a “very low risk” (< 1.4% chance of venous thromboembolism at 3-months) population.

Group 3 – Assessing Effectiveness (Metrics)

The following measurements were proposed to assess the success of efforts to reduce CTPA ordering in the emergency evaluation for PE with no particular preference given to any metric: CTPA efficiency (defined as proportion of CTs demonstrating PE over number of PE protocol CTs ordered); CTPA order rates adjusted for non-PE indications; V/Q order rates; ED length of stay; proportion of physicians compliant with the algorithm; PE imaging costs; lawsuits filed and plaintiffs paid for missed PE cases. The group did not discuss who would be responsible for collecting these metrics, how often they would be evaluated, what thresholds would define success or failure, and how the metrics would be shared with the individual clinician, department, hospital, or regional healthcare system.