Original Research Article
Risk scoring models fail to predict pulmonary embolism in trauma patients

https://doi.org/10.1016/j.amjsurg.2021.02.007Get rights and content

Highlights

  • No pulmonary embolism risk scoring model had both high sensitivity and specificity.

  • No signs or symptoms were independently associated with pulmonary embolism.

  • Area under the curve was low for all pulmonary embolism risk scoring models.

  • No risk scoring model was effective as a rule-out tool for pulmonary embolism.

Abstract

Background

We aimed to identify risk factors and risk scoring models to help identify post-traumatic pulmonary embolisms (PE).

Methods

We performed a retrospective review (2014–2019) of all adult trauma patients admitted to our Level I trauma center that received a CT pulmonary angiogram (CTPA) for a suspected PE. A systematic literature search found eleven risk scoring models, all of which were applied to these patients. Scores of patients with and without PE were compared.

Results

Of the 235 trauma patients that received CTPA, 31 (13%) showed a PE. No risk scoring model had both a sensitivity and specificity above 90%. The Wells Score had the highest area under the curve (0.65). After logistic regression, no risk scoring model variables were independently associated with PE.

Conclusions

In trauma patients with clinically suspected PE, clinical variables and current risk scoring models do not adequately differentiate patients with and without PE.

Introduction

Approximately 600,000 individuals in the US each year experience an acute pulmonary embolism (PE), of which approximately 25% are fatal.1,2 PE is also a well-known complication in trauma. Trauma patients are at a particularly high risk of PE given that traumatic injury overlaps with all three pillars of Virchow’s triad: endothelial injury, venous stasis, and hypercoagulability. Acute PEs occur in 0.35–24% of trauma patients, depending on the level of thromboprophylaxis, and are responsible for 12% of all trauma deaths.3,4

In non-trauma patients, the standard approach to a suspected PE is to first attempt to rule it out using clinical judgement or objective criteria (i.e., a risk scoring model). According to the society guidelines for both emergency medicine and internal medicine, if the pretest probability of a PE is low and followed by a normal D-dimer, PE can be ruled out.5,6 Unfortunately, in trauma patients, the value of a D-dimer in this process is dubious because traumatic injury can lead to elevated D-dimers independent of PE.7,8 Furthermore, determining pretest probability is more challenging in trauma patients because the classic PE signs and symptoms can be masked by traumatic injuries (e.g., leg swelling, dyspnea, tachypnea, hypoxemia).9,10 This leaves trauma centers with few options to rule out PEs before ordering a computed tomography pulmonary angiogram (CTPA), the gold standard of PE diagnostics11; however, CTPAs are expensive and not without risks.12, 13, 14 Two PE risk scoring models have been developed specifically for trauma patients: Trauma Embolic Scoring System (TESS) and Risk Assessment Profile. Numerous attempts to validate the accuracy and usefulness of these trauma risk scoring models have produced conflicting results. 15, 16, 17, 18, 19 Of the many non-trauma risk scoring models, the Wells score and the Revised Geneva score have been the most validated.20 The Wells score and Revised Geneva score have also been tested in trauma patients, but only in subpopulations (e.g., orthopedic trauma) or in studies focused on deep vein thrombosis (DVT). 21, 22, 23 To our knowledge, there have been no studies comparing the usefulness of a diverse array of risk scoring models in the general trauma patient population. There are two ways that an effective PE risk scoring model could help trauma centers: diagnosing PEs and ruling out PEs. If a risk scoring model helps diagnose PEs, it has the potential to identify PEs earlier; if a risk scoring model helps rule out PEs, it has the potential to reduce unnecessary CTPAs. Our hypothesis was that no risk scoring model is sufficiently accurate to diagnose or rule out PEs in trauma patients. To test this hypothesis, we sought to evaluate the value of 11 published risk scoring models by applying them to our trauma database.

Section snippets

Methods

This was a retrospective study of all adult trauma patients between 18 and 89 years of age admitted to our ACS-verified Level I trauma center from March 2014 to October 2019 that received a CTPA on clinical suspicion of a PE. Data collection in both the trauma registry and the electronic health record included patient demographics, admission physiology, injury details, diagnostic tests, and other PE risk factors.

We performed a systematic search to identify studies that used risk scoring models

Results

Of the 15,861 total trauma patients admitted to our trauma center, 235 patients met our inclusion criteria, of which 31 patients had a PE diagnosed on CTPA. Patients in our study were mostly middle aged (52 ± 19 years), white (89%), and male (67%). Average time from arrival to the CTPA was 8 ± 10 days for both the PE group and the no PE group. The only demographics, admission physiology, and injury variables associated with PE were minority race (23% vs. 9%, p = 0.02) and higher BMI (33 ± 8 vs.

Discussion

The purpose of the study was to evaluate the diagnostic value and rule-out capabilities of 11 risk scoring models for predicting PE in our trauma patients. Our hypothesis was that no risk scoring model is sufficiently accurate to diagnose or rule out PEs in trauma patients. Our study proved this hypothesis.

If an accurate PE risk scoring model is discovered for trauma patients, it has the potential to improve PE detection and reduce unnecessary CTPAs. For this reason, we designed a study to

Conclusions

In conclusion, in trauma patients with clinically suspected PE, clinical variables and current risk scoring models do not adequately differentiate patients with and without PE. Until an accurate, safe, and useful PE risk scoring model is developed, CTPA is necessary as a screening tool for PE in trauma patients.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Acknowledgements

Literature search, study design, data collection, data analysis, data interpretation, writing, and critical revision were a joint effort with Mr. Frank Buchanan, Ms. Emily Leede, Dr. Lawrence Brown, Dr. Pedro Teixeira, Dr. Jayson Aydelotte, Dr. Tatiana C. Cardenas, Dr. Thomas B. Coopwood, Dr. Marc D. Trust, Ms. Sadia Ali, and Dr. Carlos Brown.

References (34)

Cited by (0)

View full text