Clinical probability scores (CPS) are reliable, noninvasive tools that, based on history and clinical findings, determine pretest probability and assess the need to perform various diagnostic tests in patients with suspected pulmonary embolism (PE).

Different CPS models, including the Geneva revised score, have been validated for PE diagnosis.1–4 These scores, used as part of a diagnostic algorithm in combination with the determination of D-dimer (DD) levels, may help exclude PE in low risk groups and make further tests to rule out this diagnosis unnecessary.5–12 Although it is well accepted that imaging tests should only be carried out when there is a high clinical probability (CP) of PE,10,12 guidelines are clearly not followed, and multislice computed tomography angiography (MSCT) of lung or ventilation-perfusion scintigraphy (V/Q scan) is performed as a first step in PE diagnosis.9,11 Thus, several studies in recent years have shown positivity rates in MSCT of less than 10% in patients with suspected PE,13–16 while prospective clinical trials conducted two decades ago, in which pretest clinical evaluations were carried out before V/Q scan, revealed the disease in at least one third of the patients.17,18 These results suggest overuse of the technique, and possibly poor selection criteria. The probability of confirming positive PE on MSCT in patients without risk factors is extremely small (0.95%). Therefore, it seems that MSCT is probably unnecessary in this scenario,19 and the indiscriminate use of MSCT raises concerns regarding increased exposure to radiation.9

Our hypothesis is that CPS are poorly implemented in practice, and diagnostic protocols are not being applied. The objectives of this study were to determine the degree of compliance with CPS and diagnostic algorithms in clinical practice in our hospital in cases of suspected PE.

During the study period, 3924 cases of suspected PE were identified (56% female), and the diagnosis was confirmed in 360 patients (9.2%). Demographic data are shown in Table 2. Table 3 contains the most common diagnoses of patients without PE.

CPS had been recorded in the clinical history of only 24 patients (0.6%). In another 3218 (82%), this could be calculated a posteriori. Finally, data from the clinical history were insufficient for making the calculation in 682 cases (17.4%). Of the 3242 patients in whom CP was determined, 1281 (39.5%) had low probability, 1837 (56.7%) intermediate, and 124 (3.8%) high (Table 4).

As expected, the prevalence of PE varied depending on CP: 4.8% [61/1281] in the low probability group, 13.1% [241/1837] in the intermediate probability group, and 40.3% (50/124) in the high probability group (Fig. 1) (P<.001 comparing all of them). PE was diagnosed in 8 more cases among the 682 patients in whom CP calculation was not possible (Table 4). Overall prevalence of PE was 9.2% (360/3924) and incidence was 30.6 cases per 100000inhabitants/year.

Fig. 1.

Prevalence of pulmonary embolism by clinical probability.

(0.04MB).

Prevalence of PE in patients with low CP and positive DD was 7.5% (55/737), and 0.4% (2/512) in those with negative DD. In those with intermediate CP, prevalence rates were 14.8% (207/1395) and 0.3% (1/318) and, in those with high CP, the rates were 40.8% (42/103) and 16.7% (1/6), respectively (Fig. 2) (P<.001 for low and intermediate CP, between positive and negative DD, but not significant in high CP) (Table 4).

Fig. 2.

Prevalence of pulmonary embolism by clinical probability (low or intermediate) and D-dimer result. DD, D-dimer; NS, not significant; CP, clinical probability.

(0.06MB).

DD was determined in 94.7% of cases (3071/3242). Its sensitivity for the diagnosis of PE was 98.7%, and the negative predictive value was 99.2%. MSCT was not performed in 70 patients due to the risk of contrast nephropathy.

Among patients with low CPS, DD was not determined in 32 cases (2.5%). Of the 1249 in whom it was measured, results were negative in 512 (41%) and positive in 737 (59%). Among those with negative DD, MSCT was performed in 26 cases (5.1%), and this was positive in 2 cases (7.7%). No PE was demonstrated in 486 patients in whom DD was negative and MSCT was not performed. Of those patients who had positive DD (737), MSCT was performed in only 303 (41.1%), of whom 52 (17.2%) had PE. Of the 434 patients who did not have MSCT results, 3 had PE (MSCT was not performed in these 3/10 due to risk of contrast nephropathy). Of the 32 patients in whom DD was not measured, MSCT was performed in 28, of whom 4 had PE, while 4 did not undergo any other diagnostic test and PE was not demonstrated in any of them (Fig. 3).

Fig. 3.

Patient flowchart according to clinical probability, with the results of diagnostic tests for pulmonary embolism. DD, D-dimer; PE, pulmonary embolism; CPS, clinical probability score; V/Q Scan/US, ventilation-perfusion scintigraphy/perfusion or compression ultrasonography of the lower extremities; MSCT, multislice computed tomography; (−), negative; (+), positive; , patients in whom the diagnostic algorithm of pulmonary embolism was not followed; , patients diagnosed with pulmonary embolism by multislice computed tomography, but whose clinical probability score could not be calculated.

(0.52MB).

Among patients with intermediate CPS, DD was not determined in 124 patients (6.8%). In 1713 in whom DD was measured (93.2%), results were negative in 318 (18.6%) and positive in 1395 (81.4%). In the first group (negative DD), MSCT was performed in 28 cases (8.8%), all of which were negative. Of those in whom MSCT was not done (290), one patient presented PE (1 of the 4 patients in whom MSCT was not performed due to risk of contrast nephropathy). Among those with positive DD, MSCT was performed in 655 (46.9%), and was positive in 191 (29.2%). Of the 740 patients with intermediate CPS and positive DD in whom MSCT was not performed, 16 (2.2%) had PE (16/44 with risk of contrast nephropathy that did not undergo MSCT). Finally, of the 124 patients in whom DD was not obtained, MSCT was performed in 122 (98.4%), of which 32 (26.2%) had PE. Of the 2 patients who did not undergo MSCT due to risk of contrast nephropathy, one had PE (Fig. 3).

Of the 124 patients with high CPS, DD was determined in 109 (87.9%) and found positive in 103 (94.5%). MSCT was performed in only 2 of the 6 patients with negative DD, one of whom had PE. PE was not demonstrated in the 4 patients that did not undergo MSCT. MSCT was performed in 68 (66%) of the 103 patients with positive DD; results were positive in 39 (57.4%). Of the 35 patients in whom MSCT was not performed despite having high CP of PE and positive DD, 3 had PE (3/8 in whom MSCT was not performed due to risk of contrast nephropathy). Of the 15 patients without DD determination, 13 had MSCT, which was positive in 6 (46.2%). Of the 2 patients in whom MSCT and DD were not performed, one presented PE (1/2 in whom MSCT was not performed due to risk of contrast nephropathy) (Fig. 3).

Of the total 3924 patients with suspected PE, CPS and diagnostic algorithms were not followed in 2125 (54.2%): in 682 cases (32.1%) the data required to estimate CP were not collected; DD was not determined in 109 patients (5.1%) with high CP; of 852 cases (40.1%) with intermediate CP, DD was not determined in 124, MSCT was not performed despite positive DD in 696, or MSCT or other diagnostic tests were performed with negative DD (32); and, in the case of the 482 patients (22.7%) with low CP, DD was not determined in 32, MSCT was not performed despite positive DD in 424, or MSCT was performed with negative DD in 26 (Table 4 and Fig. 3).

This study confirms the hypothesis that CPS in the diagnosis of PE is rarely calculated and diagnostic algorithms are often not applied in clinical practice, since these were correctly used in only 45.8% of suspected PE.

Diagnosis of PE requires the use of several diagnostic techniques that should be used sequentially. Since the use of a validated diagnostic study is associated with a substantially decreased risk of complications,25 it is advisable that this diagnosis is carried out in a standardized manner. CPSs are intended to identify patients with intermediate or high CP, because they should receive anticoagulant therapy until the results of diagnostic tests are available.

CP calculation was not possible in 682 patients (17.4%) because the CP history data were insufficient, although 8 patients were diagnosed with PE. The correct diagnostic protocol had apparently not been applied in any of these cases.

DD determination is highly sensitive for ruling out PE if the blood concentration is <500ng/ml, at least in patients with low or intermediate CP.26 MSCT is not indicated in this case, since the probability of PE is low.1,2,6 In this manner, excessive use of MSCT that could cause an increase in cancer incidence attributable to radiation9 and the risk of contrast nephropathy are avoided.23 Therefore, this technique should only be carried out in patients with low or intermediate CP if DD levels are >500ng/ml. For this approach to allow the exclusion of PE with a sufficient safety margin, the method for DD determination should ensure a negative predictive value of >98%.27 In our series, DD sensitivity was 98.7% and its negative predictive value was 99.2% (four false negatives). Among patients with low and intermediate CP in whom DD was <500ng/ml, MSCT was performed in 5.1% (26/512) and 8.8% (28/318) of patients respectively. These results are similar to those obtained by Corwin et al. (7%).28 However, if it is taken into account that MSCT was also directly performed in 28/32 patients with low CP and in 123/124 (99.2%) with intermediate CP with no DD determination, 205 MSCT scans should not have been carried out for one or other reason (20.8% with negative DD or without results), a figure still below the 24% of patients in the series of Dunn et al.29

When CP is low or intermediate and DD >500ng/ml, the diagnostic algorithm states that MSCT or another diagnostic test should be performed to rule out the disease. In our series, this was not done in 52.5% of these patients (1120/2132; 424/737 with low and CP and 696/1395 with intermediate CP). One reason that might explain why the protocol was not followed is that positive DD is nonspecific, and an alternative diagnosis may have been established in some cases (Table 3). However, this would not apply to all cases because, although some of these diagnoses may involve elevated DD,30 most are not related to the cardiopulmonary system. Moreover, a final diagnosis was not established in many others. Corwin et al. found that in 42% of the patients (605/1431) with positive DD in their series, MSCT was not performed.28 This might imply that an unknown percentage of patients with PE were not diagnosed, a possibility that cannot be ruled out. These patients were followed for 3 months through their clinical history, but this was aimed more at confirming that PE did not recur within that time, than at ruling out the possibility that diagnosis was not PE in the first place. PE incidence in our series (30.6 cases per 100000 inhabitants/year) was lower than that reported by other authors.31–33 This could also be attributed to possible underdiagnosis, since MSCT was not performed in these cases.

Among patients with high CP for PE, the diagnostic protocol was applied in a standardized manner in only 15 patients (12.1%), because DD was determined first in 109 (87.9%).

The total number of patients in whom the diagnostic algorithm was not applied was 2125 (54.2%). In a recent study, Weiss et al. found that only 23% of the physicians surveyed used the published prediction scores for assessing pretest probability of PE, and 73% preferred an unstructured approach despite knowing the published guidelines.34 While acknowledging that this pretest probability of PE is empirically evaluated,12 this should not influence the correct application of established diagnostic protocols.

As expected, PE prevalence increased with CPS.35 Similarly, PE prevalence was significantly higher within the same CP (P<.001) when DD was positive, except in the high CP group. Possibly the reason is that only 6 patients had negative DD, compared to 103 with positive DD. Overall PE prevalence was 9.2%, slightly higher than the 6% of Corwin et al.,28 but well below the 27% in the series of Prologo et al.36 This decline in PE prevalence compared to a series of previous decades may be due to the increasingly frequent use of MSCT in emergency services, with a consequent reduction in diagnostic yield. It should be remembered that DD specificity is very low and it is not useful for diagnosing acute PE. If DD is determined in cases where suspected PE is clinically not well supported, elevated results may lead to unnecessary MSCT scans.

The main limitation of our study is its retrospective nature. CPS or compliance with the diagnostic protocol could not be established in 17.4% of cases (682) for this reason. Another limitation is the inability to determine whether any of the patients with a positive DD in whom MSCT was not performed, did actually have PE. Follow-up of the clinical history for 3 months does not exclude the diagnosis at the time that these patients came to the emergency department.

In summary, CPS for PE diagnosis is rarely calculated, and diagnostic protocols are rarely applied in clinical practice in our center. This can lead, in some cases, to the performance of unnecessary imaging techniques that may also lead to serious side effects and, in others, to a high risk of underdiagnosis, since the disease cannot be safely ruled out. If CPS were implemented as part of a diagnostic algorithm, higher diagnostic yield could be achieved, and the available resources could be optimized.

P. Sanjuán is the writer and author, involved in conception and design, data collection, analysis and interpretation of data and approved the final manuscript.

N. Rodríguez-Núñez is the co-author, involved in data collection, critical revision of the manuscript and approved the final manuscript

C. Rábade is the co-author, involved in data collection, critical revision of the manuscript and approved the final manuscript.

A. Lama is the co-author, involved in data collection, critical revision of the manuscript and approved the final manuscript.

L. Ferreiro is the co-author, involved in data collection, critical revision of the manuscript and approved the final manuscript.

F.J. González-Barcala is the co-author, involved in data collection, critical revision of the manuscript and approved the final manuscript.

J.M. Álvarez-Dobaño is the co-author, involved in data collection, critical revision of the manuscript and approved the final manuscript.

M.E. Toubes is the co-author, involved in data collection, critical revision of the manuscript and approved the final manuscript.

A. Golpe is the co-author, involved in data collection, critical revision of the manuscript and approved the final manuscript.

L. Valdés is the writer and author, involved in conception and design, analysis and interpretation of data and approved the final manuscript.

This work has not benefited from any scholarship or financial help.

The authors declare no conflicts of interest.