Wells Score for pulmonary embolism

Clinical background

Pulmonary embolism cannot be reliably distinguished from other causes of acute dyspnoea or chest pain by history and examination alone. At the same time, imaging (computed tomography pulmonary angiography, CTPA) is resource-intensive, involves a radiation dose and the risks of contrast, and has a high proportion of negative results. The Wells score serves to adjust the probability before testing: it stratifies patients into risk groups so that a negative D-dimer in the low-risk group can reliably exclude pulmonary embolism without further investigation, while high-risk patients proceed directly to imaging [1]. Without a structured pre-test estimate, the D-dimer is difficult to interpret, since its specificity is low and its value depends entirely on the pre-test probability.

Calculating the Wells score

The score is the sum of seven variables:

$$\text{Wells} = 3,I_{\text{DVT}} + 3,I_{\text{PE most likely}} + 1{.}5,I_{\text{HR}>100} + 1{.}5,I_{\text{immob/surgery}} + 1{.}5,I_{\text{previous PE/DVT}} + 1,I_{\text{haemoptysis}} + 1,I_{\text{malignancy}}$$

where each $I$ takes the value 0 or 1 and the maximum score is 12.5.

The variables are: clinical signs of DVT (3 points), pulmonary embolism is the most likely diagnosis (3 points), heart rate >100/min (1.5 points), immobilisation ≥3 days or surgery within 4 weeks (1.5 points), previous pulmonary embolism or DVT (1.5 points), haemoptysis (1 point) and malignancy treated within 6 months or palliative (1 point).

The derivation cohort consisted of patients with suspected pulmonary embolism in a prospective multicentre study in Canada [1]. Of 40 clinical variables, seven were selected by logistic regression in 80% of the cohort; the remaining 20% were used for internal validation. Two classification systems were produced: a three-level system (low, intermediate, high probability) and a two-level system (pulmonary embolism likely or unlikely). In the derivation material, 7.8% of patients with a score ≤4 had pulmonary embolism, but if the D-dimer was also negative the proportion fell to 2.2% (95% CI 1.0–4.0) in the derivation dataset and 1.7% in the validation dataset [1]. This combination was found in 46% of patients.

Interpretation in practice

The Wells score can be interpreted under two systems, and both are accepted in clinical guidelines:

In the three-level system, low probability applies below 2 points, intermediate at 2–6 and high above 6. In the two-level system, ≤4 is classified as "pulmonary embolism unlikely" and >4 as "pulmonary embolism likely". The two-level system is simpler to use within a diagnostic algorithm and is the commoner one in European practice: a score ≤4 combined with a negative D-dimer excludes pulmonary embolism without imaging, and a score >4 leads directly to CTPA whatever the D-dimer [1]. At intermediate probability in the three-level system, an age-adjusted D-dimer threshold (age × 10 µg/L for patients over 50) can increase the proportion of patients in whom imaging can be avoided [5].

Validation and performance

In a systematic review of 29 studies and 31,215 patients, the pooled prevalence of pulmonary embolism was 6% in the low-risk group, 23% in the intermediate group and 49% in the high-risk group for the three-level Wells system [2]. For the two-level system the prevalence was 8% and 34% respectively.

An individual patient data meta-analysis based on 16 studies and 28,305 patients showed that the Wells score combined with the D-dimer had a c-statistic of 0.73 (95% CI 0.70–0.75) for predicting pulmonary embolism, compared with 0.79 (95% CI 0.76–0.81) for structured clinical pre-test assessment [3]. A newly developed model with objective variables and an interaction term for age and D-dimer reached a c-statistic of 0.87 (95% CI 0.85–0.89) with good calibration (O:E ratio 0.99) [3].

In a network meta-analysis of 40 studies and 37,027 patients, the three-level Wells system was adequate for excluding pulmonary embolism (negative likelihood ratio LR− 0.34, 95% credible interval 0.25–0.45), but the two-level system markedly underperformed (LR− 0.56, 95% CI 0.45–0.68) [4]. The same analysis found that the revised Geneva score was superior to Wells in correctly assigning patients to imaging rather than to a D-dimer (LR+ 6.65 versus 5.59; diagnostic odds ratio 8.03 versus 7.40) [4].

In a direct comparative study of 300 consecutive patients there was no statistically significant difference in discrimination (AUC) between the Wells score and the revised Geneva score, and no patient with a low or intermediate Wells score combined with a normal D-dimer developed venous thromboembolism over three months of follow-up [6].

An individual patient data meta-analysis of 7,268 patients from six prospective studies showed that a Wells score ≤4 combined with a negative D-dimer had a failure rate (missed pulmonary embolism at three months) below 3% in all the subgroups examined, including patients aged ≥75 years and patients with malignancy [5]. The efficiency, that is, the proportion of patients in whom imaging could be avoided, was 28% with a fixed D-dimer threshold (≤500 µg/L) and 33% with an age-adjusted threshold [5].

Limitations

The most debated weakness of the Wells score is the variable pulmonary embolism is the most likely diagnosis, which requires a subjective clinical judgement [3, 4]. This judgement varies between assessors and is hard to standardise, which is why newer models omit it [3]. The subjective component is also one reason why the revised Geneva score, which rests solely on objective variables, performs at least as well and in some respects better [4].

The derivation cohort consisted mainly of outpatients in Canada. Performance in inpatient populations is lower, and efficiency in particular falls, since the D-dimer is more often raised in inpatients for other reasons [5].

The Wells score was not developed for pregnant women, who were excluded from the derivation material and from most validation studies [3]. In pregnant patients, imaging is generally recommended when there is clinical suspicion whatever the score, possibly with a D-dimer in low-risk cases depending on local practice.

The score should not be confused with risk stratification after pulmonary embolism has been established (PESI, sPESI), which concerns prognosis and treatment intensity rather than diagnosis. The Wells score is a pre-test instrument and does not by itself indicate diagnostic certainty.

References

1. Wells PS et al. Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the models utility with the SimpliRED D-dimer. Thromb Haemost 2000;83(3):416–20. PMID: 10744147
2. Ceriani E et al. Clinical prediction rules for pulmonary embolism: a systematic review and meta-analysis. J Thromb Haemost 2010;8(5):957–70. PMID: 20149072
3. van Es N et al. Diagnostic management of acute pulmonary embolism: a prediction model based on a patient data meta-analysis. Eur Heart J 2023;44(32):3073–3081. PMID: 37452732
4. Etemadi A et al. Comparative diagnostic accuracy of pre-test clinical probability scores for the risk stratification of patients with suspected pulmonary embolism: a systematic review and Bayesian network meta-analysis. BMC Pulm Med 2025;25:162. PMID: 40200307
5. van Es N et al. Wells Rule and d-Dimer Testing to Rule Out Pulmonary Embolism: A Systematic Review and Individual-Patient Data Meta-analysis. Ann Intern Med 2016;165(4):253–61. PMID: 27182696
6. Klok FA et al. Comparison of the revised Geneva score with the Wells rule for assessing clinical probability of pulmonary embolism. J Thromb Haemost 2008;6(1):40–4. PMID: 17973649