Revised Geneva Score for pulmonary embolism

Clinical background

Pre-test probability is the first step in the structured diagnostic work-up of suspected pulmonary embolism. Its purpose is not to make the diagnosis but to direct the next step: which patients can be investigated with a D-dimer and which should proceed directly to imaging. Without a structured probability assessment, clinicians tend to over-investigate with computed tomography pulmonary angiography, with a low yield of confirmed emboli.

The revised Geneva score for pulmonary embolism was developed to offer an alternative to the Wells score that is entirely free of subjective judgement. The Wells score contains the item "pulmonary embolism is the most likely diagnosis", which requires an overall clinical judgement that varies between assessors and is difficult to standardise. The revised Geneva score rests exclusively on objective variables: age, history, vital signs and clinical findings that can be reproduced independently of who performs the assessment [1].

Calculating the revised Geneva score

The score is a weighted sum of eight objective clinical variables:

$$\text{Revised Geneva} = \text{Age}$${>65} + 3 \cdot \text{Previous DVT/PE} + 2 \cdot \text{Surgery/fracture}{1,\text{month}} + 2 \cdot \text{Malignancy} + 3 \cdot \text{Unilateral leg pain} + 2 \cdot \text{Haemoptysis} + \text{HR}_{\text{band}} + 4 \cdot \text{DVT findings}

where $\text{HR}_{\text{band}}$ is 0 for a heart rate <75, 3 for 75 to 94, and 5 for ≥95 beats/min. The variables are binary (yes/no) apart from the heart rate, which is divided into three bands. The maximum score is 22.

The derivation cohort consisted of consecutive patients presenting to the emergency departments of three European university hospitals with suspected pulmonary embolism [1]. Variables statistically significantly associated with pulmonary embolism on univariate analysis were entered into a multivariable logistic regression model, and points were assigned according to the regression coefficients. The score was then externally validated in an independent cohort from separate diagnostic studies. In the validation cohort the prevalence of pulmonary embolism was 8 per cent in the low-risk group (0 to 3 points), 28 per cent in the intermediate group (4 to 10 points) and 74 per cent in the high-risk group (≥11 points) [1].

Interpretation in practice

The score divides patients into three probability bands. Each band has a concrete pathway for further investigation:

The key difference between the bands does not concern the use of the D-dimer as such, but the fact that high-risk patients should proceed directly to imaging. The D-dimer retains its value in both the low and the intermediate group, but the proportion of patients who actually have an embolism differs markedly between the groups, which affects the predictive value of a positive or negative D-dimer result.

For patients aged 50 years or older, an age-adjusted D-dimer threshold (age × 10 µg/L) may be considered instead of the fixed threshold of 500 µg/L. In a prospective multicentre study of 3,346 patients this increased the proportion of patients in whom pulmonary embolism could be excluded without imaging from 6.4 to 29.7 per cent among patients aged 75 years or older with a non-high clinical probability, with a 3-month failure rate of 0.3 per cent [4].

Validation and performance

A systematic review and network meta-analysis from 2025 included 40 studies with a total of 37,027 patients and compared all established probability scores for pulmonary embolism [2]. The three-level revised Geneva score and the three-level Wells score performed similarly for excluding pulmonary embolism, with negative likelihood ratios of 0.39 (95 per cent credible interval 0.27 to 0.58) and 0.34 (0.25 to 0.45) respectively. The revised Geneva score was, however, superior in correctly assigning patients to imaging rather than to a D-dimer, with a positive likelihood ratio of 6.65 (3.75 to 10.56) versus 5.59 (3.7 to 8.37) for Wells, and a diagnostic odds ratio of 8.03 (4.35 to 14.1) versus 7.4 (4.65 to 11.84). The authors conclude that the revised Geneva score's independence from subjective variables supports its preference over the three-level Wells score [2].

In a Chinese retrospective emergency department study of 3,437 patients (of whom 698 had confirmed pulmonary embolism), the revised Geneva score alone had a sensitivity of 87.2 per cent and a specificity of only 7.2 per cent, with an AUC of 0.54 [3]. Combined with the D-dimer, sensitivity rose to 94.8 per cent and the AUC to 0.69, at the cost of a specificity of 21.3 per cent. Combined with PERC, a sensitivity of 99.6 per cent was reached but with low specificity (34.8 per cent) and an AUC of 0.76 [3]. Agreement between Wells and the revised Geneva score in classifying patients as low or high probability was moderate (kappa 0.46), but increased markedly when the D-dimer or PERC was added [3]. This study's population consisted exclusively of patients of East Asian origin, which limits generalisability.

A smaller Chinese study of 454 patients in training and test sets plus 204 in a validation set compared the revised Geneva score with Wells, the YEARS algorithm and a machine learning model [5]. The revised Geneva score combined with the D-dimer showed the highest specificity among the clinical strategies (0.744), while the AUC was 0.719 for the YEARS algorithm and 0.813 for the machine learning model. The study is a single retrospective study with a small cohort, and the results should be interpreted with caution.

Limitations

The score was derived for and validated in patients presenting acutely with suspected pulmonary embolism. It was not developed for inpatients who develop suspected pulmonary embolism during admission, nor for patients with concurrent COVID-19 or other conditions that alter the baseline probability and the predictive value of the variables.

Four of the eight variables rest on historical information (previous DVT or pulmonary embolism, surgery or fracture within 1 month, active malignancy, haemoptysis) that may be difficult to obtain in patients with reduced consciousness or where there is a language barrier. Interobserver reliability for the individual variables was not studied in the derivation work [1].

The commonest misuse is to treat the score as a diagnostic test rather than a probability tool. The revised Geneva score can neither confirm nor exclude pulmonary embolism on its own. A patient with a low score may still have an embolism, and a patient with a high score may have another diagnosis. The role of the score is to guide the choice between a D-dimer and imaging, not to replace them.

Another pitfall is to use a two-level version (low/high) instead of the three-level version used by this calculator. In the meta-analysis, the two-level Wells score performed significantly worse than all three-level alternatives on every measure [2], and the same probably applies to the revised Geneva score, where the intermediate group has clear clinical value in distinguishing patients who can be investigated with a D-dimer from those who should proceed directly to CT.

A simplified version of the revised Geneva score, in which all variables score 1 point regardless of their original weight, has been proposed for easier use. This version is not what the calculator computes, and its performance differs from that of the weighted original.

References

1. Le Gal G, Righini M, Roy PM et al. Prediction of pulmonary embolism in the emergency department: the revised Geneva score. Ann Intern Med. 2006;144(3):165–71. PMID: 16461960
2. Etemadi A, Hosseini M, Rafiee H 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;25:478353. PMID: 40200307
3. Tang L, Hu Y, Min M et al. Comparisons of clinical scoring systems among suspected pulmonary embolism patients presenting to emergency department. Health Sci Rep. 2024;7(8):e70003. PMID: 39170892
4. Righini M, Van Es J, Den Exter PL et al. Age-adjusted D-dimer cutoff levels to rule out pulmonary embolism: the ADJUST-PE study. JAMA. 2014;311(11):1117–24. PMID: 24643601
5. Xi L, Kang H, Deng M et al. A machine learning model for diagnosing acute pulmonary embolism and comparison with Wells score, revised Geneva score, and Years algorithm. Chin Med J (Engl). 2024;137(6):676–682. PMID: 37828028