Padua Prediction Score for VTE risk

Clinical background

Medical inpatients are at increased risk of venous thromboembolism (VTE) during and after their hospital stay, but the absolute risk varies widely between individuals and is often low. Pharmacological thromboprophylaxis reduces the risk but carries an increased risk of bleeding, and the net benefit depends on the patient's background risk. The Padua score was developed to distinguish medical inpatients whose VTE risk is high enough to justify pharmacological prophylaxis from those whose risk is so low that prophylaxis is not worth the bleeding risk.

Calculating the Padua score

The Padua score is a weighted sum of eleven clinical variables, each of them binary (no = 0, yes = 1):

$$\text{Padua score} = 3 \cdot (\text{active cancer}) + 3 \cdot (\text{previous VTE}) + 3 \cdot (\text{reduced mobility}) + 3 \cdot (\text{thrombophilia}) + 2 \cdot (\text{recent trauma/surgery}) + (\text{age} \geq 70) + (\text{cardiac/respiratory failure}) + (\text{myocardial infarction/stroke}) + (\text{infection/rheumatological disorder}) + (\text{obesity}) + (\text{hormone treatment})$$

The threshold for high risk is $\geq 4$ points.

The score was derived in a prospective cohort of 1,180 consecutive patients admitted to a general medical ward at the University of Padua, Italy, over a two-year period [1]. Follow-up was up to 90 days and the outcome was symptomatic VTE. Of the patients, 469 (39.7%) were classified as high risk. Among high-risk patients who did not receive thromboprophylaxis, 31 of 283 (11.0%) developed VTE, compared with 2 of 711 (0.3%) low-risk patients (HR 32.0, 95% CI 4.1–251.0). Among high-risk patients who received adequate prophylaxis, 4 of 186 (2.2%) developed VTE (HR 0.13, 95% CI 0.04–0.40), and bleeding occurred in 1.6%.

Interpretation in practice

A score in itself is not a prescription. In a patient with a high score but a simultaneously high bleeding risk, the decision may be to withhold pharmacological prophylaxis and use mechanical prophylaxis instead. The score identifies the risk of VTE, not of bleeding, and the two must be weighed separately.

Validation and performance

The Padua score has been validated in more than 16 external studies according to a systematic review from 2021 [2]. The results vary considerably: the c-statistic was often poor (< 0.7), sometimes moderate (0.7–0.8) and rarely excellent (> 0.8). No risk model emerged as clearly superior to the others, and the heterogeneity between studies was too great for meta-analysis.

In a prospective head-to-head comparison of four risk models among 1,352 medical patients at three Swiss university hospitals (the RISE cohort, 2020–2022), discrimination was poor for all the models [3]. The AUC for every model was below 0.60, and after adjustment for thromboprophylaxis and study site the difference in VTE risk between high- and low-risk groups was no longer significant for any model. With the Padua score, 47% of patients were classified as high risk.

In a Japanese retrospective validation of 3,876 medical patients, the c-statistic for the Padua score was 0.64 (95% CI 0.58–0.69) [4]. Calibration was poor: the model systematically overestimated VTE risk in this population, which had a longer length of stay, more intensive care and more reduced mobility than the derivation cohort but a lower proportion with active cancer and obesity.

In a French validation of 14,660 medical patients, the AUC for the Padua score was 0.64 (95% CI 0.61–0.67) [5]. None of the three risk models tested performed better than using age alone as a predictor (AUC 0.61, 95% CI 0.58–0.64).

Limitations

The Padua score applies to non-surgical medical inpatients. It should not be used in surgical patients, who have different risk profiles and different risk models, or in intensive care patients, for whom separate models have been validated.

The score is empirically derived, that is, based on expert judgement and the literature rather than on statistical modelling of a cohort [2]. This means that the point weights are not optimised against data and that the model's transportability between populations is uncertain. External validations have shown this: in Swiss [3], Japanese [4] and French [5] populations alike, discrimination was poor and calibration inadequate.

A common error is to use the Padua score as the sole basis for a decision without considering bleeding risk. The score contains no variables for bleeding tendency, and a patient may have a high VTE risk while pharmacological prophylaxis is contraindicated because of active bleeding, thrombocytopenia or another risk factor. In such cases mechanical prophylaxis may be considered.

The body mass index, required for the obesity variable (BMI $\geq 30$), is missing from up to 44% of clinical records in some series [5], which may lead to the score being underestimated in clinical work.

References

1. Barbar S et al. A risk assessment model for the identification of hospitalized medical patients at risk for venous thromboembolism: the Padua Prediction Score. J Thromb Haemost 2010;8(11):2450–7. PMID: 20738765
2. Pandor A et al. Risk assessment models for venous thromboembolism in hospitalised adult patients: a systematic review. BMJ Open 2021;11(7):e045672. PMID: 34326045
3. Häfliger E et al. Risk Assessment Models for Venous Thromboembolism in Medical Inpatients. JAMA Netw Open 2024;7(5):e249980. PMID: 38728035
4. Arakaki D et al. External Validation of the Padua and IMPROVE-VTE Risk Assessment Models for Predicting Venous Thromboembolism in Hospitalized Adult Medical Patients: A Retrospective Single-Center Study in Japan. Ann Vasc Dis 2023;16(1):60–68. PMID: 37006863
5. Moumneh T et al. Validation of risk assessment models predicting venous thromboembolism in acutely ill medical inpatients: A cohort study. J Thromb Haemost 2020;18(6):1398–1407. PMID: 32168402