IMPROVEDD Score for venous thromboembolism (VTE)

Clinical background

Venous thromboembolism (VTE) is one of the commonest hospital-acquired complications in medical inpatients, but the decision about pharmacological prophylaxis must be balanced against bleeding risk and the varying vulnerability of individual patients. The IMPROVEDD risk score was developed to refine the established IMPROVE score by adding the D-dimer as a biomarker. Its purpose is to identify patients whose risk is high enough to justify prophylaxis and, just as importantly, to single out those whose risk is so low that prophylaxis can be avoided. The score is above all a tool for risk stratification at baseline, that is, at the time of admission, and can guide both the decision to start prophylaxis and considerations about extended prophylaxis after discharge.

Calculating the IMPROVEDD risk score

The IMPROVEDD risk score consists of the seven clinical variables of the IMPROVE score plus the D-dimer. It is calculated as:

$$\text{IMPROVEDD} = \text{IMPROVE} + 2 \cdot \mathbb{1}(\text{D-dimer} \geq 2 \times \text{ULN})$$

where the IMPROVE score in turn is:

$$\text{IMPROVE} = 3 \cdot \mathbb{1}$${\text{previous VTE}} + 2 \cdot \mathbb{1}{\text{thrombophilia}} + 2 \cdot \mathbb{1}{\text{paresis/paralysis}} + 2 \cdot \mathbb{1}{\text{cancer}} + 1 \cdot \mathbb{1}{\text{immobilisation} \geq 7\text{ days}} + 1 \cdot \mathbb{1}{\text{ICU/CCU}} + 1 \cdot \mathbb{1}_{\text{age} > 60}

The score ranges from 0 to 14. The variables refer to the patient's status at baseline, that is, at admission. The D-dimer should be measured at baseline and assessed against the local laboratory's upper limit of normal (ULN); the value must be at least twice as high to score 2 points.

The derivation cohort consisted of 7,441 medical inpatients in the APEX trial, a randomised double-blind multicentre study comparing extended betrixaban with standard enoxaparin for 35 to 42 days and 10 ± 4 days respectively [1]. The inclusion criteria comprised acute medical illness (heart failure, respiratory failure, infection, ischaemic stroke or rheumatological disease), certain age and risk factor requirements, expected immobilisation and an expected length of stay of at least 3 days. The modelled outcome was symptomatic deep vein thrombosis, non-fatal pulmonary embolism or VTE-related death up to day 77 after randomisation. Importantly, all patients in the APEX trial received thromboprophylaxis, which affects how the score should be interpreted in a population where prophylaxis is not given routinely.

Interpretation in practice

The IMPROVEDD risk score uses a threshold of 2 points, corresponding to an expected VTE risk of about 1%, which is the limit the American College of Chest Physicians (ACCP) has chosen as justifying pharmacological prophylaxis [1,5]. Patients fall into two main categories:

In the derivation cohort, patients with a score ≥2 had a VTE risk of 2.22% at day 77 compared with 0.91% for those with 0–1 points (HR 2.73, 95% CI 1.52–4.90) [1]. The decision about extended prophylaxis after discharge should also be guided by bleeding risk and the patient's mobility, not by the score alone.

Validation and performance

In the derivation cohort, the addition of the D-dimer improved discrimination measured as the AUC from 0.568 (IMPROVE alone) to 0.625 (IMPROVEDD) at day 77, a difference of 0.06 (95% CI 0.02–0.09, p = 0.0006) [1]. The net reclassification improvement (NRI) was 0.34 (continuous) and 0.13 (categorical), both statistically significant. Calibration was, however, suboptimal in several score categories: scores of 0 and ≥5 tended to overestimate the risk, while scores of 2 and 4 underestimated it [1]. This means that the absolute risk estimates for individual score levels should be interpreted with caution.

External validation has mainly been carried out in COVID-19 populations. Spyropoulos et al. validated IMPROVEDD in a retrospective cohort of 9,407 hospitalised COVID-19 patients and found an AUC of 0.702, with VTE rates of 0.4% (score 0–1), 1.3% (score 2–3) and 5.3% (score ≥4) [2]. The sensitivity for distinguishing low risk from moderate/high risk was 0.971 and the negative predictive value 0.996, while specificity was low (0.218). Goldin et al. reported, in a similar analysis of the same cohort, an AUC of 0.703 for IMPROVEDD compared with 0.635 for IMPROVE without the D-dimer, a difference of 6.8 percentage points (p < 0.0001) [3]. This improvement is in line with that seen in the derivation cohort.

A comprehensive review from 2025 concludes that the IMPROVE score has shown consistent reliability in medical patient populations, but that IMPROVEDD has mainly been validated within the APEX trial and COVID-19 cohorts, and that external validation in broader medical inpatient populations is lacking [4]. The review also points out that the D-dimer may reflect inflammation, malignancy or acute illness rather than thrombosis as such, and that the biomarker should therefore be interpreted within the multivariable model rather than in isolation.

Limitations

The IMPROVEDD risk score applies to medical inpatients. It is not validated for surgical patients, obstetric patients or patients on psychiatric wards. In trauma or major surgery, other models should be used.

All patients in the derivation cohort received thromboprophylaxis (betrixaban or enoxaparin) [1,4]. This means that the reported VTE rates reflect a population receiving prophylaxis, and the score's ability to predict VTE in patients not receiving prophylaxis has not been studied directly. Anyone using the score to decide about prophylaxis should therefore be aware that the thresholds derive from a cohort in which most patients were already being treated.

The D-dimer is non-specific and rises in inflammation, infection, malignancy and old age, all common in medical inpatients. A raised value does not confirm thrombosis but contributes a risk weighting within the model. The score should not be used as a diagnostic test in a patient with suspected VTE, but as a risk stratification tool at baseline.

Calibration is uneven across score levels, with systematic overestimation at low (0) and high (≥5) scores and underestimation at intermediate levels [1]. Clinically this means that patients with a very high score do not necessarily have a proportionally higher absolute risk than the score suggests.

Finally, external validation in general medical inpatient populations outside the COVID-19 setting is lacking. The COVID-19 cohorts in which IMPROVEDD has been validated were characterised by an exceptionally high thrombotic risk and widespread hyperinflammation, which may have given the D-dimer a stronger predictive weight than applies in other acute medical conditions [2,3,4].

References

1. Gibson CM et al. The IMPROVEDD VTE Risk Score: Incorporation of D-Dimer into the IMPROVE Score to Improve Venous Thromboembolism Risk Stratification. TH Open. 2017. PMID: 31249911
2. Spyropoulos AC et al. Validation of the IMPROVE-DD risk assessment model for venous thromboembolism among hospitalized patients with COVID-19. Res Pract Thromb Haemost. 2021. PMID: 33733028
3. Goldin M et al. External validation of the IMPROVE-DD risk assessment model for venous thromboembolism among inpatients with COVID-19. J Thromb Thrombolysis. 2021. PMID: 34146235
4. Alfehaid L et al. A comprehensive review of venous thromboembolism risk assessment models for hospitalized medical patients: comparative evidence, implementation challenges, and future directions. Front Cardiovasc Med. 2025. PMID: 41602342
5. Stevens SM et al. Antithrombotic Therapy for VTE Disease: Compendium and Review of CHEST Guidelines 2012-2021. Chest. 2024. PMID: 38458430