Our data show that hospital-acquired IDDVT is a more frequent finding than proximal DVT and its prevalence in patients with low PPS is similar than in those with high PPS, in contrast with the aim of PPS itself, such as the stratification of VTE risk. Of note, PPS cannot be used to predict and diagnose DVT.
It has already been clearly established that hospitalization is one of the major factors for the risk of VTE and hospitalization for acute medical illness is associated with an eightfold increased risk of VTE [10]. As shown by Heit et al., the overall VTE incidence rate in a cohort of patients who resided in Olmsted County, Minnesota, was of 960.5 per 10,000 person-years in hospitalized patients while it was 100 times less in community residents (7.1 per 10,000 person-years) [11]. We have already shown that DVT prevalence was 1.1% in our hospital, in line with data from a large population of US medical patients, among whom 2.0% of all patients experienced a DVT during their hospitalization [12]. In accordance with these data, past and present guidelines recommend anticoagulant thrombo-prophylaxis with low-molecular-weight heparin for acutely ill hospitalized medical patients at increased risk of thrombosis [4, 8]. At the time of the study design, the 9th American College of Chest Physicians Evidence-Based Clinical Practice Guidelines suggested PPS for VTE risk [4]. PPS was validated based on several studies with conflicting results [13]. Vardi et al. studied VTE risk among 1080 patients hospitalized because of sepsis; they showed that 71.2% of the patients had a positive PPS and this was highly associated with death and may reflect a more general co-morbidity and disease severity index [14]. The ESTIMATE study, the first study to test the PPS in a multicenter setting, showed that both Geneva Risk Score and PPS were strongly associated with the composite endpoint of symptomatic VTE or VTE-related death [15]. PPS was compared to the Caprini RAM (risk assessment model) in a Chinese case-control study by Zhou et al; the Caprini score showed greater sensitivity in identifying high risk hospitalized patients (82.3% of high risk patients according to the Caprini RAM had VTE vs 30.1% according to PPS), even if VTE risk associated to highest risk groups determined by both models was similar [16]. Also Liu et al found the Caprini RAM having a greater sensitivity and positive and negative predictive values than the Padua RAM, although PPS had a higher specificity [17]. In addition, an Automated Padua Prediction Score (APPS) to auto-calculate a VTE risk score using electronic health record was developed, showing no significant difference in average score and a similar ability in predicting VTE risk [18].
Among the many available RAMs, in 2018 the “TEVere Score” was developed by Vincentelli et al, based on VTE risk factors with higher statistical significance, and it showed a higher specificity and sensitivity (respectively 43.3 and 87.5, with accuracy 72.1) compared with the PPS [19]. Of note, a retrospective analysis on patients prospectively enrolled in the PREVENU trial, aimed at comparing the main RAMs (including the Caprini score, IMPROVE and PPS), showed that none of them performed significantly better than advanced age as a single predictor [6]. Also Wang et al. compared the PPS with nine machine learning methods, since the PPS model is not suitable for the Chinese population because of differences in race and disease spectrum; nevertheless they showed lower sensitivities to that of the PPS [20]. In our series, we expected a lower prevalence of DVT in the low risk PPS group vs those with PPS ≥ 4, whereas IDDVT prevalence was similar in patients with high and low PPS score and proximal DVT prevalence was only slightly higher in patients with high PPS vs those with low PPS. This evidence is in line with a recent review on risk assessment models for VTE in hospitalised adult patients that found a modest ability of PPS in predicting the risk of VTE [13]. In the Prevention of Venous Thromboembolism Disease in Emergency Departments (PREVENU) study on 14 660 patients hospitalized for at least 2 days in a medical ward, PPS performance was not superior than an advanced age alone in VTE risk assessment [6]. Moreover, in a multicenter retrospective cohort study including over 1 million of unselected consecutive hospitalizations across the United States, the PPS demonstrated limited predictive ability with a PPS discriminatory accuracy for VTE risk of 0.59 [21], in line with our results (discriminatory accuracy of 0.62).
Our study shows the prevalence of hospital acquired IDDVT is higher than the prevalence of proximal DVT, in opposition with our findings in outpatients [22]. Since the prevalence of inpatients DVT is higher than community-acquired DVT, such difference may be at least partially due to an elevated prevalence of isolated distal DVT. Our study suggests that patients in medical wards are at higher risk of IDDVT than outpatients. In fact, the Riete registry showed that IDDVT was more frequently associated to transient risk factors (i.e. recent travel, hospitalization, recent surgery), whereas proximal DVTs were more frequently associated with chronic states [23].
Patients receiving pharmacological thrombo-prophylaxis had a similar prevalence of DVT than those without thrombo-prophylaxis. This may be due to several reasons: 1) our participants were enrolled because of suspected DVT and this could represent a potential selection bias; 2) we performed DUS within 24h from clinical suspicion and this could have led to an early diagnosis before extension to the proximal veins; 3) we only enrolled patients complaining of symptoms and could have missed asymptomatic thrombosis; 4) the study was not prospective; 5) the use of mechanical prophylaxis in several patients could have reduced DVT prevalence in patients not receiving pharmacological thrombo-prophylaxis.
The use of pharmacological thrombo-prophylaxis was similar in patients with a high and a low PPS. This is in line with a recent meta-analysis showing that thrombo-prophylaxis prescriptions were still unsatisfactory among hospitalized medically ill patients in several countries [24]. Despite guidelines recommendations, adherence to thrombo-prophylaxis remains moderate, with almost 40% of patients at high risk according to PPS that do not receive prophylaxis [24]. These results further support a call to action for pharmacological thrombo-prophylaxis in medical patients because they are at risk not only for DVT, but also for PE [8]. The risk assessment models should aim to help clinicians selecting medical inpatients who are at increased risk of VTE and may benefit of prophylaxis. However, no risk assessment model had satisfactory performances in this setting and which risk assessment model is optimal is still uncertain. Even though not all patients may benefit from thrombo-prophylaxis, our data support the use of thrombo-prophylaxis in all medical inpatients without contraindications or high bleeding risk, as recently suggested [25].
Some limitations of the present study should be acknowledged. No inter-observer variability was assessed for IDDVT diagnosis; we did not follow-up patients with negative whole-leg ultrasonography examination, but several studies have shown that anticoagulant therapy can be safely withheld after negative complete compression ultrasound without further testing [22, 26] also in inpatients [27]. The prevalence of DVT may have been underestimated since we did not evaluate patients with asymptomatic DVT or with symptomatic DVT who were discharged before referral to our service. The study was conducted in a single academic institution and may not be representative of population in different types of hospitals. We must underline that our participants were enrolled because suspected DVT, suggesting a potential selection bias.
In summary, IDDVT is a frequent finding in inpatients and its prevalence is not related to PPS, in contrast with the aim of PPS itself. Our study supports that clinical judgment should be integrated with risk assessment models for VTE in medical inpatients.