To the best of our knowledge, this study is the first cohort study to validate the PPS in predicting the risk of VTE in Chinese internal medicine inpatients. We found that the PPS had limitations not only in the ability of identifying the VTE risk of patients, but also in the performance of discrimination. This retrospective cohort study provided new insights into the risk prediction of VTE in Chinese hospitalized medical patients by comparing PPS and Caprini RAM.
In our study, 217 (13.2%) patients were objectively confirmed with VTE during hospitalization. In the low-risk group, the rate of VTE events was 6.5% by PPS, which was obviously higher than that assessed by Caprini RAM. That is to say, 33.6% of VTE events were not recognized by the PPS in the present setting. The sensitivity of the PPS appears to be lower than that of Caprini RAM. Our findings were consistent with other researches. Zhou et al. found that Caprini RAM assessed 82.3% of VTE patients as high risk, whereas only 30.1% of VTE patients was identified as having high risk by PPS in a Chinese case-control study . It was thought that the Caprini RAM may be considered as the first choice in a general hospital because of higher sensitivity when compared with PPS . Higher specificity was noted when PPS compared with Caprini RAM in our study. However, lower Youden index demonstrated that PPS was inferior to Caprini RAM in authenticity, which is consistent with other studies[16, 17]. As Liu et al. reported previously, PPS has higher specificity (85.6% vs 73.4%) and lower Youden index (0.443 vs 0.010) in a retrospective study .
Our study suggested that the AUC of the Caprini RAM was larger than that of PPS for either all the patients or those without pharmacological prophylaxis, which was fully consistent with existing literatures even if the study design was different [5, 16, 17, 21, 22]. Chen et al found that, in Chinese hospitalized patients, the AUC of the Caprini RAM (0.779) was significantly higher than the value of PPS (0.635, p < 0.05), and concluded that Caprini RAM has a better predictive ability . Although, another Chinese case-control study showed that there was no significant difference between the AUCs of the Caprini RAM and PPS, Caprini RAM was still recommended as the first choice in a general hospital because of its incorporation of comprehensive risk factors, higher sensitivity, and potential for prediction of mortality . In addition, in terms of study design, the case-control study artificially increased the incidence of the disease and modified the tests properties. Therefore, the results of our cohort study are more convincing to reflect the real-world scenario.
The NRI and IDI were proposed by Pencina et al in 2008 to evaluate the added prediction performance of a new marker originally , and they have been advocated and adopted widely in point-based risk scores . In our study, the value of NRI was 0.094, which was calculated by measuring the net change in risk classification in VTE and non-VTE patients, representing improvement in the power of risk prediction. However, this was due to a higher proportion of patients with correct classification in the non-VTE group (0.366) and a lower proportion in the VTE group (-0.272). These indicated that the recognition ability of PPS was still inferior to Caprini RAM in the VTE group. The IDI is a measurement of improvement in differentiation, regardless of risk categories, and can be viewed as an integrated difference in Youden’s indices . In our study, although IDI was positive, there was no statistical difference between the two RAMs. To the best of our knowledge, this is the first study to employ NRI and IDI to evaluate predictive ability of VTE events by the PPS and Caprini RAM in Chinese inpatients. Considering the three indicators of AUC, NRI and IDI comprehensively, Caprini RAM had a better predictive ability than PPS. Although the current guidelines recommend that estimate hospitalized patients at risk of VTE using the PPS for nonsurgical inpatients , the predictive value of PPS in our study is still unsatisfactory.
These differences may be attributed to the flaws of the PPS, which were originally derived from a single center among Caucasian population and included 11 risk factors . The cut-off of BMI (≥ 30) might not be optimal for Asian population. Studies suggested that there was considerable variation in body fat and fat-free mass among various ethnic groups . The factor for “reduced mobility” appeared arduous to anticipate on admission, but it had a high weight (+ 3) and moderate repeatability owing to the variability of clinicians’ comprehension. And some of the thrombophilia were not routinely examined in clinical practice. Therefore, it is rational to infer that the PPS lacks an advantage in predicting VTE risk for Chinese internal medicine inpatients.
In our study, the Caprini RAM showed a better predictive ability and accuracy than PPS. Caprini RAM had been validated efficiently in more than 200 studies, especially for surgical patients [15, 27, 28]. The better prediction performance of Caprini RAM was attributed mostly to the comprehensive risk factors it incorporated. Meanwhile, it’s somewhat cumbersome to implement for medical inpatients. Since it contains information that is difficult to obtain in internal medicine, such as surgery, obstetrics and gynecology, as well as some rare thrombophilia in Chinese population. Additionally, the complex variables of Caprini RAM bring challenges to the feasibility and convenience of clinical work. Furthermore, we found that the specificity of Caprini RAM was lower than that of PPS, which may lead to the occurrence of relatively excessive prevention such as bleeding.
Our study found that the real thromboprophylaxis rate of VTE high-risk patients remains discouraging, which was similar with previously domestic researches . Most of the reasons were due to the non-standard formulation of prevention strategies, insufficient knowledge of physician and fear of bleeding risk associated with prophylaxis. Therefore, it is critical to enhance medical education, and establish VTE risk-assessment systems for Chinese internal medicine inpatients. In addition, VTE events still occurred in 38.5% of patients who received pharmacologic prophylaxis. That is to say, thromboprophylaxis was not associated with a reduction in the occurrence of VTE, which accords with another study in VTE prophylaxis among acute medically ill patients . However, in our cohort, there was no lethal VTE, and 82% of patients had isolated distal DVT. These indicated that although thromboprophylaxis could not avoid the development of VTE, it could effectively reduce the occurrence of fatal VTE. It is reported that a range of 5–10% incidence of fatal VTE in hospitalized patients without prophylaxis [6, 30].
To our knowledge, this study is the first consecutive cohort study to predict VTE risk among hospitalized medical patients. Systematic screening of hospitalized patients for the occurrence of VTE resulted in a more realistic incidence and decreased misdiagnosis of asymptomatic patients. That’s also the main reason why the overall VTE incidence of our study is higher than other researches worldwide [31–33]. Though the diagnosis and treatment of patients with asymptomatic VTE remain controversial [32, 34], correct diagnosis of these patients is essential for enhanced surveillance and management, such as ultrasonic monitoring of isolated distal DVT.
There are a few limitations in our study. Firstly, it was a single center, retrospective cohort study. As with all retrospective studies, there may be challenges in identifying risk factors and estimating the incidence of VTE due to selective bias. However, our researchers are systematically trained to maximize the quality of the data. Secondly, there is a confounder that some of the patients received thromboprophylaxis at the beginning of hospitalization, but we have adjusted it. Moreover, the patients evaluated in this study may be less representative owing to the diseases involved are limited.
In conclusion, PPS seems to be inappropriate to assess VTE risk for internal medicine inpatients in our center. By contrast, Caprini RAM showed greater advantages in predicting VTE risk in terms of sensitivity and accuracy. However, both of these two scoring systems have some deficiencies in our study. Therefore, we need to establish an accurate and more efficient VTE prediction model for Chinese populations in the future.