Noam Shohat, Joseph Caprini, Pascal-André Vendittoli.
Response/Recommendation: There is currently no validated risk score that can be used across all orthopaedic subspecialties. Most of the studies concerning risk scores have originated from joint replacement literature and generally use similar risk factors that classify patients as either high- or low-risk. Unfortunately, none of these scores have been properly externally validated. They also lack any assessment of major bleeding events. Furthermore, the impact of these risk scores on patient outcomes and decision-making remains unknown. Additional studies are required to address these major limitations.
Strength of Recommendation: Moderate.
Rationale: There are numerous risk factors for the development of venous thromboembolism (VTE) after an orthopaedic surgical procedure. These can be generally classified as either host- or surgery-related1–9. While it is important to know the individual risk factors associated with VTE, this alone does not always contribute to clinical decision-making and overall risk stratification. In the era of personalized medicine, individualized risk scores are increasingly needed, especially when taking into consideration the many existing pharmacological and non-pharmacological options for VTE prophylaxis. Improved risk stratification could help to better target these prophylactic measures for individual patients.
An ideal risk stratification model should include the following key components:
- Risk scores that are derived from data that is both granular and contemporary, taking into account recent changes in orthopaedic procedures as well as enhanced recovery protocols.
- Procedure-specific risk scores, which consider risk factors that are unique to each orthopaedic subspecialty.
- Account for modifiable factors such as unilateral versus bilateral procedures, type of anesthesia, and the use of tranexamic acid10.
- Include the modifying effect that different antithrombotic agents have on the overall risk. It is often assumed that patients with multiple risk factors and deemed to be high-risk based on risk calculators would benefit from more potent chemoprophylaxis. However, recent studies have suggested otherwise11.
- Be easy to use and interactive, allowing the clinician to assess how a patient’s individual risk can be modified.
- Predict major bleeding events (including wound-related issues), which are influenced by the type of chemoprophylaxis chosen.
- Make a clear distinction between pulmonary embolism (PE), proximal deep venous thrombosis (DVT) and distal DVT, as each may have different risk factors as well as different implications with regard to treatment.
- Include external validation, ideally in a global context, in multiple consistent studies and across various subspecialties. For example, the exponential increase in thrombotic risk with the presence of multiple comorbidities has been shown in other specialties but needs to be investigated in the arthroplasty population. Machine learning tools may be used to identify the interactions between multiple risk factors.
A systematic review of the literature was performed that included only papers concerning the development or validation of a risk stratification model. Papers that involved non-surgical treatments were excluded. Papers were also excluded if only preoperative VTE was assessed. Our literature search resulted in 513 potentially relevant papers. Of those, 426 were excluded on the basis of titles/abstracts and 87 full-text articles were reviewed. Of these 87 articles, only 10 studies met the abovementioned criteria; eight articles involved total joint arthroplasty (TJA) patients, one article involved lower extremity trauma patients, and one article involved foot and ankle patients. No randomized controlled trials were identified. Only one study evaluated patients prospectively. The number of component variables in a single risk score ranged from 5 to 25. Seven different risk scores or institutional protocols were identified. The number of patients included in the studies ranged from 217 to 1’721,806. Our assessment of these current risk stratification models (Table 1) was performed in comparison to the abovementioned ideal.
In general, the majority of the risk scores identified similar risk factors associated with VTE. While they differed in the relative weight given to each variable in the scoring system, all scores were similar in their attempt to stratify patients into risk groups under the assumption that this should guide the choice of chemoprophylaxis. The specific threshold that was chosen ultimately affects the performance (i.e., sensitivity and specificity) of the individual risk score. The Caprini risk score represents an excellent example. While this score has served the medical and surgical community for many years and has had great success outside of orthopaedics, it is not widely accepted in orthopaedics as the original threshold score of 5 would automatically place all orthopaedic patients in the high-risk group. This would obviously result in capturing all VTE events (i.e., high sensitivity), but at the cost of extremely low specificity. Acknowledging this issue, a recent study investigated the optimal threshold (maximizing sensitivity and specificity) for stratifying patients and found a score of 10 to be ideal, thus improving specificity at the expense of sensitivity. Choosing this threshold for prophylaxis would result in prescribing warfarin or other non-aspirin alternatives to nearly 40% of patients. Other scoring systems identified different, and at times arbitrary, thresholds defining “high-risk” patients, resulting in 1.8% to 9.3% of patients being included in this category. This affects the interpretation of each risk score and does not allow a direct comparison of their performance across different studies. Establishing an acceptable common threshold above which patients should be considered “high-risk” may help to standardize these risk scores. Notwithstanding, the assumption that “high-risk” patients may require more aggressive anticoagulation is still contentious, as recent studies have suggested that aspirin may be appropriate for higher-risk groups11. With advances in the field of machine learning and taking into account that current institutional data includes “high-risk” patients receiving aspirin, a more personalized approach to risk stratification may be feasible. Instead of classifying patients into groups, it may be possible to report the specific VTE probability for individual patients and allow clinicians to make their own decision with regard to the optimal chemoprophylaxis based on their assessment of what is an “acceptable” risk.
External validation was lacking for all studies. The included studies could be broadly subdivided into studies in which scoring systems were formulated and evaluated within the same cohort or a holdout of patients3,12, and other studies that set out to validate an existing score or protocol13–18. The majority of studies that showed encouraging results with external validation were conducted by the same authors that developed the score, thus leading to concerns with regard to reproducibility and generalization. Nam et. al.,18 were the only group that prospectively evaluated their institutional protocol and thus were able to take into account the influence of chemoprophylaxis in their evaluation. Using a simple protocol, they categorized patients into either “routine-” (75.4%) or “high-” (24.6%) risk groups and showed a 0.5% VTE rate in both groups. Whether their protocol simply did not capture all “high-risk” patients or whether the VTE rate was affected by the use of more aggressive anticoagulation in the “high-risk” group remains debatable. The Caprini score was the only score to be evaluated by external groups and across fields other than joint replacement, aside from one external study16 that also evaluated the VTEstimator score. Bateman et al.,16 retrospectively evaluated the Caprini and VTEstimator score in a group of 363 total knee arthroplasty (TKA) and total hip arthroplasty (THA) patients. They failed to show an association between mean scores and VTE risk. However, the study suffered from many methodological issues including the small sample size and low event rate (only 10 VTE), an inability to assess the scores adequately due to missing data, and the evaluation of scores as a continuous variable as opposed to a categorical variable for risk stratification19. Krauss et al.,13 compared a departmental protocol with the Caprini score and showed that using a threshold of 10, the latter was able to capture 7 out of 8 VTE compared to only 1 event that was captured using the former. Notably, this threshold of 10 was chosen to optimize the results of the Caprini score within that specific cohort (using the Youden index), and therefore does not reflect a true external validation of the score. More recently in a cohort of 2,155 TJA patients, Gold et al,14 failed to find an association between high Caprini scores (when evaluated continuously and categorically using 11 as the threshold) and VTE risk when taking into account chemoprophylaxis in a multivariate analysis. Outside of joint arthroplasty, two studies have evaluated the Caprini score. Saragas et al.,17 failed to show the utility of the Caprini score in a heterogeneous group of foot and ankle patients; however, they used the original 5 point threshold for stratifying patients and the sample size was too small to perform statistical analysis. Dashe et al.,15 examined a group of lower extremity fractures. The Caprini score was no different between “low-” (isolated foot and ankle) and “high-” (pelvic/acetabular) risk fractures, although the latter cohort was more prone to VTE. While the optimal cutoff to predict VTE was 11-12, the actual performance of the score could not truly be evaluated. Furthermore, the group studied included both preoperative and postoperative VTE, thereby making it difficult to interpret the results.
Another major limitation of all but two of the current risk scores is the inclusion of PE and DVT (proximal and distal) as one combined outcome. Some studies also included asymptomatic and isolated distal DVT, the importance of which remains unknown. Additionally, there is evidence suggesting that PE and DVT are two distinct entities, and hence individual risk factors may need to be weighed differently with respect to each entity. In a cohort of 1078 patients, Krauss et al.,13 were able to capture 7/394 VTE in the “high-risk” patients, while only 1/684 “low-risk” patients had a VTE. Interestingly, only 1 of the 7 “high-risk” patients had a PE, while the single patient in the “low-risk” group had a PE as well. If one were to consider PE as the primary outcome, the occurrence of PE would in fact be a sporadic event, and no association with any of the risk scores would be found. On further analysis, the PE patient in the low-risk group was later found to have a congenital thrombophilic factor that would have placed the patient in the high-risk category.
Finally, one inherent limitation to all of the current scoring systems is the failure of these scores to account for major bleeding events or wound complications. The sensitivities and specificities of the different scores affect the overall number of patients receiving aggressive anticoagulation, as mentioned above. Not only may this have a financial impact, but there may also be a direct effect on the number of major bleeding events or wound complications that could complicate VTE prophylaxis. Predicting these adverse events may be as important as predicting VTE risk. Nam et al.,18 reported on the treatment of 75.4% “routine-risk” patients with aspirin and 24.6% “high-risk” patients with warfarin. While VTE rates were 0.5% in both groups, patients in the former cohort had a lower rate of major bleeding (0.5% versus 2.0%, p=0.006) and wound complications (0.2% versus 1.2%, p=0.01) compared to the later cohort. It is clear that decisions which modify a patient’s VTE risk may also affect bleeding risk, and contemporary risk assessment tools should ideally take this into account. As such, future VTE risk scoring systems should also incorporate the risk of major bleeding events and wound complications.
Table 1: Studies included in the systematic review.
|Study (Country, Year)||Data source||Population (Sample size)||Predictors||External validation||Includes procedure-specific risk factors||Chemoprophylaxis included in the model||Endpoint that was statistically assessed||Key findings|
|Dauty (France, 2012)20||Institutional data (at least 4 centers; both public and private)||TKA (primary/ revision not mentioned) (n=272)||RAPT score||No||No||No||Symptomatic DVT||RAPT score<6 was associated with a 3.0 relative risk for DVT|
|Saragas (South Africa, 2013)17||Single institution||Foot and ankle surgeries (n=216)||Caprini score||Yes||No||No||Any documented DVT or PE||No significant difference in number of risk factors in the VTE and non-VTE groups. 90.9% of patients in the VTE group and 73.7% of patients in the non-VTE group had a total risk factor score of ≥5 (no statistical comparison provided)|
|Parvizi (USA, 2014)2||Single institution||TKA and THA (primary and revision) (n=26,391)||Knee surgery, CCI, atrial fibrillation, postoperative DVT, COPD, anemia, depression, BMI||No||Limited||No||Symptomatic PE||Patients were classified into low- (0.35%), medium- (1.4%), and high- (9.3%) risk categories|
|Nam (USA, 2015)18||Single institution||THA (Primary and revision) and hip resurfacing (n=1,859)||Institutional protocol: Age (>70), previous DVT, active cancer, hypercoagulability, multiple comorbidities, morbid obesity, family history of VTE, immobility.||Yes (Prospective)||No||No||Symptomatic DVT or PE||75.4% were categorized as routine risk and 24.6% as high risk. The cumulative rate of VTE was 0.5% in the routine and 0.5% in the high-risk cohort within 6 weeks postoperatively (p=1.00). Patients in the routine risk cohort had a lower rate of major bleeding (0.5% versus 2.0%, p=0.006) and wound complications (0.2% versus 1.2%, p=0.01)|
|Parvizi (USA, 2016)3||Nationwide Inpatient Sample||TKA and THA (primary and revision) (n=1’721,806)||VTEstimator||Yes. On institutional data. (n=25,775)||Limited||No||Any documented DVT or PE||A score above 75 was provisionally chosen to dichotomize patients into low- and high-risk. Above this threshold, the rate of VTE in the NIS group was 1.68% and 3.85% in the validation group|
|Bohl (USA, 2016)12||American College of Surgeons National Surgical Quality Improvement Program||Primary TKA and THA (n=118,473)||Age, sex, BMI, preoperative hematocrit, knee surgery.||Yes. On institutional data. (n=17,384)||No||No||Symptomatic PE||Patients with a score between 9 and 12 had an increased risk of PE in the validation group (2.6%)|
|Bateman (USA, 2017)16||Single institution||Primary TKA and THA (n=363)||Caprini and VTEstimator||Yes||No||No||Symptomatic PE and DVT||Mean Caprini and VTEstimator scores were not different between those who did and did not develop VTE|
|Dashe (USA, 2019)15||Single institution||Lower extremity fractures grouped into “low-” risk (isolated foot and ankle) and “high-” risk (pelvis and acetabulum)||Caprini score||Yes||No||No||Any documented DVT or PE before or after surgery.||Caprini score was not different between “low-” and “high-” risk fractures. The cutoff that best-predicted VTE events based on receiver-operating curves was 12 (c = 0.74) in the high-risk group, 11 (c = 0.79) in the low-risk group.|
|Krauss (USA, 2019)13||Single institution||TKA and THA (n=1,078)||Institutional protocol and Caprini score||Yes||No||Reported but not statistically evaluated||Symptomatic PE and DVT||Patients were dichotomized based on a threshold score of 10. 7/394 in the high-risk group developed VTE compared to 1/684 in the low-risk group.|
|Gold (USA, 2020)14||Single institution||Primary TKA and THA (n=2,155)||Caprini score||Yes||No||Taken into consideration for validation||Symptomatic PE and DVT||Higher Caprini scores (continuous and dichotomized with a threshold of 11) were not associated with increased VTE risk when controlling for VTE chemoprophylaxis|
TKA=Total knee arthroplasty; RAPT=Risk Assessment and Prediction Tool; DVT=Deep venous thrombosis; PE=Pulmonary embolism; VTE=Venous thromboembolism; THA=Total hips arthroplasty; CCI=Charlson comorbidity index; COPD=Chronic obstructive pulmonary disease; BMI=Body mass index; NIS=Nationwide inpatient sample.
1. Zhang Z, Shen B, Yang J, Zhou Z, Kang P, Pei F. Risk factors for venous thromboembolism of total hip arthroplasty and total knee arthroplasty: a systematic review of evidences in ten years. BMC Musculoskelet Disord. 2015;16:24. doi:10.1186/s12891-015-0470-0
2. Parvizi J, Huang R, Raphael IJ, Arnold WV, Rothman RH. Symptomatic pulmonary embolus after joint arthroplasty: stratification of risk factors. Clin Orthop Relat Res. 2014;472(3):903-912. doi:10.1007/s11999-013-3358-z
3. Parvizi J, Huang R, Rezapoor M, Bagheri B, Maltenfort MG. Individualized Risk Model for Venous Thromboembolism After Total Joint Arthroplasty. J Arthroplasty. 2016;31(9 Suppl):180-186. doi:10.1016/j.arth.2016.02.077
4. Parvizi J. Dvt prophylaxis: risk stratification solutions. Orthopaedic Proceedings. 2017;99-B(SUPP_7):40-40. doi:10.1302/1358-992X.99BSUPP_7.CCJR2016-040
5. Januel J-M, Chen G, Ruffieux C, et al. Symptomatic in-hospital deep vein thrombosis and pulmonary embolism following hip and knee arthroplasty among patients receiving recommended prophylaxis: a systematic review. JAMA. 2012;307(3):294-303. doi:10.1001/jama.2011.2029
6. Falck-Ytter Y, Francis CW, Johanson NA, et al. Prevention of VTE in orthopedic surgery patients: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest. 2012;141(2 Suppl):e278S-e325S. doi:10.1378/chest.11-2404
7. Bizien N, Noel-Savina E, Tromeur C, et al. Age is a major risk factor of venous thromboembolism (VTE). European Respiratory Journal. 2011;38(Suppl 55). Accessed August 23, 2021. https://erj.ersjournals.com/content/38/Suppl_55/p3936
8. Anderson FA, Spencer FA. Risk Factors for Venous Thromboembolism. Circulation. 2003;107(23_suppl_1):I-9. doi:10.1161/01.CIR.0000078469.07362.E6
9. Abdol Razak NB, Jones G, Bhandari M, Berndt MC, Metharom P. Cancer-Associated Thrombosis: An Overview of Mechanisms, Risk Factors, and Treatment. Cancers. 2018;10(10):380. doi:10.3390/cancers10100380
10. Warren JA, Sundaram K, Anis HK, Kamath AF, Higuera CA, Piuzzi NS. Have Venous Thromboembolism Rates Decreased in Total Hip and Knee Arthroplasty? J Arthroplasty. 2020;35(1):259-264. doi:10.1016/j.arth.2019.08.049
11. Huang RC, Parvizi J, Hozack WJ, Chen AF, Austin MS. Aspirin Is as Effective as and Safer Than Warfarin for Patients at Higher Risk of Venous Thromboembolism Undergoing Total Joint Arthroplasty. J Arthroplasty. 2016;31(9 Suppl):83-86. doi:10.1016/j.arth.2016.02.074
12. Bohl DD, Maltenfort MG, Huang R, Parvizi J, Lieberman JR, Della Valle CJ. Development and Validation of a Risk Stratification System for Pulmonary Embolism After Elective Primary Total Joint Arthroplasty. J Arthroplasty. 2016;31(9 Suppl):187-191. doi:10.1016/j.arth.2016.02.080
13. Krauss ES, Segal A, Cronin M, et al. Implementation and Validation of the 2013 Caprini Score for Risk Stratification of Arthroplasty Patients in the Prevention of Venous Thrombosis. Clin Appl Thromb Hemost. 2019;25:1076029619838066. doi:10.1177/1076029619838066
14. Gold PA, Ng TY, Coury JR, et al. Can the Caprini score predict thromboembolism and guide pharmacologic prophylaxis after primary joint arthroplasty? J Orthop. 2020;21:345-349. doi:10.1016/j.jor.2020.07.005
15. Dashe J, Parisien RL, Pina M, De Giacomo AF, Tornetta P. Is the Caprini Score Predictive of Venothromboembolism Events in Orthopaedic Fracture Patients? J Orthop Trauma. 2019;33(6):269-275. doi:10.1097/BOT.0000000000001451
16. Bateman DK, Dow RW, Brzezinski A, Bar-Eli HY, Kayiaros ST. Correlation of the Caprini Score and Venous Thromboembolism Incidence Following Primary Total Joint Arthroplasty-Results of a Single-Institution Protocol. J Arthroplasty. 2017;32(12):3735-3741. doi:10.1016/j.arth.2017.06.042
17. Saragas NP, Ferrao PNF, Saragas E, Jacobson BF. The impact of risk assessment on the implementation of venous thromboembolism prophylaxis in foot and ankle surgery. Foot Ankle Surg. 2014;20(2):85-89. doi:10.1016/j.fas.2013.11.002
18. Nam D, Nunley RM, Johnson SR, Keeney JA, Clohisy JC, Barrack RL. Thromboembolism Prophylaxis in Hip Arthroplasty: Routine and High Risk Patients. J Arthroplasty. 2015;30(12):2299-2303. doi:10.1016/j.arth.2015.06.045
19. Diaz Quintero LA, Fuentes HE, Salazar Adum JP, Tafur AJ, Kudrna JC, Caprini JA. Letter to the Editor on “Correlation of the Caprini Score and Venous Thromboembolism Incidence Following Primary Total Joint Arthroplasty-Results of a Single-Institution Protocol.” J Arthroplasty. 2018;33(8):2697-2698. doi:10.1016/j.arth.2018.03.065
20. Dauty M, Schmitt X, Menu P, Rousseau B, Dubois C. Using the Risk Assessment and Predictor Tool (RAPT) for patients after total knee replacement surgery. Ann Phys Rehabil Med. 2012;55(1):4-15. doi:10.1016/j.rehab.2011.10.006