Predictive and Associative Models to Identify Hospitalized Medical Patients at Risk for VTE

CHEST

Original Research PULMONARY VASCULAR DISEASE

Predictive and Associative Models to Identify Hospitalized Medical Patients at Risk for VTE Alex C. Spyropoulos, MD, FCCP; Frederick A. Anderson Jr, PhD; Gordon FitzGerald, PhD; Herve Decousus, MD; Mario Pini, MD; Beng H. Chong, MD, PhD; Rainer B. Zotz, MD; Jean-François Bergmann, MD; Victor Tapson, MD, FCCP; James B. Froehlich, MD, MPH; Manuel Monreal, MD; Geno J. Merli, MD; Ricardo Pavanello, MD; Alexander G. G. Turpie, MD; Mashio Nakamura, MD; Franco Piovella, MD; Ajay K. Kakkar, MBBS, PhD; and Frederick A. Spencer, MD; for the IMPROVE Investigators

Background: Acutely ill hospitalized medical patients are at risk for VTE. We assessed the incidence of VTE in the observational International Medical Prevention Registry on Venous Thromboembolism (IMPROVE) study and derived VTE risk assessment scores at admission and associative VTE scores during hospitalization. Methods: Data from 15,156 medical patients were analyzed to determine the cumulative incidence of clinically observed VTE over 3 months after admission. Multiple regression analysis identified factors associated with VTE risk. Results: Of the 184 patients who developed symptomatic VTE, 76 had pulmonary embolism, and 67 had lower-extremity DVT. Cumulative VTE incidence was 1.0%; 45% of events occurred after discharge. Factors independently associated with VTE were previous VTE, known thrombophilia, cancer, age . 60 years, lower-limb paralysis, immobilization ⱖ 7 days, and admission to an ICU or coronary care unit (first four were available at admission). Points were assigned to each factor identified to give a total risk score for each patient. At admission, 67% of patients had a score ⱖ 1. During hospitalization, 31% had a score ⱖ 2; for a score of 2 or 3, observed VTE risk was 1.5% vs 5.7% for a score ⱖ 4. Observed and predicted rates were similar for both models (C statistic, 0.65 and 0.69, respectively). During hospitalization, a score ⱖ 2 was associated with higher overall and VTE-related mortality. Conclusions: Weighted VTE risk scores derived from four clinical risk factors at hospital admission can predict VTE risk in acutely ill hospitalized medical patients. Scores derived from seven clinical factors during hospitalization may help us to further understand symptomatic VTE risk. These scores require external validation. CHEST 2011; 140(3):706–714 Abbreviations: ACCP 5 American College of Chest Physicians; CCU 5 coronary care unit; HR 5 hazard ratio; IMPROVE 5 International Medical Prevention Registry on Venous Thromboembolism; PE 5 pulmonary embolism; RAM 5 risk assessment model; RCT 5 randomized controlled trial

ill hospitalized medical patients are at risk Acutely for VTE. Although originally considered a com-

plication of surgery,1  75% of VTE-related deaths occur in nonsurgical populations.2 Medical patients are more prone to potentially fatal pulmonary embolism (PE), have more severe comorbid conditions, and have VTE prophylaxis more frequently omitted.3-5 Pharmacologic prophylaxis for ⱕ 14 days in acutely ill hospitalized medical patients is safe and effective

in reducing VTE risk.3,6,7 The American College of Chest Physicians (ACCP) has high-grade recommendations for thromboprophylaxis in this population.1 The ACCP clinical guidelines have a group-specific thromboembolic risk assessment strategy in hospitalized medical patients. This strategy is attractive because of the complex eligibility criteria of the randomized studies and the lack of evidence for the application of specific VTE risk factors in assessing a

706

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians

patient’s overall risk. However, this approach may contribute to the significant underuse and heterogeneity of ACCP-recommended thromboprophylaxis practices across medical patient groups.1,5,8 Although VTE risk models have been developed in surgical patients,9 there is limited information on individual risk factors present in hospitalized medical patients at admission and to what extent they interact to determine thromboembolic risk. Most available data on VTE risk factors in acutely ill medical patients are derived from patient subgroups within randomized controlled trials (RCTs) rather than from reallife cohorts.2,10,11 Risk assessment models (RAMs) for determining VTE risk focus predominantly on cancer patients and pregnant women.12-14 Presently, there is no weighted scoring system of individual VTE risk factors that allows physicians to estimate the probability of VTE across a range of diagnoses in acutely ill hospitalized medical patients. The International Medical Prevention Registry on Venous Thromboembolism (IMPROVE) was designed to examine VTE prophylaxis practices and clinical outcomes in patients hospitalized for an acute medical illness with a range of diagnoses. Data were collected from 15,156 patients in 12 countries and 52 hospitals.8 Manuscript received August 4, 2010; revision accepted February 9, 2011. Affiliations: From the Hamilton Health Sciences General Hospital (Drs Spyropoulos and Turpie), McMaster University, and Hamilton Health Sciences (Dr Spencer), McMaster University, Hamilton, ON, Canada; Center for Outcomes Research (Drs Anderson and FitzGerald), University of Massachusetts Medical School, Worcester, MA; INSERM, CIE3, Saint-Etienne (Dr Decousus), University Saint-Etienne, and CHU Saint-Etienne, Hôpital Nord, Service de Médecine Interne et Thérapeutique, Saint-Etienne, France; Medicina Interna II (Dr Pini), Fidenza Hospital, Parma, Italy; St. George Clinical School (Dr. Chong), University of New South Wales, Sydney, NSW, Australia; Hämostase-Institut Düsseldorf (Dr Zotz), Düsseldorf, Germany; Hôpital Lariboisiere Clinique Thérapeutique (Dr Bergmann), University Paris Diderot, Paris, France; Duke University Medical Center (Dr Tapson), Durham, NC; Vascular Medicine (Dr Froehlich), University of Michigan Health System, Ann Arbor, MI; Servicio de Medicina Interna (Dr Monreal), Hospital Germans Trias i Pujol, Badalona, Spain; Jefferson Vascular Diseases Center (Dr Merli), Departments of Surgery and Medicine, Thomas Jefferson University Hospital, Philadelphia, PA; Hospital do Coracao Clinica Medica São Paulo (Dr Pavanello), São Paulo, Brazil; Department of Cardiology (Dr Nakamura), Mie University Graduate School of Medicine, Tsu Mie, Japan; U.O. Angiologia (Dr Piovella), Fondazione IRCCS Policlinico San Matteo, Pavia, Italy; and Thrombosis Research Institute and University College London (Dr Kakkar), London, England. Funding/Support: The IMPROVE study was supported by a grant from Sanofi-Aventis to the Center for Outcomes Research at the University of Massachusetts Medical School. Correspondence to: Alex C. Spyropoulos, MD, FCCP, McMaster University, Hamilton General Hospital, Thrombosis Unit, 6th Floor, McMaster Clinic, 237 Barton St E, Hamilton, ON, L8L 2X2, Canada; e-mail: [email protected] © 2011 American College of Chest Physicians. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (http://www.chestpubs.org/ site/misc/reprints.xhtml). DOI: 10.1378/chest.10-1944 www.chestpubs.org

After determining the incidence of VTE within a 3-month period following hospital admission, we evaluated the risk factors for clinically apparent VTE and then derived a risk score to predict 3-month VTE risk in this population. Materials and Methods Patient recruitment took place between July 2002 and September 2006. The study design has been described elsewhere.8 Briefly, at the start of each month, participating hospitals systematically enrolled the first 10 eligible, acutely ill hospitalized medical patients. Key inclusion criteria were age ⱖ 18 years, admission for an acute medical illness, and ⱖ 3-day duration of hospitalization. Patients were excluded if they were enrolled in a therapeutic clinical trial or if any of the following applied: anticoagulant or thrombolytic drug use at admission or within 48 h after admission, major surgery or trauma within 3 months before admission, admission for DVT or PE (or a diagnosis of either within 24 h of admission), or if follow-up was deemed impossible. The study was developed and coordinated under the guidance of a Scientific Advisory Board by the Center for Outcomes Research (University of Massachusetts Medical School; Worcester, Massachusetts). Ethics committee approval was obtained from the University of Massachusetts Medical School on June 13, 2002 (IRB Multiple Assurance # M-1207), and followed the Declaration of Helsinki recommendations. Owing to the observational design of the study, the requirement for signed patient consent was waived by the Institutional Review Board. Standardized case report forms were used to record patient demographics, medical conditions and medications, predefined risk factors for VTE, immobilization (defined as confinement to a bed or chair for . 24 h), history of VTE, predefined risk factors for bleeding, types of VTE prophylaxis, timing and duration of prophylaxis (thromboprophylaxis had to be used for ⱖ 1 day during hospitalization), and discharge disposition. Glomerular filtration rate was used as a measure of renal function.15 Known thrombophilia was defined as a familial or acquired disorder of the hemostatic system resulting in an increased risk of thrombosis that was present at or during current admission. VTE was defined as DVT or PE that was observed clinically within 92 days of admission. Patients were considered to have DVT if they were treated for DVT and had a positive venogram or compression ultrasonography test. Patients were considered to have PE if they were treated for PE and had a positive lung scan, pulmonary angiogram, or spiral CT scan. Fatal PE was defined as PE diagnosed by autopsy or, in the absence of autopsy, when PE was considered the most likely cause of death. Data Analysis All patients meeting the enrollment criteria, including those who died during hospitalization, were considered for the initial analysis. Cumulative incidence of VTE risk from admission to 3-month follow-up was estimated using the Kaplan-Meier method. Patients without follow-up contributed data up to hospital discharge. Upper-extremity DVT and VTE in unknown locations were excluded from the main Kaplan-Meier analysis and from the analyses of timing of VTE events and mortality. We performed a second Kaplan-Meier analysis, assuming VTEs of unknown location were lower extremity, as an estimate of an upper bound for observed VTE incidence. Because length of hospital stay varied and not all patients had 3-month data, Cox multiple regression analysis was used to identify independent VTE risk factors. Our final Cox model considered CHEST / 140 / 3 / SEPTEMBER, 2011

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians

707

all univariate factors with P ⱕ .25 among VTE risk factors plus prior medical conditions. We then used backward selection with a ⱕ .05 to select final model variables. We also assessed possible interactions among the final variables (none found), and the proportional hazards assumption (not violated). We also added back the omitted variables one at a time, but none met the a ⱕ .05 criterion. For Kaplan-Meier curves and Cox models, time 0 was defined as hospital admission day. Two models were formed using these methods: a four-factor predictive model based solely on factors available at hospital admission and a seven-factor model that included factors present during hospital stay. This second model contained in-hospital factors whose timing relative to the VTE event could not be determined in our data set. In an attempt to see whether VTE prophylaxis use confounded the final model estimates, the Cox multiple regression analysis was run considering in-hospital use of VTE prophylaxis. The VTE risk score was calculated as follows: the factor with the smallest logarithmic hazard ratio (HR) (natural log of HR) was assigned one point, with other factor scores based on the size of their estimates relative to the smallest logarithmic HR. Individual factor scores were summed to give a total risk score for each patient. Model calibration was assessed by comparing model-predicted to observed VTE risk (using the Kaplan-Meier method) over the range of VTE risk scores. A poor model (ie, one no better than chance) will show no correlation between observed and modelpredicted risk. The C index was computed using an SAS macro for the Cox model.16 The mean VTE risk score and clinically observed VTE rates were described according to whether patients would have been recommended to receive prophylaxis according to the 2004 ACCP criteria (e-Appendix 1).17 Statistical analyses were performed using SAS, version 9.1 (SAS Institute Inc; Cary, North Carolina) statistical software.

Results Patient characteristics have been reported previously.8 One-half of the patients were women, the median age was 68 years, and the median length of hospital stay 7 days (Table 1). The most frequent medical conditions during hospitalization were infectious disease (32%), respiratory failure (19%), and cancer (12%). Patients with a cardiac condition represented 37% of the study population (ischemic heart disease, 12%; congestive heart failure, 11%; other cardiac condition, 14%), and patients with stroke represented 6%. In total, 4% of the population had prior VTE, and 2% had lower-extremities paralysis. In total, 13,172 patients (87%) had 3 months of follow-up data (follow-up rate, 90%, after excluding hospital deaths). Three months after admission, 184 patients (1.2%) developed VTE. Fourteen patients with upper-limb DVT and 27 with VTE at an unknown site were excluded, so the final analysis included 143 patients. Cumulative VTE incidence from admission to 3 months was 1.0% (Fig 1). A second Kaplan-Meier analysis (not shown) where the 27 VTEs of unknown location were considered as part of the analysis gave similar results to those shown in Figure 1 for the upper confidence limits (1.2% symptomatic VTE by 92 days after admission).

Most (98 of 143, 69%) patients developed VTE 1 to 30 days after hospital admission (median, 16 days) (Table 2). Of the 143 patients with VTE, 76 had PE with or without DVT, and 67 had lower-limb DVT alone; 79 (55%) and 64 (45%) had in-hospital and postdischarge VTE, respectively. Mortality in Patients With VTE Among the 79 patients with VTE diagnosed in hospital, 12 (15%) died while in the hospital, and four (5%) died after discharge. For those with VTE after discharge, 24 (38%) died within 3 months of admission. Among the 67 patients who developed DVT, 12 (18%) died within 3 months of admission, whereas 28 of 76 patients (37%) with PE died within this period (in 24 of 28 cases, PE was the primary or most likely cause of death), for an overall crude death rate of 28% (40 of 143). The crude death rate for patients without VTE was 9%. Factors Present at Admission Predictive of VTE Risk This model considers only factors known at or before admission (Table 3). Factors most strongly associated with VTE risk included previous VTE, known thrombophilia present within 3 months of admission, cancer, and age . 60 years (all P ⱕ .001) (Table 3). The final adjusted Cox multiple regression model included 15,156 patients with complete data, including 44% of patients who received in-hospital VTE prophylaxis. Factors independently associated with VTE risk in the 3-month period following hospital admission matched those found most statistically significant in Table 3: previous VTE event, known thrombophilia, cancer, and age . 60 years (C statistic, 0.65; P ⱕ .02) (Table 4). VTE Risk Score at Admission Older age and cancer were assigned one point each and known thrombophilia and previous VTE three points each (Table 4). Values for the identified risk factors can be entered into the IMPROVE VTE Table 1—Characteristics of Acutely Ill Hospitalized Medical Patients Characteristic Women, % Age, y Weight (No. 5 10,433), kg Length of hospital stay, d Immobile for ⱖ 7 d, including days immediately prior to admission (No. 5 15,125), % Time immobile (No. 5 1,169), d preadmission

Patients (N 5 15,156) 50 68 (52-79) 69 (59-81) 7 (5-13) 19 3 (1-15)

Data are presented as median (interquartile range), unless otherwise indicated. Adapted from Tapson et al.8

708

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians

Figure 1. Kaplan-Meier curve showing cumulative VTE incidence from hospital admission to 3 months of follow-up (N 5 15,156, n 5 143 patients with VTE).

risk calculator (available at http://www.outcomes. org/improve) to estimate 3-month risk. Almost 90% of patients had low risk (score of 0-1 and predicted VTE ⱕ 1.0%) (Table 5). A score ⱖ 2 indicates a considerably greater 3-month risk of ⱖ 2%. The model shows good calibration between observed and model-predicted risk over the risk score range. Factors During Hospitalization Associated With VTE The associative model considers patient factors occurring prior to and during hospital stay. Factors most strongly associated with VTE in the 3-month period following hospital admission were previous VTE event, known thrombophilia, current lower-limb paralysis, current cancer, immobilization ⱖ 7 days, stay in an ICU or coronary care unit (CCU), and age . 60 years (C statistic, 0.69; P , .05) (Table 6). When either pharmacologic or mechanical VTE prophylaxis use after admission was considered, estimates for these Table 2—Occurrence of VTE (N 5 143)a Days After Hospital Admission

In-hospital VTE (n 5 79)

Postdischarge VTE (n 5 64)

All VTE (N 5 143)

42 (53) 32 (41) 5 (6) 0

0 24 (38) 23 (36) 17 (27)

42 (29) 56 (39) 28 (20) 17 (12)

1-7 8-30 31-60 61-91

Data are presented as No. (%). aBased on timing of VTE treatment. A total of 131 patients had a date for start of treatment. For analysis of the 12 patients missing a start date, the last possible date to start the VTE treatment was assumed. For patients with VTE who died without receiving VTE treatment, the date of death was considered as the VTE treatment starting date. www.chestpubs.org

seven clinical risk factors did not change appreciably (, 10%) (e-Table 1). Associative VTE Score During Hospitalization Older age, ICU/CCU stay, and immobilization for ⱖ 7 days were assigned one point each; current cancer, current lower-limb paralysis, and known thrombophilia were each assigned two points; and previous VTE was assigned three points (Table 6). Values for the identified factors can be entered into the IMPROVE VTE risk calculator to estimate 3-month VTE rate. Almost 70% of patients had a VTE score of 0 to 1, 16% a score of 2, 9% a score of 3, and 7% a score ⱖ 4. The model also showed good calibration between observed and model-expected rates over the range of scores (Table 7). Patients with a score of 2 to 3 had a 3-month VTE symptomatic event rate of 1.5% by Kaplan-Meier method. Patients with a score ⱖ 4 had a 3-month VTE symptomatic event rate of 5.7% and symptomatic PE event rate of 3.1%. Increasing VTE scores were associated with a higher risk of overall and VTE-related death. Crude all-cause death rates were 15% (seven of 47) among patients with a score of 0 to 1 and 35% (34 of 96) among those with a score ⱖ 2. When stratified by VTE type, mortality rates for PE were 20% (six of 30) for patients with a score of 0 to 1 vs 48% (22 of 46) for those with a score of ⱖ 2. For DVT, rates were 6% (one of 17) for patients with a score of 0 to 1 vs 24% (12 of 50) for those with a score of ⱖ 2. Of the 6,898 patients at ACCP 2004-defined risk for whom thromboprophylaxis would be recommended, the observed VTE rate was 1.5%, with a mean VTE score of  2 (Table 8). In patients who were not at CHEST / 140 / 3 / SEPTEMBER, 2011

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians

709

Table 3—Patient Characteristics Associated With 3-Month VTE Risk, Unadjusted (N 5 15,156) Characteristic Priora known thrombophilia Previous VTE event Prior lower-limb paralysis Prior cardiovascular catheter Cancera Immobilization ⱖ 1 db ICU/CCU stayb Prior inflammatory bowel diseaseb Prior use of estrogens/HRT Prior infection Glomerular filtration rate,c mL/min/1.73 m2 , 60 , 30 Creatinine level . 1.2 mg/dL Prior congestive heart failure Prior respiratory failure Prior varicose veins Prior NSAID use Age, y . 60 vs ⱕ 60 . 75 vs ⱕ 75 Prior hepatic failure Immobilization ⱖ 7 dd 1-6 vs 0 d Currenta known thrombophilia Current lower-limb paralysis Current cardiovascular catheter Current cancer Current obesity Current diabetes

No.

HR

95% CI

x2

P Value

21 551 239 305 1,628 1,169 741 169 252 1,883

11.0 6.1 1.8 0.7 2.2 1.4 1.6 2.3 2.1 1.4

2.7-45 4.0-9.3 0.7-4.9 0.2-2.9 1.5-3.4 0.8-2.6 0.8-3.2 0.7-7.4 0.9-5.2 0.9-2.2

11 69 1.4 0.2 14 1.3 1.8 2.1 2.7 2.1

.001 , .001 .25 .65 , .001 .26 .18 .15 .10 .15

5,294 1,629 4,324 1,560 2,752 815 3,193

1.6 1.4 1.6 1.7 1.2 1.9 1.4

1.1-2.2 0.9-2.3 1.1-2.2 1.1-2.7 0.8-1.8 1.1-3.2 1.0-2.1

6.4 2.3 6.0 5.7 1.1 4.8 3.5

.01 .13 .01 .01 .30 .03 .06

9,646 5,398 235 2,846 3,104 42 309 1,228 1,735 2,421 3,452

2.0 1.2 0.5 2.6 1.6 8.4 4.1 2.3 2.9 1.1 0.9

1.3-2.9 0.8-1.6 0.1-3.4 1.8-3.7 1.0-2.5 2.7-26 2.2-7.6 1.5-3.6 2.0-4.3 0.7-1.7 0.6-1.3

11.7 0.8 0.6 30 4.6 13 20 14 31 0.6 0.4

.001 .38 .46 , .0001 .03 .0003 , .0001 .0002 , .0001 .55 .51

Thrombophilia was defined as a familial or acquired disorder of the hemostatic system that resulted in an increased risk of thrombosis, including antithrombin III deficiency, resistance to activated protein C, protein C and protein S deficiencies, prothrombin G20210A mutation, Factor V Leiden, and antiphospholipid syndrome. CCU 5 coronary care unit; HR 5 hazard ratio; HRT 5 hormone replacement therapy; NSAID 5 nonsteroidal antiinflammatory drug. aIn general, “prior” indicates prior medical history or present within 3 months of current hospital admission; “current” indicates present at or during current admission. Cancer denotes not newly diagnosed in the hospital. bAmong 11,555 patients for whom we could determine the characteristic preceded VTE. cGlomerular filtration rate was computed by the Modification of Diet in Renal Disease Study Group method (n 5 13,747 patients with complete information: 12% with severe and 27% with moderate renal impairment and 61% with normal renal function [reference group]). dDays immediately before admission plus days in hospital.

ACCP-defined risk, the observed VTE rate was 0.7%, with a mean score of  1. Discussion We report the incidence of VTE in a large, observational, multicenter study representative of acutely ill hospitalized medical patients and have developed Table 4—Adjusted Cox Predictive Model for 3-Month VTE and Points Assigned to Each Independent Risk Factor VTE Risk Factor

HR (95% CI)

x2

P Value

Points

Previous VTE Known thrombophilia Cancer Age . 60 y

5.0 (3.3-7.8) 5.2 (1.3-21.5)

53 5.2

, .001 .02

3 3

2.0 (1.3-3.1) 1.8 (1.2-2.7)

11 8.5

.001 .004

1 1

See Table 3 legend for expansion of abbreviation.

data-based and weighted VTE risk assessment scores that are predictive of VTE risk in medical patients at the time of hospital admission and are associated with VTE during hospitalization. Four independent risk factors during hospital admission were predictive and seven independent clinical risk factors were associative of an increased risk of VTE during the 3-month period following hospital admission. During hospitalization, there was an association between increasing VTE risk score and overall and VTE-related death. A major impediment to implementation of previous RAMs in hospitalized medical patients is that they were based on expert opinion, were cumbersome, were not subjected to rigorous evaluation, or did not account for heterogeneity among groupspecific diagnoses1,13,18; were developed in subsets of RCT patients with little or no evidence of how the various VTE risk factors interact in a quantitative

710

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians

Table 5—Calibration of Predictive Model With Risk Score and Observed VTE Events (N 5 15,156) Score

Patients, % (No.)

3-Mo Predicted VTE Risk, %a

Observed VTE Rate, % (No. VTE Events)b

Observed PE Rate, % (No. of PEs)b

33 (4,981) 56 (8,441) 8 (1,166) 1 (127) 2 (376) 0.4 (65)

0.5 1.0 1.7 3.1 5.4 11

0.5 (24) 1.0 (72) 2.1 (20) 4.0 (5) 4.7 (16) 11 (6)

0.3 (16) 0.5 (36) 1.2 (12) 1.6 (2) 1.8 (6) 8.0 (4)

0 1 2 3 4 5-8

PE 5 pulmonary embolism. aFrom the Cox regression model relating VTE to VTE risk score. Predicted risks are means for patients in a given risk score group. bKaplan-Meier method.

manner2; or were based on data derived exclusively or predominantly from cancer populations.12,13 The only model that used an independent validation cohort to predict symptomatic VTE was developed in patients with cancer undergoing chemotherapy.12 As a consequence, the ACCP guidelines do not provide a patient-specific VTE risk model for medical patients and incorporate all at-risk medical patients within a single risk grouping. Factors associated with VTE include a previous VTE event, known thrombophilia, current lower-limb paralysis, current cancer, immobilization for ⱖ 7 days, ICU/CCU stay, and patient age . 60 years. Some are predisposing risk factors unique to the individual patient, and others are extrinsic factors related to the patient’s clinical situation. Previous VTE, known thrombophilia, cancer, and advanced age are risk factors that are able to predict VTE risk at hospital admission. The remaining factors are associative of VTE risk during hospitalization, which is important because a patient’s risk status may change during the course of hospitalization. The VTE risk factors identified are consistent with those described by the most Table 6—Adjusted Cox Associative Model for 3-Month VTE and Points Assigned to Each Patient Characteristic (N 5 15,125) Patient Characteristic Previous VTE Known thrombophilia Current lowerlimb paralysis Current cancer Immobilized ⱖ 7 db ICU/CCU stay Age . 60 y a

HR (95% CI)

x2

P Value

Points

4.7 (3.0-7.2) 3.5 (1.1-11)

48 5.2

, .001 .04

3 2

3.0 (1.6-5.7)

11

.001

2

2.8 (1.9-4.2) 1.9 (1.3-2.7)

27 11

, .001 .001

2 1

.01 .01

1 1

1.8 (1.1-2.9) 1.7 (1.1-2.6)

6.1 6.3

See Table 3 legend for expansion of abbreviations. Previous VTE and age are both known to have occurred prior to 3-month VTE; the other patient factors are known to have been present at or during hospital admission. bDays immobile immediately prior to and during hospital admission. a

www.chestpubs.org

recent ACCP guidelines.1 These are also consistent with those used in the Prophylaxis in Medical Patients with Enoxaparin (MEDENOX) study, where a history of VTE, malignant disease, increased age, thrombophilia, and prolonged immobility each were identified as independent VTE risk factors,6 and those used in the Extended Clinical Prophylaxis in Acutely Ill Medical patients (EXCLAIM) study, which were older age and patient immobility.19 Consistent with results from EXCLAIM, the present study identified immobilization (defined as confinement to a bed or chair for . 24 h) as an independent risk factor for VTE. Interestingly, a previous study reported that up to 80% of patients who had bed rest for . 1 week before death had VTE at autopsy.20 Although the present study reports an underestimation of the VTE incidence due to lack of mandatory application of venography or ultrasonography, the overall 1.0% cumulative symptomatic VTE incidence (PE and lower-limb DVT) in the 3-month period following hospital admission is consistent with published studies reporting VTE rates of 1.3% to 1.9%.6,7,21,22 During hospitalization, . 31% of our study population had an associative score ⱖ 2, with an average clinically observed VTE event rate of 2.4%. Patients at ACCP-defined risk for whom thromboprophylaxis is recommended had a median risk score of nearly 2 and a clinical VTE rate of 1.5%; those not at ACCPdefined risk had a median risk score of 1 and a clinical VTE rate of 0.7%. Thus, our VTE associative model suggests that patients with a score of ⱖ 2 during hospitalization may benefit from thromboprophylaxis and that patients with a score , 2 (which represented 69% of the total patient population) may not necessarily benefit from thromboprophylaxis. In addition, patients with a score ⱖ 4 (7% of the study population) had particularly high VTE rates, with a 3-month VTE symptomatic event rate of 5.7% and a symptomatic PE event rate of 3.1%. This VTE associative model showed similar 3-month expected VTE risk and observed VTE rates (including observed PE rates). Furthermore, higher scores were associated with increasing VTE rates; a score ⱖ 2 CHEST / 140 / 3 / SEPTEMBER, 2011

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians

711

Table 7—Calibration of Associative Model With Risk Score and Observed VTE Events (N 5 15,125) Score

Patients, % (No.)

3-Mo Expected VTE Risk, %a

Observed VTE Rate, % (No. VTE Events)b

Observed PE Rate, % (No. of PEs)b

0 1 2 3 4 5-10

27 (4,029) 42 (6,350) 16 (2,420) 9 (1,335) 5 (729) 2 (262)

0.4 0.6 1.0 1.7 2.9 7.2

0.4 (14) 0.6 (33) 1.5 (31) 1.6 (18) 4.8 (30) 8.1 (17)

0.3 (11) 0.3 (19) 0.6 (13) 0.8 (9) 2.8 (17) 3.8 (7)

See Table 5 legend for expansion of abbreviation. From the Cox regression model relating VTE to VTE risk score; predicted risks are means for patients in a given risk score group. A total of 31 patients with incomplete covariate information were dropped from the final model. bKaplan-Meier method. a

also was associated with a higher rate of overall and VTE-related death. The present analysis is consistent with previous studies reporting more severe forms of VTE (PE and fatal PE) in acutely ill hospitalized medical patients.3,4 In addition, almost half of the VTE events were diagnosed after patient discharge, and 71% of patients developed VTE . 1 week after admission. A high mortality rate also was noted among patients with postdischarge VTE events (38% within 3 months). Notably, a recent US population-based study reported that more VTE events are diagnosed in the 3 months after hospitalization than during hospitalization, with  67% of events occurring during the first month.23 Limitations and Strengths This study was not designed to assess whether thromboprophylaxis use after admission had an effect on the VTE event rate or to what extent VTE prophylaxis may modify model estimates because prophylaxis was not randomly assigned to patients. Although this analysis was not aimed to assess the effect of in-hospital prophylaxis on VTE incidence, the inclusion of VTE prophylaxis in the model did not change significantly the rank order or the regression coefficients for the identified VTE risk factors, suggesting that the management strategy did not have an imporTable 8—Relationship Among ACCP Recommendations, VTE Risk Score, and Observed VTE Rates by Associative Model (N 5 15,125) Median VTE Clinically Patient IMPROVE Risk Score, Observed VTE, % Characteristic Patients, % (n/N) Points (IQR) (No. VTE Events)a Not at ACCP 54 (8,227/15,125) risk At ACCP risk 46 (6,898/15,125)

0.9 (0-1)

0.7 (50)

1.8 (1-2)

1.5 (93)

ACCP 5 American College of Chest Physicians; IMPROVE 5 International Medical Prevention Registry on Venous Thromboembolism; IQR 5 interquartile range. aKaplan-Meier method.

tant influence on the risk factors themselves or on their relative weights. Another potential limitation is the use of clinical end points rather than routinely tested VTE, which underestimated the incidence of VTE. However, we were conservative in excluding VTE where there were conflicting data or where VTE was in an unknown location. Selection bias was minimized by the consecutive enrollment of patients at the beginning of every month. Because autopsies were not performed routinely, we could not always verify that deaths were VTE related. The associative model during hospitalization was able to identify factors that were associated with an increased rate of VTE and should be confirmed by future clinical studies on the subject. The model also does not include novel risk factors, such as platelet or leukocyte counts. Strengths of the present study were its large, representative patient cohort across many hospital systems. The mean RAM-predicted VTE rates were similar to observed rates across our entire range of risk scores. The risk model explicitly excluded upperlimb DVT, which can be a potential confounder because of the different clinical characteristics and uncertainty about the need for thromboprophylaxis in patients with upper- vs lower-limb DVT.24 External validation of both the predictive and the associative models with another large, representative database of acutely ill hospitalized medical patients is an important next step for the potential clinical use of our VTE risk score. Conclusions To our knowledge, this study is the first to evaluate VTE incidence in a large, representative population of hospitalized medical patients and to identify four independent clinical risk factors predictive of VTE risk during hospital admission and seven factors associated with VTE risk during hospitalization. The weighted VTE risk scores showed an association between a higher risk score and a higher observed VTE rate and identified high-risk patient subgroups,

712

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians

including those with higher overall and VTE-related mortality. The present risk model, subject to its validation in external studies, will provide a useful adjunct to the ACCP guidelines for the assessment of VTE risk in acutely ill medical patients upon hospital admission. This would be especially helpful in optimizing risk/ benefit assessment of pharmacologic prophylaxis in conjunction with a bleeding risk score developed in this patient population,24 in selecting strategies of thromboprophylaxis in high-risk medical patients with risk factors that do not fit into existing group-specific VTE risk categories, and in the design of future RCTs for prophylactic strategies. Acknowledgments Author contributions: Drs Spyropoulos and FitzGerald had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Dr Spyropoulos: contributed to the study design, manuscript preparation, and editorial review of the manuscript. Dr Anderson: contributed to the study design, data collection and processing, manuscript preparation, and editorial review of the manuscript. Dr FitzGerald: contributed to data collection and processing, statistical analysis, and manuscript preparation. Dr Decousus: contributed to the manuscript preparation. Dr Pini: contributed to the manuscript preparation. Dr Chong: contributed to the manuscript preparation. Dr Zotz: contributed to the manuscript preparation. Dr Bergmann: contributed to the manuscript preparation. Dr Tapson: contributed to the manuscript preparation. Dr Froehlich: contributed to the manuscript preparation. Dr Monreal: contributed to the manuscript preparation. Dr Merli: contributed to the manuscript preparation. Dr Pavanello: contributed to the manuscript preparation. Dr Turpie: contributed to the manuscript preparation. Dr Nakamura: contributed to the manuscript preparation. Dr Piovella: contributed to the manuscript preparation. Dr Kakkar: contributed to the manuscript preparation. Dr Spencer: contributed to the manuscript preparation. Financial/nonfinancial disclosures: The authors have reported to CHEST the following conflicts of interest: Dr Spyropoulos has been a consultant for Sanofi-Aventis, Eisai Co Ltd, Bayer, and Boehringer-Ingelheim. Dr Anderson has received research support from Sanofi-Aventis, The Medicines Company, Procter & Gamble, and Scios Inc; has been a consultant for Sanofi-Aventis, GlaxoSmithKline, and Sage Millennium; and has served on advisory boards for Sanofi-Aventis and The Medicines Company. Dr Decousus has received research support from Sanofi-Aventis, GlaxoSmithKline, and Bayer. Dr Pini has received honoraria from Sanofi-Aventis. Dr Chong has served on advisory boards with Bayer, Boehringer-Ingelheim, and GlaxoSmithKline and has been a speaker for Bayer. Dr Zotz has been a consultant for SanofiAventis. Dr Bergmann has received honoraria from Sanofi-Aventis, GlaxoSmithKline, and AstraZeneca. Dr Tapson has received research support from and has been a consultant for Sanofi-Aventis. Dr Froehlich has received research support from Blue Cross Blue Shield of Michigan and the Gore and Mardigian Foundation; has been a consultant for Pfizer and Sanofi-Aventis; and has received honoraria from Pfizer, Sanofi-Aventis, and Merck/Schering-Plough. Dr Merli has received research support from and has served on advisory boards for Sanofi-Aventis, Bristol-Meyers Squibb, and Bayer, and has been a speaker for Sanofi-Aventis. Dr Pavanello has received research support from Sanofi-Aventis; has served on an advisory board for Bayer; and has been a speaker for Bayer, Pfizer, and Novartis. Dr Turpie has served on advisory boards for Sanofi-Aventis, Bayer, Johnson & Johnson, GlaxoSmithKline, Portola Pharmaceuticals, and Boehringer-Ingelheim. Dr Nakamura has been a consultant for Daiichi-Sankyo, GlaxoSmithKline, www.chestpubs.org

and Astellas. Dr Piovella has received research support from GlaxoSmithKline; has served on advisory boards for Bayer, Boehringer-Ingelheim, and GlaxoSmithKline; and has been a speaker for Bayer, Boehringer-Ingelheim, GlaxoSmithKline, and Sanofi-Aventis. Dr Kakkar has received research funding from and served on advisory boards with Bayer HealthCare, SanofiAventis, Boehringer-Ingelheim, Pfizer, Bristol-Meyers Squibb, and Eisai Co Ltd; has been a consultant for Bayer HealthCare, Sanofi-Aventis, Boehringer-Ingelheim, Pfizer, Bristol-Meyers Squibb, Eisai Co Ltd, ARYx Therapeutics, and Canyon Pharmaceuticals; and has received honoraria from Bayer HealthCare, Sanofi-Aventis, Boehringer-Ingelheim, Pfizer, Bristol-Meyers Squibb, Eisai Co Ltd, and GlaxoSmithKline. Dr Spencer has been a consultant for and has received grants from Sanofi-Aventis. Drs FitzGerald and Monreal have reported that no potential conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article. Role of sponsors: The authors received editorial support in the preparation of this manuscript funded by Sanofi-Aventis; however, the authors were fully responsible for content and editorial decisions for this manuscript. The sponsor was not involved in the conduct of the study or in the analysis of data. Additional information: The e-Appendix and e-Table can be found in the Online Supplement at http://chestjournal.chestpubs. org/content/140/3/706/suppl/DC1.

References 1. Geerts WH, Bergqvist D, Pineo GF, et al; American College of Chest Physicians. Prevention of venous thromboembolism: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines (8th Edition). Chest. 2008;133(suppl 6): 381S-453S. 2. Cohen AT, Alikhan R, Arcelus JI, et al. Assessment of venous thromboembolism risk and the benefits of thromboprophylaxis in medical patients. Thromb Haemost. 2005;94(4):750-759. 3. Cohen AT, Davidson BL, Gallus AS, et al; ARTEMIS Investigators. Efficacy and safety of fondaparinux for the prevention of venous thromboembolism in older acute medical patients: randomised placebo controlled trial. BMJ. 2006;332(7537): 325-329. 4. Piazza G, Seddighzadeh A, Goldhaber SZ. Double trouble for 2,609 hospitalized medical patients who developed deep vein thrombosis: prophylaxis omitted more often and pulmonary embolism more frequent. Chest. 2007;132(2):554-561. 5. Cohen AT, Tapson VF, Bergmann JF, et al; ENDORSE Investigators. Venous thromboembolism risk and prophylaxis in the acute hospital care setting (ENDORSE study): a multinational cross-sectional study. Lancet. 2008;371(9610): 387-394. 6. Samama MM, Cohen AT, Darmon JY, et al; Prophylaxis in Medical Patients with Enoxaparin Study Group. A comparison of enoxaparin with placebo for the prevention of venous thromboembolism in acutely ill medical patients. N Engl J Med. 1999;341(11):793-800. 7. Leizorovicz A, Cohen AT, Turpie AG, Olsson CG, Vaitkus PT, Goldhaber SZ; PREVENT Medical Thromboprophylaxis Study Group. Randomized, placebo-controlled trial of dalteparin for the prevention of venous thromboembolism in acutely ill medical patients. Circulation. 2004;110(7):874-879. 8. Tapson VF, Decousus H, Pini M, et al; IMPROVE Investigators. Venous thromboembolism prophylaxis in acutely ill hospitalized medical patients: findings from the International Medical Prevention Registry on Venous Thromboembolism. Chest. 2007;132(3):936-945. 9. Bahl V, Hu HM, Henke PK, Wakefield TW, Campbell DA Jr, Caprini JA. A validation study of a retrospective venous thromboembolism risk scoring method. Ann Surg. 2010;251(2): 344-350. CHEST / 140 / 3 / SEPTEMBER, 2011

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians

713

10. Alikhan R, Cohen AT, Combe S, et al; MEDENOX Study. Risk factors for venous thromboembolism in hospitalized patients with acute medical illness: analysis of the MEDENOX Study. Arch Intern Med. 2004;164(9):963-968. 11. Haas S, Spyropoulos AC. Primary prevention of venous thromboembolism in long-term care: identifying and managing the risk. Clin Appl Thromb Hemost. 2008;14(2):149-158. 12. Khorana AA, Kuderer NM, Culakova E, Lyman GH, Francis CW. Development and validation of a predictive model for chemotherapy-associated thrombosis. Blood. 2008; 111(10):4902-4907. 13. Kucher N, Koo S, Quiroz R, et al. Electronic alerts to prevent venous thromboembolism among hospitalized patients. N Engl J Med. 2005;352(10):969-977. 14. Dargaud Y, Rugeri L, Vergnes MC, et al. A risk score for the management of pregnant women with increased risk of venous thromboembolism: a multicentre prospective study. Br J Haematol. 2009;145(6):825-835. 15. Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D; Modification of Diet in Renal Disease Study Group. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Ann Intern Med. 1999;130(6):461-470. 16. Harrell FE Jr, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluating the yield of medical tests. JAMA. 1982;247(18): 2543-2546. 17. Geerts WH, Pineo GF, Heit JA, et al. Prevention of venous thromboembolism: the Seventh ACCP Conference on

18. 19.

20. 21.

22. 23. 24.

Antithrombotic and Thrombolytic Therapy. Chest. 2004; 126(suppl 3):338S-400S. Lutz L, Haas S, Hach-Wunderle V, et al. Venous thromboembolism in internal medicine: risk assessment and pharmaceutical prophylaxis [in German]. Med Welt. 2002;53:231-234. Hull RD, Schellong SM, Tapson VF, et al; EXCLAIM (Extended Prophylaxis for Venous ThromboEmbolism in Acutely Ill Medical Patients With Prolonged Immobilization) study. Extended-duration venous thromboembolism prophylaxis in acutely ill medical patients with recently reduced mobility: a randomized trial. Ann Intern Med. 2010;153(1):8-18. Gibbs NM. Venous thrombosis of the lower limbs with particular reference to bed-rest. Br J Surg. 1957;45(191):209-236. Bosson JL, Pouchain D, Bergmann JF; for the ETAPE Study Group. A prospective observational study of a cohort of outpatients with an acute medical event and reduced mobility: incidence of symptomatic thromboembolism and description of thromboprophylaxis practices. J Intern Med. 2006;260(2):168-176. Edelsberg J, Hagiwara M, Taneja C, Oster G. Risk of venous thromboembolism among hospitalized medically ill patients. Am J Health Syst Pharm. 2006;63(20)(suppl 6):S16-S22. Spencer FA, Lessard D, Emery C, Reed G, Goldberg RJ. Venous thromboembolism in the outpatient setting. Arch Intern Med. 2007;167(14):1471-1475. Spyropoulos AC. Upper vs. lower extremity deep vein thrombosis: outcome definitions of venous thromboembolism for clinical predictor rules or risk factor analyses in hospitalized patients. J Thromb Haemost. 2009;7(6):1041-1042.

714

Original Research

Downloaded from chestjournal.chestpubs.org by Kimberly Henricks on September 7, 2011 © 2011 American College of Chest Physicians