Congestive heart failure and outpatient risk of venous thromboembolism

Journal of Clinical Epidemiology 54 (2001) 810–816

Congestive heart failure and outpatient risk of venous thromboembolism: A retrospective, case-control study M.D. Howella,1, J.M. Geracib, A.A. Knowltona,* a

Cardiology Section, Department of Medicine, Houston Veterans Affairs Medical Center and Baylor College of Medicine, Houston, TX, USA b Section of General Internal Medicine and HSR&D Field Program, Department of Medicine, Houston Veterans Affairs Medical Center and Baylor College of Medicine, Houston, TX, USA Received 22 September 2000; received in revised form 20 November 2000; accepted 7 December 2000

Abstract Although CHF has been considered a risk factor for venous thromboembolism, this has not been directly studied. We hypothesized that congestive heart failure would increase the risk of venous thromboembolism in an outpatient population, and that this risk would increase as patients’ ventricular function worsened. We conducted a case-control study to examine whether CHF due to left ventricular dysfunction was an independent risk factor for acute venous thromboembolism in outpatients, once established risk factors such as recent surgery and prior venous thromboembolism are taken into account. We reviewed 106 cases of DVT and 603 controls, admitted for diabetes mellitus or infection, matched for month of admission at a VA hospital. Assignment of a diagnosis of venous thromboembolism required a definitive test, as did classification as CHF. In a logistic regression model CHF was an independent predictor of venous thromboembolism. A second logistic regression model showed that the risk of venous thromboembolism increased as the ejection fraction (EF) decreased, with an EF  20 associated with a venous thromboembolism OR of 38.3 (95% CI 9.6, 152.5). CHF is an independent risk factor for venous thromboembolism, and the risk increases markedly as the EF decreases. These results support the use of anticoagulation in selected patients with CHF. © 2001 Elsevier Science Inc. All rights reserved. Keywords: Congestive heart failure; Thromboembolism; DVT; Cardiomyopathy; Pulmonary embolism

1. Introduction Clinicians have long associated congestive heart failure (CHF) with an increased risk of developing venous thromboembolism (VTE). (For this study CHF refers to left ventricular systolic dysfunction only.) The magnitude of this risk, however, remains poorly studied, in spite of the fact that deep venous thrombosis (DVT) and pulmonary embolism (PE) are common conditions with substantial morbidity and mortality [1–7]. The association between CHF and VTE is rheologically intuitive: blood that flows slowly tends to clot, as Rudolf Virchow postulated over a century ago. Acting on this understanding, physicians such as Proctor Harvey and Clem-

* Corresponding author. Cardiology Research 151C, VA Medical Center, 2002 Holcombe Blvd., Houston, TX 77030-4211. E-mail address: [email protected] (A.A. Knowlton) 1 Current address: Department of Medicine, Beth Israel Hospital, Boston, MA

ent Finch anticoagulated hospitalized CHF patients in the 1950s, finding better outcomes among those who received dicumarol [8]. More recently, a post hoc analysis of the Studies of Left Ventricular Dysfunction (SOLVD) patients by Al-Khadra and others found a survival advantage for those patients on warfarin at study entry [9]. Finally, mounting biochemical evidence suggests a chronic hypercoagulable state in CHF, in addition to the venous stasis component [10–12]. Nonetheless, there has never been a randomized, controlled trial of anticoagulation in CHF, and most research has focused on arterial rather than VTE [13–18]. Much debate today focuses on whether to chronically anticoagulate CHF patients in the outpatient setting [18–20]. For these reasons we hypothesized that congestive heart failure would increase the risk of VTE in an outpatient population, and that this risk would increase as patients’ ventricular function worsened. Therefore, we conducted a retrospective, case-control study to examine whether CHF due to left ventricular dysfunction was an independent risk factor for acute VTE in outpatients, once established risk factors such as prior surgery and VTE are taken into account.

0895-4356/01/$ – see front matter © 2001 Elsevier Science Inc. All rights reserved. PII: S0895-4356(00)00 3 7 3 - 5

M.D. Howell et al. / Journal of Clinical Epidemiology 54 (2001) 810–816

2. Methods 2.1. Definitions Case patients were those with a PE or proximal DVT. For inclusion, we required that these patients be symptomatic, that the symptoms begin in an outpatient setting, and that an accurate, confirmatory imaging test was performed within 24 h of admission to verify clinical suspicion. For the diagnosis of DVT, we accepted ultrasound, impedence plethysmography, radionuclide venography, and contrast venography. We excluded isolated calf vein thrombosis because of its uncertain clinical significance [21,22]. For PE, we accepted contrast angiography, computed tomography (CT), and ventilation-perfusion (VQ) scintography. Patients who received VQ scans, but not angiography, were accepted if there was an explicitly stated high clinical suspicion of pulmonary embolism and an abnormal VQ scan (high, intermediate, or low probability); or if they had an intermediate or low clinical suspicion and a high probability VQ scan. PIOPED (Prospective Investigation of Pulmonary Embolism Diagnosis) data indicate that 40–96% of this group will have pulmonary embolism [23]. In the event of sudden death within 24 h of admission, an autopsy finding of pulmonary embolism was also acceptable. We chose as control patients those admitted with acute infection (e.g., pneumonia) or diabetic complications (e.g., ketoacidosis) as neither of these diagnoses have been reported to be associated with an increased risk of VTE [24– 29]. We thought that this would give us a broad cross-section of patients with as little bias as possible. Control patients were matched to cases by month of admission to control for practice characteristics of physicians, such as the propensity to order objective tests to measure LVEF. This is especially important in an academic institution where residents rotate frequently. We developed an a priori definition of CHF that required both a consistent clinical picture and an objective measurement of left ventricular ejection fraction (LVEF) of less than 45%. An isolated mention of a previous medical history of CHF was not sufficient. We accepted results from echocardiography, multiple-gated acquisition analysis (MUGA), or cardiac catheterization. Echocardiograms and MUGAs are available by open referral at our center and are readily attainable by any primary care or specialty physician. Because this was a retrospective assessment, the timing of LVEF measurement is important: we accepted any LVEF measurement prior to or within 3 months after the index stay. 2.2. Identification of patients In the absence of data indicating the true odds of VTE due to coexistent CHF, we believed that an odds ratio of 2.0 or higher would be important to clinicians. We calculated that 106 cases and 424 controls would give us 80% power to find this result at the .05 significance level, assuming a CHF

811

prevalence of 5% in control patients and 12.5% in patients with VTE. We chose these numbers based on internal estimates of general CHF prevalence [30] and literature estimates of CHF prevalence in VTE patients [24–28]. Using the Patient Treatment File, the national hospital discharge database maintained by the Department of Veterans Affairs, we identified 326 patients with diagnosis codes from the International Classification of Diseases, Ninth Revision (ICD-9-CM) suggestive of VTE (451.1, 453.8, 453.9, or 415.1, or a DRG of 128) and 644 patients with codes indicating acute infection or diabetes and its complications (038, 250.0–250.3, 250.8, 250.9, 481, 482.0–482.4, 482.8, 482.9, 483–486, 320, 590.1, 590.2, 590.8, 590.9, and 599.0). These patients were discharged from the Houston Veterans Affairs Medical Center (HVAMC) between November 1, 1993, and October 31, 1996. 2.3. Sources of data Because previous work has shown that administrative database diagnoses can be inaccurate [31,32], we reviewed the discharge summary and relevant imaging reports for each potential case and control patient. Patients without discharge summaries were excluded. For patients who did not meet the inclusion criteria noted above, the reason for exclusion was recorded. For all patients who met inclusion criteria, we used a standard form to abstract the following data: age, gender, ethnicity, date of admission, date of discharge, previous medical history, previous surgical history, medications, cardiac history and physical findings, chest radiography results, any cardiac imaging studies, and potential risk factors for deep venous thrombosis (previous VTE, malignancy, obesity, immobility, autoimmune disease, stroke, paralysis, hypercoaguable state, nephrotic syndrome, polycythemia, recent trauma, recent fracture, and recent surgery). For all case patients, the results of the patient’s venous ultrasound, contrast venography, impedance plethysmography, VQ scan, or pulmonary angiogram were recorded. For all control patients, the type or types of major illnesses that occurred during their index hospitalization were recorded. The Institutional Review Board of Baylor College of Medicine approved this protocol in advance. 2.4. Patient review for CHF status All patient data relevant to a diagnosis of CHF were abstracted onto a special form that contained no information regarding the patient’s case/control status. Items of previous medical history, which would clearly have indicated case/ control status (e.g., a familial protein S deficiency), were not included, and a history of previous VTE was also omitted. A cardiologist (A.A.K.) and an internist (J.M.G.), who were not involved in data collection, independently reviewed these forms to determine whether a patient met the criteria for congestive heart failure. Possible right heart failure and diastolic dysfunction were also determined. Right heart failure and diastolic dysfunction accounted for 1–3%

812

M.D. Howell et al. / Journal of Clinical Epidemiology 54 (2001) 810–816

of cases, and were therefore excluded from analysis. If the two reviewers disagreed on interpretation, they discussed the case and, in a few instances, reviewed previous imaging reports. 2.5. Statistical analysis For bivariate comparisons between patients with VTE and controls, we assessed statistical significance with the two-tailed chi-square test or Fisher’s exact test, as appropriate. For the purposes of analysis, we created a variable called “surgical state,” which combined recent orthopedic or general surgery, recent long-bone fracture, and recent trauma. “Recent” was defined as within 1 month. To evaluate the independent effect of each patient characteristic on the risk of developing VTE, we used multiple logistic regression, creating models in a stepwise fashion with variable entry and exit values of P  .1. All variables associated with VTE occurrence on bivariate analysis at a level P  .1 were made available to the stepwise procedure, if they occurred in five or more study patients. Odds ratios and 95% confidence intervals were calculated using standard methodology. In one model, we evaluated the contribution of CHF as an independent, dichotomous predictor of VTE. In a second model we evaluated whether the degree of left ventricular dysfunction was associated with the magnitude of VTE risk. Exploratory analyses demonstrated that VTE risk did not increase in a simple linear fashion with decreasing LVEF. Thus, patients were divided into four subgroups based on LVEF (  20%, 20–44%,  45%, and patients without a measured LVEF as the referent). We used the c statistic to evaluate the predictive accuracy of our models [33] and the Hosmer-Lemeshow to evaluate their calibration [34]. The results of the matched case-control analysis differed negligibly from those of the unmatched analysis; only the results of the unmatched analyses are presented. Statistical analyses were performed using the SAS statistical package (SAS Institute, Cary, NC). 3. Results 3.1. Patient characteristics We enrolled 141 of 326 potential case patients, and 603 of 644 potential control patients. Approximately 5% of each group was excluded because no discharge summary had been dictated. In almost all cases, these were patients with a very short length of stay incompatible with a diagnosis of DVT at our institution in the years studied. We excluded a total of 169 potential cases because they did not meet our case definition (Fig. 1). There were no significant differences in demographics between patients with VTE and controls. The VA patient population is predominantly male, with caucasians comprising approximately two thirds of the study population (Table 1). Similar proportions of case and control patients had an objective measurement of LVEF within the time limit spec-

Fig. 1. Case and control exclusions. VTE, venous thromboembolism; DVT, deep venous thrombosis.

ified. Patients with VTE were more likely to be on warfarin at the time of presentation, and a number of VTE risk factors were also more common in case patients. These risk factors included prior VTE, obesity, surgical state, and a documented hypercoaguable condition. There was a trend toward an increased prevalence of CHF in patients with VTE, but this did not reach statistical significance on bivariate analysis. Chronic obstructive pulmonary disease and diabetes mellitus were more common in controls, as expected from the way our controls were selected. Coronary artery disease, hypertension, peripheral vascular disease, and a history of atrial fibrillation were also more common in control patients. Patients with CHF were older than patients without the disease (Table 2), but otherwise similar in demographic characteristics. They were more likely to be on aspirin and warfarin, and more likely to have atrial fibrillation, coronary artery disease, hypertension, and peripheral vascular disease. Patients with and without CHF were very similar in terms of the prevalence of VTE risk factors. 3.2. CHF as a risk factor for venous thromboembolism The logistic regression model evaluating the presence or absence of CHF as a risk factor for VTE is presented in Table 3. Adjusting for other risk factors, patients with CHF were more likely to develop VTE than patients without CHF, with an adjusted odds ratio (OR) of 2.6 [95% confidence interval (CI), 1.4 to 4.7]. Additionally, the history of prior VTE was strongly associated with current VTE (OR  35.5, 95% CI 14.5, 87.0), as were the “surgical state” characteristic and obesity. In the second logistic regression model we found that the risk of VTE increased in an LVEF-dependent manner (Table 4). Patients with an LVEF  20% had an especially

M.D. Howell et al. / Journal of Clinical Epidemiology 54 (2001) 810–816 Table 1 Patient characteristics by case/control status Characteristic Age, mean (S.D.) Male (%) Race White (%) Black (%) Hispanic (%) Other (%) Medications Warfarin Aspirin Available LVEF measurement LVEF 20% LVEF 20–44% LVEF 45% or more Cardiac conditions History of atrial fibrillation History of coronary artery disease History of hypertension CHF Other comorbid conditions COPD Diabetes mellitus Peripheral vascular disease VTE risk factors Prior VTE Obesity Surgical statea Recent orthopedic surgery Recent other surgery Recent fracture Recent trauma Documented hypercoagulable state Malignancy Nephrotic syndrome Recent immobility Para-, hemi-, or quadriplegia Polycythemia Autoimmune disorder

813

Table 2 Patient characteristics by CHF status Cases (n  141) 61.0 (12.3) 99.3% 89 (63.1) 44 (31.2) 8 (5.7) 0 (0) 6.4% 22.0% 48.2% 5.7% 13.5% 30.2% 2.1% 15.6% 25.5% 18.4%

Controls (n  603) 62.4 (14.5) 98.7% 332 (55.1) 225 (37.3) 44 (7.3) 2 (0.3) 0.8%*** 20.1% 42.2% 0.7%*** 11.3% 29.1% 8.0%* 22.4% 39.5%* 13.6%

9.9% 15.6% 4.3%

22.4%** 31.8%*** 9.0%

27.0% 15.6% 14.9% 2.8% 7.1% 1.4% 5.7% 2.1% 23.4% 0.7% 17.7% 2.8% 1.4% 4.0%

1.2%*** 7.5%** 2.0%*** 0.3%* 1.5%*** 0.2%* 0.2%*** 0.2%* 20.2% 0.2% 17.7% 4.8% 0.3% 6.4%

VTE  venous thromboembolism, CHF  congestive heart failure, LVEF  left ventricular ejection fracture, COPD  chronic obstructive pulmonary disease. a “Surgical state” includes all patients with recent orthopedic or other surgery, fracture, or trauma. *P  .05, **P  .01, ***P  .001.

Characteristic

CHF (n  108) No CHF (n  636)

Age, mean (S.D.) Male (%) Race White (%) Black (%) Hispanic (%) Other (%) Medications Warfarin Aspirin Cardiac conditions History of atrial fibrillation History of coronary artery disease History of hypertension Other comorbid conditions COPD Diabetes mellitus Peripheral vascular disease VTE risk factors Prior VTE Obesity

68.8 (10.4)

61.0 (14.3)*

61 (56.5) 40 (37.0) 7 (6.5) 0

360 (56.6) 229 (36.0) 45 (7.1) 2 (0.3)

5 (4.6) 46 (42.6)

9 (1.4)* 106 (16.7)***

19 (17.6) 52 (48.2) 56 (51.9)

32 (5.0)*** 105 (16.5)*** 218 (34.3)***

24 (22.2) 32 (29.6) 25 (23.2)

124 (19.5) 182 (28.6) 35 (5.5)***

4 (3.7) 9 (8.3)

41 (6.5) 58 (9.1)

*P  .05, ***P  .001.

4. Discussion Our results substantiate clinicians’ long-held belief that patients with CHF are more likely to develop deep venous thrombosis and pulmonary embolism. For many years, inpatients with CHF have been considered at increased risk for VTE and prophylaxis is recommended for them [35,36]. These recommendations were based on expert opinion, small clinical trials of VTE prophylaxis performed before the routine use of diagnostic testing [8], and more recent trials of prophylaxis in general medical patients whose admitting diagnoses frequently included CHF [37,38]. The only previous study to quantify the risk of deep vein thrombosis in patients with CHF did find increased risk, with an odds ratio of 1.8 (1.3–2.3) [26]. However, this was a secondary analysis, assessed CHF status by interview, and included

Table 3 Logistic regression model evaluating CHF and odds of VTEa

marked increased risk of developing VTE, with an OR of 38.3 (95% CI 9.6, 152.5), compared with study patients having no available LVEF measurement. An LVEF of 20– 44% was associated with an increased risk of VTE, with an odds ratio of 2.8 (95% CI 1.4, 5.7). Subjects with a borderline or normal LVEF also had mildly increased risk compared with those without available LVEFs. This model also identified previous VTE as a very powerful predictor of current VTE, and surgical state and obesity also predicted current VTE.

Risk Factor

B-coefficient

Odds ratio (95% CI)

CHF Prior VTE Surgical stateb Obesity

0.96 3.57 2.29 1.23

2.61 (1.44, 4.73) 35.50 (14.48, 87.04) 9.86 (4.34, 22.38) 3.66 (1.85, 7.23)

VTE  venous thromboembolism, CHF  congestive heart failure, CI  confidence interval. a Controlling for diagnoses of coronary artery disease, chronic obstructive pulmonary disease, diabetes mellitus, and hypertension. b “Surgical state” includes all patients with recent orthopedic or other surgery, fracture, or trauma. Model c statistic  0.78. Hosmer Lemeshow statistic  2.58; P  .92 with 7 df.

814

M.D. Howell et al. / Journal of Clinical Epidemiology 54 (2001) 810–816

Table 4 Logistic regression model evaluating LVEF and VTEa Risk Factor

B-coefficient

Odds ratio (95% CI)

LVEF  20% LVEF 20–44% LVEF  45% Prior VTE “Surgical state”b Obesity

3.64 1.03 0.54 3.77 2.36 1.35

38.3 (9.6, 152.5) 2.8 (1.4, 5.7) 1.7 (1.03, 2.9) 43.4 (17.0, 111.1) 10.6 (4.6, 24.6) 3.9 (1.9, 7.6)

VTE  venous thromboembolism, CHF  congestive heart failure, LVEF  left ventricular ejection fraction, CI  confidence interval. a Controlling for diagnoses of coronary artery disease, chronic obstructive pulmonary disease, diabetes mellitus, and hypertension. Patients whose LVEF was not measured before or within 3 months of the index admission are defined as an odds ratio of 1.0. b “Surgical state” includes all patients with recent orthopedic or other surgery, fracture, or trauma. Model c statistic  0.80. Hosmer Lemeshow statistic  5.65; P  .58 with 7 df.

only patients without previous deep vein thrombosis. Our study was specifically designed to assess the relationship between CHF and VTE; importantly, we required objective documentation for clinical diagnoses of both CHF and VTE, given the notorious inaccuracy of clinical and historial assessment of these diseases [39–43]. Reports in the literature suggest that VTE is an underestimated cause of morbidity and mortality in patients with CHF. In one study of pediatric patients with dilated cardiomyopathy awaiting transplant the incidence of pulmonary embolism was 13.9% [44]. This is particularly striking as PE is very uncommon in the young. Grubman et al. reviewed cardiac deaths in patients with third-generation AICDs capable of storing electrograms [45]. One thousand seven hundred twenty-nine patients were followed from 2 to 4 years. Over this period of time there were 119 deaths, of which 5 were attributed to PE. However, only 15% of the time was an autopsy done, and cause of death was determined from clinical notes in the majority of the cases. Ventricular arrhythmias were not a major cause of death, but frequently electromechanical dissociation and bradyarrhythmias were present terminally, and these are compatible with PE. PE is commonly under diagnosed and overlooked as a cause of sudden death, particular in cardiac patients in the absence of autopsy. Thus, if anything, the reported 4.2% rate of PE in Grubman’s study underestimates the incidence of fatal PE. A new finding in the present study is that the increased VTE risk due to CHF is LVEF dependent, with extremely high risk at LVEFs less than 20%. This relationship is both clinically and pathophysiologically plausible. As LVEF falls, venous stasis worsens and patients’ exercise tolerance and mobility decrease. Biochemical evidence for hypercoagulability in CHF is provocative [10,12], and these markers increase as LVEF falls [12]. In addition, treatment with warfarin decreases these markers in a dose-dependent fashion [46].

Even patients with relatively preserved systolic function (LVEF  45%) had a slightly increased risk of VTE when compared with study patients without available EFs. This high EF group included patients with syndromes of diastolic dysfunction or right heart failure, conditions not included in our definition of CHF. One would expect that such patients might be prone to VTE for similar reasons as patients with left heart failure, plus the increased venous stasis seen with right-sided failure. Our study has several limitations. The case-control design is vulnerable to biases in selection of cases and in case’s recall of exposures. We believe that our study design minimized the potential for such biases. For example, we evaluated all patients who had an ICD-9 discharge code suggestive of VTE among their first four discharge diagnoses. We used cases definitions specified before data collection. CHF status was determined through independent, blinded review by two investigators not involved in data collection. These reviewers agreed on CHF status in over 95% of all study patients. Controls were a randomly selected sample of inpatients with acute infections or diabetes and its complications. We believe the controls would have an opportunity for EF measurement similar to or greater than that of the cases. We also do not believe that the medical record would preferentially note a history of CHF in cases versus controls. We suspect that the admitting and treating clinicians were more thorough in ascertaining VTE risk factors for cases than for controls. Consequently, the odds ratios for certain VTE risk factors (e.g., a history of previous VTE) may be overestimated. Clinicians may have had a higher diagnostic suspicion of VTE in patients with CHF and thus may have pursued the diagnosis of VTE more vigorously in these patients than in the control populations. Thus, it is likely that a type of observation bias, described by Hennekens and Buring as interviewer bias, is present in our study [47]. Nevertheless, these are well-established VTE risk factors, and it was essential that we control for them in our assessment. Finally, we would have preferred to have EF measurements on all study patients, as it is possible there are cases and controls with cardiac dysfunction that has not been detected. However, we did not detect a substantial difference in EF assessment between the two groups. Our results indicate a significant risk of VTE, particularly among patients with ejection fractions less than 20%. Combining our observations with previous reports in the literature, a case can be made for anticoagulation of patients with heart failure in the absence of atrial fibrillation based on three lines of evidence. In addition to the findings of the present study showing a markedly increased risk of venous thrombosis and embolism, there is also an increased risk of arterial embolism, though conflicting results have been reported on the clinical significance of this [9,15,48,49]. Moreover, there is growing evidence that aspirin, an alternative for thrombosis prevention, interferes with the action of ACE inhibitors, making aspirin less attractive as prophy-

M.D. Howell et al. / Journal of Clinical Epidemiology 54 (2001) 810–816

laxis against venous and arterial embolism in heart failure patients [50–52]. Thus, the indication for anticoagulation with warfarin is supported by: 1) a markedly increased risk of VTE; 2) risk of arterial thromboembolism; and 3) inhibition of the action of ACE inhibitors by aspirin. In conclusion, our results suggest that CHF is an independent risk factor for the development of VTE, and that this risk is especially marked in patients with an LVEF 20%. Our findings provide evidence supporting the practice of chronic anticoagulation in selected patients with CHF, particularly when taken in concert with previous work suggesting decreased all-cause mortality in CHF patients on warfarin [9]. Acknowledgments The authors thank Andrew Schafer for his continued support and guidance. The authors thank Michael Johnson, Ph.D., for assistance with programming and statistical analysis.

[16]

[17]

[18]

[19]

[20] [21] [22]

[23]

References [1] Rodeheffer RJ, Jacobsen SJ, Gersh BJ, Kottke TE, McCann HA, Bailey KR, Ballard DJ. Incidence and prevalence of congestive heart failure in Rochester, Minnesota. Mayo Clin Proc 1993;68:1143–50. [2] Clarke KW, Gray D, Hampton JR. How common is heart failure? Evidence from PACT (Prescribing Analysis and Cost) data in Nottingham. J Public Health Med 1995;17:459–64. [3] Kupari M, Lindroos M, Iivanainen AM, Heikkila J, Tilvis R. Congestive heart failure in old age: prevalence, mechanisms and 4-year prognosis in the Helsinki Ageing Study. J Int Med 1997;241:387–94. [4] Bourassa MG, Gurne O, Bangdiwala SI, Ghali JK, Young JB, Rousseau M, Johnstone DE, Yusuf S. Natural history and patterns of current practice in heart failure. J Am Coll Cardiol 1993;22:14A–19A. [5] Ho KKL, Pinsky JL, Kannel WB, Levy D. Epidemiology of heart failure: the Framingham Study. J Am Coll Cardiol 1993;22:6A–13A. [6] Dargie HJ, McMurray JJV, McDonagh TA. Heart failure—implications of the true size of the problem. J Int Med 1996;239:309–15. [7] Ho KKL, Anderson KM, Kannel WB, Grossman W, Levy D. Survival after the onset of congestive heart failure in Framingham Heart Study subjects. Circulation 1993;88:107–15. [8] Harvey WP, Finch CA. Dicumarol prophylaxis of thromboembolic disease in congestive heart failure. N Engl J Med 1950;242:208–11. [9] Al-Khadra AS, Salem DN, Rand WM, Udelson JE, Smith JJ, Konstam MA. Warfarin anticoagulation and survival: a cohort analysis from the Studies of Left Ventricular Dysfunction. J Am Coll Cardiol 1998;31:749–53. [10] Jafri SM. Hypercoagulability in heart failure. Semin Thromb Hemostasis 1997;23:543–5. [11] Jafri SM, Ozawa T, Mammen E, Levine TB, Johnsons C, Goldstein S. Platelet function, thrombin and fibrinolytic activity in patients with heart failure. Eur Heart J 1993;14:205–12. [12] Sbarouni E, Bradshaw A, Andreotti F, Tuddenham E, Oakley CM, Cleland JGF. Relationship between hemostatic abnormalities and neuroendocrine activity in heart failure. Am Heart J 1994;127:607–12. [13] Eckman MH, Levine HJ, Pauker SG. Making decisions about antithrombotic therapy in heart disease: decision analytic and cost-effectiveness issues. Chest 1995;108(Suppl 4):457S–470S. [14] Loh E, Sutton MSJ, Wun CCC, Rouleau JL, Flaker GC, Gottlieb SS, Lamas GA, Moye LA, Goldhaber SZ, Pfeffer MA. Ventricular dysfunction and the risk of stroke after myocardial infarction. N Engl J Med 1997;336:251–7. [15] Dunkman WB, Johnson GR, Carson PE, Bhat G, Farrell L, Cohn J.

[24]

[25]

[26]

[27] [28]

[29]

[30]

[31]

[32]

[33]

[34] [35] [36] [37]

815

Incidence of thromboembolic events in congestive heart failure. Circulation 1993;87(Suppl VI):VI94–VI101. Dries DL, Rosenberg YD, Waclawiw MA, Domanski MJ. Ejection fraction and risk of thromboembolic events in patients with systolic dysfunction and sinus rhythm: evidence for gender differences in the studies of left ventricular dysfunction trials. J Am Coll Cardiol 1997; 29:1074–80. Cioffi G, Pozzoli M, Forni G, Franchini M, Opasich C, Cobelli F, Tavazzi L. Systemic thromboembolism in chronic heart failure: a prospective study in 406 patients. Eur Heart J 1996;17:1381–9. Eckman MH, Levine HJ, Salem DN, Pauker SG. Making decisions about antithrombotic therapy in heart disease: decision analytic and cost-effectiveness issues. Chest 1998;114:699S–714S. Tsevat J, Eckman MH, McNutt RA, Pauker SG. Warfarin for dilated cardiomyopathy: a bloody tough pill to swallow? Med Decis Making 1989;9:162–9. Koniaris LS, Goldhaber SZ. Anticoagulation in dilated cardiomyopathy. J Am Coll Cardiol 1998;31:745–8. Ginsberg JS. Management of venous thromboembolism. N Engl J Med 1996;335:1816–28. Hirsch J, Hoak J. Management of deep vein thrombosis and pulmonary embolism: a statement for healthcare professionals. Circulation 1996;93:2212–45. The PIOPED Investigators. Value of the ventilation/perfusion scan in acute pulmonary embolism: results of the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED). JAMA 1990;263: 2753–9. Anderson FA, Wheeler HB, Goldberg RJ, Hosmer DW, Patwardhan NA, Jovanovic BJ, Forcier A, Dalen JE. Population-based perspective of the hospital incidence and case-fatality rates of deep vein thrombosis and pulmonary embolism: the Worcester DVT study. Arch Int Med 1991;151:933–8. Anderson FA, Wheeler HB, Goldberg RJ, Hosmer DW, Forcier A. Prevalence of risk factors for venous thromboembolism among hospital patients. Arch Int Med 1992;152:1660–4. Cogo A, Bernardi E, Prandoni P, Girolami B, Noventa F, Simioni P, Girolami A. Acquired risk factors for deep-vein thrombosis in symptomatic outpatients. Arch Int Med 1994;154:164–8. Anderson FA, Wheeler HB. Venous thromboembolism: risk factors and prophylaxis. Clin Chest Med 1995;16:235–51. Kniffing WD, Baron JA, Barrett J, Birkmeyer JD, Anderson FA. Epidemiology of diagnosed pulmonary embolism and deep venous thrombosis in the elderly. Arch Int Med 1994;154:861–6. Heit JA, Silverstein MD, Mohr DN, Petterson TM, O’Fallon M, Melton JL III. Risk factors for deep vein thrombosis and pulmonary embolism. Arch Int Med 2000;160:809–15. Secretary of Veterans Affairs. Annual report of the Secretary of Veterans Affairs. Washington, DC: Department of Veterans Affairs, 1995. p. 39. Jolis JG, Ancukiewicz M, DeLong ER, Pryor DB, Muhlbaier LH, Mark DB. Discordance of databases designed for claims payment vs. clinical information systems. Ann Int Med 1993;119:844–59. Romano P, Luft HS. Getting the most out of messy data: problems and approaches for dealing with large administrative data sets. In: Grady ML, Schwartz HA, editors. Medical effectiveness research data methods. Rockville, MD: US Department of Health and Human Services, 1992. pp. 57–75. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143: 29–36. Hosmer J, Lemeshow S. Applied logistic regression. New York: John Wiley and Sons, 1989. Clagett GP, Anderson FA, Heit JA, Lieberman JR, Wheeler HB. Prevention of venous thromboembolism. Chest 1998;114:531S–560S. Lederle FA. Heparin prophylaxis for medical patients? Ann Int Med 1998;128:767–70. Belch JJ, Lowe DO, Ward AG, Forbes CD, Prentice CRM. Preven-

816

[38]

[39] [40]

[41] [42]

[43] [44]

[45]

M.D. Howell et al. / Journal of Clinical Epidemiology 54 (2001) 810–816 tion of deep vein thrombosis in medical patients by low-dose heparin. Scottish Med J 1981;26:115–7. Halkin H, Goldberg J, Modan M, Modan B. Reduction of mortality in general medical patients by low-dose heparin prophylaxis. Ann Int Med 1982;96:561–5. Badgett RG, Lucey CR, Mulrow CD. Can the clinical examination diagnose left-sided heart failure in adults? JAMA 1997;277:1712–9. Remes J, Miettinen H, Reunanen A, Pyorala K. Validity of clinical diagnosis of heart failure in primary health care. Eur Heart J 1991;12: 315–21. Cleland JGF, Habib F. Assessment and diagnosis of heart failure. J Int Med 1996;239:317–25. Wells PS, Hirsh J, Lensing AWA, Foster G, Kearon C, Weitz J, D’Ovidio R, Cogo A, Prandoni P, Girolami A, Ginsberg JS. Accuracy of clinical assessment of deep-vein thrombosis. Lancet 1995; 345:1326–30. Anand SS, Wells PS, Hunt D, Brill-Edwards P, Cook D, Ginsberg JS. Does this patient have deep vein thrombosis? JAMA 1998;279:1094–9. Hsu DT, Addonizio LJ, Hordof AJ, Gersony WM. Acute pulmonary embolism in pediatric patients awaiting heart transplantation. J Am Coll Cardiol 1991;17:1621–5. Grubman EM, Pavri BB, Shipman T, Britton N, Kocovic DZ. Cardiac death and stored electrograms in patients with third-generation implantable cardioverter-defibrillators. J Am Coll Cardiol 1998;32: 1056–62.

[46] Jafri SM, Mammen EF, Masura J, Goldstein S. Effects of warfarin on markers of hypercoagulability in patients with heart failure. Am Heart J 1997;134:27–36. [47] Hennekens CH, Buring JE, Mayrent SL. Epidemiology in medicine. Boston: Little, Brown and Company, 1987. [48] Dries DL, Rosenber YD, Waclawiw MA, Domanski MJ. Ejection fraction and risk of thromboembolic events in patients with systolic dysfunction and sinus rhythm: evidence for gender differences in the Studies of Left Ventricular Dysfunction trials. J Am Coll Cardiol 1997;29:1074–80. [49] Kyrle PA, Korninger C, Gossinger H, Glogar D, Lechner K, Niessner H, Pabinger I. Prevention of arterial and pulmonary embolism by oral anticoagulants in patients with dilated cardiomyopathy. Thromb Haemost 1985;54:521–3. [50] Hall D, Zeitler H, Rudolph W. Counteraction of the vasodilator effects of enalapril by aspirin in severe heart failure. J Am Coll Cardiol 1992;20:1549–55. [51] Spaulding C, Charbonnier B, Cohen-Solal A, Juillière Y, Kromer EP, Benhamada K, Cador R, Weber S. Acute hemodynamic interaction of aspirin and ticlopidine with enalapril: results of a double-blind, randomized comparative trial. Circulation 1998;98:757–65. [52] Cleland JGF, Bulpitt CJ, Falk RH, Findlay IN, Oakley CM, Murray G, Poole-Wilson PA, Prentice CRM, Sutton GC. Is aspirin safe for patients with heart failure? Br Heart J 1995;74:215–9.