- Email: [email protected]
Clinical and economic consequences of bleeding following major orthopedic surgery
Thrombosis Research (2006) 117, 569 — 577
intl.elsevierhealth.com/journals/thre
REGULAR ARTICLE
Clinical and economic consequences of bleeding following major orthopedic surgery Montserrat Vera-Llonch, May Hagiwara, Gerry Oster* Policy Analysis Inc. (PAI), Four Davis Court, Brookline, MA 02445, USA Received 2 December 2004; received in revised form 14 April 2005; accepted 18 April 2005 Available online 23 May 2005
KEYWORDS Orthopedic surgery; VTE; Bleeding; Outcomes; Hemorrhage; Costs; Cost analysis
Abstract Background: Major orthopedic surgery patients who receive antithrombotics as prophylaxis against VTE may be at higher risk of bleeding. The clinical and economic consequences of this complication may be relevant to therapeutic decision-making. Objective: To assess the impact of major bleeding following major orthopedic surgery on length of stay (LOS) and inpatient charges. Methods: Using a database with information on ~750,000 admissions annually to 100+ US acute-care hospitals, we identified all patients who underwent major orthopedic surgery between January 1, 1998 and December 13, 2000. Patients were stratified according to whether or not they experienced major postoperative bleeding prior to hospital discharge, defined as (a) fatal bleeding; (b) nonfatal bleeding at critical site; (c) re-operation due to bleeding; and (d) overt bleeding with bleeding index (BI) z 2, where BI = number of blood units transfused plus pre-bleeding minus postbleeding hemoglobin (g/dL) values. LOS and total inpatient charges were compared between patients with and without major bleeding. Results: The incidence of major bleeding among 23,518 patients who underwent major orthopedic surgery was 2.6%. In multivariate analyses controlling for differences in baseline characteristics between patients with and without major bleeds, adjusted mean LOS was 1.8 days longer among those with major bleeding (95% CI: 1.5, 2.0) (6.1 days vs. 4.3 days for those without bleeds); adjusted mean total inpatient charges were $7593 higher (95% CI: $6622, $8646) ($25,669 vs. $18,076). Conclusion: Bleeding following major orthopedic surgery may increase length of stay and total hospital charges and should be an important consideration in choice of VTE prophylaxis. D 2005 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: +1 617 232 4400; fax: +1 617 232 1155. E-mail address: [email protected] (G. Oster). 0049-3848/$ - see front matter D 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.thromres.2005.04.018
570
Introduction Deep-vein thrombosis (DVT) and pulmonary embolism (PE)—together referred to as venous thromboembolism (VTE)—are important causes of morbidity and death among hospitalized patients, particularly those undergoing major surgery [1—4]. Major surgery confers an especially high risk of VTE due to postoperative immobility as well as the effects of surgical trauma on the coagulation system. Among common major surgical procedures, arthroplasty of the hip or knee and hip-fracture repair confer the highest risks of VTE, probably due to direct trauma to deep veins and the marked venous stasis associated with these procedures. Because of the difficulty of early diagnosis and the potentially serious consequences of VTE, prophylaxis has long been recommended in high-risk patients. Most prophylactic regimens today include an antithrombotic, usually a low-molecular-weight heparin (LMWH) or warfarin. While the efficacy of antithrombotic therapy is well established [5], patients who receive chemoprophylaxis may be at increased risk of postoperative bleeding, which may differ by prophylaxis received. In meta-analyses of prophylactic regimens among patients undergoing hip replacement, pooled rates of major bleeding were estimated to be 3.3% for oral anticoagulants and 5.3% for enoxaparin, although the trials that were reviewed used somewhat different definitions of this outcome [6,7]. In patients undergoing total knee replacement, rates of bleeding have been reported to be 0.9% for warfarin and 2.8% for LMWH [8]. Most recently, in a meta-analysis of data from four randomized controlled trials in patients undergoing elective hip or knee replacement or hip fracture repair who were randomized to receive fondaparinux or enoxaparin, the rate of major bleeding was reported to be 2.7% for the former and 1.7% for the latter [9]. Bleeding following major orthopedic surgery may have important clinical consequences, extending stays in hospital and increasing costs of care. We examine this issue in this study using a large US inpatient database.
Materials and methods Overview Using a large computerized inpatient database, we compared the length of stay in hospital and total inpatient charges between major orthopedic sur-
M. Vera-Llonch et al. gery patients who developed postoperative major bleeding vs. those who did not. Data were obtained from the MQ-Profile Database v3.3 (bMQ-ProQ) (Cardinal Information Corporation, Marlborough, MA), which contains clinical and administrative information on approximately 750,000 inpatient admissions annually to over 100 US acute-care hospitals. The study sample consisted of all patients in MQPro who underwent major orthopedic surgery (total hip replacement [THR], major knee surgery [MKS], or hip fracture repair [HFR]) between January 1, 1998 and December 31, 2000 (the most recent 3year period for which data were available at the time the study was conducted). Major postoperative bleeding was defined as the occurrence of (a) fatal bleeding; (b) nonfatal bleeding at a critical site; (c) re-operation due to bleeding; or (d) overt bleeding with a bleeding index (BI) z 2, where BI is calculated as the number of whole blood or packed red blood cell units transfused plus the difference between pre-bleeding and post-bleeding hemoglobin (g/dL) values. This definition of major postoperative bleeding is consistent with the one employed in a recent Phase III clinical trials program of fondaparinux [10—12]. We then compared length of stay in hospital and total inpatient charges between patients who experienced major postoperative bleeding vs. those who did not using both univariate and multivariate techniques. Findings also were examined for each of the constituent measures of the composite endpoint (i.e., fatal bleeding, non-fatal bleeding at a critical site, re-operation due to bleeding, and overt bleeding with BI z 2).
Data source MQ-Pro is maintained by Cardinal Information Corporation (CIC), which sells and distributes software (Atlask) to US acute-care hospitals for the collection of detailed clinical and administrative data. Hospitals use this information for a variety of quality-improvement activities; data also are submitted by each hospital to CIC for use in proprietary databases (including MQ-Pro), which are used primarily by hospitals for benchmarking studies. MQ-Pro is the largest of these databases and contains clinical and administrative data on approximately 750,000 inpatient admissions annually to over 100 US acute-care hospitals. Some data elements in MQ-Pro (e.g., patient demographics, diagnoses, admission and discharge dates, charges) are extracted from hospital information systems. Other information is collected manually via chart review by trained abstractors
Clinical and economic consequences of bleeding following major orthopedic surgery using a standardized glossary. Hospitals participating in MQ-Pro must collect data on all patients admitted to their facility, ensuring complete data capture and minimizing potential selection or reporting biases. While collection of data for all patients is mandatory, some data elements are optional. For example, facilities may choose to collect detailed clinical information for the first 5 days of each admission or for the first 2 days only. Similarly, some facilities choose to collect detailed information on transfusions while others do not. For all admissions in the database, MQ-Pro includes information on patient demographics (e.g., age, gender), admission source, medical history, all documented diagnosis codes (in ICD-9CM format), length of stay, billed total and ancillary hospital charges, and discharge disposition. The database also includes a listing of all procedures performed during the hospital stay (coded in ICD-9CM format) along with the day the procedure began. Information also is collected for each admission on over 400 key clinical findings (bKCFsQ). KCFs are collected from admission through the second or fifth day in hospital, depending upon the collection option chosen by the facility. All KCFs are dated and include observations obtained through physician assessments, procedures, pathology, imaging and electrocardiograms (EKGs). KCFs also are used to assess clinical severity on admission; admission severity groups (ASGs) (range, 0—4) are derived based on the probability of in-hospital mortality, which is calculated using disease-specific logistic regression models. The predictive capabilities of these models have been shown to be comparable to those of other severity adjustment methodologies (e.g., APACHE II, Disease Staging) [13—16]. In addition to KCFs, the database includes information on the values of laboratory tests performed during hospitalization, including hematology, coagulation, chemistry, urine, blood gases, gastrointestinal and miscellaneous tests. This information is recorded on a daily basis. If multiple values are available for any given day, the most extreme value (i.e., highest, lowest, or furthest from some benchmark) is typically recorded. Daily values for selected vital signs also are recorded. A number of optional data elements also are included in MQ-Pro. Some facilities collect information on unplanned clinical events during hospitalization (beventsQ), including (but not limited to) barrest,Q bbleeding,Q and bcardiac events.Q Events are coded using a proprietary system. Information on each such event includes its nature (e.g., transfusion, return to operating room [OR], transfer to intensive care unit [ICU]) and the day it
571
occurred. Facilities also may choose to collect detailed information on transfusions. For all hospitals that collect such data, information is available on the type of transfusion (e.g., whole blood, packed red cells, plasma, platelets), the day of transfusion and the number of units administered. Hospitals also may collect information on pre- and post-transfusion values for selected coagulation parameters (e.g., hemoglobin). Other optional data elements include information on time spent in special-care units and whether patients received selected procedures (e.g., mechanical ventilation, thrombolytic therapy).
Patient selection All patients in MQ-Pro who had undergone THR (ICD9-CM procedure codes 81.51, 81.53), MKS (81.54, 81.55), or HFR (79.15, 79.35, 81.4, 81.52) between January 1, 1998 and December 31, 2000 were identified. As we were principally interested in the clinical and economic impact of bleeding associated with VTE prophylaxis, patients undergoing treatment for VTE (based on secondary diagnosis codes of DVT [ICD-9-CM diagnosis codes 451.1, 451.2, and 453.8] and/or PE [415.1]) were excluded from the study sample, as were those with multiple trauma (ICD-9CM 800.XX-804.XX [skull fracture], 805.XX-809.XX [neck and trunk fracture], 810.XX-819.XX [upper limb fracture], 850.XX-854.XX [intracranial injury], 860.XX-869.XX [internal injury of thorax, abdomen, and pelvis], and 828.XX [multiple fractures involving both lower limbs, lower with upper limb, and lower limb(s) with rib(s) and sternum]). Patients with evidence of major bleeding prior to surgery also were excluded, as were those with a transfer KCF of bleeding and those aged 18 years or younger. All remaining patients in the sample were then stratified according to whether or not they experienced major bleeding during the postoperative period, which was defined as beginning on the day of the surgical procedure (inclusive of that date) and ending on the day of hospital discharge or death. Consistent with the definition used in the fondaparinux Phase III clinical trials programs, major bleeding was defined as follows: Fatal bleeding: A KCF of bbleedingQ or an beventQ due to bleeding during the postoperative period with or without a discharge ICD-9-CM diagnosis code of bleeding, and evidence of death at discharge, without evidence of stroke (ICD-9-CM 436) or myocardial infarction (410.X). Nonfatal bleeding at a critical site: A KCF of bbleedingQ with KCF body-site modifiers suggesting bleeding at a critical site (i.e., intracranial [KCF
572 modifier, brain], retroperitoneal [abdomen], intraocular [eye], pericardium [heart], or adrenal [endocrine]), or a discharge diagnosis code suggestive of bleeding at a critical site (ICD-9-CM 432.X [hemorrhage, intracranial], 360.43 [hemorrhage, eye], 364.41 [hemorrhage, anterior chamber], 362.81 [hemorrhage, fundus], 336.1 [hemorrhage, spinal cord], 459.0 [hemorrhage, retroperitoneal], 423.0 [hemopericardium], or 255.4 [hemorrhage, adrenal capsule]) and discharge status alive. Re-operation due to bleeding: An beventQ labeled breturn to ORQ due to bleeding during the postoperative period, or one or more of the following procedure codes (in ICD-9-CM format) after the date of the initial surgical procedure: 80.85 (repeat operation for hemorrhage control: hip replacement and fracture repair), 80.15 (repeat operation for drainage of hematoma; hip replacement and fracture repair), 80.86 (repeat operation, for hemorrhage control, knee surgery), or 80.16 (repeat operation, drainage of hematoma, knee surgery). Overt bleeding associated with BI z 2: A KCF of bbleeding,Q or an beventQ due to bleeding, with or without a discharge diagnosis code of bleeding, and a bleeding index (BI) z 2, where BI is calculated as the number of whole blood or packed red blood cell units transfused plus the difference between the pre-bleeding and post-bleeding hemoglobin (g/dL) values during the postoperative period. Evidence of transfusion, type of blood products (whole blood and packed red blood cells) and number of units transfused were identified based on information on interventions in the study database. Pre- and post-bleeding hemoglobin values were ascertained from daily hemoglobin laboratory values dated immediately prior and subsequent to the date of the bleeding event. The criteria that were used to identify patients with major bleeding were implemented in order of presentation above and were deemed to be mutually exclusive (e.g., a patient who underwent reoperation due to bleeding who also met the criteria for a fatal bleed was included in the latter but not the former category).
Study measures The measures of interest were total length of stay in hospital and total inpatient charges.
Analyses Demographic and clinical characteristics of study subjects were examined, including age, gender, ASG and the presence of selected comorbidities
M. Vera-Llonch et al. (coagulation disorders [ICD-9-CM 286.0, 286.1, 286.2, 286.4, 286.5, 286.7], heart disease [393398, 401—405, 410—414, 415—417, 420—429], and hepatic and/or renal disorders [570—573, 580— 599]). Differences between patients with and without major bleeding were tested using standard univariate techniques (i.e., t-tests for continuous measures, contingency tables [chi-square and Fisher’s exact tests, as appropriate] for categorical measures). An alpha level of 0.05 was used to assess statistical significance. Differences between THR, MKS, and HFR patients were not tested for statistical significance. Length of stay in hospital and total inpatient charges were compared between patients with and without major bleeding on both a univariate (i.e., unadjusted) basis, as well as controlling for differences in baseline demographic and clinical characteristics using multivariate least-squares regression. The dependent variable in these regression models was either length of stay in hospital or total inpatient charges, as appropriate. The dependent variable was log-transformed prior to analysis to correct for skewness. Explanatory (i.e., independent) variables included age, gender, ASG and the presence of coagulation disorders, heart disease and hepatic and/or renal disorders. Adjusted least-squares means were generated for each group and were retransformed to their natural values using Duan’s smearing estimator [17]. Ninety-five percent confidence intervals were calculated for both unadjusted and adjusted estimates of mean length of stay in hospital and mean inpatient charges. Findings also were examined for each of the constituent categories of the composite measure (i.e., fatal bleeding, nonfatal bleeding at a critical site, re-operation due to bleeding, and overt bleeding associated with a BI z 2). All analyses were undertaken using PC-SAS for Windows (version 8.0).
Results A total of 25,132 patients were identified in MQ-Pro database who underwent THR, MKS, or HFR between January 1, 1998 and December 31, 2000. A total of 262 patients were excluded from the sample who had secondary diagnoses of VTE (DVT and/or PE). We also excluded patients for whom a date of service for the principal procedure could not be determined (n = 36), those with evidence of bleeding prior to hospital admission (n = 1) or the date of their surgical procedure (n = 224), those with multiple trauma (n = 858) and those aged 18 years or younger (n = 230). Finally, three patients
Clinical and economic consequences of bleeding following major orthopedic surgery Table 1
Characteristics of study subjects, by surgical procedure
Parameter
Surgical procedure Total hip replacement (n = 5940)
Age (mean F S.D.) Male gender (n, %) Comorbid conditions (n, %) Coagulation disorder Heart disease Hepatic and/or renal disorders Admission severity group (ASG) (n, %) 0 1 2 3 4
Major knee surgery (n = 11,030)
67.8 (13.1) 2432 (40.9) — 4 (0.1) 3495 (58.8) 442 (7.4) 329 5295 291 25 0
68.3 (10.4) 3912 (35.5) — 10 (0.1) 7107 (64.4) 596 (5.4)
(5.5) (89.1) (4.9) (0.4) (0.0)
were excluded due to unknown gender. The resulting study sample therefore consisted of 23,518 patients (THR, 5940; MKS, 11,030; HFR, 6548). Selected demographic and clinical characteristics of study subjects are presented in Table 1. Mean (F S.D.) age across all surgical procedures was 70.3 (F 13.8) years; HFR patients were older than those undergoing THR or MKS. Slightly less than two-thirds of study subjects were women; the proportion was highest among those undergoing HFR. The overall prevalence of coagulation disorders was 0.1%; heart disease, 63%; and hepatic and/or renal disorders, 9.6%. Hepatic and/or renal disorders were more prevalent among patients undergoing HFR, probably reflective of their older age. The majority of patients were in ASG 1, with the exception of those undergoing HFR, who were most likely to be in ASG 2. The overall incidence of major bleeding was 2.6% (fatal bleeding, 0.1%; non-fatal bleeding at a critical site, 0.2%; re-operation due to bleeding, 0.7%; and overt bleeding associated with a BI z 2, 1.7%) (Table 2). Rates of major bleeding among THR, MKS, and HFR patients were 4.9%, 1.3% and
Table 2
573
98 10,882 50 0 0
(0.9) (98.7) (0.5) (0.0) (0.0)
Hip fracture repair (n = 6548)
Total (n = 23,518)
76.0 (17.3) 1926 (29.4) — 12 (0.2) 4211 (64.3) 1222 (18.7)
70.3 (13.8) 8270 (35.2) — 26 (0.1) 14,813 (63.0) 2260 (9.6)
399 1710 3949 489 1
826 17,887 4290 514 1
(6.1) (26.1) (60.3) (7.5) (0.0)
(3.5) (76.1) (18.2) (2.2) (0.0)
2.8%, respectively. The incidence of overt bleeding associated with a BI z 2 was highest among THR patients (3.8%) and lowest among those undergoing MKS (0.5%). The rate of fatal bleeding was lowest among MKS patients. Differences in demographic and clinical characteristics between patients with and without major bleeding are reported in Table 3. Patients with major bleeding were more likely to be men ( p b 0.0001), more likely to have renal and/or hepatic disorders ( p b 0.0001), and more likely to be sicker on admission than those without major bleeding ( p b 0.0001). Findings with respect to age, however, were inconsistent across the three surgical procedures. THR and MKS patients who experienced major bleeding were significantly older than those who did not ( p b 0.0001 and p = 0.02, respectively), while HFR patients who experienced major bleeding were younger ( p = 0.025). Mean length of stay in hospital was 3.1 days (95% CI: 2.9, 3.4) longer for patients with major bleeding (7.9 days vs. 4.8 days for those without bleeding). The difference in length of stay was of similar magnitude across all three surgical procedures
Incidence of major bleeding among subjects undergoing major orthopedic surgery, by surgical procedure
Parameter
No major bleeding (n, %) Any major bleeding (n, %) Fatal bleeding (n, %) Nonfatal bleeding at critical site (n, %) Bleeding associated with re-operation at the operative site (n, %) Overt bleeding with BI z 2 (n, %) BI = bleeding index.
Surgical procedure Total hip replacement (n = 5940)
Major knee surgery (n = 11,030)
Hip fracture repair (n = 6548)
Total (n = 23,518)
5647 293 6 10 53
10,890 140 3 11 73
6367 181 12 18 34
22,904 614 21 39 160
(95.1) (4.9) (0.1) (0.2) (0.9)
224 (3.8)
(98.7) (1.3) (0.03) (0.1) (0.7)
53 (0.5)
(97.2) (2.8) (0.2) (0.3) (0.5)
117 (1.8)
(97.4) (2.6) (0.1) (0.2) (0.7)
394 (1.7)
574
Table 3
Characteristics of subjects with and without major bleeding, by surgical procedure
Parameter
Surgical procedure Total hip replacement
Age (mean F S.D.) Male gender (n, %) Comorbid conditions (n, %) Coagulation disorders Heart disease Hepatic and/or renal disorders Admission severity group (ASG) 0 1 2 3 4
Major knee surgery
No major bleeding (n = 6647)
Major bleeding (n = 293)
p-value
67.7 (13.0) 2303 (40.8)
70.8 (13.3) 129 (44.0)
b0.0001 0.271
—
—
3307 (68.7) 384 (6.8)
188 (64.2) 68 (19.8)
0.057 b0.0001
316 (6.7) 6069 (89.7) 262 (4.5) 20 (0.4) 0 (0.0)
13 (4.4) 236 (80.6) 39 (13.3) 6 (1.7) 0 (0.0)
b0.0001
4 (0.1)
No major bleeding (n = 10,890)
Hip fracture repair
Total
Major bleeding (n = 140)
p-value
No major bleeding (n = 6367)
Major bleeding (n = 181)
p-value
70.4 (9.9) 78 (66.7)
0.021 b0.0001
76.1 (17.2) 1862 (29.1)
72.8 (19.7) 74 (40.9)
0.025 b0.001
2 (1.4)
b0.001
12 (0.2)
—
—
7020 (64.5) 670 (6.2)
87 (62.1) 26 (18.7)
0.569 b0.0001
4091 (64.3) 1168(18.3)
120 (66.3) 64 (29.8)
0.571 b0.0001
96 (0.9) 10,762 (98.7) 42 (0.4) 0 (0.0) 0 (0.0)
2 (1.4) 130 (92.9) 8 (6.7) 0 (0.0) 0 (0.0)
b0.0001
381(7.0) 1666 (26.2) 3844 (60.4) 476 (7.5) 1 (0.0)
18 (9.9) 46 (24.9) 106 (68.0) 13 (7.2) 0 (0.0)
0.303
68.3 (10.4) 3834 (36.2)
8 (0.1)
No major bleeding (n = 22,904)
Major bleeding (n = 614)
p-value
70.3 (13.7) 7989(34.9)
71.3 (14.8) 281 (46.8)
0.115 b0.0001
24 (0.1)
2 (0.3)
0.147
14,418 (62.9) 2122 (9.3)
396 (64.3) 138 (22.5)
0.484 b0.0001
793 (3.5) 17,476 (76.3) 4138 (18.1) 496 (2.2) 1 (0.0)
33 (6.4) 411 (66.9) 162 (24.8) 18 (2.9) 0 (0.0)
b0.0001
M. Vera-Llonch et al.
Clinical and economic consequences of bleeding following major orthopedic surgery
575
Table 4 Adjusted length of stay (mean, 95% CI) in hospital among subjects with or without major bleeding, surgical procedure and type of bleeding Parameter
Surgical procedure
No major bleeding (n = 22,904) Any major bleeding (n = 614) Fatal bleeding (n = 21) Nonfatal bleeding at critical site (n = 39) Bleeding associated with re-operation at the operative site (n = 160) Overt bleeding with BI z 2 (n = 394)
Total hip replacement (n = 5940)
Major knee surgery (n = 11,030)
Hip fracture repair (n = 6548)
Total (n = 23,518)
4.2 5.7 4.7 4.1 8.2
3.9 5.6 5.4 5.4 5.6
5.4 7.3 8.1 5.5 9.4
4.3 6.1 6.3 4.9 7.2
(4.1, (5.3, (1.4, (2.8, (6.6,
4.2) 6.1) 14.3) 5.8) 10.0)
5.7 (5.3, 6.1)
(3.8, (5.0, (0.0, (3.7, (4.8,
3.9) 6.2) 17.5) 8.3) 6.6)
6.1 (5.2, 7.1)
(5.3, (6.7, (5.4, (4.4, (7.7,
5.4) 7.9) 12.9) 6.9) 11.6)
7.1 (6.5, 7.8)
(4.3, (5.8, (4.0, (4.1, (6.4,
4.3) 6.3) 9.5) 5.8) .8.0)
6.2 (5.9, 6.5)
BI = bleeding index.
(THR, 3.0 days [95% CI: 2.6, 3.3]; MKS, 3.2 days [2.9, 3.5]; HFR, 2.9 days [2.3, 3.5]). Length of stay also was consistently longer for patients who experienced any of the individual outcomes that were part of the composite endpoint. Patients with overt bleeding associated with a BI z 2, for example, had hospital stays that were 2.8 days longer on average than those of patients without major bleeding. In multivariate analyses controlling for differences in selected covariates (i.e., age, gender, ASG, and presence of coagulation disorders, heart disease and hepatic and/or renal disease) between patients with and without major bleeding, the covariate-adjusted difference in hospital length of stay was estimated to be 1.8 days (95% CI: 1.5, 2.0) (6.1 days vs. 4.3 days, respectively) (Table 4). Similar differences were noted for each of the surgical procedures (THR, 1.5 days [95% CI: 1.2, 2.0]; MKS, 1.7 days [1.2, 2.4]; HFR, 1.9 days [1.4, 2.5]). On an adjusted basis, patients with overt bleeding associated with a BI z 2 were found to stay
in hospital an average of 1.9 additional days compared with those without major bleeding. Mean total inpatient charges were $10,675 higher among patients with major bleeding (95% CI: $9853, $11,599) ($30,349 vs. $19,674 for those without bleeding). The difference was of similar magnitude across all three surgical procedures (THR, $10,915 [95% CI: $9817, $13,013]; MKS, $9852 [$8419, $11,386]; HFR, $10,200 [($8300, $13,100]). The difference in mean total charges for patients with overt bleeding associated with a BI z 2 compared with those without major bleeding was $9869. In multivariate analyses controlling for differences in covariates (i.e., age, gender, ASG and presence of coagulation disorders, heart disease and hepatic and/or renal disease) between patients with and without major bleeding, the adjusted difference in total inpatient charges was estimated to be $7593 (95% CI: $6622, $8646) ($25,669 vs. $18,076) (Table 5). Corresponding
Table 5 Adjusted total inpatient charges (mean, 95% CI) among subjects with and without major bleeding, by surgical procedure and type of bleeding Parameter
No major bleeding (n = 22,904) Any major bleeding (n = 614) Fatal bleeding (n = 21) Nonfatal bleeding at critical site (n = 39) Bleeding associated with re-operation at the operative site (n = 160) Overt bleeding with BI z 2 (n = 394) BI = bleeding index.
Surgical procedure Total hip replacement (n = 5940)
Major knee surgery (n = 11,030)
Hip fracture repair (n = 6548)
Total (n = 23,518)
$20,060 ($19,881, $28,333 ($27,033, $38,981 ($24,078, $21,073 ($15,293, $36,135 ($31,989,
$40,832)
$19,029 ($18,898, $19,141) $24,340 ($22,218, $26,477) $51,497 ($0, $90,388) $24,264 ($19,014, $31,370) $22,767 ($20,241, $25,762)
$15,116 ($14,938, $21,766 ($20,130, $27,514 ($16,565, $15,546 ($12,404, $28,308 ($22,935,
$18,076 ($17,981, $25,669 ($24,713, $34,106 ($24,594, $19,399 ($16,505, $27,195 ($24,735,
$27,688 ($26,270, $29,210)
$26,957 ($23,729, $30,691)
$21,987 ($20,173, $23,972)
$20,246) $29,794) $66,095) $27,847)
$15,289) $23,561) $46,908) $19,479) $35,584)
$18,168) $26,748) $46,374) $22,521) $29,582)
$25,909 ($24,812, $27,078)
576 estimates by surgical procedure were $8273 ($6907, $9770) for THR, $5311 ($3196, $7496) for MKS and $6651 ($5030, $8412) for HFR. On an adjusted basis, the difference in mean total charges for patients with overt bleeding associated with a BI z 2 compared with those without major bleeding was $7833.
Discussion Our findings suggest that major bleeding following THR, MKS, or HFR is associated with significantly longer stays in hospital and higher costs of inpatient care. On average, patients who develop postoperative major bleeding spend an additional 3.1 days in hospital (1.8 days after multivariate adjustment) compared with those without this complication; the difference in length of stay is of similar magnitude across all three surgical procedures. Total inpatient charges are $10,675 higher ($7593 after multivariate adjustment); once again, findings are consistent for THR, MKS, and HFR patients. Our results also suggest that all major bleeds are clinically and economically relevant, even those that do not result in death, bleeding into a critical organ, or the need for reoperation; patients with overt bleeds with a BI z 2 spent 2.8 additional days in hospital (1.9 after multivariate adjustment) and incurred $9869 ($7833) in additional inpatient charges. While the clinical relevance of the BI was labeled uncertain in a meta-analysis of data from the fondaparinux clinical trials program [9], it also has been suggested that the BI may be a ba valid and sensitive surrogate marker of the potential to cause major clinically important bleedingQ [18]. Examples of the use of the BI as a study measure include the Coronary Revascularization Using Integrilin and Single Bolus Enoxaparin Study (where it was the primary endpoint) [19] and a 5-year randomized controlled trial of endometrial coagulation vs. resection in the treatment of heavy dysfunctional bleeding [20]. Strengths of our study include the use of a large, multi-hospital database containing clinical and economic information on over 23,000 major orthopedic surgery patients. Our sample size is thus substantially larger than those of earlier studies [21—28]. Our study is also the first to report costs of bleeding-related care based on direct examination of clinical and economic outcomes for a large sample of US hospital admissions. The degree of clinical detail in the study database allowed us to utilize a definition of major bleeding consistent with that employed in
M. Vera-Llonch et al. the recent Phase III clinical trials program of fondaparinux and permitted adjustment for differences in demographic and clinical characteristics between patients who did and did not develop major bleeding. The incidence of major bleeding in our study (2.6%) also is consistent with that reported in the published literature [29—36]. A few limitations of our study should be noted. For one, as the collection of beventQ and btransfusionQ data is optional in MQ-Pro, a potential for selection bias exists, as only hospitals that collect this information were included in our analyses. For this reason, we compared severity of illness on admission (using ASG groups), mean length of stay in hospital, and mean inpatient charges for admissions to hospitals that collected this information vs. admissions to those that did not. The two groups did not differ with respect to these characteristics (data not presented). We therefore do not believe that the associated bias is substantial. Other limitations of our study include the potential for errors in the ICD-9-CM and proprietary coding methodologies. For example, the accuracy of the assignment of day of service to bleeding events might have influenced our estimates of the risk of major bleeding. As noted above, however, our estimates of the overall incidence of this complication—as well as that of each bleeding subtype—are consistent with published estimates. We cannot rule out the possibility that the effects of systematic differences between patients who experienced major bleeding and those who did not were not fully accounted for in our multivariate analyses and, therefore, that our estimates of the costs of major bleeding are biased as a result of confounding. We were unable to examine the impact of major bleeding on health-care costs (vs. charges), as this information was unavailable in the study database. Finally, the study database did not contain any information on medication use, which precluded examination of outcomes in relation to anti-thrombotic agent received. Despite these limitations, we believe that our findings are important, as they suggest that bleeding following major orthopedic surgery may significantly increase length of stay and total hospital charges.
Acknowledgements Funding for this research was provided by Aventis, part of the Sanofi-Aventis Group.
Clinical and economic consequences of bleeding following major orthopedic surgery
References [1] Clagett GP, Anderson FA, Heit J, et al. Prevention of venous thromboembolism. Chest 1992;102(Supplement 4): 391S – 407S. [2] Clagget GP, Anderson FA, Heit J, et al. Prevention of venous thromboembolism. Chest 1995;108(Supplement 4): 312S – 34S. [3] National Institutes of Health Consensus Development Conference. Prevention of venous thromboembolism. NIH consensus development. JAMA 1986;256:744 – 9. [4] Haake DA, Berkman SA. Venous disease after hip surgery. Risk factors, prophylaxis, and diagnosis. Clin Orthop 1989; 242:212 – 31. [5] Geerts WH, Heit JH, Clagget P, et al. Prevention of venous thromboembolism. Chest 2001;119:132S – 75S. [6] Mohr DN, Silverstein MD, Murtaugh PA, et al. Prophylactic agents for venous thrombosis in elective hip surgery metaanalysis of studies using venographic assessment. Arch Intern Med 1993;153:2221 – 8. [7] Imperiale TF, Speroff T. A meta-analysis of methods to prevent VTE following total hip replacement. JAMA 1994; 271:1780 – 5. [8] Hull R, Raskob G, Pineo G, et al. A comparison of subcutaneous low-molecular-weight heparin with warfarin sodium for prophylaxis against deep-vein thrombosis after hip or knee implantation. N Engl J Med 1993; 329:1370 – 6. [9] Turpie AG, Bauer KA, Eriksson BI, et al. Fondaparinux vs. enoxaparin for the prevention of venous thromboembolism in major orthopedic surgery: a meta-analysis of 4 randomized double-blind studies. Arch Intern Med 2002;162(16): 1806 – 8. [10] Lassen MR, Bauer KA, Eriksson BI, et al. Post-operative fondaparinux vs. preoperative enoxaparin for prevention of venous thromboembolism in elective hip-replacement surgery: a randomized double-blind comparison. Lancet 2002; 359:1715 – 20. [11] Turpie AGG, Bauer KA, Eriksson BI, et al. Post-operative fondaparinux vs. post-operative enoxaparin for prevention of venous thromboembolism after elective hip-replacement surgery: a randomized double-blind trial. Lancet 2002; 1721 – 6. [12] Turpie AGG, Gallus AS, Hoek JA, et al. A synthetic pentasaccharide for the prevention of deep-vein thrombosis after total hip replacement. N Engl J Med 2001;344(9): 619 – 25. [13] Atlas Scoring: A Technical White Paper. Marlborough, MA: MediQual Systems (Cardinal Information Corporation), February 1996. [14] Steen PM. Approaches to predictive modeling. Ann Thorac Surg 1994;58:1836 – 40. [15] Iezzoni LI, Ash AS, Coffman GA, Moskowitz MA. Predicting in-hospital mortality: a comparison of severity measurement approaches. Med Care 1992;30:347 – 59. [16] Iezzoni LI. The risks of risk adjustment. JAMA 1997;278: 1600 – 7. [17] Duan N. Smearing estimate: a nonparametric retransformation method. J Am Stat Assoc 1983;78(383):605 – 10. [18] Raskob GE, Hirsch J. Controversies in timing of the first dose of anticoagulant prophylaxis against venous thromboembolism after major orthopedic surgery. Chest 2003;124: 379S – 85S. [19] Bhatt DL, Lee BI, Casterella PJ, et al. Safety of concomitant therapy with eptifibatide and enoxaparin in patients undergoing percutaneous coronary intervention:
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
577
eesults of the coronary revascularization using integrilin and single bolus enoxaparin study. J Am Coll Cardiol 2003; 41(1):988 – 92. Boujida VH, Philipsen T, Pelle J. Five-year follow-up of endometrial ablation: endometrial coagulation vs. endometrial resection. Obstet Gynecol 2002;99(6):982 – 8. Mandani MM, Weingarten CM, Stevenson JF. Thromboembolic prophylaxis in moderate-risk patients undergoing elective abdominal surgery: decision and cost-effective analysis. Pharmacotherapy 1996;16:1111 – 27. Shorr AF, Ramage AS. Enoxaparin for thromboprophylaxis after major trauma: potential cost implications. Crit Care Med 2001;29:1659 – 65. Hull RD, Pineo GE, Raskob GE, et al. The economic impact of treating deep vein thrombosis with low molecular weight heparin: outcome of therapy and health economics aspects. Haemostasis 1998;28(Suppl 3):8 – 16. Hull RD, Raskob GE, Pineo GF, et al. Subcutaneous lowmolecular-weight heparin vs. warfarin for prophylaxis of deep vein thrombosis after hip or knee implantation. An economic perspective. Arch Intern Med 1997;157:298 – 303. Marchetti M, Pistorio A, Barone M, Serafini S, Barosi G. Lowmolecular-weight heparin vs. warfarin for secondary prophylaxis of venous thromboembolism: a cost-effectiveness analysis. Am J Med 2001;111:130 – 9. Friedman RJ, Dunsworth GA. Cost analyses of extended prophylaxis with enoxaparin after arthroplasty. Clin Orthop 2000;370:171 – 82. Paiement GD, Wessinger SJ, Harris WH. Cost-effectiveness of prophylaxis in total hip replacement. Am J Surg 1991; 161:519 – 24. Haawkins DW, Langley PC, Krueger KP. A pharmacoeconomic assessment of enoxaparin and warfarin as prophylaxis for deep vein thrombosis in patients undergoing knee replacement surgery. Clin Ther 1998;20:182 – 95. Eriksson BI, Wille-Jorgensen P, Kalebo P, et al. A comparison of recombinant hirudin with a low-molecular-weight-heparin to prevent thromboembolic complications after total hip replacement. N Engl J Med 1997;337:1329 – 35. Leclerc JR, Gearts WH, Desjardins L, et al. Prevention of venous thromboembolism after knee arthroplasty: a randomized, double-blind trial comparing enoxaparin with warfarin. Ann Intern Med 1996;124:616 – 28. Spiro TE, Johnson CJ, Christie MJ, et al. Efficacy and safety of enoxaparin to prevent deep venous thrombosis after hip replacement surgery. Ann Intern Med 1994;121:81 – 9. Hull R, Raskob G, Pineo G, et al. A comparison of subcutaneous low-molecular-weight heparin with warfarin sodium for prophylaxis against DVT after hip or knee implantation. N Engl J Med 1993;329:1370 – 6. Leclerc JR, Geerts WH, Desjardins L, et al. Prevention of deep vein thrombosis after major knee surgery: a randomized, double-blind trial comparing a low molecular weight heparin fragment (enoxaparin) to placebo. Thromb Haemost 1992;67(4):417 – 23. Levine MH, Hirsch J, Gent M. Prevention of deep vein thrombosis after elective hip surgery: a randomized trial comparing low molecular weight heparin with standard unfractionated heparin. Ann Intern Med 1991;114:545 – 51. Leyvraz PF, Bachmann F, Hoek J, et al. Prevention of deep vein thrombosis after hip replacement: randomized comparison between unfractionated heparin and low molecular weight heparin. BMJ 1991;303:543 – 8. Turpie AGG, Levine MN, Hirsch J, et al. A randomized controlled trial of low-molecular weight heparin (enoxaparin) to prevent deep-vein thrombosis in patients undergoing elective hip surgery. N Engl J Med 1986;31:925 – 9.
intl.elsevierhealth.com/journals/thre
REGULAR ARTICLE
Clinical and economic consequences of bleeding following major orthopedic surgery Montserrat Vera-Llonch, May Hagiwara, Gerry Oster* Policy Analysis Inc. (PAI), Four Davis Court, Brookline, MA 02445, USA Received 2 December 2004; received in revised form 14 April 2005; accepted 18 April 2005 Available online 23 May 2005
KEYWORDS Orthopedic surgery; VTE; Bleeding; Outcomes; Hemorrhage; Costs; Cost analysis
Abstract Background: Major orthopedic surgery patients who receive antithrombotics as prophylaxis against VTE may be at higher risk of bleeding. The clinical and economic consequences of this complication may be relevant to therapeutic decision-making. Objective: To assess the impact of major bleeding following major orthopedic surgery on length of stay (LOS) and inpatient charges. Methods: Using a database with information on ~750,000 admissions annually to 100+ US acute-care hospitals, we identified all patients who underwent major orthopedic surgery between January 1, 1998 and December 13, 2000. Patients were stratified according to whether or not they experienced major postoperative bleeding prior to hospital discharge, defined as (a) fatal bleeding; (b) nonfatal bleeding at critical site; (c) re-operation due to bleeding; and (d) overt bleeding with bleeding index (BI) z 2, where BI = number of blood units transfused plus pre-bleeding minus postbleeding hemoglobin (g/dL) values. LOS and total inpatient charges were compared between patients with and without major bleeding. Results: The incidence of major bleeding among 23,518 patients who underwent major orthopedic surgery was 2.6%. In multivariate analyses controlling for differences in baseline characteristics between patients with and without major bleeds, adjusted mean LOS was 1.8 days longer among those with major bleeding (95% CI: 1.5, 2.0) (6.1 days vs. 4.3 days for those without bleeds); adjusted mean total inpatient charges were $7593 higher (95% CI: $6622, $8646) ($25,669 vs. $18,076). Conclusion: Bleeding following major orthopedic surgery may increase length of stay and total hospital charges and should be an important consideration in choice of VTE prophylaxis. D 2005 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: +1 617 232 4400; fax: +1 617 232 1155. E-mail address: [email protected] (G. Oster). 0049-3848/$ - see front matter D 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.thromres.2005.04.018
570
Introduction Deep-vein thrombosis (DVT) and pulmonary embolism (PE)—together referred to as venous thromboembolism (VTE)—are important causes of morbidity and death among hospitalized patients, particularly those undergoing major surgery [1—4]. Major surgery confers an especially high risk of VTE due to postoperative immobility as well as the effects of surgical trauma on the coagulation system. Among common major surgical procedures, arthroplasty of the hip or knee and hip-fracture repair confer the highest risks of VTE, probably due to direct trauma to deep veins and the marked venous stasis associated with these procedures. Because of the difficulty of early diagnosis and the potentially serious consequences of VTE, prophylaxis has long been recommended in high-risk patients. Most prophylactic regimens today include an antithrombotic, usually a low-molecular-weight heparin (LMWH) or warfarin. While the efficacy of antithrombotic therapy is well established [5], patients who receive chemoprophylaxis may be at increased risk of postoperative bleeding, which may differ by prophylaxis received. In meta-analyses of prophylactic regimens among patients undergoing hip replacement, pooled rates of major bleeding were estimated to be 3.3% for oral anticoagulants and 5.3% for enoxaparin, although the trials that were reviewed used somewhat different definitions of this outcome [6,7]. In patients undergoing total knee replacement, rates of bleeding have been reported to be 0.9% for warfarin and 2.8% for LMWH [8]. Most recently, in a meta-analysis of data from four randomized controlled trials in patients undergoing elective hip or knee replacement or hip fracture repair who were randomized to receive fondaparinux or enoxaparin, the rate of major bleeding was reported to be 2.7% for the former and 1.7% for the latter [9]. Bleeding following major orthopedic surgery may have important clinical consequences, extending stays in hospital and increasing costs of care. We examine this issue in this study using a large US inpatient database.
Materials and methods Overview Using a large computerized inpatient database, we compared the length of stay in hospital and total inpatient charges between major orthopedic sur-
M. Vera-Llonch et al. gery patients who developed postoperative major bleeding vs. those who did not. Data were obtained from the MQ-Profile Database v3.3 (bMQ-ProQ) (Cardinal Information Corporation, Marlborough, MA), which contains clinical and administrative information on approximately 750,000 inpatient admissions annually to over 100 US acute-care hospitals. The study sample consisted of all patients in MQPro who underwent major orthopedic surgery (total hip replacement [THR], major knee surgery [MKS], or hip fracture repair [HFR]) between January 1, 1998 and December 31, 2000 (the most recent 3year period for which data were available at the time the study was conducted). Major postoperative bleeding was defined as the occurrence of (a) fatal bleeding; (b) nonfatal bleeding at a critical site; (c) re-operation due to bleeding; or (d) overt bleeding with a bleeding index (BI) z 2, where BI is calculated as the number of whole blood or packed red blood cell units transfused plus the difference between pre-bleeding and post-bleeding hemoglobin (g/dL) values. This definition of major postoperative bleeding is consistent with the one employed in a recent Phase III clinical trials program of fondaparinux [10—12]. We then compared length of stay in hospital and total inpatient charges between patients who experienced major postoperative bleeding vs. those who did not using both univariate and multivariate techniques. Findings also were examined for each of the constituent measures of the composite endpoint (i.e., fatal bleeding, non-fatal bleeding at a critical site, re-operation due to bleeding, and overt bleeding with BI z 2).
Data source MQ-Pro is maintained by Cardinal Information Corporation (CIC), which sells and distributes software (Atlask) to US acute-care hospitals for the collection of detailed clinical and administrative data. Hospitals use this information for a variety of quality-improvement activities; data also are submitted by each hospital to CIC for use in proprietary databases (including MQ-Pro), which are used primarily by hospitals for benchmarking studies. MQ-Pro is the largest of these databases and contains clinical and administrative data on approximately 750,000 inpatient admissions annually to over 100 US acute-care hospitals. Some data elements in MQ-Pro (e.g., patient demographics, diagnoses, admission and discharge dates, charges) are extracted from hospital information systems. Other information is collected manually via chart review by trained abstractors
Clinical and economic consequences of bleeding following major orthopedic surgery using a standardized glossary. Hospitals participating in MQ-Pro must collect data on all patients admitted to their facility, ensuring complete data capture and minimizing potential selection or reporting biases. While collection of data for all patients is mandatory, some data elements are optional. For example, facilities may choose to collect detailed clinical information for the first 5 days of each admission or for the first 2 days only. Similarly, some facilities choose to collect detailed information on transfusions while others do not. For all admissions in the database, MQ-Pro includes information on patient demographics (e.g., age, gender), admission source, medical history, all documented diagnosis codes (in ICD-9CM format), length of stay, billed total and ancillary hospital charges, and discharge disposition. The database also includes a listing of all procedures performed during the hospital stay (coded in ICD-9CM format) along with the day the procedure began. Information also is collected for each admission on over 400 key clinical findings (bKCFsQ). KCFs are collected from admission through the second or fifth day in hospital, depending upon the collection option chosen by the facility. All KCFs are dated and include observations obtained through physician assessments, procedures, pathology, imaging and electrocardiograms (EKGs). KCFs also are used to assess clinical severity on admission; admission severity groups (ASGs) (range, 0—4) are derived based on the probability of in-hospital mortality, which is calculated using disease-specific logistic regression models. The predictive capabilities of these models have been shown to be comparable to those of other severity adjustment methodologies (e.g., APACHE II, Disease Staging) [13—16]. In addition to KCFs, the database includes information on the values of laboratory tests performed during hospitalization, including hematology, coagulation, chemistry, urine, blood gases, gastrointestinal and miscellaneous tests. This information is recorded on a daily basis. If multiple values are available for any given day, the most extreme value (i.e., highest, lowest, or furthest from some benchmark) is typically recorded. Daily values for selected vital signs also are recorded. A number of optional data elements also are included in MQ-Pro. Some facilities collect information on unplanned clinical events during hospitalization (beventsQ), including (but not limited to) barrest,Q bbleeding,Q and bcardiac events.Q Events are coded using a proprietary system. Information on each such event includes its nature (e.g., transfusion, return to operating room [OR], transfer to intensive care unit [ICU]) and the day it
571
occurred. Facilities also may choose to collect detailed information on transfusions. For all hospitals that collect such data, information is available on the type of transfusion (e.g., whole blood, packed red cells, plasma, platelets), the day of transfusion and the number of units administered. Hospitals also may collect information on pre- and post-transfusion values for selected coagulation parameters (e.g., hemoglobin). Other optional data elements include information on time spent in special-care units and whether patients received selected procedures (e.g., mechanical ventilation, thrombolytic therapy).
Patient selection All patients in MQ-Pro who had undergone THR (ICD9-CM procedure codes 81.51, 81.53), MKS (81.54, 81.55), or HFR (79.15, 79.35, 81.4, 81.52) between January 1, 1998 and December 31, 2000 were identified. As we were principally interested in the clinical and economic impact of bleeding associated with VTE prophylaxis, patients undergoing treatment for VTE (based on secondary diagnosis codes of DVT [ICD-9-CM diagnosis codes 451.1, 451.2, and 453.8] and/or PE [415.1]) were excluded from the study sample, as were those with multiple trauma (ICD-9CM 800.XX-804.XX [skull fracture], 805.XX-809.XX [neck and trunk fracture], 810.XX-819.XX [upper limb fracture], 850.XX-854.XX [intracranial injury], 860.XX-869.XX [internal injury of thorax, abdomen, and pelvis], and 828.XX [multiple fractures involving both lower limbs, lower with upper limb, and lower limb(s) with rib(s) and sternum]). Patients with evidence of major bleeding prior to surgery also were excluded, as were those with a transfer KCF of bleeding and those aged 18 years or younger. All remaining patients in the sample were then stratified according to whether or not they experienced major bleeding during the postoperative period, which was defined as beginning on the day of the surgical procedure (inclusive of that date) and ending on the day of hospital discharge or death. Consistent with the definition used in the fondaparinux Phase III clinical trials programs, major bleeding was defined as follows: Fatal bleeding: A KCF of bbleedingQ or an beventQ due to bleeding during the postoperative period with or without a discharge ICD-9-CM diagnosis code of bleeding, and evidence of death at discharge, without evidence of stroke (ICD-9-CM 436) or myocardial infarction (410.X). Nonfatal bleeding at a critical site: A KCF of bbleedingQ with KCF body-site modifiers suggesting bleeding at a critical site (i.e., intracranial [KCF
572 modifier, brain], retroperitoneal [abdomen], intraocular [eye], pericardium [heart], or adrenal [endocrine]), or a discharge diagnosis code suggestive of bleeding at a critical site (ICD-9-CM 432.X [hemorrhage, intracranial], 360.43 [hemorrhage, eye], 364.41 [hemorrhage, anterior chamber], 362.81 [hemorrhage, fundus], 336.1 [hemorrhage, spinal cord], 459.0 [hemorrhage, retroperitoneal], 423.0 [hemopericardium], or 255.4 [hemorrhage, adrenal capsule]) and discharge status alive. Re-operation due to bleeding: An beventQ labeled breturn to ORQ due to bleeding during the postoperative period, or one or more of the following procedure codes (in ICD-9-CM format) after the date of the initial surgical procedure: 80.85 (repeat operation for hemorrhage control: hip replacement and fracture repair), 80.15 (repeat operation for drainage of hematoma; hip replacement and fracture repair), 80.86 (repeat operation, for hemorrhage control, knee surgery), or 80.16 (repeat operation, drainage of hematoma, knee surgery). Overt bleeding associated with BI z 2: A KCF of bbleeding,Q or an beventQ due to bleeding, with or without a discharge diagnosis code of bleeding, and a bleeding index (BI) z 2, where BI is calculated as the number of whole blood or packed red blood cell units transfused plus the difference between the pre-bleeding and post-bleeding hemoglobin (g/dL) values during the postoperative period. Evidence of transfusion, type of blood products (whole blood and packed red blood cells) and number of units transfused were identified based on information on interventions in the study database. Pre- and post-bleeding hemoglobin values were ascertained from daily hemoglobin laboratory values dated immediately prior and subsequent to the date of the bleeding event. The criteria that were used to identify patients with major bleeding were implemented in order of presentation above and were deemed to be mutually exclusive (e.g., a patient who underwent reoperation due to bleeding who also met the criteria for a fatal bleed was included in the latter but not the former category).
Study measures The measures of interest were total length of stay in hospital and total inpatient charges.
Analyses Demographic and clinical characteristics of study subjects were examined, including age, gender, ASG and the presence of selected comorbidities
M. Vera-Llonch et al. (coagulation disorders [ICD-9-CM 286.0, 286.1, 286.2, 286.4, 286.5, 286.7], heart disease [393398, 401—405, 410—414, 415—417, 420—429], and hepatic and/or renal disorders [570—573, 580— 599]). Differences between patients with and without major bleeding were tested using standard univariate techniques (i.e., t-tests for continuous measures, contingency tables [chi-square and Fisher’s exact tests, as appropriate] for categorical measures). An alpha level of 0.05 was used to assess statistical significance. Differences between THR, MKS, and HFR patients were not tested for statistical significance. Length of stay in hospital and total inpatient charges were compared between patients with and without major bleeding on both a univariate (i.e., unadjusted) basis, as well as controlling for differences in baseline demographic and clinical characteristics using multivariate least-squares regression. The dependent variable in these regression models was either length of stay in hospital or total inpatient charges, as appropriate. The dependent variable was log-transformed prior to analysis to correct for skewness. Explanatory (i.e., independent) variables included age, gender, ASG and the presence of coagulation disorders, heart disease and hepatic and/or renal disorders. Adjusted least-squares means were generated for each group and were retransformed to their natural values using Duan’s smearing estimator [17]. Ninety-five percent confidence intervals were calculated for both unadjusted and adjusted estimates of mean length of stay in hospital and mean inpatient charges. Findings also were examined for each of the constituent categories of the composite measure (i.e., fatal bleeding, nonfatal bleeding at a critical site, re-operation due to bleeding, and overt bleeding associated with a BI z 2). All analyses were undertaken using PC-SAS for Windows (version 8.0).
Results A total of 25,132 patients were identified in MQ-Pro database who underwent THR, MKS, or HFR between January 1, 1998 and December 31, 2000. A total of 262 patients were excluded from the sample who had secondary diagnoses of VTE (DVT and/or PE). We also excluded patients for whom a date of service for the principal procedure could not be determined (n = 36), those with evidence of bleeding prior to hospital admission (n = 1) or the date of their surgical procedure (n = 224), those with multiple trauma (n = 858) and those aged 18 years or younger (n = 230). Finally, three patients
Clinical and economic consequences of bleeding following major orthopedic surgery Table 1
Characteristics of study subjects, by surgical procedure
Parameter
Surgical procedure Total hip replacement (n = 5940)
Age (mean F S.D.) Male gender (n, %) Comorbid conditions (n, %) Coagulation disorder Heart disease Hepatic and/or renal disorders Admission severity group (ASG) (n, %) 0 1 2 3 4
Major knee surgery (n = 11,030)
67.8 (13.1) 2432 (40.9) — 4 (0.1) 3495 (58.8) 442 (7.4) 329 5295 291 25 0
68.3 (10.4) 3912 (35.5) — 10 (0.1) 7107 (64.4) 596 (5.4)
(5.5) (89.1) (4.9) (0.4) (0.0)
were excluded due to unknown gender. The resulting study sample therefore consisted of 23,518 patients (THR, 5940; MKS, 11,030; HFR, 6548). Selected demographic and clinical characteristics of study subjects are presented in Table 1. Mean (F S.D.) age across all surgical procedures was 70.3 (F 13.8) years; HFR patients were older than those undergoing THR or MKS. Slightly less than two-thirds of study subjects were women; the proportion was highest among those undergoing HFR. The overall prevalence of coagulation disorders was 0.1%; heart disease, 63%; and hepatic and/or renal disorders, 9.6%. Hepatic and/or renal disorders were more prevalent among patients undergoing HFR, probably reflective of their older age. The majority of patients were in ASG 1, with the exception of those undergoing HFR, who were most likely to be in ASG 2. The overall incidence of major bleeding was 2.6% (fatal bleeding, 0.1%; non-fatal bleeding at a critical site, 0.2%; re-operation due to bleeding, 0.7%; and overt bleeding associated with a BI z 2, 1.7%) (Table 2). Rates of major bleeding among THR, MKS, and HFR patients were 4.9%, 1.3% and
Table 2
573
98 10,882 50 0 0
(0.9) (98.7) (0.5) (0.0) (0.0)
Hip fracture repair (n = 6548)
Total (n = 23,518)
76.0 (17.3) 1926 (29.4) — 12 (0.2) 4211 (64.3) 1222 (18.7)
70.3 (13.8) 8270 (35.2) — 26 (0.1) 14,813 (63.0) 2260 (9.6)
399 1710 3949 489 1
826 17,887 4290 514 1
(6.1) (26.1) (60.3) (7.5) (0.0)
(3.5) (76.1) (18.2) (2.2) (0.0)
2.8%, respectively. The incidence of overt bleeding associated with a BI z 2 was highest among THR patients (3.8%) and lowest among those undergoing MKS (0.5%). The rate of fatal bleeding was lowest among MKS patients. Differences in demographic and clinical characteristics between patients with and without major bleeding are reported in Table 3. Patients with major bleeding were more likely to be men ( p b 0.0001), more likely to have renal and/or hepatic disorders ( p b 0.0001), and more likely to be sicker on admission than those without major bleeding ( p b 0.0001). Findings with respect to age, however, were inconsistent across the three surgical procedures. THR and MKS patients who experienced major bleeding were significantly older than those who did not ( p b 0.0001 and p = 0.02, respectively), while HFR patients who experienced major bleeding were younger ( p = 0.025). Mean length of stay in hospital was 3.1 days (95% CI: 2.9, 3.4) longer for patients with major bleeding (7.9 days vs. 4.8 days for those without bleeding). The difference in length of stay was of similar magnitude across all three surgical procedures
Incidence of major bleeding among subjects undergoing major orthopedic surgery, by surgical procedure
Parameter
No major bleeding (n, %) Any major bleeding (n, %) Fatal bleeding (n, %) Nonfatal bleeding at critical site (n, %) Bleeding associated with re-operation at the operative site (n, %) Overt bleeding with BI z 2 (n, %) BI = bleeding index.
Surgical procedure Total hip replacement (n = 5940)
Major knee surgery (n = 11,030)
Hip fracture repair (n = 6548)
Total (n = 23,518)
5647 293 6 10 53
10,890 140 3 11 73
6367 181 12 18 34
22,904 614 21 39 160
(95.1) (4.9) (0.1) (0.2) (0.9)
224 (3.8)
(98.7) (1.3) (0.03) (0.1) (0.7)
53 (0.5)
(97.2) (2.8) (0.2) (0.3) (0.5)
117 (1.8)
(97.4) (2.6) (0.1) (0.2) (0.7)
394 (1.7)
574
Table 3
Characteristics of subjects with and without major bleeding, by surgical procedure
Parameter
Surgical procedure Total hip replacement
Age (mean F S.D.) Male gender (n, %) Comorbid conditions (n, %) Coagulation disorders Heart disease Hepatic and/or renal disorders Admission severity group (ASG) 0 1 2 3 4
Major knee surgery
No major bleeding (n = 6647)
Major bleeding (n = 293)
p-value
67.7 (13.0) 2303 (40.8)
70.8 (13.3) 129 (44.0)
b0.0001 0.271
—
—
3307 (68.7) 384 (6.8)
188 (64.2) 68 (19.8)
0.057 b0.0001
316 (6.7) 6069 (89.7) 262 (4.5) 20 (0.4) 0 (0.0)
13 (4.4) 236 (80.6) 39 (13.3) 6 (1.7) 0 (0.0)
b0.0001
4 (0.1)
No major bleeding (n = 10,890)
Hip fracture repair
Total
Major bleeding (n = 140)
p-value
No major bleeding (n = 6367)
Major bleeding (n = 181)
p-value
70.4 (9.9) 78 (66.7)
0.021 b0.0001
76.1 (17.2) 1862 (29.1)
72.8 (19.7) 74 (40.9)
0.025 b0.001
2 (1.4)
b0.001
12 (0.2)
—
—
7020 (64.5) 670 (6.2)
87 (62.1) 26 (18.7)
0.569 b0.0001
4091 (64.3) 1168(18.3)
120 (66.3) 64 (29.8)
0.571 b0.0001
96 (0.9) 10,762 (98.7) 42 (0.4) 0 (0.0) 0 (0.0)
2 (1.4) 130 (92.9) 8 (6.7) 0 (0.0) 0 (0.0)
b0.0001
381(7.0) 1666 (26.2) 3844 (60.4) 476 (7.5) 1 (0.0)
18 (9.9) 46 (24.9) 106 (68.0) 13 (7.2) 0 (0.0)
0.303
68.3 (10.4) 3834 (36.2)
8 (0.1)
No major bleeding (n = 22,904)
Major bleeding (n = 614)
p-value
70.3 (13.7) 7989(34.9)
71.3 (14.8) 281 (46.8)
0.115 b0.0001
24 (0.1)
2 (0.3)
0.147
14,418 (62.9) 2122 (9.3)
396 (64.3) 138 (22.5)
0.484 b0.0001
793 (3.5) 17,476 (76.3) 4138 (18.1) 496 (2.2) 1 (0.0)
33 (6.4) 411 (66.9) 162 (24.8) 18 (2.9) 0 (0.0)
b0.0001
M. Vera-Llonch et al.
Clinical and economic consequences of bleeding following major orthopedic surgery
575
Table 4 Adjusted length of stay (mean, 95% CI) in hospital among subjects with or without major bleeding, surgical procedure and type of bleeding Parameter
Surgical procedure
No major bleeding (n = 22,904) Any major bleeding (n = 614) Fatal bleeding (n = 21) Nonfatal bleeding at critical site (n = 39) Bleeding associated with re-operation at the operative site (n = 160) Overt bleeding with BI z 2 (n = 394)
Total hip replacement (n = 5940)
Major knee surgery (n = 11,030)
Hip fracture repair (n = 6548)
Total (n = 23,518)
4.2 5.7 4.7 4.1 8.2
3.9 5.6 5.4 5.4 5.6
5.4 7.3 8.1 5.5 9.4
4.3 6.1 6.3 4.9 7.2
(4.1, (5.3, (1.4, (2.8, (6.6,
4.2) 6.1) 14.3) 5.8) 10.0)
5.7 (5.3, 6.1)
(3.8, (5.0, (0.0, (3.7, (4.8,
3.9) 6.2) 17.5) 8.3) 6.6)
6.1 (5.2, 7.1)
(5.3, (6.7, (5.4, (4.4, (7.7,
5.4) 7.9) 12.9) 6.9) 11.6)
7.1 (6.5, 7.8)
(4.3, (5.8, (4.0, (4.1, (6.4,
4.3) 6.3) 9.5) 5.8) .8.0)
6.2 (5.9, 6.5)
BI = bleeding index.
(THR, 3.0 days [95% CI: 2.6, 3.3]; MKS, 3.2 days [2.9, 3.5]; HFR, 2.9 days [2.3, 3.5]). Length of stay also was consistently longer for patients who experienced any of the individual outcomes that were part of the composite endpoint. Patients with overt bleeding associated with a BI z 2, for example, had hospital stays that were 2.8 days longer on average than those of patients without major bleeding. In multivariate analyses controlling for differences in selected covariates (i.e., age, gender, ASG, and presence of coagulation disorders, heart disease and hepatic and/or renal disease) between patients with and without major bleeding, the covariate-adjusted difference in hospital length of stay was estimated to be 1.8 days (95% CI: 1.5, 2.0) (6.1 days vs. 4.3 days, respectively) (Table 4). Similar differences were noted for each of the surgical procedures (THR, 1.5 days [95% CI: 1.2, 2.0]; MKS, 1.7 days [1.2, 2.4]; HFR, 1.9 days [1.4, 2.5]). On an adjusted basis, patients with overt bleeding associated with a BI z 2 were found to stay
in hospital an average of 1.9 additional days compared with those without major bleeding. Mean total inpatient charges were $10,675 higher among patients with major bleeding (95% CI: $9853, $11,599) ($30,349 vs. $19,674 for those without bleeding). The difference was of similar magnitude across all three surgical procedures (THR, $10,915 [95% CI: $9817, $13,013]; MKS, $9852 [$8419, $11,386]; HFR, $10,200 [($8300, $13,100]). The difference in mean total charges for patients with overt bleeding associated with a BI z 2 compared with those without major bleeding was $9869. In multivariate analyses controlling for differences in covariates (i.e., age, gender, ASG and presence of coagulation disorders, heart disease and hepatic and/or renal disease) between patients with and without major bleeding, the adjusted difference in total inpatient charges was estimated to be $7593 (95% CI: $6622, $8646) ($25,669 vs. $18,076) (Table 5). Corresponding
Table 5 Adjusted total inpatient charges (mean, 95% CI) among subjects with and without major bleeding, by surgical procedure and type of bleeding Parameter
No major bleeding (n = 22,904) Any major bleeding (n = 614) Fatal bleeding (n = 21) Nonfatal bleeding at critical site (n = 39) Bleeding associated with re-operation at the operative site (n = 160) Overt bleeding with BI z 2 (n = 394) BI = bleeding index.
Surgical procedure Total hip replacement (n = 5940)
Major knee surgery (n = 11,030)
Hip fracture repair (n = 6548)
Total (n = 23,518)
$20,060 ($19,881, $28,333 ($27,033, $38,981 ($24,078, $21,073 ($15,293, $36,135 ($31,989,
$40,832)
$19,029 ($18,898, $19,141) $24,340 ($22,218, $26,477) $51,497 ($0, $90,388) $24,264 ($19,014, $31,370) $22,767 ($20,241, $25,762)
$15,116 ($14,938, $21,766 ($20,130, $27,514 ($16,565, $15,546 ($12,404, $28,308 ($22,935,
$18,076 ($17,981, $25,669 ($24,713, $34,106 ($24,594, $19,399 ($16,505, $27,195 ($24,735,
$27,688 ($26,270, $29,210)
$26,957 ($23,729, $30,691)
$21,987 ($20,173, $23,972)
$20,246) $29,794) $66,095) $27,847)
$15,289) $23,561) $46,908) $19,479) $35,584)
$18,168) $26,748) $46,374) $22,521) $29,582)
$25,909 ($24,812, $27,078)
576 estimates by surgical procedure were $8273 ($6907, $9770) for THR, $5311 ($3196, $7496) for MKS and $6651 ($5030, $8412) for HFR. On an adjusted basis, the difference in mean total charges for patients with overt bleeding associated with a BI z 2 compared with those without major bleeding was $7833.
Discussion Our findings suggest that major bleeding following THR, MKS, or HFR is associated with significantly longer stays in hospital and higher costs of inpatient care. On average, patients who develop postoperative major bleeding spend an additional 3.1 days in hospital (1.8 days after multivariate adjustment) compared with those without this complication; the difference in length of stay is of similar magnitude across all three surgical procedures. Total inpatient charges are $10,675 higher ($7593 after multivariate adjustment); once again, findings are consistent for THR, MKS, and HFR patients. Our results also suggest that all major bleeds are clinically and economically relevant, even those that do not result in death, bleeding into a critical organ, or the need for reoperation; patients with overt bleeds with a BI z 2 spent 2.8 additional days in hospital (1.9 after multivariate adjustment) and incurred $9869 ($7833) in additional inpatient charges. While the clinical relevance of the BI was labeled uncertain in a meta-analysis of data from the fondaparinux clinical trials program [9], it also has been suggested that the BI may be a ba valid and sensitive surrogate marker of the potential to cause major clinically important bleedingQ [18]. Examples of the use of the BI as a study measure include the Coronary Revascularization Using Integrilin and Single Bolus Enoxaparin Study (where it was the primary endpoint) [19] and a 5-year randomized controlled trial of endometrial coagulation vs. resection in the treatment of heavy dysfunctional bleeding [20]. Strengths of our study include the use of a large, multi-hospital database containing clinical and economic information on over 23,000 major orthopedic surgery patients. Our sample size is thus substantially larger than those of earlier studies [21—28]. Our study is also the first to report costs of bleeding-related care based on direct examination of clinical and economic outcomes for a large sample of US hospital admissions. The degree of clinical detail in the study database allowed us to utilize a definition of major bleeding consistent with that employed in
M. Vera-Llonch et al. the recent Phase III clinical trials program of fondaparinux and permitted adjustment for differences in demographic and clinical characteristics between patients who did and did not develop major bleeding. The incidence of major bleeding in our study (2.6%) also is consistent with that reported in the published literature [29—36]. A few limitations of our study should be noted. For one, as the collection of beventQ and btransfusionQ data is optional in MQ-Pro, a potential for selection bias exists, as only hospitals that collect this information were included in our analyses. For this reason, we compared severity of illness on admission (using ASG groups), mean length of stay in hospital, and mean inpatient charges for admissions to hospitals that collected this information vs. admissions to those that did not. The two groups did not differ with respect to these characteristics (data not presented). We therefore do not believe that the associated bias is substantial. Other limitations of our study include the potential for errors in the ICD-9-CM and proprietary coding methodologies. For example, the accuracy of the assignment of day of service to bleeding events might have influenced our estimates of the risk of major bleeding. As noted above, however, our estimates of the overall incidence of this complication—as well as that of each bleeding subtype—are consistent with published estimates. We cannot rule out the possibility that the effects of systematic differences between patients who experienced major bleeding and those who did not were not fully accounted for in our multivariate analyses and, therefore, that our estimates of the costs of major bleeding are biased as a result of confounding. We were unable to examine the impact of major bleeding on health-care costs (vs. charges), as this information was unavailable in the study database. Finally, the study database did not contain any information on medication use, which precluded examination of outcomes in relation to anti-thrombotic agent received. Despite these limitations, we believe that our findings are important, as they suggest that bleeding following major orthopedic surgery may significantly increase length of stay and total hospital charges.
Acknowledgements Funding for this research was provided by Aventis, part of the Sanofi-Aventis Group.
Clinical and economic consequences of bleeding following major orthopedic surgery
References [1] Clagett GP, Anderson FA, Heit J, et al. Prevention of venous thromboembolism. Chest 1992;102(Supplement 4): 391S – 407S. [2] Clagget GP, Anderson FA, Heit J, et al. Prevention of venous thromboembolism. Chest 1995;108(Supplement 4): 312S – 34S. [3] National Institutes of Health Consensus Development Conference. Prevention of venous thromboembolism. NIH consensus development. JAMA 1986;256:744 – 9. [4] Haake DA, Berkman SA. Venous disease after hip surgery. Risk factors, prophylaxis, and diagnosis. Clin Orthop 1989; 242:212 – 31. [5] Geerts WH, Heit JH, Clagget P, et al. Prevention of venous thromboembolism. Chest 2001;119:132S – 75S. [6] Mohr DN, Silverstein MD, Murtaugh PA, et al. Prophylactic agents for venous thrombosis in elective hip surgery metaanalysis of studies using venographic assessment. Arch Intern Med 1993;153:2221 – 8. [7] Imperiale TF, Speroff T. A meta-analysis of methods to prevent VTE following total hip replacement. JAMA 1994; 271:1780 – 5. [8] Hull R, Raskob G, Pineo G, et al. A comparison of subcutaneous low-molecular-weight heparin with warfarin sodium for prophylaxis against deep-vein thrombosis after hip or knee implantation. N Engl J Med 1993; 329:1370 – 6. [9] Turpie AG, Bauer KA, Eriksson BI, et al. Fondaparinux vs. enoxaparin for the prevention of venous thromboembolism in major orthopedic surgery: a meta-analysis of 4 randomized double-blind studies. Arch Intern Med 2002;162(16): 1806 – 8. [10] Lassen MR, Bauer KA, Eriksson BI, et al. Post-operative fondaparinux vs. preoperative enoxaparin for prevention of venous thromboembolism in elective hip-replacement surgery: a randomized double-blind comparison. Lancet 2002; 359:1715 – 20. [11] Turpie AGG, Bauer KA, Eriksson BI, et al. Post-operative fondaparinux vs. post-operative enoxaparin for prevention of venous thromboembolism after elective hip-replacement surgery: a randomized double-blind trial. Lancet 2002; 1721 – 6. [12] Turpie AGG, Gallus AS, Hoek JA, et al. A synthetic pentasaccharide for the prevention of deep-vein thrombosis after total hip replacement. N Engl J Med 2001;344(9): 619 – 25. [13] Atlas Scoring: A Technical White Paper. Marlborough, MA: MediQual Systems (Cardinal Information Corporation), February 1996. [14] Steen PM. Approaches to predictive modeling. Ann Thorac Surg 1994;58:1836 – 40. [15] Iezzoni LI, Ash AS, Coffman GA, Moskowitz MA. Predicting in-hospital mortality: a comparison of severity measurement approaches. Med Care 1992;30:347 – 59. [16] Iezzoni LI. The risks of risk adjustment. JAMA 1997;278: 1600 – 7. [17] Duan N. Smearing estimate: a nonparametric retransformation method. J Am Stat Assoc 1983;78(383):605 – 10. [18] Raskob GE, Hirsch J. Controversies in timing of the first dose of anticoagulant prophylaxis against venous thromboembolism after major orthopedic surgery. Chest 2003;124: 379S – 85S. [19] Bhatt DL, Lee BI, Casterella PJ, et al. Safety of concomitant therapy with eptifibatide and enoxaparin in patients undergoing percutaneous coronary intervention:
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
577
eesults of the coronary revascularization using integrilin and single bolus enoxaparin study. J Am Coll Cardiol 2003; 41(1):988 – 92. Boujida VH, Philipsen T, Pelle J. Five-year follow-up of endometrial ablation: endometrial coagulation vs. endometrial resection. Obstet Gynecol 2002;99(6):982 – 8. Mandani MM, Weingarten CM, Stevenson JF. Thromboembolic prophylaxis in moderate-risk patients undergoing elective abdominal surgery: decision and cost-effective analysis. Pharmacotherapy 1996;16:1111 – 27. Shorr AF, Ramage AS. Enoxaparin for thromboprophylaxis after major trauma: potential cost implications. Crit Care Med 2001;29:1659 – 65. Hull RD, Pineo GE, Raskob GE, et al. The economic impact of treating deep vein thrombosis with low molecular weight heparin: outcome of therapy and health economics aspects. Haemostasis 1998;28(Suppl 3):8 – 16. Hull RD, Raskob GE, Pineo GF, et al. Subcutaneous lowmolecular-weight heparin vs. warfarin for prophylaxis of deep vein thrombosis after hip or knee implantation. An economic perspective. Arch Intern Med 1997;157:298 – 303. Marchetti M, Pistorio A, Barone M, Serafini S, Barosi G. Lowmolecular-weight heparin vs. warfarin for secondary prophylaxis of venous thromboembolism: a cost-effectiveness analysis. Am J Med 2001;111:130 – 9. Friedman RJ, Dunsworth GA. Cost analyses of extended prophylaxis with enoxaparin after arthroplasty. Clin Orthop 2000;370:171 – 82. Paiement GD, Wessinger SJ, Harris WH. Cost-effectiveness of prophylaxis in total hip replacement. Am J Surg 1991; 161:519 – 24. Haawkins DW, Langley PC, Krueger KP. A pharmacoeconomic assessment of enoxaparin and warfarin as prophylaxis for deep vein thrombosis in patients undergoing knee replacement surgery. Clin Ther 1998;20:182 – 95. Eriksson BI, Wille-Jorgensen P, Kalebo P, et al. A comparison of recombinant hirudin with a low-molecular-weight-heparin to prevent thromboembolic complications after total hip replacement. N Engl J Med 1997;337:1329 – 35. Leclerc JR, Gearts WH, Desjardins L, et al. Prevention of venous thromboembolism after knee arthroplasty: a randomized, double-blind trial comparing enoxaparin with warfarin. Ann Intern Med 1996;124:616 – 28. Spiro TE, Johnson CJ, Christie MJ, et al. Efficacy and safety of enoxaparin to prevent deep venous thrombosis after hip replacement surgery. Ann Intern Med 1994;121:81 – 9. Hull R, Raskob G, Pineo G, et al. A comparison of subcutaneous low-molecular-weight heparin with warfarin sodium for prophylaxis against DVT after hip or knee implantation. N Engl J Med 1993;329:1370 – 6. Leclerc JR, Geerts WH, Desjardins L, et al. Prevention of deep vein thrombosis after major knee surgery: a randomized, double-blind trial comparing a low molecular weight heparin fragment (enoxaparin) to placebo. Thromb Haemost 1992;67(4):417 – 23. Levine MH, Hirsch J, Gent M. Prevention of deep vein thrombosis after elective hip surgery: a randomized trial comparing low molecular weight heparin with standard unfractionated heparin. Ann Intern Med 1991;114:545 – 51. Leyvraz PF, Bachmann F, Hoek J, et al. Prevention of deep vein thrombosis after hip replacement: randomized comparison between unfractionated heparin and low molecular weight heparin. BMJ 1991;303:543 – 8. Turpie AGG, Levine MN, Hirsch J, et al. A randomized controlled trial of low-molecular weight heparin (enoxaparin) to prevent deep-vein thrombosis in patients undergoing elective hip surgery. N Engl J Med 1986;31:925 – 9.