Current Medical Research & Opinion 0300-7995 doi:10.1185/03007995.2014.916159
2014, 1–10
Article FT-0118.R1/916159 All rights reserved: reproduction in whole or part not permitted
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
Original article Is rivaroxaban associated with lower inpatient costs compared to warfarin among patients with non-valvular atrial fibrillation? Franc¸ois Laliberte´ Dominic Pilon Groupe d’analyse Lte´e, Montre´al, QC, Canada
Monika K. Raut Winnie W. Nelson William H. Olson
Abstract Background: Warfarin has been the mainstay treatment used by patients with a moderate-to-high risk of stroke due to non-valvular atrial fibrillation (NVAF). Unlike rivaroxaban, laboratory monitoring to allow the attainment of the prothrombin time international normalized ratio goal is required with warfarin, thereby potentially increasing a patient’s hospitalization costs.
Janssen Scientific Affairs LLC, Raritan, NJ, USA
Guillaume Germain Groupe d’analyse Lte´e, Montre´al, QC, Canada
Jeff R. Schein Janssen Scientific Affairs LLC, Raritan, NJ, USA
Patrick Lefebvre Groupe d’analyse Lte´e, Montre´al, QC, Canada Address for correspondence: Franc¸ois Laliberte´ MA, Groupe d’analyse Lte´e, 1000 rue de la Gauchetie`re Ouest, Bureau 1200, Montre´al, Que´bec H3B 4W5, Canada. Tel.: +1 514 394 4488; Fax: +1 514 394 4461; [email protected] Keywords: Anticoagulant agents – Atrial fibrillation – Healthcare costs – Rivaroxaban – Warfarin Accepted: 15 April 2014; published online: 2 May 2014 Citation: Curr Med Res Opin 2014; 1–10
Objective: To compare hospitalization costs between hospitalized NVAF patients using rivaroxaban versus warfarin in a real-world setting. Methods: A retrospective claims analysis was conducted using the Premier Perspective Comparative Hospital Database from November 2010 to September 2012. The study included adult patients hospitalized for NVAF after November 2011. Patients using rivaroxaban during hospitalization were matched with up to four warfarin users by propensity score analyses. Hospitalization costs were compared between the matched cohorts using generalized estimating equations. A sub-analysis was performed for patients who were first administered their treatment on day three or later of their hospital stay. Sensitivity analyses were conducted on matched cohorts with a primary diagnosis of AF. Results: The matched cohorts’ (2809 rivaroxaban and 11,085 warfarin users) characteristics were well balanced. The mean age of cohorts was 71 years and 49% of patients were female. The average hospitalization cost of rivaroxaban users was $11,993 compared to $13,255 for warfarin users. The cost difference was significantly lower by $1284 (P50.001). Patients who were administered rivaroxaban treatment on day three or after incurred significantly lower hospitalization costs (cost difference: $4350; P50.001) compared to warfarin users. Rivaroxaban users with a primary diagnosis of AF also had significantly lower costs compared to warfarin users. Limitations: These included possible inaccuracies or omissions in diagnoses, completeness of baseline characteristics, and a study population that included patients newly initiated on and patients who continued anticoagulant therapy. Conclusion: Hospitalization costs for rivaroxaban were significantly lower than those for warfarin in NVAF patients treated with rivaroxaban.
! 2014 Informa UK Ltd www.cmrojournal.com
Hospitalization costs of NVAF using rivaroxaban vs warfarin Laliberte´ et al.
1
Current Medical Research & Opinion
2014
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
Introduction Atrial fibrillation (AF) is a growing healthcare problem in the United States with an estimated prevalence ranging between 2.7 and 6.1 million in 20101. The incremental cost burden of patients with AF compared to patients without AF was estimated at $26 billion in the United States in 2010, with more than 50% of this amount being attributed to hospitalization costs2–4. For decades, warfarin and other vitamin-K antagonists (VKAs) have been the only oral anticoagulants used by patients with a moderate-to-high risk of stroke due to nonvalvular atrial fibrillation (NVAF) in the prevention of strokes and systemic embolisms5. Warfarin is a high-risk medication with narrow therapeutic ranges. Patient variability in drug metabolism, comorbidities, and concomitant use of interacting drugs significantly impacts the anticoagulant effect, making it difficult to maintain patients within the defined anticoagulation range. Patients with AF taking warfarin are almost 50% of the time outside the INR target range of 2–36. Warfarin is also known to increase the risk of bleeding especially when the drug is above its target therapeutic ranges7,8. Moreover, in a prospective observational study, warfarin ranked third (after NSAIDs and diuretics) among the drugs or drug groups that were the most commonly associated with hospitalizations caused by adverse drug reactions9, and the average cost to manage hospitalizations with warfarin complications was estimated at $10,81910. Unlike warfarin, recently approved target-specific oral anticoagulants, such as rivaroxaban, apixaban, and dabigatran, have the advantage of not requiring international normalized ratio (INR) monitoring11. Among patients with NVAF, these target-specific oral anticoagulants have been found to be effective in preventing stroke or systemic embolism12–14 and to have similar risk of major bleeding compared to warfarin15. In addition, among hospitalized patients with NVAF, recent studies have reported that patients receiving target-specific oral anticoagulants had a shorter hospital length of stay (LOS) compared to those receiving warfarin16–18. The purpose of this study was to compare hospitalization costs among a sample of patients with NVAF using rivaroxaban and a matched sample of patients with NVAF using warfarin in a real-world setting.
Methods Data source Health insurance claims from the Premier Perspective Comparative Hospital Database were used to conduct the analysis. The Premier database is a United States retrospective database that contains more than 45 million 2
Hospitalization costs of NVAF using rivaroxaban vs warfarin Laliberte´ et al.
inpatient discharges from over 600 acute-care hospitals. In contrast to data recorded by insurance companies, patients’ medical information available in the Premier database comes from records collected for billing purposes at the hospital level. Data elements include demographics (e.g., age, gender, marital status, race, payer type), visitlevel information (e.g., primary and secondary diagnoses), hospital characteristics (e.g., urban, teaching, number of beds, region), and detailed drug use information (e.g., drug name, dosage strength, dispensed quantity). The Premier data used in the current analysis covered the period from November 2010 to September 2012. Premier data are de-identified and fully compliant with all HIPAA privacy and security requirements to protect participant anonymity and confidentiality.
Study design A retrospective matched-cohort design was used to compare the cost of hospitalization between rivaroxaban and warfarin users in an NVAF population. Patients included in the study had to have at least one AF diagnosis (ICD-9CM: 427.31) during a hospitalization (index hospitalization) after November 2011, the month in which rivaroxaban was approved for patients with NVAF, be at least 18 years of age, and have a dispensing of rivaroxaban or warfarin during the index hospitalization. Patients were excluded from the study if they had an observed valvular heart disease diagnosis or a valve replacement procedure, or if they used dabigatran or both rivaroxaban and warfarin during the index hospitalization. The observation period spanned from the first admission day of the index hospitalization to the hospitalization discharge. Records up to 12 months before the index hospitalization and on the day of hospital admission were used to evaluate patient and hospital baseline characteristics.
Study endpoints The outcome measure of the study was the hospitalization cost difference between rivaroxaban and warfarin users during the index hospitalization. Using the cost variables provided by Premier representing each hospital’s estimate of expenses incurred to provide services, including prorated fixed costs of maintaining facilities (i.e., overhead costs such as utilities and insurance) and variable perpatient care costs (i.e., direct costs such as materials, labor and expenses related to the production of service), the difference in costs between treatments was divided into (a) fixed costs and (b) variable costs. A subgroup analysis was conducted on patients who were administered their first rivaroxaban or warfarin treatment on day three or later during the hospitalization. We assumed these patients were more likely to be newly initiated on therapy. www.cmrojournal.com ! 2014 Informa UK Ltd
Current Medical Research & Opinion
A sensitivity analysis was also conducted on patients with a primary diagnosis of AF and on patients with a primary diagnosis of AF who were first administered rivaroxaban or warfarin on day 3 or later.
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
Statistical analysis To minimize confounding bias, we performed propensity score matching. Patients in the rivaroxaban cohort were matched with up to four patients of the warfarin cohort based on propensity score calipers of 5% (i.e., finding patients with similar propensity scores in the two groups inside a 5% range) and exact matching factors. Exact matching factors used were pre-treatment use of parenteral anticoagulant therapy (defined as the use of unfractionated heparin, low molecular weight heparins, or synthetic heparinoid like agents [i.e., fondaparinux]), rivaroxaban or warfarin first administered on the third day or later, admitting physician specialty (i.e., internal medicine, cardiology, family/general medicine, surgery general, and other/unknown specialty), and admission site (i.e., emergency room, intensive care unit, and other sites). The up to 4 matching ratio was chosen to increase statistical power. The propensity scores were calculated with a multivariate logistic regression model including the following baseline characteristics (Table 1): age, gender, marital status, ethnicity, primary payer type, region of hospital, hospital characteristics (i.e., urban, large [500 beds], teaching), admitting physician specialty, admission source (i.e., physician referral, clinical referral, transfer from hospital), admission site (i.e., emergency room, intensive care unit), degree of severity (i.e., minor, moderate, major, extreme) of the admission diagnosis-related group defined by 3M’s algorithm APR-DRG grouper code19, pre-treatment administration of parenteral anticoagulant therapy, rivaroxaban or warfarin first administered on the third day or later, CHADS2 score, CHA2DS2VASC score, year of index date, and specific comorbidities (i.e., cardiovascular disease, venous thromboembolism, systemic embolism, major bleeding, chronic kidney disease, cancer, chronic obstructive pulmonary disease, current smoker, total knee replacement, total hip replacement, bleeding disorders, human immunodeficiency virus). The CHADS2 and CHA2DS2VASC scores and the specific comorbidities were not reported in Table 1 since they are potentially incomplete due to the nature of the database (i.e., only information available during hospitalizations occurring at the same hospital was available). The available information was nevertheless used in the propensity score matching. For the analysis of matched patients with a primary diagnosis of AF, the same variables were used for the propensity score calculation and pre-treatment administration of parenteral anticoagulant therapy, and ! 2014 Informa UK Ltd www.cmrojournal.com
2014
rivaroxaban or warfarin first administered on the third day or later were used as exact matching factors. A separate matching was conducted for patients with a primary diagnosis of AF who were first administered rivaroxaban or warfarin on day 3 or later based only on the propensity score. For these analyses patients were matched exactly 1 to 4. Baseline characteristics were summarized by means, medians, and standard deviations (SDs) for continuous data and relative frequencies for categorical data. Differences in baseline characteristics between the two cohorts were assessed using standardized differences. Baseline characteristics were considered well balanced if the standardized difference was below 10%20–22. Hospitalization cost differences between rivaroxaban and warfarin users among NVAF patients were calculated using generalized estimating equation models with a gamma distribution, log link, and adjusting for the matched design. All costs were inflation-adjusted to 2013 $US based on the medical care component of the Consumer Price Index. Statistical significance was assessed at -level of 0.05 or less. All statistical analyses were conducted using SAS 9.3 (SAS Institute Inc., Cary, NC, USA).
Results Figure 1 presents the patient disposition flow chart. A total of 2809 eligible patients on rivaroxaban and 95,955 patients on warfarin were identified. All of the patients treated with rivaroxaban were matched with 11,085 warfarin users. Baseline characteristics for the matched cohorts are presented in Table 1. All baseline characteristics between cohorts were well balanced with a standardized difference 510%. The mean age of the matched cohorts was 71 years and 49% of patients were female. About three out of four patients in each cohort had a moderate or major disease severity level. In addition, the mean number of days between the admission and the administration of warfarin or rivaroxaban therapy was 1.7 days for both cohorts. Table 2 presents the baseline characteristics for the sub-analysis in which patients were administered their first anticoagluant therapy on the third day or later. In this subgroup, a total of 932 rivaroxaban users were matched to 3702 warfarin users, and patients’ baseline characteristics were well balanced. Baseline characteristics for patients with a primary diagnosis of AF were also well balanced (Appendices 1 and 2). Figure 2 presents descriptive statistics on the hospitalization costs and the adjusted cost difference between the matched cohorts during the index hospitalization. Total inpatient stay costs were, on average, $1284 lower per patient in the rivaroxaban cohort than those in the warfarin cohort (mean: $11,993 vs. $13,255; 95% CI ¼ 759, Hospitalization costs of NVAF using rivaroxaban vs warfarin Laliberte´ et al.
3
Current Medical Research & Opinion
2014
Table 1. Patient demographics and clinical characteristics – matched cohorts. Characteristics
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
Matching factorsa Demographicsb Age, mean [median] (SD) Gender, female, n (%) Marital statusb, n (%) Married Single Other/unknown Ethnicityb, n (%) White Black Other/unknown Primary payer typeb, n (%) Medicare Managed care Commercial indemnity Medicaid Other/unknown Region of hospitalb, n (%) South Midwest Northeast West Unknown Hospital characteristicsb, n (%) Urban (vs. rural) Large (500 beds) Teaching Admitting physician specialtyb, n (%) Internal medicine (internist) Cardiology Family/general medicine Surgery general Other/unknown Admission sourceb, n (%) Physician referral Clinical referral Transfer from hospital Other source Admission siteb,c, n (%) Emergency room Intensive care unit Other site Degree of severityd, n (%) Minor Moderate Major Extreme Treatment patterns, n (%) First administration of rivaroxaban or warfarin on day 3 or latere Pre-treatment administration of parenteral therapyf Unfractionated heparing Low molecular weight heparing Non-matching factors Days to rivaroxaban/warfarin administratione, mean [median] (SD) Administration of other anticoagulant therapyh, n (%)
Rivaroxaban Cohort (N ¼ 2809)
Warfarin Cohort (N ¼ 11,085)
Standardized Difference
71.0 [73] (12.0) 1372 (48.8)
71.1 [72] (11.6) 5390 (48.6)
0.4% 0.4%
1495 (53.2) 1097 (39.1) 217 (7.7)
5832 (52.6) 4429 (40.0) 824 (7.4)
1.2% 1.8% 1.1%
2357 (83.9) 132 (4.7) 320 (11.4)
9367 (84.5) 522 (4.7) 1196 (10.8)
1.6% 0.0% 1.9%
2011 (71.6) 496 (17.7) 138 (4.9) 57 (2.0) 107 (3.8)
7993 (72.1) 1834 (16.5) 540 (4.9) 259 (2.3) 459 (4.1)
1.1% 3.0% 0.2% 2.1% 1.7%
1546 (55.0) 638 (22.7) 324 (11.5) 296 (10.5) 5 (0.2)
6016 (54.3) 2574 (23.2) 1271 (11.5) 1201 (10.8) 23 (0.2)
1.5% 1.2% 0.2% 1.0% 0.7%
2298 (81.8) 802 (28.6) 935 (33.3)
9138 (82.4) 3174 (28.6) 3759 (33.9)
1.6% 0.2% 1.3%
1208 (43.0) 520 (18.5) 247 (8.8) 63 (2.2) 771 (27.4)
4832 (43.6) 2063 (18.6) 967 (8.7) 245 (2.2) 2978 (26.9)
1.2% 0.3% 0.2% 0.2% 1.3%
2231 (79.4) 266 (9.5) 140 (5.0) 172 (6.1)
8831 (79.7) 978 (8.8) 569 (5.1) 707 (6.4)
0.6% 2.2% 0.7% 1.1%
1397 (49.7) 352 (12.5) 1060 (37.7)
5562 (50.2) 1388 (12.5) 4135 (37.3)
0.9% 0.0% 0.9%
457 (16.3) 1306 (46.5) 862 (30.7) 184 (6.6)
1676 (15.1) 5177 (46.7) 3436 (31.0) 796 (7.2)
3.2% 0.4% 0.7% 2.5%
932 (33.2) 1490 (53.0) 744 (26.5) 936 (33.3)
3702 (33.4) 5919 (53.4) 3078 (27.8) 3958 (35.7)
0.5% 0.7% 2.9% 5.0%
1.7 [1] (2.4) 45 (1.6)
1.7 [1] (3.0) 169 (1.5)
1.0% 0.6%
SD: standard deviation. a Additional propensity score matching factors not reported in this table include the following variables evaluated in the up to 12 months before the admission (index hospitalization) and with records on the admission: year of index date, CHADS2 score, CHA2DS2VASC score, and specific comorbidities (i.e., cardiovascular disease, venous thromboembolism, systemic embolism, major bleeding, chronic kidney disease, cancer, chronic obstructive pulmonary disease, current smoker, total knee replacement, total hip replacement, bleeding disorders, human and immunodeficiency virus). b Evaluated at the first day of the index hospitalization. c For patients with both ICU and ER admissions on the first day, the admission site was classified as ICU. d Severity level for the admission diagnosis-related group, as defined by 3M’s algorithm (APR-DRG grouper codes). e Evaluated during the index hospitalization. f Including unfractionated heparin, low molecular weight heparins, or synthetic heparinoid like agents (i.e., fondaparinux). g Not included as a matching factor. h Including thrombin inhibitors hirudin type (i.e, bivalirudin, desirudin, and lepirudin), or thrombin inhibitors selective direct and reversible (i.e., argatroban) during the index hospitalization.
4
Hospitalization costs of NVAF using rivaroxaban vs warfarin Laliberte´ et al.
www.cmrojournal.com ! 2014 Informa UK Ltd
Current Medical Research & Opinion
2014
Patients with ≥1 AF diagnosis during a hospitalization after 1 November 2011 N = 380,371
Rivaroxaban Cohort
Warfarin Cohort
Use of rivaroxaban therapy during the first AF hospitalization (i.e., index hospitalization) N = 3944
Use of warfarin therapy during the first AF hospitalization (i.e., index hospitalization) N = 133,148
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
1135 patients were excluded:
37,193 patients were excluded:
0 were
2014, 1–10
Article FT-0118.R1/916159 All rights reserved: reproduction in whole or part not permitted
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
Original article Is rivaroxaban associated with lower inpatient costs compared to warfarin among patients with non-valvular atrial fibrillation? Franc¸ois Laliberte´ Dominic Pilon Groupe d’analyse Lte´e, Montre´al, QC, Canada
Monika K. Raut Winnie W. Nelson William H. Olson
Abstract Background: Warfarin has been the mainstay treatment used by patients with a moderate-to-high risk of stroke due to non-valvular atrial fibrillation (NVAF). Unlike rivaroxaban, laboratory monitoring to allow the attainment of the prothrombin time international normalized ratio goal is required with warfarin, thereby potentially increasing a patient’s hospitalization costs.
Janssen Scientific Affairs LLC, Raritan, NJ, USA
Guillaume Germain Groupe d’analyse Lte´e, Montre´al, QC, Canada
Jeff R. Schein Janssen Scientific Affairs LLC, Raritan, NJ, USA
Patrick Lefebvre Groupe d’analyse Lte´e, Montre´al, QC, Canada Address for correspondence: Franc¸ois Laliberte´ MA, Groupe d’analyse Lte´e, 1000 rue de la Gauchetie`re Ouest, Bureau 1200, Montre´al, Que´bec H3B 4W5, Canada. Tel.: +1 514 394 4488; Fax: +1 514 394 4461; [email protected] Keywords: Anticoagulant agents – Atrial fibrillation – Healthcare costs – Rivaroxaban – Warfarin Accepted: 15 April 2014; published online: 2 May 2014 Citation: Curr Med Res Opin 2014; 1–10
Objective: To compare hospitalization costs between hospitalized NVAF patients using rivaroxaban versus warfarin in a real-world setting. Methods: A retrospective claims analysis was conducted using the Premier Perspective Comparative Hospital Database from November 2010 to September 2012. The study included adult patients hospitalized for NVAF after November 2011. Patients using rivaroxaban during hospitalization were matched with up to four warfarin users by propensity score analyses. Hospitalization costs were compared between the matched cohorts using generalized estimating equations. A sub-analysis was performed for patients who were first administered their treatment on day three or later of their hospital stay. Sensitivity analyses were conducted on matched cohorts with a primary diagnosis of AF. Results: The matched cohorts’ (2809 rivaroxaban and 11,085 warfarin users) characteristics were well balanced. The mean age of cohorts was 71 years and 49% of patients were female. The average hospitalization cost of rivaroxaban users was $11,993 compared to $13,255 for warfarin users. The cost difference was significantly lower by $1284 (P50.001). Patients who were administered rivaroxaban treatment on day three or after incurred significantly lower hospitalization costs (cost difference: $4350; P50.001) compared to warfarin users. Rivaroxaban users with a primary diagnosis of AF also had significantly lower costs compared to warfarin users. Limitations: These included possible inaccuracies or omissions in diagnoses, completeness of baseline characteristics, and a study population that included patients newly initiated on and patients who continued anticoagulant therapy. Conclusion: Hospitalization costs for rivaroxaban were significantly lower than those for warfarin in NVAF patients treated with rivaroxaban.
! 2014 Informa UK Ltd www.cmrojournal.com
Hospitalization costs of NVAF using rivaroxaban vs warfarin Laliberte´ et al.
1
Current Medical Research & Opinion
2014
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
Introduction Atrial fibrillation (AF) is a growing healthcare problem in the United States with an estimated prevalence ranging between 2.7 and 6.1 million in 20101. The incremental cost burden of patients with AF compared to patients without AF was estimated at $26 billion in the United States in 2010, with more than 50% of this amount being attributed to hospitalization costs2–4. For decades, warfarin and other vitamin-K antagonists (VKAs) have been the only oral anticoagulants used by patients with a moderate-to-high risk of stroke due to nonvalvular atrial fibrillation (NVAF) in the prevention of strokes and systemic embolisms5. Warfarin is a high-risk medication with narrow therapeutic ranges. Patient variability in drug metabolism, comorbidities, and concomitant use of interacting drugs significantly impacts the anticoagulant effect, making it difficult to maintain patients within the defined anticoagulation range. Patients with AF taking warfarin are almost 50% of the time outside the INR target range of 2–36. Warfarin is also known to increase the risk of bleeding especially when the drug is above its target therapeutic ranges7,8. Moreover, in a prospective observational study, warfarin ranked third (after NSAIDs and diuretics) among the drugs or drug groups that were the most commonly associated with hospitalizations caused by adverse drug reactions9, and the average cost to manage hospitalizations with warfarin complications was estimated at $10,81910. Unlike warfarin, recently approved target-specific oral anticoagulants, such as rivaroxaban, apixaban, and dabigatran, have the advantage of not requiring international normalized ratio (INR) monitoring11. Among patients with NVAF, these target-specific oral anticoagulants have been found to be effective in preventing stroke or systemic embolism12–14 and to have similar risk of major bleeding compared to warfarin15. In addition, among hospitalized patients with NVAF, recent studies have reported that patients receiving target-specific oral anticoagulants had a shorter hospital length of stay (LOS) compared to those receiving warfarin16–18. The purpose of this study was to compare hospitalization costs among a sample of patients with NVAF using rivaroxaban and a matched sample of patients with NVAF using warfarin in a real-world setting.
Methods Data source Health insurance claims from the Premier Perspective Comparative Hospital Database were used to conduct the analysis. The Premier database is a United States retrospective database that contains more than 45 million 2
Hospitalization costs of NVAF using rivaroxaban vs warfarin Laliberte´ et al.
inpatient discharges from over 600 acute-care hospitals. In contrast to data recorded by insurance companies, patients’ medical information available in the Premier database comes from records collected for billing purposes at the hospital level. Data elements include demographics (e.g., age, gender, marital status, race, payer type), visitlevel information (e.g., primary and secondary diagnoses), hospital characteristics (e.g., urban, teaching, number of beds, region), and detailed drug use information (e.g., drug name, dosage strength, dispensed quantity). The Premier data used in the current analysis covered the period from November 2010 to September 2012. Premier data are de-identified and fully compliant with all HIPAA privacy and security requirements to protect participant anonymity and confidentiality.
Study design A retrospective matched-cohort design was used to compare the cost of hospitalization between rivaroxaban and warfarin users in an NVAF population. Patients included in the study had to have at least one AF diagnosis (ICD-9CM: 427.31) during a hospitalization (index hospitalization) after November 2011, the month in which rivaroxaban was approved for patients with NVAF, be at least 18 years of age, and have a dispensing of rivaroxaban or warfarin during the index hospitalization. Patients were excluded from the study if they had an observed valvular heart disease diagnosis or a valve replacement procedure, or if they used dabigatran or both rivaroxaban and warfarin during the index hospitalization. The observation period spanned from the first admission day of the index hospitalization to the hospitalization discharge. Records up to 12 months before the index hospitalization and on the day of hospital admission were used to evaluate patient and hospital baseline characteristics.
Study endpoints The outcome measure of the study was the hospitalization cost difference between rivaroxaban and warfarin users during the index hospitalization. Using the cost variables provided by Premier representing each hospital’s estimate of expenses incurred to provide services, including prorated fixed costs of maintaining facilities (i.e., overhead costs such as utilities and insurance) and variable perpatient care costs (i.e., direct costs such as materials, labor and expenses related to the production of service), the difference in costs between treatments was divided into (a) fixed costs and (b) variable costs. A subgroup analysis was conducted on patients who were administered their first rivaroxaban or warfarin treatment on day three or later during the hospitalization. We assumed these patients were more likely to be newly initiated on therapy. www.cmrojournal.com ! 2014 Informa UK Ltd
Current Medical Research & Opinion
A sensitivity analysis was also conducted on patients with a primary diagnosis of AF and on patients with a primary diagnosis of AF who were first administered rivaroxaban or warfarin on day 3 or later.
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
Statistical analysis To minimize confounding bias, we performed propensity score matching. Patients in the rivaroxaban cohort were matched with up to four patients of the warfarin cohort based on propensity score calipers of 5% (i.e., finding patients with similar propensity scores in the two groups inside a 5% range) and exact matching factors. Exact matching factors used were pre-treatment use of parenteral anticoagulant therapy (defined as the use of unfractionated heparin, low molecular weight heparins, or synthetic heparinoid like agents [i.e., fondaparinux]), rivaroxaban or warfarin first administered on the third day or later, admitting physician specialty (i.e., internal medicine, cardiology, family/general medicine, surgery general, and other/unknown specialty), and admission site (i.e., emergency room, intensive care unit, and other sites). The up to 4 matching ratio was chosen to increase statistical power. The propensity scores were calculated with a multivariate logistic regression model including the following baseline characteristics (Table 1): age, gender, marital status, ethnicity, primary payer type, region of hospital, hospital characteristics (i.e., urban, large [500 beds], teaching), admitting physician specialty, admission source (i.e., physician referral, clinical referral, transfer from hospital), admission site (i.e., emergency room, intensive care unit), degree of severity (i.e., minor, moderate, major, extreme) of the admission diagnosis-related group defined by 3M’s algorithm APR-DRG grouper code19, pre-treatment administration of parenteral anticoagulant therapy, rivaroxaban or warfarin first administered on the third day or later, CHADS2 score, CHA2DS2VASC score, year of index date, and specific comorbidities (i.e., cardiovascular disease, venous thromboembolism, systemic embolism, major bleeding, chronic kidney disease, cancer, chronic obstructive pulmonary disease, current smoker, total knee replacement, total hip replacement, bleeding disorders, human immunodeficiency virus). The CHADS2 and CHA2DS2VASC scores and the specific comorbidities were not reported in Table 1 since they are potentially incomplete due to the nature of the database (i.e., only information available during hospitalizations occurring at the same hospital was available). The available information was nevertheless used in the propensity score matching. For the analysis of matched patients with a primary diagnosis of AF, the same variables were used for the propensity score calculation and pre-treatment administration of parenteral anticoagulant therapy, and ! 2014 Informa UK Ltd www.cmrojournal.com
2014
rivaroxaban or warfarin first administered on the third day or later were used as exact matching factors. A separate matching was conducted for patients with a primary diagnosis of AF who were first administered rivaroxaban or warfarin on day 3 or later based only on the propensity score. For these analyses patients were matched exactly 1 to 4. Baseline characteristics were summarized by means, medians, and standard deviations (SDs) for continuous data and relative frequencies for categorical data. Differences in baseline characteristics between the two cohorts were assessed using standardized differences. Baseline characteristics were considered well balanced if the standardized difference was below 10%20–22. Hospitalization cost differences between rivaroxaban and warfarin users among NVAF patients were calculated using generalized estimating equation models with a gamma distribution, log link, and adjusting for the matched design. All costs were inflation-adjusted to 2013 $US based on the medical care component of the Consumer Price Index. Statistical significance was assessed at -level of 0.05 or less. All statistical analyses were conducted using SAS 9.3 (SAS Institute Inc., Cary, NC, USA).
Results Figure 1 presents the patient disposition flow chart. A total of 2809 eligible patients on rivaroxaban and 95,955 patients on warfarin were identified. All of the patients treated with rivaroxaban were matched with 11,085 warfarin users. Baseline characteristics for the matched cohorts are presented in Table 1. All baseline characteristics between cohorts were well balanced with a standardized difference 510%. The mean age of the matched cohorts was 71 years and 49% of patients were female. About three out of four patients in each cohort had a moderate or major disease severity level. In addition, the mean number of days between the admission and the administration of warfarin or rivaroxaban therapy was 1.7 days for both cohorts. Table 2 presents the baseline characteristics for the sub-analysis in which patients were administered their first anticoagluant therapy on the third day or later. In this subgroup, a total of 932 rivaroxaban users were matched to 3702 warfarin users, and patients’ baseline characteristics were well balanced. Baseline characteristics for patients with a primary diagnosis of AF were also well balanced (Appendices 1 and 2). Figure 2 presents descriptive statistics on the hospitalization costs and the adjusted cost difference between the matched cohorts during the index hospitalization. Total inpatient stay costs were, on average, $1284 lower per patient in the rivaroxaban cohort than those in the warfarin cohort (mean: $11,993 vs. $13,255; 95% CI ¼ 759, Hospitalization costs of NVAF using rivaroxaban vs warfarin Laliberte´ et al.
3
Current Medical Research & Opinion
2014
Table 1. Patient demographics and clinical characteristics – matched cohorts. Characteristics
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
Matching factorsa Demographicsb Age, mean [median] (SD) Gender, female, n (%) Marital statusb, n (%) Married Single Other/unknown Ethnicityb, n (%) White Black Other/unknown Primary payer typeb, n (%) Medicare Managed care Commercial indemnity Medicaid Other/unknown Region of hospitalb, n (%) South Midwest Northeast West Unknown Hospital characteristicsb, n (%) Urban (vs. rural) Large (500 beds) Teaching Admitting physician specialtyb, n (%) Internal medicine (internist) Cardiology Family/general medicine Surgery general Other/unknown Admission sourceb, n (%) Physician referral Clinical referral Transfer from hospital Other source Admission siteb,c, n (%) Emergency room Intensive care unit Other site Degree of severityd, n (%) Minor Moderate Major Extreme Treatment patterns, n (%) First administration of rivaroxaban or warfarin on day 3 or latere Pre-treatment administration of parenteral therapyf Unfractionated heparing Low molecular weight heparing Non-matching factors Days to rivaroxaban/warfarin administratione, mean [median] (SD) Administration of other anticoagulant therapyh, n (%)
Rivaroxaban Cohort (N ¼ 2809)
Warfarin Cohort (N ¼ 11,085)
Standardized Difference
71.0 [73] (12.0) 1372 (48.8)
71.1 [72] (11.6) 5390 (48.6)
0.4% 0.4%
1495 (53.2) 1097 (39.1) 217 (7.7)
5832 (52.6) 4429 (40.0) 824 (7.4)
1.2% 1.8% 1.1%
2357 (83.9) 132 (4.7) 320 (11.4)
9367 (84.5) 522 (4.7) 1196 (10.8)
1.6% 0.0% 1.9%
2011 (71.6) 496 (17.7) 138 (4.9) 57 (2.0) 107 (3.8)
7993 (72.1) 1834 (16.5) 540 (4.9) 259 (2.3) 459 (4.1)
1.1% 3.0% 0.2% 2.1% 1.7%
1546 (55.0) 638 (22.7) 324 (11.5) 296 (10.5) 5 (0.2)
6016 (54.3) 2574 (23.2) 1271 (11.5) 1201 (10.8) 23 (0.2)
1.5% 1.2% 0.2% 1.0% 0.7%
2298 (81.8) 802 (28.6) 935 (33.3)
9138 (82.4) 3174 (28.6) 3759 (33.9)
1.6% 0.2% 1.3%
1208 (43.0) 520 (18.5) 247 (8.8) 63 (2.2) 771 (27.4)
4832 (43.6) 2063 (18.6) 967 (8.7) 245 (2.2) 2978 (26.9)
1.2% 0.3% 0.2% 0.2% 1.3%
2231 (79.4) 266 (9.5) 140 (5.0) 172 (6.1)
8831 (79.7) 978 (8.8) 569 (5.1) 707 (6.4)
0.6% 2.2% 0.7% 1.1%
1397 (49.7) 352 (12.5) 1060 (37.7)
5562 (50.2) 1388 (12.5) 4135 (37.3)
0.9% 0.0% 0.9%
457 (16.3) 1306 (46.5) 862 (30.7) 184 (6.6)
1676 (15.1) 5177 (46.7) 3436 (31.0) 796 (7.2)
3.2% 0.4% 0.7% 2.5%
932 (33.2) 1490 (53.0) 744 (26.5) 936 (33.3)
3702 (33.4) 5919 (53.4) 3078 (27.8) 3958 (35.7)
0.5% 0.7% 2.9% 5.0%
1.7 [1] (2.4) 45 (1.6)
1.7 [1] (3.0) 169 (1.5)
1.0% 0.6%
SD: standard deviation. a Additional propensity score matching factors not reported in this table include the following variables evaluated in the up to 12 months before the admission (index hospitalization) and with records on the admission: year of index date, CHADS2 score, CHA2DS2VASC score, and specific comorbidities (i.e., cardiovascular disease, venous thromboembolism, systemic embolism, major bleeding, chronic kidney disease, cancer, chronic obstructive pulmonary disease, current smoker, total knee replacement, total hip replacement, bleeding disorders, human and immunodeficiency virus). b Evaluated at the first day of the index hospitalization. c For patients with both ICU and ER admissions on the first day, the admission site was classified as ICU. d Severity level for the admission diagnosis-related group, as defined by 3M’s algorithm (APR-DRG grouper codes). e Evaluated during the index hospitalization. f Including unfractionated heparin, low molecular weight heparins, or synthetic heparinoid like agents (i.e., fondaparinux). g Not included as a matching factor. h Including thrombin inhibitors hirudin type (i.e, bivalirudin, desirudin, and lepirudin), or thrombin inhibitors selective direct and reversible (i.e., argatroban) during the index hospitalization.
4
Hospitalization costs of NVAF using rivaroxaban vs warfarin Laliberte´ et al.
www.cmrojournal.com ! 2014 Informa UK Ltd
Current Medical Research & Opinion
2014
Patients with ≥1 AF diagnosis during a hospitalization after 1 November 2011 N = 380,371
Rivaroxaban Cohort
Warfarin Cohort
Use of rivaroxaban therapy during the first AF hospitalization (i.e., index hospitalization) N = 3944
Use of warfarin therapy during the first AF hospitalization (i.e., index hospitalization) N = 133,148
Curr Med Res Opin Downloaded from informahealthcare.com by Library of Health Sci-Univ of Il on 06/17/14 For personal use only.
1135 patients were excluded:
37,193 patients were excluded:
0 were
