Comparison of hospital length of stay and hospitalization costs among patients with non-valvular atrial fibrillation treated with apixaban or warfarin: An early view

Abstract

Objective: To quantify and compare hospital length of stay (LOS) and costs between hospitalized non-valvular atrial fibrillation (NVAF) patients treated with either apixaban or warfarin via a large claims database.

Methods: Adult patients hospitalized with AF were selected from the Premier Perspective Claims Database (01JAN2013-31MARCH2014). Patients with evidence of valvular heart disease, valve replacement procedures, or pregnancy during the index hospitalization were excluded. Patients treated with apixaban or warfarin during hospitalization were identified. Propensity score matching (PSM) was performed to control for baseline imbalances between patients treated with apixaban or warfarin. Primary outcomes were hospital LOS (days), post-medication administration LOS, and index hospitalization costs, and were compared using paired t-tests in the matched sample.

Results: Before PSM, 2894 apixaban and 124,174 warfarin patients were identified. Patients treated with warfarin were older and sicker compared to those treated with apixaban. After applying PSM, a total of 2886 patients were included in each cohort, and baseline characteristics were balanced. The mean (standard deviation [SD] and median) hospital LOS was significantly (p = 0.002) shorter for patients treated with apixaban for 5.1 days (5.7 and 3) compared to warfarin for 5.5 days (4.8 and 4). The trend appeared consistent in the hospital LOS from point of apixaban or warfarin administration to discharge (4.5 vs 4.7 days, p = 0.051). Patients administered apixaban incurred significantly lower hospitalization costs compared to those administered warfarin ($11,262 vs $12,883; p < 0.001).

Conclusions: Among NVAF patients, apixaban treatment was associated with significantly shorter hospital LOS and lower costs when compared to warfarin treatment.

Keywords:

• Apixaban
• Warfarin
• Non-valvular atrial fibrillation
• Non-VKA oral anticoagulants
• Length of stay

View correction statement:

Erratum

Introduction

Atrial fibrillation (AF) is the most common heart dysrthythmia diagnosed in the US¹. AF prevalence has increased with an increasingly aging population in the US, and is expected to continue this trend substantially in the coming decades^2,3. In 2010 there were an estimated 5.2 million cases of AF and the number of AF cases is projected to increase to 12.1 million cases in 2030³. Furthermore, AF carries a significant financial burden, costing the US healthcare system ∼$26 billion annually, with hospitalizations accounting for a majority of the costs (52%)⁴. Non-valvular AF (NVAF) refers to AF that is not associated with rheumatic mitral stenosis, mechanical or bioprosthetic heart valves, or mitral valve repair⁵. NVAF affects ∼85–95% of the overall population with AF^6,7. NVAF has long been identified as a significant risk factor for disabling or fatal ischemic stroke and systemic embolism⁸. Specifically, NVAF is associated with a 5-fold increase in stroke risk compared to those without NVAF, and AF-related strokes tend to be more severe than non-AF-related strokes^9,10.

Until recently, vitamin K antagonists (VKAs), like warfarin, were the main drugs recommended for the reduction of risk of stroke in NVAF patients with a CHADS₂ score ≥2. However, due to warfarin’s narrow therapeutic index and multiple drug–drug and drug–food interactions, requiring regular INR monitoring with dose adjustments, only about half of eligible AF patients in the US receive warfarin therapy^9,11. Non-VKA oral anticoagulants (NOACs), including dabigatran, apixaban, and rivaroxaban, have demonstrated similar or superior stroke risk reduction without the need for routine monitoring and dose adjustments, compared to warfarin, and are alternative treatments for NVAF patients¹². Apixaban has been shown to be superior to dose-adjusted warfarin in terms of reduction in the rates of stroke and systemic embolism, major bleeding, and all-cause mortality¹³. Previous studies have reported that NVAF patients treated with NOACs had significantly shorter hospital length of stay (LOS) compared to those treated with warfarin^14–16. Moreover, NOACs are becoming increasingly popular, particularly among patients with lower CHADS₂ and HAS-BLED scores¹⁷. By June 2013, more than 60% of patients newly prescribed oral anticoagulants were prescribed dabigatran, rivaroxaban, or apixaban¹⁷. Apixaban was approved in December 2012; therefore, insufficient evidence exists regarding clinical and economic outcomes among NVAF patients in real-world settings. The goal of this study was to estimate and compare index hospitalization costs and hospital LOS for real-world, hospitalized NVAF patients treated with either warfarin or apixaban.

Methods

This study used data from the Premier Perspective Claims Database (PCD) from January 2012 through March 2014 and identified patients for the study population from January 2013 through March 2014. Premier is a large US hospital-based clinical and economic database developed for quality and utilization. This database contains detailed billing and coding information for more than 2.5 million daily service records for an average of more than 5 million de-identified hospital discharges per year from 2000 to the present. The hospitals included are nationally representative based on geographic location, bed size, and teaching hospital status. Data elements include hospital characteristics, and patient demographics, diagnoses, and procedures during the inpatient admission. Patients’ medical information is derived from records utilized for billing purposes at the hospital level. Additionally, the Premier PCD cost variables reflect billable amounts, including drugs, and hospital costs, which include all supplies, labour, and depreciation of equipment. It is the sum of the fixed (overhead) and variable (direct) costs contained in the dataset. Premier data are de-identified and fully compliant with all Health Insurance Portability and Accountability Act privacy and security requirements.

Study sample

The study period ranged from January 2012–March 2014. Adult patients (≥18 years) were selected from the Premier PCD during the identification period (January 2013–March 2014) if they were diagnosed with AF (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM] code: 427.31) during an inpatient admission. The first AF hospitalization during the identification period was defined as the index hospitalization.

In order to evaluate NVAF patients, those with evidence of mitral valvular heart disease (ICD-9-CM codes: 093.21, 394.xx, 369.xx, 424.0, 746.5, 746.6), valve replacement procedures (ICD-9 procedure codes: 35.05-.09, 35.20–35.28, 35.97), or pregnancy (ICD-9-CM diagnosis codes: 630-679, V22–V24, V27, V28, V61.6, V61.7, 792.3, 796.5 or ICD-9 procedure codes: 72–75.99) were excluded. Patients with a discharge date after February 28, 2014 were excluded. Patients were categorized into cohorts based on anticoagulant treatment received during the index hospitalization. If patients were administered apixaban during the index hospitalization then they were placed in the Apixaban Cohort and if they were administered warfarin they were placed in the Warfarin Cohort. If patients were administered both apixaban and warfarin or other NOACs (dabigatran or rivaroxaban) during the index hospitalization they were excluded. Patients were followed until re-admission or end of the study period.

Covariates

Demographic and clinical characteristics were collected during hospitalizations 12 months prior to and during the NVAF index hospitalization. Covariates included gender, age, region, race, pre-index Charlson Comorbidity Index (CCI) score, CHADS₂ score (congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism), CHA₂DS₂-VASc score (CHADS₂ plus vascular disease, age 65–74 years, sex category; Supplementary Material, Table 1), HAS-BLED score (hypertension, abnormal renal and liver function, stroke, bleeding, labile International Normalized Ratios [INRs], elderly age [≥65 years], drug therapy and alcohol intake; Supplementary Material, Table 2), and pre-index comorbid conditions^18–21. The CCI score predicts the 10-year mortality for a patient who may have a range of 22 co-morbid conditions¹⁸. The CHADS₂ and CHA₂DS₂-VASc scores predict stroke risk²⁰. The HAS-BLED score can range from 0–9, with a score ≥3 associated with a high bleeding risk^21,22. Co-morbid conditions included diabetes, congestive heart failure, hypertension, stroke, vascular disease, renal insufficiency, venous thromboembolism, acute coronary syndrome, drug and alcohol use, chronic kidney disease, chronic obstructive pulmonary disease, and anemia.

Table 1. Baseline characteristics of hospitalized NVAF patients treated with apixaban and warfarin.

	Apixaban (n = 2894)		Warfarin(n = 124,174)		p-value
	n/Mean	%/SD	n/Mean	%/SD
Age	73.2	11.7	75.1	11.0	<0.001
18–64	622	21.5%	21,047	16.9%	<0.001
65–74	808	27.9%	30,798	24.8%	<0.001
75+	1464	50.6%	72,329	58.2%	<0.001
Gender
Male	1462	50.5%	65,659	52.9%	0.012
Female	1432	49.5%	58,507	47.1%	0.012
Race
White	2423	83.7%	96,156	77.4%	<0.001
Black	149	5.1%	9,320	7.5%	<0.001
Other	322	11.1%	18,698	15.1%	<0.001
Insurance coverage
Managed	511	17.7%	13,021	10.5%	<0.001
Government	2300	79.5%	107,851	86.9%	<0.001
Indigent/Charity	10	0.3%	387	0.3%	0.747
Other	73	2.5%	2915	2.3%	0.539
Admission type
Emergency	2007	69.4%	87,928	70.8%	0.088
Urgent	377	13.0%	16,569	13.3%	0.621
Elective	479	16.6%	18,891	15.2%	0.048
Trauma center	9	0.3%	329	0.3%	0.635
Unknown	22	0.8%	457	0.4%	0.001
Discharge type
Expired	43	1.5%	2,449	2.0%	0.062
Discharged	2843	98.2%	120,988	97.4%	0.007
Other	8	0.3%	737	0.59%	0.028
Hospital characteristics
US Region
Northeast	397	13.7%	29,773	24.0%	<0.001
Midwest	740	25.6%	30,006	24.2%	0.081
South	1553	53.7%	46,799	37.7%	<0.001
West	204	7.0%	17,596	14.2%	<0.001
Population served
Urban	2325	80.3%	105,987	85.4%	<0.001
Rural	569	19.7%	18,187	14.6%	<0.001
Hospital setting
Teaching	1139	39.4%	49,473	39.8%	0.599
Non-teaching	1755	60.6%	74,701	60.2%	0.599
Hospital bed size
<100	155	5.4%	6,304	5.1%	0.499
100–199	350	12.1%	16,684	13.4%	0.036
200–299	569	19.7%	22,429	18.1%	0.027
300–499	983	34.0%	43,232	34.8%	0.343
≥500	837	28.9%	35,525	28.6%	0.713
Any hospitalization 12-months prior to index hospitalization	370	12.8%	30,793	24.8%	<0.001
Baseline Comorbidity
Deyo-Charlson Comorbidity Index	2.1	2.0	2.8	2.2	<0.001
CHADS2 Score	2.4	1.4	2.7	1.3	<0.001
0 = low risk	159	5.5%	3,533	2.8%	<0.001
1 = moderate risk	585	20.2%	16,806	13.5%	<0.001
2 = high risk	887	30.6%	36,167	29.1%	0.075
<2 = high risk	1263	43.6%	67,668	54.5%	<0.001
CHA2DS2-VASc Score	3.9	1.8	4.3	1.7	<0.001
0 = low risk	69	2.4%	1,001	0.8%	<0.001
1 = moderate risk	194	6.7%	4,339	3.5%	<0.001
2 = high risk	335	11.6%	11,298	9.1%	<0.001
<2 = high risk	2296	79.3%	107,536	86.6%	<0.001
HAS-BLED Score	3.1	1.2	3.3	1.2	<0.001
Congestive Heart Failure	1064	36.8%	59,130	47.6%	<0.001
Diabetes Mellitus	942	32.6%	49,606	39.9%	<0.001
Hypertension	2373	82.0%	103,907	83.7%	<0.001
Ischemic Stroke	294	10.2%	11,281	9.1%	0.047
Transient Ischemic Attack	79	2.7%	2,771	2.2%	0.074
Stroke	594	20.5%	27,886	22.5%	<0.001
Vascular Disease	1209	41.8%	62,920	50.7%	<0.001
Renal Insufficiency	830	28.7%	51,027	41.1%	<0.001
Venous Thromboembolism	24	0.8%	8,498	6.8%	<0.001
Acute Coronary Syndrome	241	8.3%	10,010	8.1%	0.603
Drug and Alcohol Use	411	14.2%	14,283	11.5%	<0.001
Chronic Kidney Disease	582	20.1%	39,476	31.8%	<0.001
Chronic Obstructive Pulmonary  Disease	649	22.4%	33,646	27.1%	<0.001
Anemia	646	22.3%	44,060	35.5%	<0.001

CHADS₂, congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism; CHA₂DS₂-VASc, CHADS₂ plus vascular disease, age 65–74 years, sex category; NVAF, non-valvular atrial fibrillation; SD, standard deviation.

Table 2. Baseline characteristics of hospitalized NVAF patients treated with apixaban and warfarin after 1:1 PSM.

	Apixaban		Warfarin		Std difference
	(n = 2886)		(n = 2886)
	n/Mean	%/SD	n/Mean	%/SD
Age	73.3	11.6	73.3	11.4	0.00
18–64	614	21.3%	589	20.4%	0.02
65–74	808	28.0%	823	28.5%	0.01
75+	1464	50.7%	1474	51.1%	0.01
Gender
Male	1454	50.4%	1494	51.8%	0.03
Female	1432	49.6%	1392	48.2%	0.03
Race
White	2415	83.7%	2421	83.9%	0.01
Black	149	5.2%	145	5.0%	0.01
Other	322	11.2%	320	11.1%	0.00
Insurance Coverage
Managed	503	17.4%	495	17.2%	0.01
Government	2300	79.7%	2308	80.0%	0.01
Indigent/Charity	10	0.3%	12	0.4%	0.01
Other	73	2.5%	71	2.5%	0.00
Admission Type
Emergency	2000	69.3%	2014	69.8%	0.01
Urgent	377	13.1%	364	12.6%	0.01
Elective	479	16.6%	477	16.5%	0.00
Trauma center	9	0.3%	11	0.4%	0.01
Unknown	21	0.7%	20	0.7%	0.00
Discharge Type
Discharged	2835	98.2%	2821	97.7%	0.03
Expired	43	1.5%	46	1.6%	0.01
Other	8	0.3%	19	0.7%	0.06
Hospital characteristics
Region
Northeast	397	13.8%	407	14.1%	0.01
Midwest	738	25.6%	738	25.6%	0.00
South	1547	53.6%	1553	53.8%	0.00
West	204	7.1%	188	6.5%	0.02
Population served
Urban	2318	80.3%	2342	81.2%	0.02
Rural	568	19.7%	544	18.8%	0.02
Hospital setting
Teaching	1134	39.3%	1133	39.3%	0.00
Non-teaching	1752	60.7%	1753	60.7%	0.00
Hospital bed size
<100	155	5.4%	148	5.1%	0.01
100–199	350	12.1%	354	12.3%	0.00
200–299	569	19.7%	559	19.4%	0.01
300–499	980	34.0%	958	33.2%	0.02
≥500	832	28.8%	867	30.0%	0.03
Any hospitalization 12 months prior to index hospitalization	370	12.8%	361	12.5%	0.01
Baseline Comorbidity
Deyo-Charlson Comorbidity Index score	2.1	2.0	2.2	2.0	0.03
CHADS2 score	2.4	1.3	2.4	1.3	0.01
0 = low risk	151	5.2%	167	5.8%	0.02
1 = moderate risk	585	20.3%	541	18.7%	0.04
2 = high risk	887	30.7%	882	30.6%	0.00
<2 = high risk	1263	43.8%	1296	44.9%	0.02
CHADS2-VASc score	3.9	1.8	3.9	1.7	0.00
0 = low risk	61	2.1%	70	2.4%	0.02
1 = moderate risk	194	6.7%	180	6.2%	0.02
2 = high risk	335	11.6%	320	11.1%	0.02
<2 = high risk	2296	79.6%	2316	80.2%	0.02
HAS-BLED Score	3.1	1.2	3.1	1.2	0.02
Congestive Heart Failure	1064	36.9%	1081	37.5%	0.01
Diabetes Mellitus	942	32.6%	946	32.8%	0.00
Hypertension	2373	82.2%	2376	82.3%	0.00
Ischemic Stroke	294	10.2%	285	9.9%	0.01
Transient Ischemic Attack	79	2.7%	57	2.0%	0.05
Stroke	594	20.6%	565	19.6%	0.03
Vascular Disease	1209	41.9%	1220	42.3%	0.01
Renal Insufficiency	829	28.7%	846	29.3%	0.01
Venous Thromboembolism	24	0.8%	28	1.0%	0.01
Acute Coronary Syndrome	241	8.4%	261	9.0%	0.02
Drug and Alcohol Use	407	14.1%	400	13.9%	0.01
Chronic Kidney Disease	582	20.2%	601	20.8%	0.02
Chronic Obstructive Pulmonary Disease	649	22.5%	665	23.0%	0.01
Anemia	646	22.4%	654	22.7%	0.01
Hemodialysis	16	0.6%	17	0.6%	0.00
Thrombocytopenia	132	4.6%	146	5.1%	0.02

CHADS₂, congestive heart failure, hypertension, age ≥75years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism; CHADS₂-VASc, CHADS₂ plus vascular disease, age 65–74 years, sex category; HAS-BLED, hypertension, abnormal renal and liver function, stroke, bleeding, labile International Normalized Ratios, elderly age (≥65 years), drug therapy and alcohol intake; NVAF, non-valvular atrial fibrillation; PSM, propensity score matching; SD, standard deviation; Std. Difference: standardized difference.

Outcome variables

The outcome variables included hospital LOS, index hospitalization costs, and re-admission rates. Hospital LOS was defined as the number of days from hospital admission to hospital discharge. Post-medication administration LOS was conducted in which hospital LOS was measured from first apixaban or warfarin administration to hospital discharge. Index hospitalization costs were calculated for each patient during the index hospitalization and included all billable costs (e.g. procedures, medications, room and board, and hospital costs) incurred during the hospitalization. Hospitalization costs were adjusted to 2014 US dollars using the Consumer Price Index (CPI) for Medical Care Services. Hospital re-admissions within 30 days of the index hospitalization were flagged. Re-admissions were only captured if the patient was re-admitted to the same hospital in which the index hospitalization took place, as re-admissions to other hospitals are not captured in the database.

Statistical analysis

The cohorts were compared for baseline and outcome variables. Percentages and standard deviations (SDs) were calculated for dichotomous and polychotomous variables and p-values were calculated using the Chi-square test. Means, medians, and SDs were provided for continuous variables, and p-values were calculated using the student t-test. The level of significance was set to α = 0.05.

Propensity score matching (PSM) was performed to balance baseline characteristics between the two cohorts^23,24. An unconditional logistic regression model was fitted to determine the baseline variables associated with the treatment cohorts, designated as 1 for apixaban and 0 for warfarin. Using this model, a propensity score was developed for each patient and represented the probability of being a member of the Apixaban Cohort. Covariates in the propensity model included age, gender, race, insurance coverage, patient type (inpatient, other), admission type, hospital characteristics (region, population served [urban, rural], hospital setting, hospital bed size), CCI score, CHA₂DS₂-VASc score, HAS-BLED score, co-morbid conditions, VTE, hemodialysis or thrombocytopenia during index hospitalization, and any hospitalization during 12 months prior to index hospitalization. Each patient in the Apixaban Cohort was matched to a patient in the Warfarin Cohort with the closest propensity score, using the nearest neighbor technique with a caliper of ±0.01²⁵. The balance of measured covariates between the matched cohorts was assessed using absolute standardized differences, which are not influenced by sample size. An absolute standardized difference of ≤0.1 indicates a negligible difference in measured variables between groups²⁶. After matching, index hospitalization costs, hospital LOS, and 30-day re-admission were compared between the two matched cohorts using paired t-tests and McNemar tests. A sensitivity analysis was conducted by comparing post-medication administration LOS. Additionally, costs were stratified by LOS and CHADS₂ score. All statistical analyses were conducted using SAS 9.3 (SAS Institute Inc., Cary, NC).

Results

Descriptive results

During their first AF hospitalization, 130,241 patients were treated with apixaban or warfarin. Of those, 219 patients were excluded because they were administered both apixaban and warfarin during the hospitalization, and 2954 patients were excluded because they were treated with other NOACs. After all of the criteria were applied, 127,068 patients in the study sample were identified, 2894 apixaban patients and 124,174 warfarin patients. Patients treated with warfarin were older (age ≥75 years: 58.2% vs 50.6%, p < 0.001), less likely to be female (47.1% vs 49.5%, p = 0.012), white (77.4% vs 83.7%, p < 0.001), and to have managed care insurance (10.5% vs 17.7%, p < 0.001) compared to those treated with apixaban. Of the apixaban and warfarin patients, 12.8% and 24.8% were hospitalized within 12 months prior to their index hospitalization, respectively. Patients treated with apixaban had significantly lower CHADS₂, CHA₂DS₂-VASc, and HAS-BLED scores (Table 1). Among patients treated with apixaban and warfarin, 1.5% and 2.0% died during their hospitalization, respectively. AF was the most common discharge diagnosis for patients in both cohorts. Other common discharge diagnoses included septicemia, systolic heart failure, and cerebral embolism.

Before matching, patients treated with apixaban had a shorter hospital LOS (5.0 vs 6.4 days, p < 0.001) and lower index hospitalization costs ($11,239 vs $15,094, p < 0.001). Patients in the apixaban cohort were less likely to be re-admitted within 30 days compared to patients in the warfarin cohort (10.2% vs 12.2%, p = 0.002).

Multivariate results

After 1:1 PSM, the demographic and clinical characteristics between the two cohorts were well balanced, and 2886 patients were matched in each cohort. The mean CHA₂DS₂-VASc score was 4.0 for both cohorts, indicating a high stroke risk. There were no significant differences in baseline characteristics between the two cohorts (Table 2).

After matching, there was a significant difference in hospital LOS between the two cohorts. Patients in the Apixaban Cohort had a significantly shorter average hospital LOS compared to those in the Warfarin Cohort (5.1 vs 5.5 days, p = 0.002; Figure 1). There was a 1-day difference in the median hospital LOS (3 vs 4 days). Furthermore, significantly more patients treated with warfarin had a hospital LOS from 3–14 days compared to those treated with apixaban (Table 3). The mean index hospital LOS from point of apixaban or warfarin administration followed the same trend (4.5 vs 4.7 days, p = 0.051); however, it was not statistically significant.

Figure 1. Average hospital LOS for NVAF patients. LOS, length of stay; NVAF, non-valvular atrial fibrillation.

Table 3. Outcomes of hospitalized NVAF patients treated with apixaban and warfarin after 1:1 PSM.

	Apixaban (n = 2886)		Warfarin (n = 2886)		p-value
	n/Mean	%/SD	n/Mean	%/SD
Index Hospitalization Length of Stay
Hospital Length of Stay (in days)	5.1	5.7	5.5	4.8	0.002
Median	3.0		4.0
0–2	939	32.5%	712	24.7%	<0.001
3–7	1444	50.0%	1582	54.8%	<0.001
8–14	360	12.5%	448	15.5%	0.001
15–30	128	4.4%	130	4.5%	0.899
>30	15	0.5%	14	0.5%	0.848
Hospital LOS from point of apixaban or warfarin administration to discharge	4.5	5.5	4.7	4.6	0.051
Index Hospitalization Health Care Costs	$11,262	$14,048	$12,883	$15,702	<0.001
Hospital Length of Stay (in days)
0–2	$5033	$5400	$5797	$6,189	0.041
3–7	$9554	$9255	$10,179	$7,176	<0.001
8–14	$21,008	$13,610	$22,386	$13,424	0.004
15–30	$39,381	$18,574	$42,764	$34,660	0.505
>30	$91,774	$49,843	$97,273	$69,917	0.981
CHADS2 Score
0–1 (low-to-moderate risk)	$9855	$11,275	$12,713	$14,996	0.004
≥2 (high-risk)	$11,744	$14,850	$12,938	$15,927	0.006
All-cause re-admission
Patients with 30 days of follow-up	2839	98.4%	2841	98.1%	0.424
Re-admission within 30 days	291	10.3%	306	10.8%	0.493

NVAF, non-valvular atrial fibrillation; PSM, propensity score matching; SD, standard deviation; CHADS₂, congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism.

In turn, patients treated with apixaban had significantly lower hospitalization costs compared to those treated with warfarin ($11,262 vs $12,883, p < 0.001; Figure 2). This trend remained apparent when hospitalization costs were stratified by LOS and CHADS₂ score. Patients in both the low-to-moderate and high stroke risk groups incurred significantly lower hospitalization costs when treated with apixaban compared to warfarin. Additionally, patients treated with apixaban who had a hospital LOS of 0–2 ($5033 vs $5797, p = 0.040), 3–7 ($9554 vs $10,179, p < 0.001), and 8–14 days ($21,008 vs $22,386, p < 0.001) incurred significantly lower hospitalization costs compared to those treated with warfarin. The trend remained consistent for longer hospital stays (>14 days), but this was not statistically significant. After matching the two cohorts, there was no significant difference in the 30-day re-admission rate (10.3% [n = 291] vs 10.8% [n = 306]).

Figure 2. Average index hospitalization costs for NVAF patients.

Discussion

In this retrospective claims study, apixaban treatment resulted in significantly lower hospital LOS and hospitalization costs compared to warfarin treatment among NVAF patients. This study provides real-world evidence to demonstrate the economic benefits of apixaban that may be a result of more favorable clinical outcomes, as observed in clinical trials; however, this hypothesis still needs to be validated. Due to apixaban’s recent market entry, further research needs to be completed with a larger sample size.

Patients treated with apixaban had significantly shorter hospital LOS, which resulted in significantly lower hospitalization costs. Before matching, apixaban patients were younger and healthier than warfarin patients; however, after PSM, the baseline clinical and demographic characteristics were balanced between the two cohorts. PSM controls confounders, and because the cohorts are balanced on all observable covariates, it is assumed that they are also closely balanced for unobservable characteristics. After matching, patients treated with apixaban had an estimated 1 day shorter median LOS. This shorter LOS for apixaban patients resulted in ∼$1600 per patient lower estimated index hospitalization costs, including pharmacy costs, compared to warfarin patients.

Additionally, the trend remained apparent when hospitalization costs were stratified by LOS and stroke risk. About 80% of the study population had a LOS between 0–7 days. Among patients with a LOS between 0–2 and 3–7 days, those treated with apixaban incurred, on average, $764 and $624, respectively, less during the index hospitalization compared to those treated with warfarin. Hospitalization costs increased significantly with each additional day of hospitalization. Furthermore, among patients with a CHADS₂ score ≥2, those treated with apixaban incurred more than $1100 lower hospitalization costs compared to those treated with warfarin. In the study population, there was no significant difference in the 30-day re-admission rate; however, this was an early view study due to apixaban’s market entry in January 2013. After apixaban has been on the market for a longer period of time, further research needs to be conducted to compare all-cause and major bleeding-related re-admission rates.

Few studies have evaluated hospital LOS and costs among patients treated with NOACs compared to warfarin using real-world data. A similar analysis of hospitalized NVAF patients treated with apixaban or warfarin was conducted by Farr et al.²⁷ using the MarketScan Hospital Discharge Database. They found that the mean (SD) and median hospital LOS were significantly shorter in apixaban patients (4.5 [4.2] and 3 days) compared to warfarin patients (5.4 [5.0] and 4 days; p < 0.001). Previous studies have also shown that patients treated with other NOACs incur significantly lower hospitalization costs and have shorter hospital stays. In a retrospective study using the Premier database, rivaroxaban users had significantly shorter hospital LOS (0.8 days) and lower hospitalization costs, estimated to be $1284, compared to warfarin users^15,28. However, a sub-group analysis showed that, when patients were stratified by whether or not they were administered parenteral anticoagulants, hospital LOS was significantly shorter only among patients administered parental anticoagulants in addition to rivaroxaban²⁹. Further analysis in the Humana database showed that patients administered rivaroxaban were significantly less likely to have an AF-related hospitalization (0.40 vs 0.57, p = 0.022), lower AF-related hospitalization costs ($2872 vs $4147, p = 0.020), and shorter hospital LOS (2.1 vs 3.0, p = 0.014) compared to matched warfarin patients^30,31. In another retrospective claims study examining treatment-naïve, newly-diagnosed AF patients, those treated with dabigatran had an estimated 0.7 days shorter stay compared to warfarin and $2000 lower estimated hospitalization costs¹⁴. The results from previous retrospective studies are similar to the hospital LOS and cost differences reported in this study.

Due to apixaban’s recent market entry, there is limited information regarding how apixaban is used in real-world clinical settings. Numerous cost-effectiveness Markov models have been developed to compare treatment costs between NOACs and warfarin. Using data from clinical trials, the models determined that NOACs produced greater quality-adjusted life expectancy than warfarin. Furthermore, apixaban provided the greatest monetary and quality-adjusted life year value compared to dabigatran, rivaroxaban, and warfarin^32,33. Medical costs for clinical events among patients who were prescribed any NOACs were estimated to be lower compared to warfarin. Apixaban use showed the largest difference, with a $485 cost reduction³⁴. Additionally, it was determined that apixaban was the most cost-effective treatment for stroke risk reduction^35–37. When retrospective claims data were used to predict medical costs, patients treated with apixaban had medical cost reductions³⁸. Most of the Markov models are based on clinical trials; therefore, results pertaining to the comparison of real-world economic outcomes in a clinical setting need further assessment.

Like any retrospective claims study, our study has several limitations. Propensity score methods do not address confounding in unobserved characteristics such as unmeasured health status and socioeconomic status. Other measures, such as AF duration, pre-index oral anticoagulation use, over-the-counter medication use, and patient health behavior could not be measured; therefore, these estimates may be biased due to residual confounding. Baseline characteristics were captured for 12 months prior to and during the index hospitalization through inpatient settings; therefore, the comorbidity indices and frequency of comorbid conditions may be under-estimated. In addition, as is the case for claims-based studies, there may be coding errors or diagnoses entered for administrative processing rather than clinical completeness.

The Premier PCD contains information from a large number of hospitals across the US; however, the results may not be generalizable to the entire US population of NVAF patients. The Premier PCD represents only hospital costs and does not represent medical care received outside a hospital facility. In addition, re-admissions recorded in the database are only those occurring at the same hospital in which the index hospitalization took place; therefore, re-admission rates may be under-estimated. Furthermore, the identification period in this study started in January 2013 at the same time as apixaban entered the market, which resulted in a smaller sample size and shorter follow-up period for apixaban patients. As an early-view analysis, the results may evolve over time.

Conclusion

Results of this early view study indicate that NVAF patients treated with apixaban had significantly shorter hospital LOS and lower index hospitalization costs compared to those treated with warfarin. These findings demonstrate that apixaban treatment can lower costs compared to the standard treatment, and may have important clinical and economic implications for hospitals, payers, and patients.

Transparency

Declaration of funding

Study funding provided by Pfizer Inc. and Bristol-Myers Squibb Company.

Declaration of financial/other relationships

JM, WT, and JT are employees of Pfizer Inc. KP was an employee of Pfizer Inc. at the time of the study. LV and PS are employees of Bristol-Myers Squibb Company. AB and JK were employees of Bristol-Myers Squibb Company at the time of the study. LX, AK, and OB are employees of STATinMED Research, which received funding from Pfizer Inc. and Bristol-Myers Squibb Company to conduct this study and develop this manuscript. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

Statistical programming for this study was provided by Juan Du, MS, of STATinMED Research and funded by Pfizer Inc. and Bristol-Myers Squibb Company.