Benefits and Harms of Standard Versus Reduced‐Dose Direct Oral Anticoagulant Therapy for Older Adults With Multiple Morbidities and Atrial Fibrillation

Abstract

Background

Dose reduction of direct oral anticoagulant (DOAC) medications is inconsistently applied to older adults with multiple morbidities, potentially due to perceived harms and unknown benefits of standard dosing.

Methods and Results

Using 2013 to 2017 US Medicare claims linked to Minimum Data Set records, we conducted a retrospective cohort study. We identified DOAC initiators (apixaban, dabigatran, rivaroxaban) aged ≥65 years with nonvalvular atrial fibrillation residing in a nursing home. We estimated inverse‐probability of treatment weights for DOAC dose using propensity scores. We examined safety (hospitalization for major bleeding) and effectiveness outcomes (all‐cause mortality, thrombosis [myocardial infarction, stroke, systemic embolism, venous thromboembolism]). We estimated hazard ratios (HRs) and 95% CIs using cause‐specific hazard‐regression models. Of 21 878 DOAC initiators, 48% received reduced dosing. The mean age of residents was 82.0 years, 66% were female, and 31% had moderate/severe cognitive impairment. After estimating inverse‐probability of treatment weights, standard dosing was associated with a higher rate of bleeding (HR, 1.18 [95% CI, 1.03–1.37]; 9.4 versus 8.0 events per 100 person‐years). Standard‐dose therapy was associated with the highest rates of bleeding among those aged >80 years (9.1 versus 6.7 events per 100 person‐years) and with a body mass index <30 kg/m² (9.4 versus 7.4 events per 100 person‐years). There was no association of dosing with mortality (HR, 0.99 [95% CI, 0.96–1.06]) or thrombotic events (HR, 1.16 [95% CI, 0.96–1.41]).

Conclusions

In this nationwide study of nursing home residents with nonvalvular atrial fibrillation, we found a higher rate of bleeding and little difference in effectiveness of standard versus reduced‐dose DOAC treatment. Our results support the use of reduced‐dose DOACs for many older adults with multiple morbidities.

Nonstandard Abbreviations and Acronyms

DOAC

direct oral anticoagulant

IPTW

inverse‐probability of treatment weights

MDS

Minimum Data Set

NH

nursing home

NVAF

nonvalvular atrial fibrillation

PY

person‐years

Clinical Perspective.

What Is New?

• In this nationwide cohort study cohort of 21 878 older nursing home residents with nonvalvular atrial fibrillation, standard‐ versus reduced‐dose direct oral anticoagulant therapy was not associated with reductions in mortality or thrombotic events.

• Those on standard dosing experienced 1.4 more major bleeds per 100 person‐years than those on reduced dosing, with highest relative bleeding rates among those aged ≥80 years or who were not obese.

• Over the 1‐year follow‐up, the overall rate of major bleeds in this population was 8.6 per 100 person‐years and 5.7 per 100 PY for thrombotic events.

What Are the Clinical Implications?

• Older adults with multiple morbidities and nonvalvular atrial fibrillation in nursing homes are at very high risk of both major bleeds and thrombotic events.

• Given the potential harms and unclear benefits of standard direct oral anticoagulant dosing, our results support the use of reduced‐dose direct oral anticoagulants for older adults with multiple morbidities.

Direct oral anticoagulants (DOACs) can prevent serious thrombotic events like stroke in residents with nonvalvular atrial fibrillation (NVAF) ¹ , ² and have a lower risk of major bleeding events than warfarin. ¹ , ² , ³ , ⁴ , ⁵ , ⁶ However, older adults with multiple morbidities who have a high risk of both stroke and major bleeding were not well represented in the pivotal DOAC trials. ³ , ⁵ , ⁶ The ELDERCARE‐AF (Edoxaban Low‐Dose for Elder Care AF Residents) trial showed that, compared with placebo, low‐dose DOAC therapy with edoxaban was superior for preventing stroke or systemic embolism in older adults with multiple morbidities without significantly increasing risk of major bleeding. ⁷ Guidelines have since recommended against withholding anticoagulation for those with NVAF solely because of age or the risk of falls. ⁸ However, additional studies of the benefits and harms of DOAC therapy in older adults with multiple morbidities are needed. ⁹

The optimal dosing regimen of DOACs remains a major question for clinicians given the high risk of bleeding in this population. Dose reduction is recommended for DOAC therapy based on specific resident characteristics that increase bleeding risk (eg, renal impairment, low body weight, or age ≥80 years). ¹⁰ However, clinicians often prescribe reduced‐dose DOACs to older adults with multiple morbidities and NVAF who do not meet dose‐reduction criteria but are at a high risk of bleeding, especially nursing home (NH) residents. Whereas about 25% of community‐dwelling older adults have misaligned dosing according to currently approved labeling, ¹¹ this proportion is as high as 56% for NH residents. ¹² NH residents are thus an ideal real‐world population in which to answer questions on the benefits and risk of DOAC dosing strategies for older adults with multiple morbidities.

The objective of this study was to examine the association of standard versus reduced DOAC dose on major bleeding events, thrombotic events, and mortality among a national population of NH residents. In addition, we examined whether the benefits and harms of dosing strategies differed by age, stroke risk, or body mass index (BMI).

METHODS

Study Design and Data Sources

We conducted a new‐user, active comparator cohort study that included all Medicare Fee‐for‐Service beneficiaries who were long‐stay NH residents (>100 days in a facility) between 2013 and 2017. We leveraged national Medicare claims and Medicare Provider Analysis and Review data linked to Minimum Data Set (MDS) 3.0 claims and facility data (Certification and Survey Provider Enhanced Reports). The MDS is a federally mandated clinical assessment conducted at the time of NH admission and quarterly thereafter that contains more than 400 items, including cognition and function. ¹³ Medicare Provider Analysis and Review data contain claims for inpatient and short‐stay NH stays. Prescription drug information was obtained via Medicare Part D claims. Linked Medicare enrollment files provided information on resident demographics and enrollment history. Certification and Survey Provider Enhanced Reports data were used for facility‐level information. Brown University's institutional review board approved the study and waived the requirement for informed consent (Protocol #2009002821). No person‐level and limited aggregate‐level data can be shared by our team directly per the terms of the Data Use Agreements established by the Centers for Medicare and Medicaid Services. Statistical code used to generate results from this study is available from the corresponding author upon reasonable request. We used the Strengthening the Reporting of Observational Studies in Epidemiology cohort checklist when writing our report. ¹⁴

Study Population

We identified all long‐stay NH residents aged 65 years or older at their first Part D claim for a DOAC (apixaban, dabigatran, or rivaroxaban) between January 1, 2013, and December 31, 2017. Betrixaban and edoxaban were not included due to low use (<1% of residents initiating a DOAC). The study design is depicted in Figure 1. The date of first DOAC dispensing was the index date. We then required residents to have a diagnosis of NVAF within 120 days before or 14 days after (inclusive of the DOAC prescription date). AF was identified using the principal and secondary diagnostic codes in an inpatient or carrier claim (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD‐9‐CM] codes 427.31 and 427.32; ICD, Tenth Revision, CM [ICD‐10‐CM] codes I48.x). We excluded residents with a previous DOAC claim or without continuous coverage with Medicare Part A, B, and D in the 120 days before index. We also excluded residents with active cancer; a recent deep vein thrombosis or venous thromboembolism; a mechanical/prosthetic heart valve or valvular disease; or hip or knee arthroplasty (all measured within the 120 days before the index date). Residents were not excluded if they had a claim for warfarin before DOAC initiation. Finally, we excluded <1% of those without documentation of height, weight, and cognitive/physical function score in all MDS assessments in the 120 days before the index date or who were missing NH facility information. Detailed exclusion criteria are available in Table S1.

Figure 1. Study design diagram.

*Select covariates (eg, stroke history) were measured in all available lookback to maximize sensitivity. DOAC indicates direct oral anticoagulant (apixaban, dabigatran, rivaroxaban); and NVAF, nonvalvular atrial fibrillation.

Exposures and Contrast of Interest

Because DOAC dosing recommendations are not strictly followed in routine practice, ¹² we defined standard or reduced DOAC dose based on the daily dose of the first dispensing. We calculated the mean daily DOAC dose by dividing the total dose (DOAC strength in mg multiplied by the units dispensed) by the days supply. The thresholds used for reduced daily dose were apixaban (<10 mg), dabigatran (<300 mg), and rivaroxaban (<20 mg). Thus, reduced dosing was defined as a daily dose less than the standard recommended dose for AF regardless of whether labeling recommendations on dose reduction were applied to specific patients based (eg, based on renal dysfunction). The contrast of interest for outcomes was initiation of standard versus reduced dose DOAC therapy, and the average treatment effect was the estimand of interest. ¹⁵

Outcomes and Follow‐Up

Outcomes were captured via Medicare Part A hospitalization claims. The primary safety outcome was a hospitalization for major intracranial or extracranial bleeding. Our primary effectiveness outcomes were mortality and a composite measure comprising hospitalization for thrombotic events: acute myocardial infarction, systemic embolism, venous thromboembolism, ischemic stroke, and transient ischemic attack. Outcome definitions are presented in Table S2. All outcome definitions have positive predictive values >80%. ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ We followed residents from the date of first DOAC use until the first occurrence of an outcome of interest or censoring (mortality; disenrollment from Medicare Part A; end of the study period [December 31, 2017]; or 365 days from the index date).

Covariates

We used the MDS in combination with Medicare enrollment files, MEDPAR claims, and Certification and Survey Provider Enhanced Reports data to obtain covariate information on resident demographics, medical conditions, functioning, and medication use. We defined 93 covariates that might be strongly related to the risk of bleeding, thrombosis, mortality, and the likelihood of DOAC dose reduction based on product recommendations, ¹⁰ previous literature, ¹² , ²⁵ and general subject matter knowledge of our team. In brief, covariates included stroke risk (via calculated CHA₂DS₂‐VASC score ²⁶ ), bleeding risk (via calculated Anticoagulation and Risk Factors in Atrial Fibrillation score), comorbidities, demographics, BMI, functional status, cognitive status, and medication use (eg, history of warfarin use). In addition, we calculated the Combined Comorbidity Score for each resident using ICD diagnosis codes and MDS diagnoses, ²⁷ , ²⁸ and we used Certification and Survey Provider Enhanced Reports data for facility characteristics (eg, number of beds, ownership). We used a 120‐day lookback window from the index date to measure covariates, except for select comorbidities (eg, stroke) for which we used all available lookback information from initial Medicare Part A enrollment to maximize sensitivity (Table S3). ²⁹

Statistical Analysis

We calculated propensity scores for standard (versus reduced) dosing for the index DOAC prescription using a logistic regression model containing the aforementioned covariates. A complete list of variables used to construct propensity score models, with details on their measurement, is presented in Table S3. Using propensity scores, we created stabilized inverse probability of treatment weights (IPTW) for each resident based on their propensity score and the dosing regimen they received. These IPTW effectively weight the study population such that the exposure (DOAC dose) becomes independent of all measured covariates, reducing bias from confounding. ³⁰ , ³¹ In addition, we calculated stabilized inverse probability of censoring weights using the same methods and variables as IPTW to account for potential selection bias from differential loss to follow up between exposure groups. ³² , ³³ We created 1 combined weight (the inverse probability weight) for each individual by multiplying the IPTW*inverse probability of censoring weights; the inverse probability weight was the final weight used in all analyses. No truncation of weights was conducted. We assessed balance in measured covariates before and after weighting using standardized mean differences, <0.10 suggesting a negligible difference between exposure groups. ³⁴

We estimated crude and weighted incidence rates of each event per 100 person‐years of follow‐up time. Then, we used weighted Cox proportional hazards models to estimate the hazard ratio (HR) for each outcome for standard versus low DOAC dose, using cause‐specific models to account for the competing risk of mortality. ³⁵ Robust 95% CIs were calculated for HRs using the sandwich estimator. ³⁶ , ³⁷ All analyses were completed using SAS version 9.4 (SAS Institute, Cary, NC) and Stata version 17 (StataCorp, College Station, TX).

Subgroup Analyses

Because resident characteristics might modify the association of DOAC dosing and the outcomes of interest, we estimated associations by subgroups defined by stroke risk (CHA₂DS₂‐VASC score≤4 versus >4), age (≤80 versus >80 years), and obesity status (ie, BMI >30 versus ≤30 kg/m²) after reestimating the inverse probability weight. All analyses considered effect measure modification on the multiplicative scale.

Sensitivity and Stability Analyses

We conducted several sensitivity and stability analyses to assess the robustness of our findings. First, we then calculated an E‐value to investigate the impact of potential unmeasured confounding. ³⁸ , ³⁹ Second, because we hypothesized that prescribers specializing in cardiology may be more likely to prescribe standard dose DOAC treatment, we constructed inverse probability weight that contained an indicator for prescriber type (cardiologist, generalist [eg, family medicine, internist, geriatrician], other) for residents where provider information was available. Third, we conducted an as‐treated analysis where we censored patients if they switched between dosing regimens (eg, from standard dosing to reduced dosing). Next, because some residents may have a near‐zero probability of receiving standard DOAC doses (eg, recent major bleeding event), we reconstructed IPTW to obtain the average treatment effect only among those who were treated with standard dosing (average treatment effect among the treated). ³¹ Finally, we conducted a fourth subgroup analysis to explore whether outcomes differed by documentation of renal impairment within the MDS.

RESULTS

Study Population

We identified 32 540 long‐stay NH residents with a DOAC prescription and NVAF diagnosis. A detailed exclusion flow diagram is available in Figure S1. After exclusions, 21 878 residents remained (48% reduced DOAC dose; 66% female; mean age 82.0 [SD 7.9] years; 48% apixaban; 44% rivaroxaban; 8% dabigatran). Table S4 presents the average daily doses for initial prescriptions by specific DOAC prescribed. Before weighting, residents who received reduced dose DOAC therapy were older and had a lower BMI, higher Anticoagulation and Risk Factors in Atrial Fibrillation score, and a greater prevalence of renal disease (Table S5). The proportion of residents with prior warfarin use was not different between dosing groups. After weighting, all measured characteristics were balanced between DOAC dose groups (Table 1; Table S6 presents all measured characteristics). ⁴⁰ , ⁴¹ , ⁴² , ⁴³ The mean IPTW was 1.00 (SD 0.38; first percentile: 0.58, 99th percentile: 2.5), and the mean inverse probability of censoring weights was 1.00 (SD 0.73; first percentile: 0.46, 99th percentile: 4.1). Around 10% of residents in both dosing groups had only 1 dispensing of a DOAC.

Table .

Characteristics of Residents Using Standard Dose vs Reduced Dose Direct Oral Anticoagulants After Inverse Probability Weighting^*

	DOAC dose categories (N=21 878)		aSMD†
	Reduced dose, n=10 514	Standard dose, n=11 364
Age, y (mean [SD])	82.3 (8.0)	81.7 (7.8)	0.08
Age >80 y	60.3%	60.6%	0.01
Female sex	66.0%	66.8%	0.02
Race or ethnicity			0.05
White	83.0%	83.5%
Black	11.8%	11.3%
Other	5.2%	5.3%
Facility is in urban area	74.8%	74.7%	0.00
Facility has >120 beds	47.8%	48.6%	0.01
Facility owner			0.02
For profit	75.8%	75.0%
Nonprofit or government owned	24.2%	25.0%
Year of DOAC initiation			0.00
2013	4.8%	4.6%
2014	12.8%	13.1%
2015	18.0%	18.1%
2016	30.5%	30.5%
2017	33.9%	33.9%
Body weight <60 kg	22.4%	22.9%	0.01
Body mass index category			0.00
Underweight	4.7%	4.9%
Normal	34.2%	34.1%
Overweight	27.1%	27.7%
Obese	33.9%	33.3%
Anticoagulation and Risk Factors in Atrial Fibrillation score (mean [SD])‡	2.5 (1.6)	2.5 (1.6)	0.03
CHA₂DS₂‐VASc score (mean [SD])§	5.3 (1.5)	5.3 (1.5)	0.00
Activities of daily living score categoryǁ			0.07
Independent	5.2%	5.6%
Limited assistance	20.6%	21.1%
Extensive assistance	57.5%	57.6%
Dependent	16.7%	15.8%
Days between long‐stay date and DOAC index date (mean [SD])	329.2 (369.7)	321.3 (365.8)	0.04
Combined comorbidity score 27 , 28 (mean [SD])	6.9 (3.1)	6.8 (3.1)	0.02
Number of unique medications			0.00
≤5	25.5%	26.3%
6–10	29.3%	28.1%
11 or more	45.1%	44.6%
Cognitive function score 40 , ¶			0.00
Cognitively intact	42.6%	42.9%
Mildly impaired	25.0%	25.0%
Moderately impaired	25.6%	26.1%
Severely impaired	6.9%	6.0%
Changes in health, End‐stage disease and Signs and Symptoms score, 41 (mean [SD])#	0.89 (0.88)	0.87 (0.87)	0.02
Emergency department visit in the prior 4 mo	45.6%	44.1%	0.03
Hospitalizations in prior 4 mo			0.00
1	32.8%	33.4%
2+	19.0%	18.7%
Anemia	38.6%	39.6%	0.01
History of major bleeding**	29.8%	30.3%	0.01
Coronary artery disease	31.9%	32.3%	0.00
Diabetes	44.5%	43.2%	0.03
Fracture	11.9%	12.2%	0.01
Hip fracture	9.1%	8.7%	0.01
Acute myocardial infarction**	11.7%	12.1%	0.01
Heart failure**	13.7%	13.2%	0.01
Hypertension	91.4%	91.4%	0.00
History of stroke or transient ischemic attack**	35.1%	35.2%	0.00
Renal insufficiency/failure or end‐stage renal disease††	39.7%	39.3%	0.00
Dialysis	2.6%	2.5%	0.01
Medication use before DOAC initiation
Warfarin	34.2%	34.6%	0.01
Angiotensin‐converting enzyme inhibitors	21.5%	22.3%	0.02
Angiotensin receptor antagonists (blockers)	11.3%	11.4%	0.00
Antiplatelets	9.1%	9.1%	0.00
Beta‐adrenergic antagonists (blockers)	52.1%	51.8%	0.01
Histamine‐2 receptor antagonists	11.5%	11.6%	0.00
Nonsteroidal anti‐inflammatory drugs	8.0%	8.1%	0.01
Proton pump inhibitors	31.2%	31.1%	0.00
Statins	41.9%	42.3%	0.01
Specialty of DOAC prescriber‡‡			0.00
Cardiology	3.5%	3.5%
Generalist	74.0%	74.0%
Otherl§§	14.9%	14.8%
Missing	7.6%	7.7%

aSMD indicates absolute standardized mean difference; and DOAC, direct oral anticoagulant.

^*Weighted by combined inverse probability of treatment weights×inverse probability of censoring weights.

^† An aSMD ≤0.10 indicates little difference between exposure groups. ⁴³

^‡ Score ranges from 0 (lowest risk) to 10 (highest risk).

^§ A point system used to stratify stroke risk; scores range from 0 to 9, with 0=lowest stroke risk and 9=highest stroke risk.

^ǁ Scale ranges from 0 to 28 (higher scores indicate more impairment). ⁴²

^¶ Minimum Data Set 3.0 Cognitive Function Scale score. ⁴⁰ Ranges from 0 to 4, with 0=intact cognition and 4=severe cognitive impairment.

^# Predicts resident mortality and ranges from 0 (low risk) to 5 (high instability and high mortality risk).

^**Measured by all available claims.

^†† As documented in the Minimum Data Set item M3I1500: renal insufficiency, renal failure, or end‐stage renal disease.

^‡‡ For the 20 878 residents with this information available.

^§§ Asian, American Indian or Alaskan Native, Native Hawaiian or other Pacific Islander race, or Other race, AND Hispanic or Latino ethnicity.

Safety Outcome

In total, 1231 major bleeding events occurred during a mean (SD) follow‐up of 0.65 (SD 0.36) years (overall event rate 8.6 per 100 person‐years [PY]). After weighting, the HR for major bleeding events for those with standard‐ versus reduced‐dose therapy was 1.18 (95% CI, 1.03–1.37; weighted rates for standard versus reduced dose: 9.4 versus 8.0 events per 100 PY) (Figure 2). Detailed event rates, unadjusted analyses, and follow‐up time by treatment group are presented in Table S7. In subgroup analyses, standard dose therapy was associated with a higher rate of major bleeding among residents older than 80 years of age (HR, 1.37 [95% CI, 1.13–1.67]; 9.1 versus 6.7 bleeds per 100 PY) than those 80 years or younger, and with a BMI <30 versus ≥30 kg/m² (HR, 1.27 [95% CI, 1.06–1.51]); 9.4 versus 7.4 bleeds per 100 PY) (Figure 3). Other subgroup results were consistent with the primary results.

Figure 2. Association of standard‐ vs reduced‐dose direct oral anticoagulants on outcomes.

Detailed event rates, individuals at risk, and follow‐up times are presented in Table S6. IPTW indicates inverse probability treatment‐ and inverse probability censor‐weighted.

Figure 3. Association of standard‐ vs reduced‐dose direct oral anticoagulants for outcomes in subgroups.

Panels describe associations of direct oral anticoagulant dose with the following outcomes, respectively: (A) Major bleeding, (B) Mortality, and (C) Thrombotic events (effectiveness). Obese BMI defined as a BMI ≥30 kg/m². High stroke risk: CHA₂DS₂‐VASC score>4; CHA₂DS₂‐VASC is a point system used to stratify stroke risk; scores range from 0 to 9, with 0=lowest stroke risk and 9=highest stroke risk. Low stroke risk: CHA₂DS₂‐VASC score ≤4. BMI indicates body mass index; HR, hazard ratio; and IPTW, inverse probability treatment‐ and inverse probability censor‐weighted.

No major changes in results were identified in sensitivity or stability analyses (Tables S8 and S9), though the association between standard dosing and major bleeding was increased when censoring upon switch between DOAC dose (HR, 1.27 [95% CI, 1.09–1.49]). The E‐value for the CI for bleeding was 1.21, signifying that an unmeasured confounder would have to increase the relative risk of bleeding by 21% independently of all measured confounders and increase the probability of being prescribed reduced DOAC dosing by 21%.

Effectiveness Outcomes

Overall, 6150 residents died during an average of 0.67 (SD 0.36) years of follow‐up (overall event rate 41.8 per 100 PY). After weighting, standard dose therapy was not associated with all‐cause mortality (HR, 0.99 [95% CI, 0.92–1.06]; weighted rates for standard versus reduced dosing 45.1 versus 45.6 deaths per 100 PY) (Figure 2). For the composite effectiveness outcome (acute myocardial infarction, systemic embolism, stroke, transient ischemic attack, venous thromboembolism), there were 820 thrombotic events over a mean follow‐up of 0.67 (SD 0.34) years (overall event rate 5.7 per 100 PY). After weighting, there was no significant association between DOAC dose and thrombotic events (HR, 1.16 [95% CI, 0.96–1.41]; weighted rates for standard versus reduced dosing: 6.4 versus 5.5 events per 100 PY). Subgroup results for mortality outcomes were similar to the primary analysis (Figure 3). In contrast, for thrombotic risk, HRs were much closer to the null value when estimating the association of standard versus reduced dose for those <80 years of age, with a BMI <30 kg/m² or who had a high stroke risk (CHA₂DS₂‐VASc >4). HRs were higher and CIs much less precise for those with a BMI ≥30 kg/m², who were <80 years of age, or who had a calculated low stroke risk.

No meaningful changes in results were identified in sensitivity or stability analyses. The E‐values for the upper limit of the HR point estimate for mortality was 1.10, and the E‐value for the HR point estimate for thrombotic events was 1.59.

DISCUSSION

We conducted a large retrospective cohort study of frail older adults residing in NHs who initiated DOAC therapy for NVAF at standard or reduced doses. After accounting for measured confounding, we found an 18% increased relative risk of major bleeding associated with standard versus reduced dosing (1.5 more major bleeding events per 100 PY). Harms appeared greatest in older residents (>80 years) and residents with a BMI <30 kg/m², with 2.0 to 3.4 more major bleeds per 100 PY for those with standard dosing in these groups. No associations between mortality or thrombosis benefits were observed. Taken together, our results suggest that providers may consider the use of reduced‐dose DOACs for many NH residents and other frail older adults given the potential lack of benefit for thrombotic events and increased risk of bleeding. However, the results of our study should not be interpreted as encouraging further dose reductions among individuals who are already on doses lower than the current labeling‐recommended regimens.

The proportion of NH residents with NVAF who are treated with DOACs has increased more than 5‐fold from 2011 to 2016 (from 8.6% to 48.2%). ²⁵ , ⁴⁴ We found that bleeding events requiring hospitalization were more common (8.6 bleeding events per 100 PY) than thrombotic events (5.2 per 100 PY), regardless of DOAC dose. The high rate of thrombotic events in our study (5.7 events per 100 PY) confirms that many residents may benefit from anticoagulation, ⁷ particularly as thrombotic events have the potential to cause severe morbidity and mortality. However, bleeding rates in this study were even greater (8.6 events per 100 PY); bleeding events were 3 to 4 times higher than the rates observed in the landmark DOAC trials for apixaban and rivaroxaban (2.3–3.6 events per 100 PY). ³ , ⁶ Mortality was extremely high in our study of NH residents (41.8 events per 100 PY), and it is possible that some of these deaths were attributable to undiagnosed thrombotic events. Given the high mortality and bleeding risk in this population, standard‐dose DOAC treatment decisions must be grounded in shared decision‐making with the resident and caregivers.

Few data exist that directly compare the effects of standard versus reduced DOAC dosing, particularly in older adults with multiple morbidities who would not otherwise meet the criteria for dose reduction. DOAC dose‐adjustment criteria were derived based on pharmacokinetic studies that identified factors that may result in increased concentrations of drug in the bloodstream. ³ , ⁵ , ⁶ These criteria were used in the Phase III landmark trials to determine exposures in the DOAC arms. A post hoc analysis of the landmark trial of apixaban suggested that those who received standard dosing and had 1 dose‐reduction criteria had similar effects compared with warfarin as the apixaban arm as a whole. This post hoc analysis excluded subjects who had received dosages misaligned with labeling recommendations. With the greater use of DOACs, and these prior comparisons with warfarin, the present study is significant because it fills the gap on the real‐world benefits and harms of standard‐ versus reduced‐dose DOAC therapy outside of the bounds of pharmacokinetically derived dosing recommendations.

Limitations and Strengths

There were limitations to our study. First, although not significant, the unexpected increase in the risk of thrombotic events with standard versus reduced DOAC dosing paradoxically contrasts with the observed null effect on mortality. This suggests the possibility of residual confounding despite similar calculated CHA₂DS₂‐VASc scores and many other clinical factors between groups. Those who were prescribed a standard DOAC dose may have had more stroke risk factors that were not captured through our data. Nevertheless, there was no suggestion in any subgroup that standard dose versus reduced dose was associated was fewer thrombotic events, the primary indication for the medication. Second, we had no data on serum creatinine or other laboratory measures of renal function, such that we are unable to rule out residual confounding by renal function. Without this information, we also were unable to determine whether a resident's dose was aligned with prescribing recommendations, though the application of these dose reductions is known to be inconsistent among older adults in NHs. ²⁵ However, we did attempt to maximize the sensitivity of our renal impairment measure by including 2 indicators for renal impairment as measured through (1) the MDS or (2) ICD codes for chronic renal impairment. In addition, all measured characteristics, including functional assessments and other clinical assessments not typically measured in population‐based administrative data, were balanced between groups after IPTW. Results were also robust to several stability analyses. However, future studies might consider the use of electronic health record data to supplement claims in providing clinical measures like renal function for analyses.

Finally, this study was in NH residents, a subpopulation of older adults in the United States, and thus results may not be generalizable to all older adults with multiple morbidities in the community. Our population was younger than those in the ELDER‐AF trial (mean age: 82 versus 87 years), which has been used to justify the use of anticoagulation in frail older adults. ⁷ However, our NH population had a higher burden of multimorbidity and was at higher risk of stroke and bleeding than that of ELDER‐AF. Also, the frequency of use of edoxaban in the source population for our cohort was too low to include these patients, and thus results may not be generalizable to patients using edoxaban. Finally, due to study power limitations, we were unable to directly compare the risk of bleeding with standard therapy between DOACs, and each DOAC's bleeding risk profile likely differs. ⁴⁵ , ⁴⁶

CONCLUSIONS

In this large cohort study of NH residents with NVAF, we found little benefit of standard‐ versus reduced‐dose DOAC therapy. Bleeding events occurred more commonly than thrombotic events, and standard‐dose DOACs were associated with greater risk of bleeding. Our results suggest that for older adults with multiple morbidities or their caregivers who elect anticoagulation treatment, reduced‐dose DOAC therapy may offer a similar benefit with fewer harms, particularly for patients >80 years of age and those with a BMI <30 kg/m².

Sources of Funding

This research was funded by the National Institute on Aging (RF1AG061221 and R24AG064025). K.N.H. and A.R.Z. were also supported, in part, by other grants funded by the National Institute on Aging (R01AG078759 [K.N.H.]; R01AG065722, R01AG077620, R21AG061632 [A.R.Z.]). Dr Berry was also supported in part by National Institute on Aging grant K24 AG070106.

Disclosures

K.N. Hayes has received grant funding paid directly to Brown University for collaborative research from Insight Therapeutics, Sanofi, and Genentech for research on complex insulin regimens and influenza outbreak control. K.N. Hayes has also served as a consultant for the Canadian Agency for Drugs and Technologies in Health. A.R. Zullo received investigator‐initiated support from Sanofi Pasteur for research related to the epidemiology of infections and vaccinations in nursing homes. The remaining authors have no disclosures to report.

Supporting information

Data S1