Use of warfarin in long-term care: a systematic review

AF was the primary indication for warfarin therapy in two studies (n = 618), accounting for 74% [24] and 58% [25] of residents receiving warfarin, respectively. Stroke prevention without AF accounted for 12% [25], while DVT accounted for 20% [24], PE for 6% [24], and either DVT or PE for 26% [25] of residents receiving warfarin.

Eight studies (n = 22,573) [7‐13, 19] reported rates of warfarin use across all residents having a condition for which warfarin was indicated. These only included residents having AF or previous stroke; no other indications were described. Four of these studies (n = 2396) [7, 10, 12, 13] also reported the rates of use for warfarin combined with antiplatelet therapy. Rates for warfarin use alone ranged from 17% [9] to 57% [10], while the rate of either warfarin or antiplatelet use ranged from 62% [13] to 80% [10].

Of the eight studies above, five (n = 10,308) [7‐11] reported rates of warfarin use among LTC residents with AF that ranged from 17% [9] to 57%[10]. Abdel-Latif and colleagues [7] found that 46% of 117 residents with AF in six LTC facilities had been prescribed warfarin; 79% received either warfarin, aspirin, or clopidogrel. In a study of 265 LTC residents with AF in Canada from 2001-2002, Lau et al.[10] found that warfarin was prescribed for 57%; among residents who were considered optimal candidates (high risk of stroke and low risk of bleeding according to the criteria used in this study [10]), the warfarin prescribing rate was 60%. Using 1993 data, Lackner et al. [9], in a study of five LTC facilities, found that only 17% of patients with non-valvular AF received warfarin; among residents with AF with ≥ 1 additional risk factor for stroke (besides AF) and no contraindication to warfarin use, only 20% received warfarin [9]. Gurwitz et al. [8], using data from 1993 to 1995, found rates of warfarin use of 32% in 413 residents with AF. Finally, in a study of 21 LTC facilities from 1997 to 1998, McCormick et al. [11] reported that 42% of residents with AF received warfarin, and only 53% of ideal AF candidates for warfarin therapy (those having no bleeding risk factors) received it.

Three studies measured the use of warfarin in LTC facilities among stroke survivors (Table 1 Part A) [12, 13, 19]. Lapane et al. [19] evaluated, among stroke survivors, whether introduction of a prospective payment system (PPS) that required nursing homes to bear the cost for warfarin monitoring had shifted utilization from warfarin to antiplatelet agents. Comparing data from 1997 and 2000, Lapane et al. [19] found that the use of warfarin for the secondary prevention of ischemic stroke did not change significantly following the introduction of the PPS, from 22% in 1997 to 23% in 2000, while antiplatelet use did increase from 41% to 48% over the same period (likely due to the introduction of clopidogrel). Sloane et al. [13] examined 1997-1998 data from residential care/assisted living facilities in the United States and found that 38% of the 435 residents with a history of stroke (type not specified) received neither warfarin nor an antiplatelet agent. Quilliam et al. [12] analyzed records of 53,829 survivors of stroke (either ischemic or hemorrhagic) in all nursing homes in five states from 1992 to 1995 (SAGE database) and found that 67% were not receiving warfarin or any antiplatelet medication for stroke prevention; among residents recently hospitalized for ischemic stroke, 52% did not receive either of these agents.

Among residents with AF in LTC facilities, both Abdel-Latif et al. [7] (OR = 4.93, 95% confidence interval [CI] = 2.11-11.49 [correction provided by these authors]) and Gurwitz et al. [8] (OR = 1.87, 95% CI = 1.20-2.91), found a positive association between having a history of stroke and receiving warfarin. However, Lau et al. [10] found no significant association between "previous stroke or transient ischemic attack" and warfarin in residents with AF.

Both Quilliam et al. [12] (OR = 2.04, 95% CI = 1.95-2.14) and Hughes et al. [18] (OR = 1.76, 95% CI = 1.50-2.06) found that stroke survivors with AF were twice as likely to receive or be initiated on warfarin or antiplatelet therapy, compared with stroke survivors without AF.

Although both Lau et al. [10] and McCormick et al. [11] evaluated the association between degree of overall stroke risk and use of warfarin in LTC residents with AF, neither study found a significant association. In one study, [12] coronary artery disease was significantly associated (OR = 1.06, 95% CI = 1.02-1.11) with use of warfarin or antiplatelets among stroke survivors, but in another study [10] it was not associated with use of warfarin in residents with AF. While two studies found no significant association between hypertension and use of warfarin in residents with AF [7, 10], two other studies found that in stroke survivors hypertension was positively associated with initiation [18] (OR = 1.23, 95% CI = 1.09-1.39) and use [12] (OR = 1.27, 95% CI = 1.22-1.32) of warfarin or antiplatelets.

Both Hughes et al. [18] (OR = 1.22, 95% CI = 1.02-1.46) and Quilliam et al. [12] (OR = 1.11, 95% CI = 1.05-1.18) found that stroke survivors with depression were more likely to initiate or receive warfarin or antiplatelet therapy.

Both Gurwitz et al. [8], studying residents with AF (OR = 0.46, 95% CI = 0.22-0.94), and Quilliam et al. [12], studying stroke survivors (OR = 0.86, 95% CI = 0.82-0.91), found that residents ≥ 85 years were less likely to be prescribed either warfarin [8] or warfarin or antiplatelets [12]. However studies of residents with AF [10] and stroke survivors [13, 18], respectively, found no significant correlation between age ≥ 85 years and use of these agents. Hughes et al. [18] evaluated stroke survivors in their first year after nursing home admission: residents ≥ 85 years were most likely to discontinue warfarin or antiplatelet therapy (OR = 1.23, 95% CI = 1.05-1.43).

Gurwitz et al. [8] reported that AF residents with dementia were less likely than those without it to receive warfarin therapy (OR = 0.59, 95% CI = 0.38-0.90). Quilliam et al. [12] further found that among residents who were stroke survivors, those with Alzheimer's disease (OR = 0.77, 95% CI = 0.70-0.85) or dementia (OR = 0.84, 95% CI = 0.80-0.88) were less likely to receive warfarin or antiplatelets; the likelihood of anticoagulant or antiplatelet therapy also decreased among residents with moderate (OR = 0.93, 95% CI = 0.88-0.97) or severe (OR = 0.63, 95% CI = 0.60-0.67) cognitive impairment [12]. Hughes et al. [18] also found that stroke survivors with severe cognitive impairment were less likely to initiate such therapy (OR = 0.64, 95% CI = 0.52-0.80). In contrast, Sloane et al. [13] found no significant association between severe cognitive impairment and use of warfarin or antiplatelets in stroke survivors.

Both Quilliam et al. [12] (OR = 0.69, 95% CI = 0.64-0.75) and Hughes et al. [18] (OR = 0.73, 95% CI = 0.56-0.96) found that stroke survivors who had dependent physical function were less likely to receive warfarin or antiplatelets than those with independent function. However, Sloane et al. [13] found no significant association between physical dependency and receiving these agents among stroke survivors, and Gurwitz et al. [8] found no significant association between severe physical impairment and the use of warfarin in LTC residents with AF.

Abdel Latif et al. [7] found a negative association (OR = 0.18, 95% CI = 0.03-0.91) between previous gastrointestinal (GI) bleed and warfarin use in residents with AF. Quilliam et al. [12] also found a negative relationship (OR = 0.57, 95% CI = 0.52-0.62) between GI bleeding and warfarin or antiplatelet use in stroke survivors. In that study, Quilliam et al. [12] also found a negative association (OR = 0.64, 95% CI = 0.58-0.71) between peptic ulcer disease and these agents, although Lau et al. [10] found no significant association between peptic ulcer disease and warfarin use in AF. Evaluating bleeding risk factors among AF residents, Lau et al. [10] observed no significant association between warfarin treatment and overall bleeding risk or any single risk factor for bleeding, with the exception of recent surgery. However, McCormick et al. [11] reported that the odds of warfarin treatment were significantly lower (OR = 0.51, 95% CI = 0.29-0.94) for residents with AF and ≥ 2 bleeding risk factors.

Among residents with a history of stroke, four studies noted an association between race/ethnicity and being prescribed warfarin or antiplatelet therapy. Christian et al. [17] found that non-Hispanic blacks with a recent hospitalization for ischemic stroke and an indication for warfarin received warfarin less often (7.6% lower rate) than non-Hispanic white residents. In a study by Hughes et al. [18], in the year after nursing home admission, non-Hispanic black stroke survivors were less likely than non-Hispanic whites (OR = 0.62, 95% CI = 0.49-0.78) to be initiated on warfarin or antiplatelet therapy. Quilliam et al. [12] found that black stroke survivors were less likely than whites (OR = 0.80, 95% CI = 0.75-0.85) to receive warfarin or any antiplatelets. Although Abdel-Latif et al. [7] reported that non-Caucasian stroke survivors with AF were less likely than Caucasians to be prescribed warfarin therapy, these findings appear to be bivariate and were not significant in their multivariate model (OR was not reported).

For residents within residential care/assisted living facilities, Sloane et al. [13] explored the association between resident and facility characteristics and warfarin or antiplatelet prescribing for stroke survivors. Although bivariate analysis found that several facility characteristics, including larger facilities and those with registered nurses or licensed practical nurses, were associated with non-prescribing, multivariate analysis found no independent association for resident or facility characteristics. After adjusting for facility and resident characteristics, Hughes et al. [18] found that LTC facilities with white-only residents (OR = 0.74, 95% CI = NA) or the presence of full-time physicians (OR = 0.76, 95% CI = 0.63-0.93) were less likely, and that those that were part of a chain (OR = 1.20, 95% CI = 1.01-1.42), were more likely, to initiate warfarin or antiplatelets in stroke survivors. Moreover, facilities with > 10% non-white residents (OR = 1.22, 95% CI = 1.03-1.43), in an urban location (OR = 1.38, 95% CI = 1.16-1.65), having physician extenders on staff (OR = 1.21, 95% CI = 1.0-1.47), or having special care units (OR = 1.33, 95% CI = 1.02-1.73) were more likely to discontinue active warfarin or antiplatelet therapy.

Quality assessment ratings are listed for each of the 14 studies in Table 1. Methodological quality of each study was adequate for all but four areas. One or more limitations or concerns were noted for nine studies [7, 9, 11, 13, 17, 18, 21, 24, 25]. These included failing to adequately account for potential confounders, failing to consistently describe statistical error for the point estimates reported or to adequately model the relationship between stroke and bleeding risk with warfarin use. In the last case, typical problems noted were either omitting risk factors or else failing to score risks using a validated scoring method or consensus guidelines. Four studies [7, 9, 17, 21] had small sample sizes: subject counts for these studies would not have been able to support a finding of 95% significance for an OR of 1.5. Four studies [17‐19, 24] had limitations in length of follow-up.

Although the three physician surveys described above were innovative, their methodological quality was somewhat low in critical areas. While all three studies introduced new self-administered questionnaires, only one analyzed data regarding non-responders [28], while two [27, 28] did not mention steps taken to develop content validity. None of the surveys provided details regarding analysis of item response for the purpose of assessing construct validity. Finally, all three studies had limitations with analysis: failure either to adequately report statistical error or to cross-tabulate responses with subject attributes.

The quality of warfarin management and monitoring (Table 4 Part A) was measured in seven studies (n = 10,718) [8, 9, 11, 22, 25, 30, 31], six of which (n = 10,681) [8, 9, 11, 22, 25, 30] evaluated the percentage of time or person-days spent by residents in therapeutic range, based on INR results. In these six studies, INR levels were in the target range (2.0-3.0) for a mean 48% of the time (range = 37-55%). Four of these studies [8, 11, 22, 25] also reported percentage of time or person-days in subtherapeutic or supratherapeutic ranges. INR levels were in the subtherapeutic range (< 2.0) for a mean of 38% of the time (range = 35-45%) and in the supratherapeutic range (> 3.0) for a mean of 14% of the time (range = 11-16%). One study, Karki et al. [31], reported only mean INRs for residents and only for those whose INR values indicated either adequate or poor INR control.

Two studies that evaluated the effectiveness of warfarin medication management systems [23, 24] (n = 175, Table 4 Part B) found post-interventional time in therapeutic range considerably higher than reported in the six studies above. Papaioannou et al. [24] studied the effects of an electronic decision-support system for warfarin dosing and found that time in therapeutic range was 65% (pre-intervention) to 69% (post-intervention; the post-interventional change was non-significant). In a study of warfarin management by nurse practitioners, Allen et al. [23] measured 85% of INR draws within therapeutic range. These measurements were post-intervention only; no pre-intervention or control results were reported.

Three studies [22, 30, 31] (Table 4 Part A) evaluated the association of resident characteristics with INR (n = 302). Aspinall et al. [30] noted that residents who began warfarin therapy before the study period and those who did not have a history of stroke were more likely to maintain INR values within the therapeutic range. Karki et al. [31] observed that a group of LTC residents with INR fluctuations > 0.99 between blood samplings received more medications known to interact with warfarin vs a group with INR fluctuations ranging from 0.5-0.99. Verhovsek et al. [22] found that residents receiving interacting medications spent less time in therapeutic range. Time in therapeutic range varied significantly among prescribing physicians [22].

The three studies cited above [22, 30, 31] (Table 4 Part A) also evaluated the use of potentially interacting medications among residents receiving warfarin for any condition (n = 302). Aspinall et al. [30] reported that 89% of residents were receiving a medication that potentially interacts with warfarin. The most common agents were omeprazole (51% of residents), simvastatin (45%), aspirin (34%), citalopram (18%), and levothyroxine (13%). Karki et al. [31] reported that 46% of LTC residents were prescribed a potentially interacting agent during the study period. Verhovsek et al. [22] found that 79% of residents were prescribed more than one warfarin-interacting medication during a 12-month period; most common were acetaminophen (40% of residents), citalopram (25%), aspirin (16%), diltiazem (11%), and simvastatin (10%).

Overall, in the nine warfarin management and monitoring studies described above (Table 4 Parts A and B), methodological quality was low to moderate. Low power was common: five studies [9, 22, 23, 30, 31] had an insufficient number of subjects to find 95% significance for an odds ratio of 1.5. Four studies [22‐24, 30] had either inadequate descriptions of resident follow-up or insufficient follow-up periods to achieve the stated study purpose. Seven studies [9, 11, 22‐25, 31] had deficiencies either in adequately analyzing or reporting study findings or in accounting for potential confounders.

Except for limitations in one study, which neither adequately controlled for confounding nor explored the association of poor INR control with warfarin-related adverse events or resident characteristics [25], methodological quality for these two adverse event studies was acceptable.