- Email: [email protected]
Determinants of Compliance with Anticoagulation: A Case-Control Study
Determinants of Compliance with Anticoagulation: A Case-Control Study Julia H. Arnsten, MD, MPH, Joel M. Gelfand, BA, Daniel E. Singer, MD, Boston, Massachusetts
BACKGROUND: The number of patients for whom long-term anticoagulation is indicated has increased dramatically over the past decade. Good patient compliance is necessary to safely realize the benefits of anticoagulation, yet barriers to compliance with anticoagulation therapy have not been studied. METHODS: We conducted a case-control study in the Anticoagulation Therapy Unit (ATU) at Massachusetts General Hospital. Forty-three patients who had been discharged from the ATU for noncompliance (cases) and 89 randomly selected compliant ATU controls were interviewed. Noncompliant cases had selfdiscontinued warfarin or were taking warfarin with inadequate monitoring of international normalized ratio (INR) levels. Telephone interviews assessed sociodemographic features, indication for anticoagulation, patient satisfaction, and health beliefs. RESULTS: Noncompliant cases were more likely to be younger (mean 53.7 years versus 68.7 years, P õ0.0001), male (odds ratio [OR] 3.5, 95% confidence interval [CI] 1.5, 8.2), and nonwhite (OR 6.4, 95% CI 1.9, 21.9), and less likely to have had a stroke or transient ischemic attack (OR 0.2, 95% CI 0.1, 0.7). In open-ended questioning, cases were more likely to report that they did not know why warfarin had been prescribed (OR 4.4, 95% CI 1.4, 14.2). Noncompliant cases were more likely not to have a regular physician (OR 11.1, 95% CI 3.6, 50.0); among patients with a regular physician,
From the General Internal Medicine Unit, Massachusetts General Hospital, Boston Massachusetts (JHA, DES); Harvard Medical School, Boston, Massachusetts (JHA, JMG, DES); Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts (DES). The research reported in this paper was supported in part by a grant from the Eliot B. Shoolman Fund, Massachusetts General Hospital, Boston, Massachusetts. Dr. Arnsten was a fellow of the Harvard General Internal Medicine Fellowship Program, U.S. Department of Health and Human Services National Research Services Award Grant #5T32PE1100107. Portions of this research were presented at the 1995 Annual Meeting of the Society for General Internal Medicine, San Diego, CA, May 1995. Requests for reprints should be addressed to Julia H. Arnsten, MD, MPH, Centennial Building, 3rd floor, Montefiore Medical Center, Bronx, NY 10467. Manuscript submitted May 29, 1996 and accepted in revised form February 11, 1997.
noncompliant cases were more likely to feel dissatisfied. Examination of health beliefs revealed that noncompliant cases felt more burdened by taking warfarin, and perceived fewer health benefits. CONCLUSIONS: Patients who are noncompliant with warfarin share distinctive clinical characteristics. Notably, younger, male patients who have not experienced a thromboembolic event are more likely to forego INR testing or to stop anticoagulation therapy completely. Improved patient education, physician involvement, and ease of monitoring may improve compliance, particularly among younger male patients. Am J Med. 1997;103:11–17. Q 1997 by Excerpta Medica, Inc.
O
ver the past 10 years, the number of patients for whom long-term anticoagulation therapy is indicated has increased dramatically. Recent large randomized trials have demonstrated that anticoagulants prevent strokes in patients with nonrheumatic atrial fibrillation1 – 5 and reduce the risk of mortality and cardiovascular morbidity after acute myocardial infarction.6 – 8 However, anticoagulation is a demanding therapy that is underutilized in actual clinical practice. Despite a two-thirds reduction in the risk of stroke among anticoagulated patients, only a minority of patients with nonrheumatic atrial fibrillation are prescribed warfarin.9,10 One frequently stated concern among physicians is poor patient compliance with anticoagulants, especially among the elderly.11 In the randomized trials in atrial fibrillation, noncompliance with anticoagulant regimens ranged from 10% to 26%.1,2 In general populations, noncompliance is likely to be much more common. Good compliance with both medication dosage and blood test monitoring is necessary to safely realize the benefits of anticoagulation, yet barriers to compliance with anticoagulation therapy have not been studied. Building on our earlier work on the impact of warfarin on quality of life,12 we constructed a case-control study to identify determinants of noncompliance with anticoagulation therapy.
PATIENTS AND METHODS Study Design We conducted this study in the Anticoagulation Therapy Unit (ATU) at Massachusetts General Hos-
Q1997 by Excerpta Medica, Inc. All rights reserved.
/ 220f 5048 Mp
11
0002-9343/97/$17.00 PII S0002-9343(97)00048-X
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
11
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
pital, a large urban teaching hospital. Cases were patients who had demonstrated such severe noncompliance with chronic anticoagulation therapy that it was no longer considered safe. Controls were compliant patients attending the ATU. In-depth telephone surveys were completed with all patients. The study was approved by the Human Subjects Committee of the Institutional Review Board at Massachusetts General Hospital.
Setting The Massachusetts General Hospital ATU managed long-term anticoagulation for approximately 2,300 outpatients during the period of this study. The ATU is staffed by nurses who use a computer-based system to track patients’ prothrombin time tests and warfarin dose. All patients attend an introductory session, detailing the risks and benefits of anticoagulation and the importance of frequent blood test monitoring. Patients then use a convenient laboratory for ongoing prothrombin time tests, and are notified by ATU nurses of results, dosage changes, and timing of future tests. Patients who do not have at least one prothrombin time test per month receive reminder postcards and telephone calls, and ultimately are discharged from the ATU if they do not respond to these reminders. Patients discharged in this manner include the most extreme noncompliers with ATU regimens, as well as patients with whom the ATU has lost contact. These patients were the source of cases for this study. Patient Selection Cases. The criteria for selection of cases were: (1) standard indication for chronic anticoagulation; (2) discharge from the ATU for apparent failure to comply with ATU monitoring policies; and (3) patient report of self-discontinuation of warfarin therapy, or continuation of warfarin therapy without proper international normalized ratio (INR) monitoring. Between January 1991 and December 1994, 156 patients were discharged from the ATU for noncompliance. Between July 1994 and January 1995, we attempted to complete telephone interviews with these 156 patients. Sixty-one patients could not be located after six attempts, using home and office phone numbers listed in the ATU database, and contacting directory information in the region of the patient’s last known address. We were able to locate 95 patients (61%) and complete interviews with 69 patients (44%). The reasons for incomplete interviews with located patients were death (n Å 6), lack of an English-speaking interpreter (n Å 12), patient refusal (n Å 6), and severe cognitive impairment (n Å 2). 12
Among the 69 patients we interviewed who were discharged from the ATU because of apparent noncompliance, we found that many were not truly noncompliant, but were simply no longer in contact with the ATU. Twenty-six patients (38%) reported that they had been directed to discontinue warfarin by their primary physician, or were still taking warfarin but no longer used the ATU system. These patients were not included in the study. Forty-three patients (62%) reported that they had self-discontinued warfarin therapy against medical advice, or were continuing to take warfarin with improper monitoring. Improper monitoring was defined as prothrombin time tests less frequently than once every 2 months. These 43 patients constituted our cases of severe noncompliance with anticoagulation therapy. Although this is an unusually low proportion of noncompliant patients (43 of 2,300, or 1.9%), discharge from the ATU is a last resort, resulting from far more stringent and less inclusive criteria for noncompliance than were used in the randomized trials of anticoagulation. Controls. The criteria for selection of controls were: (1) active enrollment in the ATU in June 1994, and (2) standard indication for chronic anticoagulation. Using a case to control ratio of 1:3 and a total of 43 cases, we randomly selected 129 controls from among 2,308 eligible patients. We located 99 patients (77%) and completed telephone interviews with 89 patients (69%). As with the cases, we made six attempts to reach each patient. The reasons for incomplete interviews with located patients were lack of an English-speaking interpreter (n Å 2), patient refusal (n Å 2), and severe cognitive impairment (n Å 6).
Data Collection Scripted telephone interviews, conducted by one of two trained interviewers, assessed sociodemographic features, indication for chronic anticoagulation, patient satisfaction, and health beliefs relating to anticoagulation. Sociodemographic variables included age, race, gender, education, marital status, income, type of health insurance, and employment status. Primary indication for chronic anticoagulation was verified by reviewing the notes of the ATU nurses. Patient satisfaction was assessed using four previously validated questions,13 in which patients are asked to rate the quality of information, skill and competence, willingness to listen, and degree of concern exhibited by their physicians on a five-point scale. Health beliefs were assessed with seven questions adapted from an instrument previously used to measure quality of life in anticoagulated patients.12 The content of this instrument was originally developed by asking patients and staff of the ATU to identify problems with taking warfarin. The scale offered
July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
12
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
TABLE I Demographic Characteristics of Cases and Controls
Variable Age (mean yrs { SD) Sex: male Race: non-white* Education: attended some college† Employment: working full-time Retired Marital status: unmarried No health insurance‡
Cases (n Å 43) n (%)
Controls (n Å 89) n (%)
Odds Ratio (95% CI)
P value
53.7 { 13 34 (79) 10 (24) 25 (60) 26 (60) 9 (21) 24 (56) 9 (21)
68.7 { 11 46 (52) 4 (5) 37 (42) 19 (21) 60 (67) 36 (40) 4 (5)
NA 3.5 (1.5, 8.2) 6.4 (1.9, 21.9) 2.0 (1.0, 4.4) 5.6 (2.5, 12.5) 0.1 (0.1, 0.3) 1.9 (0.9, 3.9) 5.6 (1.6, 19.3)
õ0.0001 0.003 0.001 0.06 õ0.0001 õ0.0001 0.1 0.003
* Data missing on four subjects. † Data missing on one subject. ‡ Data missing on one subject.
(OR) and their 95% confidence intervals (CI) were obtained from simple logistic regression analyses. Multiple logistic regression models were used to account for the potential confounding effects of demographic features and clinical characteristics and to test interaction terms. The results from the seven-item quality of life scale were analyzed by dichotomizing the responses to each item and then generating odds ratios from logistic regression analyses. All reported P values are two-tailed.
RESULTS Figure. Age distribution of noncompliant cases and compliant controls. Categories are based on quintiles of age among controls, with the youngest quintile divided into two deciles. Noncompliant cases were more likely to be younger. Using age ú78 years as a reference, OR for age õ53 years, 45.0, 95% CI, 2.5, 386.0; OR for age 53– 60 years, 22.0, 95% CI, 2.4, 197.0; OR for age 61–67 years, 4.0, 95% CI, 0.4, 39.3; OR for age 68–72 years, 6.4, 95% CI, 0.3, 28.2; OR for age 73–77 years, 0.9, 95% CI, 0.2, 76.7; Mantel-Haenszel test of trend P õ 0.0001.
four responses to the following seven questions: (1) How much does taking warfarin restrict your physical activity? (2) How much do you worry about bleeding while taking warfarin? (3) How much of a problem are the regular blood tests you have to monitor warfarin? (4) How much has taking warfarin affected your lifestyle? (5) How much do you feel warfarin has benefited your health? (6) How much do you feel warfarin will protect your future health? and (7) To what extent do you feel warfarin is preventing blood clots?
Statistical Analysis Data were analyzed using the SAS-PC system.14 Clinical characteristics of the two groups were compared by chi-square tests for categorical variables, Student’s t tests for continuous variables, and Mantel-Haenszel trend tests for ordinal variables. Adjusted odds ratios
Demographic Features The mean age of noncompliant cases was 53.7 years, significantly younger than the mean age of 68.7 years among controls (P õ0.0001)(Table I). The striking effect of age on compliance is displayed in the Figure, in which the entire sample is divided into six groups based on quintiles of age among the controls. Because the majority of noncompliant cases were clustered in the youngest quintile, we split this group into two deciles. Using the oldest group as a reference category, noncompliant cases were far more likely to be in the youngest two age groups (for age õ53 years, OR 45.0, 95% CI 2.5, 386.0; for age 53 to 60 years, OR 22.0, 95% CI 2.4, 197.0). This age effect was essentially unchanged when the analysis was expanded to include patients who could not be reached for interviews. Specifically, when all subjects, including those not interviewed (n Å 285, 156 cases and 129 controls), were examined, the age difference between cases and controls remained highly significant (mean ages of 54.7 versus 69.4, P Å 0.0001), and the odds ratio for being noncompliant was 14.1 for age õ53 years (95% CI 5.6, 34.8) and 9.5 for age 53 to 60 years (95% CI 3.6, 25.5), compared to patients older than 78 years. In univariate analysis, men were more likely to be noncompliant (OR 3.5, 95% CI 1.5, 8.2), as were subjects who were non-white (OR 6.4, 95% CI 1.9, 21.9), July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
13
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
13
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
TABLE II Patient Self-Report of Primary Indication for Anticoagulation
Indication Don’t know Atrial fibrillation Valve replacement Prior revascularization procedure Myocardial infarction Congestive heart failure Thromboembolism Other (PVD, hypercoaguable, RHD) More than one indication Stroke/TIA
Cases (n Å 43) n (%)
Controls (n Å 89) n (%)
9 (21) 7 (16) 5 (12) 3 (7) 5 (12) 2 (5) 5 (12) 2 (5) 2 (5) 3 (7)
5 (6) 11 (13) 12 (13) 3 (3) 2 (2) 1 (1) 16 (18) 5 (6) 10 (11) 24 (27)
Odds Ratio (95% CI) 4.4 (1.4, 14.2)
5.7 (1.1, 31.3)
0.2 (0.1, 0.7)
P Value 0.007 NS NS NS 0.02 NS NS NS NS 0.008
PVD Å peripheral vascular disease; RHD Å rheumatic heart disease; TIA Å transient ischemic attack.
working full-time (OR 5.6, 95% CI 2.5, 12.5), or uninsured (OR 5.6, 95% CI 1.6, 19.2). There were no significant differences between cases and controls in income, marital status, or education. As with age, data on sex were available for all patients, including those who could not be reached for interviews. When the sex of all subjects (n Å 285) was considered, the odds ratio for being noncompliant was 2.9 (95% CI 1.7, 4.9) for men compared to women. In a logistic regression model containing terms for age (coded as a continuous variable), sex, and race, there remained a strong association between younger age and poor compliance (P Å 0.0001), and between male sex and poor compliance (P Å 0.006). The association between non-white race and poor compliance was marginally significant (P Å 0.06) in this model. Because employment and health insurance status are generally associated with age, we constructed a second logistic regression model containing terms for these three variables. This analysis demonstrated that lack of health insurance and working full time were not independently predictive of noncompliance. There were no significant interactions among the variables in either model.
Indication for Anticoagulation Accuracy of self-reported indication for anticoagulation was confirmed by reviewing the ATU nurses’ notes. Overall, 21% of cases and 6% of controls did not know why warfarin had been prescribed (OR 4.4, 95% CI 1.4, 14.2) (Table II). Among patients who were able to state the reason they were taking warfarin, there was 100% agreement with the nurses’ notes. The distribution of indications for anticoagulation was similar for atrial fibrillation, valve replacement, revascularization, congestive heart failure, thromboembolism, hypercoagulable state, and rheumatic valve disease. Patients who reported a history of 14
stroke or transient ischemic event were much less likely to be noncompliant (OR 0.2, 95% CI 0.1, 0.7). After adjusting for age in a logistic model, history of stroke or transient ischemic attack (TIA) remained marginally associated with better compliance (OR 0.3, P Å 0.06). Among cases, more patients reported a history of myocardial infarction, but this effect was not independent of age in logistic regression analysis.
Patient Satisfaction With Physician’s Care Twenty-eight percent of cases (n Å 12) reported not having a regular physician compared to 3% of controls (OR 11.1, 95% CI 3.6, 50.0) (Table III). Among patients with a regular physician (n Å 117), there were significant differences in patient satisfaction between cases and controls. Noncompliant cases were more likely to feel that their doctor was not very concerned about them (OR 3.1, 95% CI 1.2, 7.8). In addition, noncompliant cases were more likely to rate the skill and competence of their physician as ‘‘good,’’ ‘‘fair,’’ or ‘‘poor,’’ rather than ‘‘excellent’’ or ‘‘very good’’ (OR 0.2, 95% CI 0.1, 0.8), and more likely to similarly rate their physicians’ willingness to listen to their concerns (OR 0.4, 95% CI 0.1, 1.0). Perceived Benefits and Barriers Relating to Anticoagulation We asked all patients seven questions assessing perceived health benefits and barriers relating to anticoagulation (see ‘‘Patients and Methods’’) (Table IV). To generate odds ratios, responses were dichotomized by grouping the responses ‘‘some’’ and ‘‘a great deal,’’ and the responses ‘‘not at all’’ and ‘‘don’t know.’’ This grouping was based on the assumption that patients who did not know whether their quality of life was affected in the specified way were not experiencing significant impact. Analyzed
July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
14
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
TABLE III Measures of Patient Satisfaction*
Variable No regular physician Quality of information from physician rated as excellent or very good† Skill and competence of physician rated as excellent or very good‡ Willingness of physician to listen to patient concerns rated as excellent or very good Patient feels physician is not very concerned about them§
Cases (n Å 43) n (%)
Controls (n Å 89) n (%)
Odds Ratio (95% CI)
P Value
12 (28)
3 (3)
11.1 (3.6, 50.0)
õ0.0001
24 (77)
73 (86)
0.6 (0.2, 1.6)
NS
25 (81)
81 (95)
0.2 (0.1, 0.8)
0.013
22 (71)
75 (87)
0.4 (0.1, 1.0)
0.04
12 (39)
14 (17)
3.1 (1.2, 7.8)
0.01
* For assessments of patient satisfaction with physicians, only patients who reported that they had a regular physician were included (n Å 117). † Data missing on one subject. ‡ Data missing on one subject. § Data missing on three subjects.
TABLE IV Perceived Benefits and Barriers to Anticoagulation
Variable Taking warfarin benefits my health Taking warfarin prevents blood clots Taking warfarin protects my future health Taking warfain affects my lifestyle I worry about bleeding while taking warfarin Taking warfarin restricts my physical activity Regular blood tests to monitor warfarin are a problem
Cases (n Å 43) n (%)
Controls (n Å 89) n (%)
Odds Ratio* (95% CI)
P Value
23 (53) 30 (70) 28 (65) 23 (53) 21 (49) 13 (30) 26 (60)
71 (80) 74 (83) 76 (85) 28 (31) 27 (30) 13 (15) 30 (34)
0.3 (0.1, 0.6) 0.5 (0.2, 1.1) 0.3 (0.1, 0.7) 2.5 (1.2, 5.3) 2.2 (1.0, 4.6) 2.5 (1.1, 6.0) 3.0 (1.4, 6.3)
0.002 0.08 0.008 0.015 0.04 0.03 0.004
* Odds ratios are for two group comparisons (‘‘some/a great deal’’ versus ‘‘not at all/don’t know’’).
in this manner, cases were less likely to feel that taking warfarin benefited their health (OR 0.3, 95% CI 0.1, 0.6), prevented blood clots (OR 0.5, 95% CI 0.2, 1.1), or protected their future health (OR 0.3, 95% CI 0.1, 0.7). Cases were more likely to report that taking warfarin affected their lifestyle (OR 2.5, 95% CI 1.2, 5.3) and restricted their physical activity (OR 2.5, 95% CI 1.1, 6.0), that they worried about bleeding while taking warfarin (OR 2.2, 95% CI 1.0, 4.6), and that regular blood tests were a problem (OR 3.0, 95% CI 1.4, 6.3). The results were not significantly different when we repeated this analysis with ‘‘don’t know’’ responses excluded.
DISCUSSION Warfarin can be a highly effective preventive therapy but careful patient compliance with medication dose and prothrombin time testing is essential to optimizing its impact. Poor compliance with anticoagulants has been documented in randomized trials1,2 and is a frequently cited concern of physicians.11 Nonetheless, ours was the first study to explore pa-
tient features associated with noncompliance with anticoagulation. In general, previous compliance studies have focused on rates of medication adherence in cross-sectional samples of patients at a particular clinic. Few studies have attempted to identify those patients who nearly or entirely drop out of care and therefore are not included in cross-sectional surveys.15 Our large computer-supported anticoagulation therapy unit allowed us to assemble a sizable number of patients no longer adhering to medical recommendations and contrast them with compliant controls. Our results reveal that patients who are noncompliant with warfarin therapy have a distinctive clinical profile. In particular, noncompliant patients are an average of 15 years younger than compliant patients. Patients younger than 53 years were many times more likely to be noncompliant than patients in the oldest age quintile (ú78 years). In addition, noncompliant patients were more likely to be male and non-white. These findings are consistent with work by Monane et al,16 which demonstrated that July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
15
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
15
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
compliance with medications for congestive heart failure was greatest in the oldest elderly (ú85 years), in women, and in white patients. Our findings contrast with previous surveys of physicians’ attitudes toward anticoagulation in atrial fibrillation.11,17,18 In 1991, Kutner et al11 found that only 18% of physicians would use warfarin in elderly patients with atrial fibrillation; among the reasons cited for not prescribing warfarin was the belief that anticoagulation is difficult in the elderly because of poor compliance. In a recent survey by McCrory et al,18 physicians were significantly less likely to anticoagulate 75-year-old patients than 65- or 55-year-old patients with atrial fibrillation, regardless of the clinical scenario. Most physicians in this survey (67% to 75%) agreed with the statement that ‘‘anticoagulant treatment is more difficult in elderly patients.’’ In our study, older age was strongly predictive of better compliance with anticoagulation therapy and it was younger patients, particularly younger men, who felt that anticoagulation therapy was unduly burdensome. Indications for anticoagulation were similar among cases and controls with one striking exception: compliant controls were more likely to have had a stroke or TIA. Although this finding is explained in part by the age difference between cases and controls, it also suggests that patients are more likely to comply with warfarin after they have personally experienced a thromboembolic event. Noncompliant cases differed from compliant controls in many expected ways, but the magnitude of this difference was previously unknown. Cases were less likely to know why warfarin had been prescribed, and more likely to report dissatisfaction with their physicians. A small number of past studies have examined the effect of the doctor/patient relationship on compliance and concluded, as we did, that compliance is correlated with effective physician communication and patient satisfaction with medical care.19 – 23 In addition, we found that patients who are noncompliant with warfarin therapy perceive markedly different benefits and barriers to anticoagulation than compliant patients. In particular, noncompliant cases considered regular blood tests to be a big problem, felt that their life was adversely affected by taking warfarin, and failed to perceive health benefits from anticoagulation therapy. Patients who were noncompliant were characterized by poor or absent contact with our anticoagulation therapy unit. Despite intensive efforts to complete telephone interviews with the patients in our sample, this study remains limited by the significant fraction of patients we were unable to contact. Patients who completed interviews may have had systematically different opinions from those we did not 16
interview. We note, however, that this potential bias does not apply to our finding that younger age and male sex were associated with noncompliance, as age and gender were available for all potential cases and controls. There are other limitations that affect the interpretation of our results. Because current attitudes among cases may have reflected rationalization after being noncompliant rather than features existing before noncompliance, causal inferences cannot be drawn. Although our surveys were conducted using a highly structured script, our interviewers were not blinded and this may have contributed to recall bias. Our case definition included two categories of noncompliance: patients who had self-discontinued warfarin and patients who were continuing to take warfarin but without proper INR monitoring. In fact, these two populations may be different; our study did not have the power to discern this difference. Older patients who had already demonstrated noncompliance with other medical regimens may not have been prescribed warfarin by their primary physicians and thus would not have been included in our ATU database. Finally, the associations we observed, although consistent with prior compliance studies, require further validation in a prospective study. In conclusion, this first analytic study of patient features associated with noncompliance with anticoagulation therapy reinforces the importance of patient education, physician involvement, and ease of prothrombin time monitoring. Most strikingly, it points out that older patients are actually more likely to be compliant with warfarin dosing and blood tests. Conversely, younger patients, particularly male patients, are at highest risk of foregoing instructions meant to ensure safe anticoagulation. For younger patients, interventions to reduce the burden of frequent blood tests, such as home prothrombin time monitoring,24 may lead to more optimal anticoagulation management. However, prospective studies of home prothrombin time monitoring in diverse patient populations are necessary before its impact on compliance can be assessed.
ACKNOWLEDGMENTS We gratefully acknowledge the assistance of Mary A. Sheehan, RN, Robert A. Hughes, MD, and the staff of the Anticoagulation Therapy Unit at Massachusetts General Hospital.
REFERENCES 1. The Boston Area Anticoagulation trial for Atrial Fibrillation Investigators. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation. NEJM. 1990;323:1505–1511. 2. Petersen P, Godtfredsen J, Boysen G, Andersen ED, Andersen B. Placebocontrolled, randomised trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: The Copenhagen AFASAK study. Lancet. 1989;1:175–179.
July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
16
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL 3. Stroke Prevention in Atrial Fibrillation Investigators. Stroke prevention in atrial fibrillation study: Final results. Circulation. 1991;84:527–539. 4. Connolly SJ, Laupacis A, Gent M, Roberts RS, Cairns JA, Joyner C. Canadian atrial fibrillation anticoagulation (CAFA) study. J Am Coll Cardiol. 1991;18:349– 355. 5. Ezekowitz MD, Bridgers SL, James KE, et al. Warfarin in the prevention of stroke associated with nonrheumatic atrial fibrillation. NEJM. 1992;327:1406– 1412. 6. Smith P, Arnesen H, Holme I. The effect of warfarin on mortality and reinfarction after myocardial infarction. NEJM. 1990;323:147–152. 7. Anticoagulants in the Secondary Prevention of Events in Coronary Thrombosis (ASPECT) Research Group. Effect of long-term anticoagulant treatment on mortality and cardiovascular morbidity after myocardial infarction. Lancet. 1994;343:499–503. 8. van Bergen PF, Jonker JJ, van Hout BA, et al. Costs and effects of long-term oral anticoagulant treatment after myocardial infarction. JAMA. 1995;273:925– 928. 9. Stafford RS, Singer DE. National trends in warfarin use in atrial fibrillation. Arch Intern Med. 1996;156:2537–2541. 10. Furberg CD, Psaty BM, Manolio TA, Gardin JM, Smith VE, Rautaharju PM. Prevalence of atrial fibrillation in elderly subjects (the Cardiovascular Health Study). Am J Cardiol. 1994;74:236–241. 11. Kutner M, Nixon G, Silverstone F. Physicians’ attitudes toward oral anticoagulants and antiplatelet agents for stroke prevention in elderly patients with atrial fibrillation. Arch Intern Med. 1991;51:1950–1953. 12. Lancaster TR, Singer DE, Sheehan MA, et al. The impact of long-term warfarin therapy on quality of life: evidence from a randomized trial. Arch Intern Med. 1991;151:1944–1949. 13. Stewart AL, Ware JE, eds. Measuring Functioning and Well-Being: The Medical Outcomes Study Approach. Durham: Duke Univ Press, 1992.
14. SAS Proprietary Software Release 6.08, SAS Institute Inc., Cary, NC, 1989. 15. Sackett DL, Snow JC. The magnitude of compliance and non-compliance. In: Haynes RB, Taylor DW, Sackett DL, eds. Compliance in Health Care. Baltimore: The Johns Hopkins University Press, 1981;11–23. 16. Monane M, Bohn RL, Gurwitz JH, Glynn RJ, Avorn J. Noncompliance with congestive heart failure therapy in the elderly. Arch Intern Med. 1994;154:433–437. 17. Chang HJ, Bell JR, Deroo DB, Kirk JW, Wasson JH, Dartmouth Primary Care Coop Project. Physician variation in anticoagulating patients with atrial fibrillation. Arch Intern Med. 1990;150:81–84. 18. McCrory DC, Matchar DB, Samsa G, Sanders LL, Pritchett EL. Physician attitudes about anticoagulation for nonvalvular atrial fibrillation in the elderly. Arch Intern Med. 1995;155:277–281. 19. Nagy VT, Wolfe GR. Cognitive predictors of compliance in chronic disease patients. Med Care. 1984;22:912–921. 20. Hulka BS. Patient-clinician interactions and compliance. In: Haynes RB, Taylor DW, Sackett DL, eds. Compliance in Health Care. Baltimore: The Johns Hopkins University Press, 1981;63–77. 21. Kravitz RL, Hays RD, Sherbourne CD, et al. Recall of recommendations and adherence to advice among patients with chronic medical conditions. Arch Intern Med. 1993;153:1869–1878. 22. Sherbourne CD, Hays RD, Ordway L, DiMatteo MR, Kravitz RL. Antecedents of adherence to medical recommendations: Results from the Medical Outcomes Study. J Behav Med. 1992;15:447–468. 23. DiMatteo MR, Sherbourne CD, Hays RD, et al. Physicians’ characteristics influence patient adherence to medical recommendations: Results from the Medical Outcomes Study. Health Psychol. 1993;12:93–102. 24. Ansell JE, Patel N, Ostrovsky D, Nozzolillo E, Peterson AM, Fish L. Longterm patient self-management of oral anticoagulation. Arch Intern Med. 1995;155:2185–2189.
July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
17
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
17
BACKGROUND: The number of patients for whom long-term anticoagulation is indicated has increased dramatically over the past decade. Good patient compliance is necessary to safely realize the benefits of anticoagulation, yet barriers to compliance with anticoagulation therapy have not been studied. METHODS: We conducted a case-control study in the Anticoagulation Therapy Unit (ATU) at Massachusetts General Hospital. Forty-three patients who had been discharged from the ATU for noncompliance (cases) and 89 randomly selected compliant ATU controls were interviewed. Noncompliant cases had selfdiscontinued warfarin or were taking warfarin with inadequate monitoring of international normalized ratio (INR) levels. Telephone interviews assessed sociodemographic features, indication for anticoagulation, patient satisfaction, and health beliefs. RESULTS: Noncompliant cases were more likely to be younger (mean 53.7 years versus 68.7 years, P õ0.0001), male (odds ratio [OR] 3.5, 95% confidence interval [CI] 1.5, 8.2), and nonwhite (OR 6.4, 95% CI 1.9, 21.9), and less likely to have had a stroke or transient ischemic attack (OR 0.2, 95% CI 0.1, 0.7). In open-ended questioning, cases were more likely to report that they did not know why warfarin had been prescribed (OR 4.4, 95% CI 1.4, 14.2). Noncompliant cases were more likely not to have a regular physician (OR 11.1, 95% CI 3.6, 50.0); among patients with a regular physician,
From the General Internal Medicine Unit, Massachusetts General Hospital, Boston Massachusetts (JHA, DES); Harvard Medical School, Boston, Massachusetts (JHA, JMG, DES); Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts (DES). The research reported in this paper was supported in part by a grant from the Eliot B. Shoolman Fund, Massachusetts General Hospital, Boston, Massachusetts. Dr. Arnsten was a fellow of the Harvard General Internal Medicine Fellowship Program, U.S. Department of Health and Human Services National Research Services Award Grant #5T32PE1100107. Portions of this research were presented at the 1995 Annual Meeting of the Society for General Internal Medicine, San Diego, CA, May 1995. Requests for reprints should be addressed to Julia H. Arnsten, MD, MPH, Centennial Building, 3rd floor, Montefiore Medical Center, Bronx, NY 10467. Manuscript submitted May 29, 1996 and accepted in revised form February 11, 1997.
noncompliant cases were more likely to feel dissatisfied. Examination of health beliefs revealed that noncompliant cases felt more burdened by taking warfarin, and perceived fewer health benefits. CONCLUSIONS: Patients who are noncompliant with warfarin share distinctive clinical characteristics. Notably, younger, male patients who have not experienced a thromboembolic event are more likely to forego INR testing or to stop anticoagulation therapy completely. Improved patient education, physician involvement, and ease of monitoring may improve compliance, particularly among younger male patients. Am J Med. 1997;103:11–17. Q 1997 by Excerpta Medica, Inc.
O
ver the past 10 years, the number of patients for whom long-term anticoagulation therapy is indicated has increased dramatically. Recent large randomized trials have demonstrated that anticoagulants prevent strokes in patients with nonrheumatic atrial fibrillation1 – 5 and reduce the risk of mortality and cardiovascular morbidity after acute myocardial infarction.6 – 8 However, anticoagulation is a demanding therapy that is underutilized in actual clinical practice. Despite a two-thirds reduction in the risk of stroke among anticoagulated patients, only a minority of patients with nonrheumatic atrial fibrillation are prescribed warfarin.9,10 One frequently stated concern among physicians is poor patient compliance with anticoagulants, especially among the elderly.11 In the randomized trials in atrial fibrillation, noncompliance with anticoagulant regimens ranged from 10% to 26%.1,2 In general populations, noncompliance is likely to be much more common. Good compliance with both medication dosage and blood test monitoring is necessary to safely realize the benefits of anticoagulation, yet barriers to compliance with anticoagulation therapy have not been studied. Building on our earlier work on the impact of warfarin on quality of life,12 we constructed a case-control study to identify determinants of noncompliance with anticoagulation therapy.
PATIENTS AND METHODS Study Design We conducted this study in the Anticoagulation Therapy Unit (ATU) at Massachusetts General Hos-
Q1997 by Excerpta Medica, Inc. All rights reserved.
/ 220f 5048 Mp
11
0002-9343/97/$17.00 PII S0002-9343(97)00048-X
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
11
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
pital, a large urban teaching hospital. Cases were patients who had demonstrated such severe noncompliance with chronic anticoagulation therapy that it was no longer considered safe. Controls were compliant patients attending the ATU. In-depth telephone surveys were completed with all patients. The study was approved by the Human Subjects Committee of the Institutional Review Board at Massachusetts General Hospital.
Setting The Massachusetts General Hospital ATU managed long-term anticoagulation for approximately 2,300 outpatients during the period of this study. The ATU is staffed by nurses who use a computer-based system to track patients’ prothrombin time tests and warfarin dose. All patients attend an introductory session, detailing the risks and benefits of anticoagulation and the importance of frequent blood test monitoring. Patients then use a convenient laboratory for ongoing prothrombin time tests, and are notified by ATU nurses of results, dosage changes, and timing of future tests. Patients who do not have at least one prothrombin time test per month receive reminder postcards and telephone calls, and ultimately are discharged from the ATU if they do not respond to these reminders. Patients discharged in this manner include the most extreme noncompliers with ATU regimens, as well as patients with whom the ATU has lost contact. These patients were the source of cases for this study. Patient Selection Cases. The criteria for selection of cases were: (1) standard indication for chronic anticoagulation; (2) discharge from the ATU for apparent failure to comply with ATU monitoring policies; and (3) patient report of self-discontinuation of warfarin therapy, or continuation of warfarin therapy without proper international normalized ratio (INR) monitoring. Between January 1991 and December 1994, 156 patients were discharged from the ATU for noncompliance. Between July 1994 and January 1995, we attempted to complete telephone interviews with these 156 patients. Sixty-one patients could not be located after six attempts, using home and office phone numbers listed in the ATU database, and contacting directory information in the region of the patient’s last known address. We were able to locate 95 patients (61%) and complete interviews with 69 patients (44%). The reasons for incomplete interviews with located patients were death (n Å 6), lack of an English-speaking interpreter (n Å 12), patient refusal (n Å 6), and severe cognitive impairment (n Å 2). 12
Among the 69 patients we interviewed who were discharged from the ATU because of apparent noncompliance, we found that many were not truly noncompliant, but were simply no longer in contact with the ATU. Twenty-six patients (38%) reported that they had been directed to discontinue warfarin by their primary physician, or were still taking warfarin but no longer used the ATU system. These patients were not included in the study. Forty-three patients (62%) reported that they had self-discontinued warfarin therapy against medical advice, or were continuing to take warfarin with improper monitoring. Improper monitoring was defined as prothrombin time tests less frequently than once every 2 months. These 43 patients constituted our cases of severe noncompliance with anticoagulation therapy. Although this is an unusually low proportion of noncompliant patients (43 of 2,300, or 1.9%), discharge from the ATU is a last resort, resulting from far more stringent and less inclusive criteria for noncompliance than were used in the randomized trials of anticoagulation. Controls. The criteria for selection of controls were: (1) active enrollment in the ATU in June 1994, and (2) standard indication for chronic anticoagulation. Using a case to control ratio of 1:3 and a total of 43 cases, we randomly selected 129 controls from among 2,308 eligible patients. We located 99 patients (77%) and completed telephone interviews with 89 patients (69%). As with the cases, we made six attempts to reach each patient. The reasons for incomplete interviews with located patients were lack of an English-speaking interpreter (n Å 2), patient refusal (n Å 2), and severe cognitive impairment (n Å 6).
Data Collection Scripted telephone interviews, conducted by one of two trained interviewers, assessed sociodemographic features, indication for chronic anticoagulation, patient satisfaction, and health beliefs relating to anticoagulation. Sociodemographic variables included age, race, gender, education, marital status, income, type of health insurance, and employment status. Primary indication for chronic anticoagulation was verified by reviewing the notes of the ATU nurses. Patient satisfaction was assessed using four previously validated questions,13 in which patients are asked to rate the quality of information, skill and competence, willingness to listen, and degree of concern exhibited by their physicians on a five-point scale. Health beliefs were assessed with seven questions adapted from an instrument previously used to measure quality of life in anticoagulated patients.12 The content of this instrument was originally developed by asking patients and staff of the ATU to identify problems with taking warfarin. The scale offered
July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
12
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
TABLE I Demographic Characteristics of Cases and Controls
Variable Age (mean yrs { SD) Sex: male Race: non-white* Education: attended some college† Employment: working full-time Retired Marital status: unmarried No health insurance‡
Cases (n Å 43) n (%)
Controls (n Å 89) n (%)
Odds Ratio (95% CI)
P value
53.7 { 13 34 (79) 10 (24) 25 (60) 26 (60) 9 (21) 24 (56) 9 (21)
68.7 { 11 46 (52) 4 (5) 37 (42) 19 (21) 60 (67) 36 (40) 4 (5)
NA 3.5 (1.5, 8.2) 6.4 (1.9, 21.9) 2.0 (1.0, 4.4) 5.6 (2.5, 12.5) 0.1 (0.1, 0.3) 1.9 (0.9, 3.9) 5.6 (1.6, 19.3)
õ0.0001 0.003 0.001 0.06 õ0.0001 õ0.0001 0.1 0.003
* Data missing on four subjects. † Data missing on one subject. ‡ Data missing on one subject.
(OR) and their 95% confidence intervals (CI) were obtained from simple logistic regression analyses. Multiple logistic regression models were used to account for the potential confounding effects of demographic features and clinical characteristics and to test interaction terms. The results from the seven-item quality of life scale were analyzed by dichotomizing the responses to each item and then generating odds ratios from logistic regression analyses. All reported P values are two-tailed.
RESULTS Figure. Age distribution of noncompliant cases and compliant controls. Categories are based on quintiles of age among controls, with the youngest quintile divided into two deciles. Noncompliant cases were more likely to be younger. Using age ú78 years as a reference, OR for age õ53 years, 45.0, 95% CI, 2.5, 386.0; OR for age 53– 60 years, 22.0, 95% CI, 2.4, 197.0; OR for age 61–67 years, 4.0, 95% CI, 0.4, 39.3; OR for age 68–72 years, 6.4, 95% CI, 0.3, 28.2; OR for age 73–77 years, 0.9, 95% CI, 0.2, 76.7; Mantel-Haenszel test of trend P õ 0.0001.
four responses to the following seven questions: (1) How much does taking warfarin restrict your physical activity? (2) How much do you worry about bleeding while taking warfarin? (3) How much of a problem are the regular blood tests you have to monitor warfarin? (4) How much has taking warfarin affected your lifestyle? (5) How much do you feel warfarin has benefited your health? (6) How much do you feel warfarin will protect your future health? and (7) To what extent do you feel warfarin is preventing blood clots?
Statistical Analysis Data were analyzed using the SAS-PC system.14 Clinical characteristics of the two groups were compared by chi-square tests for categorical variables, Student’s t tests for continuous variables, and Mantel-Haenszel trend tests for ordinal variables. Adjusted odds ratios
Demographic Features The mean age of noncompliant cases was 53.7 years, significantly younger than the mean age of 68.7 years among controls (P õ0.0001)(Table I). The striking effect of age on compliance is displayed in the Figure, in which the entire sample is divided into six groups based on quintiles of age among the controls. Because the majority of noncompliant cases were clustered in the youngest quintile, we split this group into two deciles. Using the oldest group as a reference category, noncompliant cases were far more likely to be in the youngest two age groups (for age õ53 years, OR 45.0, 95% CI 2.5, 386.0; for age 53 to 60 years, OR 22.0, 95% CI 2.4, 197.0). This age effect was essentially unchanged when the analysis was expanded to include patients who could not be reached for interviews. Specifically, when all subjects, including those not interviewed (n Å 285, 156 cases and 129 controls), were examined, the age difference between cases and controls remained highly significant (mean ages of 54.7 versus 69.4, P Å 0.0001), and the odds ratio for being noncompliant was 14.1 for age õ53 years (95% CI 5.6, 34.8) and 9.5 for age 53 to 60 years (95% CI 3.6, 25.5), compared to patients older than 78 years. In univariate analysis, men were more likely to be noncompliant (OR 3.5, 95% CI 1.5, 8.2), as were subjects who were non-white (OR 6.4, 95% CI 1.9, 21.9), July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
13
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
13
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
TABLE II Patient Self-Report of Primary Indication for Anticoagulation
Indication Don’t know Atrial fibrillation Valve replacement Prior revascularization procedure Myocardial infarction Congestive heart failure Thromboembolism Other (PVD, hypercoaguable, RHD) More than one indication Stroke/TIA
Cases (n Å 43) n (%)
Controls (n Å 89) n (%)
9 (21) 7 (16) 5 (12) 3 (7) 5 (12) 2 (5) 5 (12) 2 (5) 2 (5) 3 (7)
5 (6) 11 (13) 12 (13) 3 (3) 2 (2) 1 (1) 16 (18) 5 (6) 10 (11) 24 (27)
Odds Ratio (95% CI) 4.4 (1.4, 14.2)
5.7 (1.1, 31.3)
0.2 (0.1, 0.7)
P Value 0.007 NS NS NS 0.02 NS NS NS NS 0.008
PVD Å peripheral vascular disease; RHD Å rheumatic heart disease; TIA Å transient ischemic attack.
working full-time (OR 5.6, 95% CI 2.5, 12.5), or uninsured (OR 5.6, 95% CI 1.6, 19.2). There were no significant differences between cases and controls in income, marital status, or education. As with age, data on sex were available for all patients, including those who could not be reached for interviews. When the sex of all subjects (n Å 285) was considered, the odds ratio for being noncompliant was 2.9 (95% CI 1.7, 4.9) for men compared to women. In a logistic regression model containing terms for age (coded as a continuous variable), sex, and race, there remained a strong association between younger age and poor compliance (P Å 0.0001), and between male sex and poor compliance (P Å 0.006). The association between non-white race and poor compliance was marginally significant (P Å 0.06) in this model. Because employment and health insurance status are generally associated with age, we constructed a second logistic regression model containing terms for these three variables. This analysis demonstrated that lack of health insurance and working full time were not independently predictive of noncompliance. There were no significant interactions among the variables in either model.
Indication for Anticoagulation Accuracy of self-reported indication for anticoagulation was confirmed by reviewing the ATU nurses’ notes. Overall, 21% of cases and 6% of controls did not know why warfarin had been prescribed (OR 4.4, 95% CI 1.4, 14.2) (Table II). Among patients who were able to state the reason they were taking warfarin, there was 100% agreement with the nurses’ notes. The distribution of indications for anticoagulation was similar for atrial fibrillation, valve replacement, revascularization, congestive heart failure, thromboembolism, hypercoagulable state, and rheumatic valve disease. Patients who reported a history of 14
stroke or transient ischemic event were much less likely to be noncompliant (OR 0.2, 95% CI 0.1, 0.7). After adjusting for age in a logistic model, history of stroke or transient ischemic attack (TIA) remained marginally associated with better compliance (OR 0.3, P Å 0.06). Among cases, more patients reported a history of myocardial infarction, but this effect was not independent of age in logistic regression analysis.
Patient Satisfaction With Physician’s Care Twenty-eight percent of cases (n Å 12) reported not having a regular physician compared to 3% of controls (OR 11.1, 95% CI 3.6, 50.0) (Table III). Among patients with a regular physician (n Å 117), there were significant differences in patient satisfaction between cases and controls. Noncompliant cases were more likely to feel that their doctor was not very concerned about them (OR 3.1, 95% CI 1.2, 7.8). In addition, noncompliant cases were more likely to rate the skill and competence of their physician as ‘‘good,’’ ‘‘fair,’’ or ‘‘poor,’’ rather than ‘‘excellent’’ or ‘‘very good’’ (OR 0.2, 95% CI 0.1, 0.8), and more likely to similarly rate their physicians’ willingness to listen to their concerns (OR 0.4, 95% CI 0.1, 1.0). Perceived Benefits and Barriers Relating to Anticoagulation We asked all patients seven questions assessing perceived health benefits and barriers relating to anticoagulation (see ‘‘Patients and Methods’’) (Table IV). To generate odds ratios, responses were dichotomized by grouping the responses ‘‘some’’ and ‘‘a great deal,’’ and the responses ‘‘not at all’’ and ‘‘don’t know.’’ This grouping was based on the assumption that patients who did not know whether their quality of life was affected in the specified way were not experiencing significant impact. Analyzed
July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
14
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
TABLE III Measures of Patient Satisfaction*
Variable No regular physician Quality of information from physician rated as excellent or very good† Skill and competence of physician rated as excellent or very good‡ Willingness of physician to listen to patient concerns rated as excellent or very good Patient feels physician is not very concerned about them§
Cases (n Å 43) n (%)
Controls (n Å 89) n (%)
Odds Ratio (95% CI)
P Value
12 (28)
3 (3)
11.1 (3.6, 50.0)
õ0.0001
24 (77)
73 (86)
0.6 (0.2, 1.6)
NS
25 (81)
81 (95)
0.2 (0.1, 0.8)
0.013
22 (71)
75 (87)
0.4 (0.1, 1.0)
0.04
12 (39)
14 (17)
3.1 (1.2, 7.8)
0.01
* For assessments of patient satisfaction with physicians, only patients who reported that they had a regular physician were included (n Å 117). † Data missing on one subject. ‡ Data missing on one subject. § Data missing on three subjects.
TABLE IV Perceived Benefits and Barriers to Anticoagulation
Variable Taking warfarin benefits my health Taking warfarin prevents blood clots Taking warfarin protects my future health Taking warfain affects my lifestyle I worry about bleeding while taking warfarin Taking warfarin restricts my physical activity Regular blood tests to monitor warfarin are a problem
Cases (n Å 43) n (%)
Controls (n Å 89) n (%)
Odds Ratio* (95% CI)
P Value
23 (53) 30 (70) 28 (65) 23 (53) 21 (49) 13 (30) 26 (60)
71 (80) 74 (83) 76 (85) 28 (31) 27 (30) 13 (15) 30 (34)
0.3 (0.1, 0.6) 0.5 (0.2, 1.1) 0.3 (0.1, 0.7) 2.5 (1.2, 5.3) 2.2 (1.0, 4.6) 2.5 (1.1, 6.0) 3.0 (1.4, 6.3)
0.002 0.08 0.008 0.015 0.04 0.03 0.004
* Odds ratios are for two group comparisons (‘‘some/a great deal’’ versus ‘‘not at all/don’t know’’).
in this manner, cases were less likely to feel that taking warfarin benefited their health (OR 0.3, 95% CI 0.1, 0.6), prevented blood clots (OR 0.5, 95% CI 0.2, 1.1), or protected their future health (OR 0.3, 95% CI 0.1, 0.7). Cases were more likely to report that taking warfarin affected their lifestyle (OR 2.5, 95% CI 1.2, 5.3) and restricted their physical activity (OR 2.5, 95% CI 1.1, 6.0), that they worried about bleeding while taking warfarin (OR 2.2, 95% CI 1.0, 4.6), and that regular blood tests were a problem (OR 3.0, 95% CI 1.4, 6.3). The results were not significantly different when we repeated this analysis with ‘‘don’t know’’ responses excluded.
DISCUSSION Warfarin can be a highly effective preventive therapy but careful patient compliance with medication dose and prothrombin time testing is essential to optimizing its impact. Poor compliance with anticoagulants has been documented in randomized trials1,2 and is a frequently cited concern of physicians.11 Nonetheless, ours was the first study to explore pa-
tient features associated with noncompliance with anticoagulation. In general, previous compliance studies have focused on rates of medication adherence in cross-sectional samples of patients at a particular clinic. Few studies have attempted to identify those patients who nearly or entirely drop out of care and therefore are not included in cross-sectional surveys.15 Our large computer-supported anticoagulation therapy unit allowed us to assemble a sizable number of patients no longer adhering to medical recommendations and contrast them with compliant controls. Our results reveal that patients who are noncompliant with warfarin therapy have a distinctive clinical profile. In particular, noncompliant patients are an average of 15 years younger than compliant patients. Patients younger than 53 years were many times more likely to be noncompliant than patients in the oldest age quintile (ú78 years). In addition, noncompliant patients were more likely to be male and non-white. These findings are consistent with work by Monane et al,16 which demonstrated that July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
15
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
15
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL
compliance with medications for congestive heart failure was greatest in the oldest elderly (ú85 years), in women, and in white patients. Our findings contrast with previous surveys of physicians’ attitudes toward anticoagulation in atrial fibrillation.11,17,18 In 1991, Kutner et al11 found that only 18% of physicians would use warfarin in elderly patients with atrial fibrillation; among the reasons cited for not prescribing warfarin was the belief that anticoagulation is difficult in the elderly because of poor compliance. In a recent survey by McCrory et al,18 physicians were significantly less likely to anticoagulate 75-year-old patients than 65- or 55-year-old patients with atrial fibrillation, regardless of the clinical scenario. Most physicians in this survey (67% to 75%) agreed with the statement that ‘‘anticoagulant treatment is more difficult in elderly patients.’’ In our study, older age was strongly predictive of better compliance with anticoagulation therapy and it was younger patients, particularly younger men, who felt that anticoagulation therapy was unduly burdensome. Indications for anticoagulation were similar among cases and controls with one striking exception: compliant controls were more likely to have had a stroke or TIA. Although this finding is explained in part by the age difference between cases and controls, it also suggests that patients are more likely to comply with warfarin after they have personally experienced a thromboembolic event. Noncompliant cases differed from compliant controls in many expected ways, but the magnitude of this difference was previously unknown. Cases were less likely to know why warfarin had been prescribed, and more likely to report dissatisfaction with their physicians. A small number of past studies have examined the effect of the doctor/patient relationship on compliance and concluded, as we did, that compliance is correlated with effective physician communication and patient satisfaction with medical care.19 – 23 In addition, we found that patients who are noncompliant with warfarin therapy perceive markedly different benefits and barriers to anticoagulation than compliant patients. In particular, noncompliant cases considered regular blood tests to be a big problem, felt that their life was adversely affected by taking warfarin, and failed to perceive health benefits from anticoagulation therapy. Patients who were noncompliant were characterized by poor or absent contact with our anticoagulation therapy unit. Despite intensive efforts to complete telephone interviews with the patients in our sample, this study remains limited by the significant fraction of patients we were unable to contact. Patients who completed interviews may have had systematically different opinions from those we did not 16
interview. We note, however, that this potential bias does not apply to our finding that younger age and male sex were associated with noncompliance, as age and gender were available for all potential cases and controls. There are other limitations that affect the interpretation of our results. Because current attitudes among cases may have reflected rationalization after being noncompliant rather than features existing before noncompliance, causal inferences cannot be drawn. Although our surveys were conducted using a highly structured script, our interviewers were not blinded and this may have contributed to recall bias. Our case definition included two categories of noncompliance: patients who had self-discontinued warfarin and patients who were continuing to take warfarin but without proper INR monitoring. In fact, these two populations may be different; our study did not have the power to discern this difference. Older patients who had already demonstrated noncompliance with other medical regimens may not have been prescribed warfarin by their primary physicians and thus would not have been included in our ATU database. Finally, the associations we observed, although consistent with prior compliance studies, require further validation in a prospective study. In conclusion, this first analytic study of patient features associated with noncompliance with anticoagulation therapy reinforces the importance of patient education, physician involvement, and ease of prothrombin time monitoring. Most strikingly, it points out that older patients are actually more likely to be compliant with warfarin dosing and blood tests. Conversely, younger patients, particularly male patients, are at highest risk of foregoing instructions meant to ensure safe anticoagulation. For younger patients, interventions to reduce the burden of frequent blood tests, such as home prothrombin time monitoring,24 may lead to more optimal anticoagulation management. However, prospective studies of home prothrombin time monitoring in diverse patient populations are necessary before its impact on compliance can be assessed.
ACKNOWLEDGMENTS We gratefully acknowledge the assistance of Mary A. Sheehan, RN, Robert A. Hughes, MD, and the staff of the Anticoagulation Therapy Unit at Massachusetts General Hospital.
REFERENCES 1. The Boston Area Anticoagulation trial for Atrial Fibrillation Investigators. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation. NEJM. 1990;323:1505–1511. 2. Petersen P, Godtfredsen J, Boysen G, Andersen ED, Andersen B. Placebocontrolled, randomised trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: The Copenhagen AFASAK study. Lancet. 1989;1:175–179.
July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
16
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
COMPLIANCE WITH ANTICOAGULATION/ARNSTEN ET AL 3. Stroke Prevention in Atrial Fibrillation Investigators. Stroke prevention in atrial fibrillation study: Final results. Circulation. 1991;84:527–539. 4. Connolly SJ, Laupacis A, Gent M, Roberts RS, Cairns JA, Joyner C. Canadian atrial fibrillation anticoagulation (CAFA) study. J Am Coll Cardiol. 1991;18:349– 355. 5. Ezekowitz MD, Bridgers SL, James KE, et al. Warfarin in the prevention of stroke associated with nonrheumatic atrial fibrillation. NEJM. 1992;327:1406– 1412. 6. Smith P, Arnesen H, Holme I. The effect of warfarin on mortality and reinfarction after myocardial infarction. NEJM. 1990;323:147–152. 7. Anticoagulants in the Secondary Prevention of Events in Coronary Thrombosis (ASPECT) Research Group. Effect of long-term anticoagulant treatment on mortality and cardiovascular morbidity after myocardial infarction. Lancet. 1994;343:499–503. 8. van Bergen PF, Jonker JJ, van Hout BA, et al. Costs and effects of long-term oral anticoagulant treatment after myocardial infarction. JAMA. 1995;273:925– 928. 9. Stafford RS, Singer DE. National trends in warfarin use in atrial fibrillation. Arch Intern Med. 1996;156:2537–2541. 10. Furberg CD, Psaty BM, Manolio TA, Gardin JM, Smith VE, Rautaharju PM. Prevalence of atrial fibrillation in elderly subjects (the Cardiovascular Health Study). Am J Cardiol. 1994;74:236–241. 11. Kutner M, Nixon G, Silverstone F. Physicians’ attitudes toward oral anticoagulants and antiplatelet agents for stroke prevention in elderly patients with atrial fibrillation. Arch Intern Med. 1991;51:1950–1953. 12. Lancaster TR, Singer DE, Sheehan MA, et al. The impact of long-term warfarin therapy on quality of life: evidence from a randomized trial. Arch Intern Med. 1991;151:1944–1949. 13. Stewart AL, Ware JE, eds. Measuring Functioning and Well-Being: The Medical Outcomes Study Approach. Durham: Duke Univ Press, 1992.
14. SAS Proprietary Software Release 6.08, SAS Institute Inc., Cary, NC, 1989. 15. Sackett DL, Snow JC. The magnitude of compliance and non-compliance. In: Haynes RB, Taylor DW, Sackett DL, eds. Compliance in Health Care. Baltimore: The Johns Hopkins University Press, 1981;11–23. 16. Monane M, Bohn RL, Gurwitz JH, Glynn RJ, Avorn J. Noncompliance with congestive heart failure therapy in the elderly. Arch Intern Med. 1994;154:433–437. 17. Chang HJ, Bell JR, Deroo DB, Kirk JW, Wasson JH, Dartmouth Primary Care Coop Project. Physician variation in anticoagulating patients with atrial fibrillation. Arch Intern Med. 1990;150:81–84. 18. McCrory DC, Matchar DB, Samsa G, Sanders LL, Pritchett EL. Physician attitudes about anticoagulation for nonvalvular atrial fibrillation in the elderly. Arch Intern Med. 1995;155:277–281. 19. Nagy VT, Wolfe GR. Cognitive predictors of compliance in chronic disease patients. Med Care. 1984;22:912–921. 20. Hulka BS. Patient-clinician interactions and compliance. In: Haynes RB, Taylor DW, Sackett DL, eds. Compliance in Health Care. Baltimore: The Johns Hopkins University Press, 1981;63–77. 21. Kravitz RL, Hays RD, Sherbourne CD, et al. Recall of recommendations and adherence to advice among patients with chronic medical conditions. Arch Intern Med. 1993;153:1869–1878. 22. Sherbourne CD, Hays RD, Ordway L, DiMatteo MR, Kravitz RL. Antecedents of adherence to medical recommendations: Results from the Medical Outcomes Study. J Behav Med. 1992;15:447–468. 23. DiMatteo MR, Sherbourne CD, Hays RD, et al. Physicians’ characteristics influence patient adherence to medical recommendations: Results from the Medical Outcomes Study. Health Psychol. 1993;12:93–102. 24. Ansell JE, Patel N, Ostrovsky D, Nozzolillo E, Peterson AM, Fish L. Longterm patient self-management of oral anticoagulation. Arch Intern Med. 1995;155:2185–2189.
July 1997 The American Journal of MedicineT Volume 103
/ 220f 5048 Mp
17
Monday Jun 30 09:44 AM
EL–AJM (v. 102, no. 6)
5408
17