Figure 1. Study activities and intervention. AUDIT-C indicates 3-item alcohol consumption measure based on the Alcohol Use Disorders Identification Test9; EHR, electronic health record; PCP, primary care provider; PHQ, Patient Health Questionnaire10; SAQ, Seattle Angina Questionnaire8; SOSQ, Seattle Outpatient Satisfaction Questionnaire11; and VR-12, Veterans RAND 12-Item Health Survey.12
Figure 2. Provider and patient recruitment. SAQ indicates Seattle Angina Questionnaire.8
Fihn SD, Bucher JB, McDonell M, Diehr P, Rumsfeld JS, Doak M, Dougherty C, Gerrity M, Heidenreich P, Larsen G, Lee PI, Lucas L, McBryde C, Nelson K, Plomondon ME, Stadius M, Bryson C. Collaborative Care Intervention for Stable Ischemic Heart Disease. Arch Intern Med. 2011;171(16):1471-1479. doi:10.1001/archinternmed.2011.372
Author Affiliations: Departments of Medicine (Drs Fihn, Nelson, and Bryson), Biostatistics (Dr Diehr), and Biobehavioral Nursing and Health Systems (Dr Dougherty), University of Washington, and Virginia Mason Medical Center, Department of Primary Care (Dr Bucher), and Veterans Affairs (VA) Puget Sound Health Care System (Drs Bucher, Diehr, Dougherty, Nelson, Stadius, and Bryson and Ms McDonell), Seattle; VA Eastern Colorado Health Care System, University of Colorado, Denver (Drs Rumsfeld, McBryde, and Plomondon); Portland VA Medical Center and Department of Medicine, Oregon Health and Sciences University, Portland (Drs Doak, Gerrity, Larsen, and Lucas); and VA Palo Alto Health Care System and Department of Medicine, Stanford University, Palo Alto, California (Drs Heidenreich and Lee). Dr Fihn and Ms McDonell are now with the Office of Informatics and Analytics, Veterans Health Administration, Washington, DC.
Background Accumulating evidence suggests that collaborative models of care enhance communication among primary care providers, improving quality of care and outcomes for patients with chronic conditions. We sought to determine whether a multifaceted intervention that used a collaborative care model and was directed through primary care providers would improve symptoms of angina, self-perceived health, and concordance with practice guidelines for managing chronic stable angina.
Methods We conducted a prospective trial, cluster randomized by provider, involving patients with symptomatic ischemic heart disease recruited from primary care clinics at 4 academically affiliated Department of Veterans Affairs health care systems. Primary end points were changes over 12 months in symptoms on the Seattle Angina Questionnaire, self-perceived health, and concordance with practice guidelines.
Results In total, 183 primary care providers and 703 patients participated in the study. Providers accepted and implemented 91.6% of 701 recommendations made by collaborative care teams. Almost half were related to medications, including adjustments to β-blockers, long-acting nitrates, and statins. The intervention did not significantly improve symptoms of angina or self-perceived health, although end points favored collaborative care for 10 of 13 prespecified measures. While concordance with practice guidelines improved 4.5% more among patients receiving collaborative care than among those receiving usual care (P < .01), this was mainly because of increased use of diagnostic testing rather than increased use of recommended medications.
Conclusion A collaborative care intervention was well accepted by primary care providers and modestly improved receipt of guideline-concordant care but not symptoms or self-perceived health in patients with stable angina.
Collaborative models of care are multifaceted interventions that typically include planned interactive communication between primary care providers (PCPs) and specialists.1- 3 Among patients with depression or diabetes mellitus, collaborative care has improved processes and outcomes of treatment,2,3 with effect sizes that rival many pharmacologic therapies.1 Moreover, care organized in this fashion may become a key component of the patient-centered medical home.4
Ischemic heart disease (IHD) is another prevalent condition that requires complex pharmacotherapy and is best managed according to evidence-based guidelines.5 Although treatments for IHD decrease mortality and improve quality of life, many patients are not treated according to well-publicized guidelines.6,7
We implemented a multifaceted intervention based on a collaborative care model and directed through patients' PCPs. We sought to determine whether the intervention would improve symptoms of angina, self-perceived health, and concordance with practice guidelines for managing chronic stable angina among patients with IHD who reported angina or sublingual nitroglycerin use weekly or more often.
The Collaborative Cardiac Care Project was a prospective trial, cluster randomized by provider, involving providers and their patients with IHD at medical center–and community-based primary care clinics of 4 academically affiliated Department of Veterans Affairs (VA) health care systems. Providers were stratified according to sex, site (VA facility), clinic type (medical center vs community clinic), and provider type (staff physician, nurse practitioner or physician assistant, or trainee). Providers were enrolled from October 7, 2004, through March 30, 2005, and patients consented to study participation between July 29, 2005, and March 9, 2007. The study protocol was approved by the institutional review boards at participating sites. Providers and patients furnished written informed consent.
Primary care providers caring for patients with IHD in general medicine clinics were asked to participate. Providers who were expected to leave the clinic within 1 year and physicians who participated in the collaborative care teams were excluded.
Potential participants with IHD were identified using VA administrative databases. Patients assigned to participating providers and confirmed to have had at least 1 visit in the past year were sent screening questionnaires to determine severity of angina using the anginal frequency subscale of the Seattle Angina Questionnaire (SAQ).8 The SAQ scores range from 0 to 100, with higher scores indicating less frequent angina. A score of 70 or below represents angina or nitroglycerin use weekly or more often. Patients scoring 70 or below and confirmed to be suitable participants by their PCPs were mailed invitations to join the study, along with consent forms and baseline questionnaires.
A collaborative care team at each site, composed of a cardiologist (J.S.R., P.H., G.L., and M.S.), general internist (S.D.F., M.D., M.G., P.I.L., L.L., C.M., K.N., and C.B.), research assistant (M.M. and M.E.P. and others), and clinical nurse specialist (C.D.) or pharmacist, met twice monthly to review patients' records, develop diagnostic and treatment plans, and conduct progress evaluations (Figure 1). Research assistants (M.M. and M.E.P.) interviewed patients by telephone about symptoms and medications, including adherence, activities, and care received outside of the VA. Relevant clinical notes, tests, laboratory results, and medications were summarized before each review. Using evidence-based guidelines5,13 and clinical judgment, the collaborative care teams made recommendations for medication adjustments, cardiac diagnostic testing, laboratory tests, referrals, consultations, or other assessments. Recommendations were transmitted in the form of unsigned orders and notes in the electronic health record to PCPs, who were asked to sign, modify, or reject. Patients who received recommendations were reviewed more frequently to assess implementation of and response to recommendations.
Collaborative care patients were mailed educational materials about IHD and were invited to attend an educational session at their VA facility. Control patients received usual care.
All patients were mailed follow-up questionnaires at 4, 8, and 12 months composed of the following: the SAQ, consisting of 19 items on anginal frequency, physical limitation, treatment satisfaction, disease perception, and anginal stability; the Veterans RAND 12-Item Health Survey,12 including a single embedded question on self-rated health14; the Seattle Outpatient Satisfaction Questionnaire11; a 3-item alcohol consumption measure based on the Alcohol Use Disorders Identification Test9; and the Patient Health Questionnaire, a measure of depression.10 A change of 5 to 10 points on the 100-point SAQ is considered clinically important.15 Deaths were identified using the VA Vital Status File.
In addition to careful management of coexisting conditions, such as hypertension and diabetes mellitus, effective treatment of chronic stable angina includes (1) antiplatelet therapy, (2) lipid-lowering therapy, (3) antianginal treatment, and (4) testing for diagnosis and risk stratification.5 For each of these guideline-based criteria, we assigned a score of 1 if a patient's treatment was concordant with the guideline and 0 if not. Scores were summed to yield a guideline concordance score, ranging from 0 to 4.
Guideline-concordant antiplatelet therapy was defined as any dose of aspirin or clopidogrel bisulfate or a contraindication to antiplatelet therapy. In accord with guidelines in effect during the study,13 patients with a low-density lipoprotein cholesterol concentration of less than 100 mg/dL without lipid-lowering therapy and patients with an untreated low-density lipoprotein cholesterol concentration of 100 mg/dL or higher who were receiving therapy were considered concordant (to convert cholesterol concentration to millimoles per liter, multiply by 0.0259). Guidelines for antianginal therapy specify a stepwise addition of β-blockers, calcium channel blockers, and long-acting nitrates.5 A computerized version of criteria mapping methods16 was used to classify each patient's antianginal therapy as guideline concordant or not. Each patient was evaluated for contraindications to β-blockers (eg, bradycardia), and if none was found and an acceptable β-blocker regimen based on the VA National Formulary was prescribed, care was rated as concordant. The same logic was applied to calcium channel blockers and then long-acting nitrates. If at any point the patient had no contraindication and was not prescribed the medication in question, care was judged not to be concordant. Concordance with guidelines for diagnostic testing was broadly defined as any prior exercise stress test, cardiac stress imaging, or coronary angiogram.
Using VA pharmacy records, we defined guideline-concordant care as medication filled within 180 days before the date of interest. Diagnostic testing information was obtained using a formal medical record abstraction tool. Guideline concordance was assessed on the date of the baseline questionnaire and at 12 months. Because adherence to long-term cardiovascular medications tends to decline over time, we calculated the odds of improving or remaining the same on the guideline concordance score.17,18
Provider satisfaction was assessed at baseline and at 12 months using the Seattle Provider Satisfaction Survey.19 This survey measures the following 6 domains: communications, clinic environment, organizational factors, interaction with specialists, patient characteristics, and philosophy of practice.
Change in the SAQ anginal frequency score was the primary end point. Initial sample size estimates were determined based on detection of a 5- to 10-point change in this measure. The dependent variable was the change between the baseline value and the 12-month value, adjusted by regression for the frequency of angina at baseline. The unit of analysis was the patient. Change in anginal frequency was regressed on treatment group, adjusting for characteristics of patients, including baseline age, sex, SAQ anginal frequency score, non-VA cardiology care, sex and type of provider, VA facility site, treatment at a community clinic, and interaction terms. Similar regression analyses were performed for all end points, adjusted by regression for the baseline values.
To adjust for clustering of patients within providers, we used cross-sectional time-series generalized estimating equations regression. We included an exchangeable covariance matrix using XTGEE in STATA (StataCorp LP, College Station, Texas).20
Because of attrition, some patients did not have 12-month symptoms of angina and self-perceived health measures. To permit an intent-to-treat analysis, we substituted 8- or 4-month values when available, and for the few patients with no follow-up data, we assumed that their 12-month data were missing at random and imputed them via the STATA program for imputation using chained equations.20 Missing data were imputed from all variables in the relevant analysis regression equation plus the baseline SAQ physical limitation score, baseline SAQ anginal frequency score, self-rated health, and (when available) change from baseline to the latest value for each variable using multiple imputation (MICOMBINE routine in STATA).20 For patients who died during the study period, we used 8- or 4-month values when available and imputed values for those who could have responded to follow-up questionnaires based on dates of death. As an alternative approach, we used a version of the Veterans RAND 12-Item Health Survey that accounts for death by transforming the physical component summary score and mental component summary score to variables, which are interpreted as the estimated probability of being healthy 1 year later.21 Guideline concordance analyses excluded patients who died during the study period, as 12-month data were unavailable.
Of 450 PCPs, 343 met inclusion criteria, and 283 of these consented to participate in the study. One hundred had no consenting patients, leaving 183 in the study group (91 in the collaborative care group and 92 in the control group) (Figure 2).
Using VA administrative databases, we identified 25 125 patients with IHD, of whom 17 485 were assigned to consenting providers and were sent screening questionnaires to determine eligibility. Of these, 5700 (32.6%) did not respond, 3951 (22.6%) refused participation, 379 (2.2%) reported that they did not have angina, 341 (2.0%) were deceased, and 81 (0.5%) had no valid mailing address or had transferred care elsewhere.
Of 7033 respondents to the screening questionnaire, 1693 were eligible for the study and were mailed baseline questionnaires and consent forms. Of these, 329 did not respond to multiple mailings, 311 refused participation, 298 had transferred care, and 52 had died. The study group consisted of 703 patients (359 in the control group and 344 in the collaborative care group).
Of 703 patients, 113 had no symptoms of angina or self-perceived health measures at 12 months. The 4- or 8-month value was carried forward for 84 persons, and the 12-month measure was imputed for the remaining 29. Thirty-three patients died during the study, 14 in the collaborative care group and 19 in the control group (P = .44). Of these, 20 had at least 1 symptom of angina and self-perceived health measurement after baseline; 5 could have responded to follow-up surveys based on dates of death, and we imputed their symptoms of angina and self-perceived health values; 8 died before the first follow-up survey. The transformed physical component summary and mental component summary scores to estimate the probability of being healthy, while including a value for death, produced the same results as analyses using the untransformed Veterans RAND 12-Item Health Survey variables.
The mean age of patients was 68 years, 97.7% were male, 87.6% received most of their care through the VA, and 18.9% received at least some non-VA cardiology care. Patients in the collaborative care and control groups had similar baseline characteristics, including SAQ anginal frequency score (52.8 vs 53.8, P = .46) (Table 1). The collaborative care group had a higher mean guideline concordance score at baseline (3.0 vs 2.9, P = .01), fewer patients reporting prior myocardial infarction (51.2% vs 61.8%, P = .01), and lower mean body mass index (calculated as weight in kilograms divided by height in meters squared) than the control group (30.2 vs 31.4, P = .03). Collaborative care and usual care providers did not differ by sex, provider type, or VA facility clinic type.
The collaborative care teams conducted a total of 895 reviews and made 701 evaluation and treatment recommendations (Table 2). Twenty-six patients (7.6%) in the collaborative group were reviewed only once because medications were prescribed by outside providers or because they died before a second review; 138 patients (40.1%) were reviewed twice, 134 (39.0%) three times, and 46 (13.4%) more than 3 times (the maximum number of reviews was 5). During the study period, 147 (42.7%) had 1 recommendation, 53 (15.4%) had 2 recommendations, 48 (14.0%) had 3 recommendations, 66 (19.2%) had 4 or more recommendations, and 30 patients (8.7%) had no recommendations made by the collaborative care teams. Primary care providers accepted (signed) and implemented 91.6% (642 of 701) of recommendations.
Forty-seven percent (328 of 701) of recommendations were related to medications. Adjustments to β-blockers, long-acting nitrates, and statins were most frequent. Of 701 recommendations, 21.8% (n = 153) were for referrals, consultations, or other assessments; 14.3% (n = 100) for cardiac diagnostic testing; and 8.7% (n = 61) for laboratory tests. Other recommendations (eg, PCP call or follow-up visit with the patient to review symptoms or compliance) constituted 8.4% (n = 59). Of 59 recommendations that PCPs did not implement, 22.0% (n = 13) were to add or increase β-blockers, 15.3% (n = 9) were to add or increase long-acting nitrates, 15.3% (n = 9) were for cardiac diagnostic testing, 13.6% (n = 8) were to add or change lipid-lowering therapy, 11.9% (n = 7) were referrals, 10.2% (n = 6) were to add or discontinue short-acting nitrates, 10.2% (n = 6) were changes to other medications, and 1.7% (n = 1) were for laboratory testing.
The intervention did not significantly improve symptoms of angina or self-perceived health, although results favored collaborative care for 10 of 13 measures (Table 3). Patients in the collaborative care group improved 0.18 more points (4.5%) than the control group on the guideline concordance score between baseline and 12 months (P < .01). The adjusted odds of improving or remaining the same on the guideline concordance score were 1.90 (95% confidence interval, 1.22-3.10) times higher for the collaborative care group compared with the control group (P < .01). Table 4 lists the full regression model for angina symptoms and guideline concordance score. Patients at VA facility sites 1 and 2 were more likely to exhibit improvement in guideline concordance scores than those at site 3. However, as summarized in Table 5, increased cardiac diagnostic testing accounted for most of the improvement observed in guideline concordance score rather than increased use of recommended medications. Baseline SAQ anginal frequency score and guideline concordance score were uncorrelated. Change in provider satisfaction from baseline to 12 months did not differ between the collaborative care group and the control group.
Two adverse events occurred in the collaborative care group. A patient with Canadian Cardiovascular Society class 3 angina and heart failure whose pharmacologic therapy for angina was optimized according to practice guidelines was subsequently admitted for acute myocardial infarction complicated by ventricular tachycardia and femoral artery occlusion. Ultimately, he was discharged home on the medication regimen recommended by the collaborative care team. Another patient underwent catheterization as recommended by the collaborative care team and was found to have 3-vessel disease. Angioplasty of the right coronary artery resulted in dissection and myocardial infarction. Following discharge, he was readmitted for syncope and underwent pacemaker placement. Both events were reviewed by the local institutional review boards, which recommended no changes in study procedures.
The collaborative care intervention had no significant effects on symptoms of angina or self-perceived health but significantly improved receipt of guideline-concordant care in patients with stable angina. The collaborative care model was well received by PCPs, who implemented 91.6% of recommendations.
Although prior investigations in this population revealed suboptimal management of chronic stable angina,6 patients in the present study were generally well managed at baseline, as evidenced by mean guideline concordance scores of 3.0 (collaborative care group) and 2.9 (control group) on a scale ranging from 0 to 4. This suggests that earlier guideline dissemination had been effective and that the study may have been limited by a ceiling effect. A similar intervention in a less well-managed population may have produced a larger effect.
The intervention was a facilitated specialist consultation, with systematic selection of patients experiencing significant anginal symptoms. Compared with more intensive case management models, the intervention was low intensity, including a mean of 2.5 reviews per patient. This may have limited the achievable effect.
Improvement in guideline-concordant care has been shown to be correlated with improved cardiovascular outcomes.23,24 However, the improvement in guideline concordance observed in the present study was related mainly to increased diagnostic testing in the collaborative care group, which did not translate to improved symptoms of angina or self-perceived health.
A positive finding was the avid uptake (91.6%) of recommendations by PCPs. Interventions that are highly effective in altering physicians' practice are uncommon. Previous efforts to change provider performance using promulgation of guidelines, continuing medical education, audit and feedback, and reminder systems have been modestly effective at best.25 The fact that evidence-based recommendations were provided by respected colleagues via the electronic health record likely contributed to the high level of acceptance in this study, suggesting that the intervention facilitated communication of clinical information between PCPs and specialists. This finding is relevant in light of emerging concepts about the structure of primary care. The patient-centered medical home has been widely endorsed as a strategy to increase access and quality of care.26 In this model, comprehensive primary care is delivered by a team of individuals led by a PCP and supported by systems that improve quality and efficiency through enhanced care coordination and patient centeredness.26 Communication and collaboration are essential among PCPs and subspecialists, although there has been little research on how to facilitate these interactions.1,27
Systems that facilitate communication without overburdening PCPs are necessary adjuncts to the patient-centered medical home model. Our intervention exploited the order entry feature of the electronic health record to efficiently and effectively communicate recommendations from specialists to PCPs, who could then easily sign, modify, or reject the orders. This approach preserved the central role of the PCP, who (as the hub of information flow and decision making) typically possesses the best overall understanding of a patient's multiple and interacting problems.
Strengths of the study include a population-based approach to patient selection based on clinical criteria, multisite randomized design, a sophisticated hierarchical analysis, systematic enrollment of providers and patients, and use of the electronic health record to implement the intervention. However, limitations to our analysis include substantial nonresponse to the invitation to participate in the study related to the 2-stage mail-in consent process mandated by the institutional review boards. In addition, misclassification of anginal pain may have occurred, despite classification of angina being based on the SAQ, a reliable, responsive, and validated measure of angina.8 Last, generalization beyond integrated care systems similar to the VA is limited.
In this study, a collaborative care intervention was acceptable and feasible for PCPs. It modestly improved receipt of guideline-concordant care but not symptoms or self-perceived health in patients with chronic stable angina.
Correspondence: Stephan D. Fihn, MD, MPH, Veterans Affairs Puget Sound Health Care System, 1100 Olive Way, Ste 1400, Seattle, WA 98101 (firstname.lastname@example.org).
Accepted for Publication: May 24, 2011.
Author Contributions: The authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Fihn, McDonell, Diehr, Rumsfeld, Gerrity, Lee, McBryde, and Stadius. Acquisition of data: Fihn, McDonell, Rumsfeld, Doak, Dougherty, Gerrity, Larsen, Lee, Lucas, McBryde, Nelson, Plomondon, and Bryson. Analysis and interpretation of data: Fihn, Baucher, McDonell, Diehr, Dougherty, Gerrity, Heidenreich, and Nelson. Drafting of the Manuscript: Fihn, Bucher, McDonell, Diehr, Lucas, Nelson, and Bryson. Critical revision of the manuscript for important intellectual content: Fihn, McDonell, Diehr, Rumsfeld, Doak, Dougherty, Gerrity, Heidenreich, Larsen, Lee, McBryde, Nelson, Plomondon, Stadius, and Bryson. Statistical expertise: Diehr. Obtained funding: Fihn, McDonell, and Gerrity. Administrative, technical, and material support: McDonell, Plomondon, and Bryson. Study Supervision: Fihn, Rumsfeld, Gerrity, Larsen, Stadius, and Bryson.
Financial Disclosure: None reported.
Funding/Support: This work was supported by grant IHI 02-062 from the Department of Veterans Affairs, Veterans Health Administration, Health Services Research and Development Service (Drs Fihn, Rumsfeld, Heidenreich, and Gerrity). Dr Bryson was supported by Career Development Award RCD 03-177 from the same funding source.
Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs.