Volume 118, Issue 12 , Pages 1414.e13-1414.e19, December 2005
Bridging the quality gap in diabetic hyperlipidemia: A practice-based intervention
Article Outline
Abstract
Purpose
Dyslipidemia treatment dramatically decreases coronary heart disease risk in diabetes, yet only a minority of these patients are screened or achieve optimal low-density lipoprotein (LDL) cholesterol levels. Our aim was to increase the percentage of diabetic patients in whom lipid management was achieved through electronic and direct educational detailing.
Methods
The study cohort comprised 884 diabetic patients at 12 primary care practices. Practice sites were randomized to one of three intervention groups: electronic educational detailing, direct (face-to-face) educational detailing, or control. Direct and electronic detailing were performed over a 12-month period. All sites were notified of our goal to enhance lipid testing among diabetic patients. Chart abstraction was performed 15 months after the start of the intervention. For the entire population (n
=884), the proportion of patients with lipid testing was calculated, and changes from pre- to postintervention were compared across groups. We compared pre- and postintervention LDL-cholesterol changes between groups using least square means to account for site variation.
Results
Favorable provider actions increased significantly with the intervention (+22% compared with +6% in controls, P=.01). By logistic regression, electronic detailing increased the likelihood of lipid testing (odds ratio 3.0, confidence interval 1.6-5.7), as did direct detailing (odds ratio 1.8, confidence interval 0.9-3.7) in patients with no preintervention LDL test (n=432). Lipid testing tended to increase to a greater extent at intervention sites (+23% for the combination of electronic and direct detailing vs +11% for controls, P=.06).
Conclusions
Brief educational detailing either through direct or electronic communication favorably impacts provider behavior regarding dyslipidemia care for diabetic patients.
Keywords: Diabetes , Hyperlipidemia , Quality , Technology , Detailing
Diabetic patients have a substantially increased risk for cardiovascular mortality, as evidenced by their identical risk for an initial myocardial infarction as nondiabetic patients with established coronary heart disease.1, 2 Because of this formidable risk, the latest iteration of the National Cholesterol Education Program 3 guidelines elevated the status of diabetes to that of a coronary heart disease risk equivalent, decreasing the low-density lipoprotein (LDL)-cholesterol treatment goal to less than 100 mg/dL.3 Recently, the American College of Physicians suggested that lipid-lowering therapy should be universally prescribed to adult patients with diabetes and just one other cardiovascular risk factor such as age greater than 55 years, smoking, or hypertension.4 Yet, despite these national guidelines, a major treatment gap exists; lipid-lowering therapy continues to be underused in patients with diabetes.5 For example, we previously reported that only 16% of diabetic patients with established coronary heart disease, from practices in a large metropolitan area, achieved target LDL-cholesterol levels.6 Despite advances in cardiovascular care that have significantly decreased event rates and mortality in the general population, mortality rates remain high and unchanged among diabetic patients.7 This is disturbing given the unequivocal benefits of lipid-lowering therapy observed in large randomized controlled trials and the burgeoning prevalence of diabetes.8, 9
Against this background of potential quality improvement opportunities, we sought to define modes of provider communication that might better facilitate the integration of, and adherence to, diabetic lipid-lowering guidelines in clinical practice. We hypothesized that a physician-directed treatment algorithm, disseminated through direct detailing or by electronic transmission, would improve the percentage of diabetic patients in whom lipid testing was performed.
Methods
The Colorado Multiple Institutional Review Board approved our study. We sought to test the hypothesis that a physician-directed intervention would increase the number of annual lipid profiles ordered. Secondary end points included the proportion of patients in whom a favorable provider action was taken to address dyslipidemia (increase in lipid tests, lipid-lowering drug started, or the dose increased) as well as the change in LDL-cholesterol levels among patients at control and intervention sites.
Study sample
Forty-five sites were initially contacted, which ultimately resulted in 12 participating sites in the Denver-metro area. Practices were contacted on the basis of their previous participation in a diabetes initiative sponsored by the Colorado Foundation for Medical Care, Colorado’s official medical quality improvement organization. To be an eligible site, each practice had to care for a minimum of 100 Medicare beneficiaries with diabetes and have the capability of accurately identifying their Medicare subscribers with diabetes. We included only patients aged more than 40 years to focus on type 2 diabetes mellitus. A total of 884 patients were evaluated during the study period.
Intervention
A stratified randomization scheme was performed on the basis of the type of practice (family medicine, internal medicine, or academic medicine). The recruited practices were randomly assigned to either control (5 sites, 323 patients) or one of two intervention groups: electronic detailing intervention (4 sites, 415 patients) and direct detailing (3 sites, 146 patients). In both intervention arms, we provided diabetic/lipid-based education to physicians on a quarterly basis over the course of 12 months. We also briefed the nonphysician office staff members about the intervention. Control sites were made aware of the overall goal of the initiative to improve lipid management in diabetes. However, no other formalized interaction was provided for the control group.
Specifically, in both intervention arms, we provided physician and nonphysician office staff with a one-page educational guideline focused on the care of diabetic patients with hyperlipidemia. Contents of the educational materials were derived from the National Cholesterol Education Program guidelines and validated by opinions from local lipid experts. The primary message was to encourage physicians to order lipid profiles in diabetic patients. The material also urged physicians to be more interactive with their patients: to discuss lipid profile results, provide patients with information on lipid disorders, and stimulate patients to ask their providers questions about their level of lipid control.
Sites randomized to electronic detailing received on a quarterly basis one-page facsimiles and e-mails, highlighting 8 to 10 concise and practical treatment recommendations for managing hyperlipidemia in diabetic patients. In contrast, sites randomized to direct detailing received quarterly, on-site, face-to-face educational sessions (15-30 minutes) conducted by a local physician opinion leader. The focus of these sessions was to reinforce current lipid treatment guidelines and answer specific questions regarding the treatment of hyperlipidemia in diabetic patients.
Data abstraction
At the completion of a 15-month evaluation period (12 months of intervention plus an additional 3 months of follow-up), trained data abstractors were sent to each participating office. All patient visits from 1 year before to 15 months after the initiation of the intervention were included.
Data regarding visit date, date of birth, ethnicity, comorbid conditions including hypertension, medications, and smoking behavior were abstracted from patient charts. Ascertainment of lipid profiles was the primary focus of the abstraction. The chart abstractors recorded levels of total cholesterol, triglycerides, LDL, and high-density lipoprotein (HDL). The last lipid profile value before the initiation of the intervention (if available) was considered as the baseline. The postintervention LDL-cholesterol value was defined as the last LDL-cholesterol value recorded in the 15 months after the initiation of the intervention (if available).
When lipid profile results contained a “not available” designation for the LDL-cholesterol because of an elevated triglyceride level, we derived the value using the following well-accepted formula: LDL-cholesterol = total cholesterol − HDL-cholesterol − (triglycerides divided by 5) when triglyceride levels were less than 600 mg/dL.10 For triglyceride values between 600 and 700 mg/dL, the following adjustment was used: triglycerides divided by 6, and for triglycerides between 700 and 800 mg/dL: triglycerides divided by 7, and so on. This calculation method was considered adequate for estimation of LDL-cholesterol in the study population.
The lipid-lowering medications prescribed at each visit during the study period were abstracted from the charts. They were categorized as HMG-CoA reductase inhibitors (statins), niacin, fibrates, and bile acid sequestering agents.
Statistical analysis
SAS software version 8.2 (SAS Institute, Cary, NC) was used for all statistical analyses. Two-sided P values of less than .05 were considered significant. Continuous variables were summarized as mean and standard deviation, and categoric variables were summarized as n and percentage per group. All comparisons were performed by the three intervention groups (control, direct, and electronic). Secondarily, comparisons were performed to test the difference between the control group and any type of intervention (direct and electronic). Because the randomization unit was a study site, analyses at the patient level were often adjusted for potential site-specific confounders.
The primary and secondary outcomes were based on LDL values obtained and reported on chart review of eligible patients at each study site. The baseline LDL value was defined as the last LDL value recorded in the year before the initiation of the study intervention. The follow-up LDL value was defined as the last LDL value recorded in the 15-month follow-up period. LDL was calculated from total cholesterol, HDL, and triglycerides using the formula described above.
Three analysis populations were analyzed depending on the data available for each patient: all patients enrolled in the study, n=884; patients with no preintervention LDL, n=432; and patients with both pre- and postintervention LDL values, n=359. Table 1, Table 2 describe the population at the site and patient level, respectively.
Table 1. Study site demographics (n=884)
| Site | Intervention | Type⁎ | Patients (n) | Preintervention LDL (%) | Postintervention LDL (%) | ||
|---|---|---|---|---|---|---|---|
| LDL test⁎ | LDL goal⁎ | LDL test⁎ | LDL goal⁎ | ||||
| 1 | Electronic | Internal medicine | 131 | 54.2 | 19.8 | 83.2 | 43.5 |
| 2 | Control | Internal medicine | 43 | 23.3 | 9.3 | 30.2 | 9.3 |
| 3 | Control | Internal medicine | 78 | 44.9 | 12.8 | 61.5 | 33.3 |
| 4 | Electronic | Internal medicine | 13 | 61.5 | 38.5 | 84.6 | 46.2 |
| 5 | Direct | Internal medicine | 33 | 66.7 | 15.2 | 87.9 | 42.4 |
| 6 | Electronic | Family medicine | 46 | 71.7 | 23.9 | 80.4 | 41.3 |
| 7 | Control | Family medicine | 44 | 75.0 | 29.5 | 88.6 | 36.4 |
| 8 | Control | Family medicine | 118 | 70.3 | 33.1 | 87.3 | 44.1 |
| 9† | Direct | Family medicine | 45 | 57.8 | 22.2 | 71.1 | 28.9 |
| 10 | Electronic | Academic | 225 | 47.6 | 20.4 | 69.3 | 40.9 |
| 11 | Control | Academic | 40 | 32.5 | 15.0 | 35.0 | 25.0 |
| 12† | Direct | Internal medicine | 68 | 16.2 | 7.4 | 58.8 | 20.6 |
⁎ LDL test = percentage of patients who had a lipid profile, and LDL goal = percentage of patients with an LDL <100 mg/dL. |
† Two sites were automatically put in the direct intervention group as a replacement for two sites randomized to the direct intervention group who dropped out. |
Table 2. Patient demographics (n=884)
| Control (n=323) | Electronic (n=415) | Direct (n=146) | |
|---|---|---|---|
| Sites | 5 | 4 | 3 |
| Age, mean (SD) | 66.0(12.0) | 61.6(11.0) | 65.3(14.3) |
| Male, n (%) | 152(47.1) | 187(45.1) | 61(41.8) |
| Race, n (%) | |||
| White | 110(34.1) | 101(24.3) | 52(35.6) |
| African American | 26(8.1) | 6(1.5) | 17(11.6) |
| Hispanic | 12(3.7) | 212(51.1) | 52(35.6) |
| American Indian | 1(0.3) | 0(0.0) | 1(0.7) |
| Asian | 63(19.5) | 2(0.5) | 2(1.4) |
| Other | 2(0.6) | 3(0.7) | 0(0.0) |
| UTD | 109(33.8) | 91(21.9) | 22(15.1) |
| Health coverage, n (%) | |||
| Private/private+other | 209(64.7) | 177(42.7) | 68(46.6) |
| Medicaid/Medicare/other government | 109(33.8) | 205(49.4) | 72(49.3) |
| Uninsured/other/UTD | 5(1.6) | 33(8.0) | 6(4.1) |
| Smoking status, n(%) | |||
| Current | 35(10.8) | 75(18.1) | 15(10.3) |
| Former/never | 157(48.6) | 204(49.2) | 91(62.3) |
| UTD | 131(40.6) | 136(32.8) | 40(27.4) |
The primary analysis was done at the study site level for all patents in the study (n=884). The percentage of patients at each site with an LDL test pre- and postintervention was calculated. The difference from pre- to postintervention in the average percentage per treatment group was compared across treatment groups using the Kruskal-Wallis test (Table 3). A higher difference indicates a larger percent increase in LDL tests from pre- to postintervention.
Table 3. Change in percentage of patients with low-density lipoprotein cholesterol evaluated from pre- to postintervention (n=884)
| Control (sites=5) | Electronic (sites=4) | Direct (sites=3) | P value⁎ | |
|---|---|---|---|---|
| Preintervention, mean (SD) | 49.2(22.8) | 58.8(10.4) | 46.9(27.0) | |
| Postintervention, mean (SD) | 60.5(27.7) | 79.4(6.9) | 72.6(14.6) | |
| Change (post-pre), mean (SD) | 11.3(6.4) | 20.6(8.6) | 25.7(15.2) | .17 |
| Control (sites=5) | Electronic/direct (sites=7) | |||
|---|---|---|---|---|
| Change (post-pre), mean (SD) | 11.3(6.4) | 22.8(11.0) | .06 | |
⁎ P value is a Kruskal-Wallis test. |
The proportion of patients experiencing at least one of the predefined provider actions was calculated per site for the pre- and postintervention periods. Favorable provider actions were defined as: having an LDL value recorded, having a new lipid medication prescribed, and having an increase in the dosage of the lipid-lowering medication.
The difference in proportion from pre- to postintervention was tested for its association with the interventions using the Kruskal-Wallis test. This site level analysis includes all study subjects (n=884).
Results
The type of intervention performed and characteristics of the 12 study sites are listed in Table 1. Table 1 also presents the percentage of patients with available LDL-cholesterol data by practice site, ranging from 16% to 75% of patients preintervention and from 30% to 89% of patients at postintervention. Direct detailing was provided to 3 sites, electronic detailing was provided to 4 sites, and an additional 5 sites served as controls. Among the 884 patients, 253 (29%) were seen in private family medicine practice sites, 366 (41%) were treated in private internal medicine sites, and 265 (30%) were treated in academic medicine practices. The sociodemographic characteristics of the patients are depicted in Table 2.
Our principal outcome was the change from baseline in the proportion of patients receiving lipid testing at each site. We observed a favorable trend for a larger increase in the frequency of lipid testing at intervention sites (+23% combination of electronic and direct detailing) compared with control sites (+11%), although the difference did not achieve statistical significance (P=.06) (Table 3).
Among those patients who did not have a preintervention lipid panel drawn, we observed a 3.0 (1.62-5.66) times greater likelihood of lipid testing in the electronic group compared with the control group using logistic regression (P=.0005), even after adjusting for type of site and race. There was a moderate difference between the direct group and control group (P=.09); however, there was no evidence of a difference between the electronic and direct groups (P=.17). Table 4 shows the odd ratios and 95% confidence intervals for the entire logistic model. Patients in academic and internal medicine sites were significantly less likely than patients in family medicine sites to have a favorable change in lipid testing. Patients with private insurance were more likely to have a favorable change in lipid testing than those who were governmentally insured (Medicaid or Medicare).
Table 4. Logistic model of the probability of post–low-density lipoprotein test among patients with no pre–low-density lipoprotein test (n=432)
| Covariate | Odds ratio (95% confidence interval) |
|---|---|
| Treatment (P=.002) | |
| Electronic vs control | 3.0(1.62-5.66) |
| Direct vs control | 1.8(0.90-3.65) |
| Type of site (P=.002) | |
| Academic vs family medicine | 0.2(0.10-0.53) |
| Internal vs family medicine | 0.4(0.22-0.87) |
| Health insurance (P=.09) | |
| Uninsured/other/UTD vs | |
| Medicaid/Medicare/other government | 1.6(0.59-4.11) |
| Private/private+other vs | |
| Medicaid/Medicare/other government | 1.7(1.04-2.81) |
For the population of patients with both pre- and post-LDL-cholesterol measurements (n=359), a mixed effects model was used to compare the percent change in LDL-cholesterol levels from pre- to postintervention while adjusting for the study site. We observed a decrease in LDL-cholesterol levels in all groups (electronic 111 to 97 mg/dL, direct 115 to 104 mg/dL, and control 109 to 101 mg/dL), although there were no differences in magnitude between the groups (P=.4). Last, the change in the proportion of patients experiencing at least one of the predefined favorable provider actions for the pre- and postintervention periods was significantly greater in the combined direct and electronic groups versus controls (P=.01) (Table 5). Of note, there were 160 patients who never had an LDL test during the pre- or postintervention period despite having at least one visit in each of the time periods.
Table 5. Provider action (n=884): Change from pre- to postintervention in percentage of patients per site with at least one favorable provider action
| Control (sites=5) | Electronic (sites=4) | Direct (sites=3) | P value⁎ | |
|---|---|---|---|---|
| Preintervention action, mean (SD) | 56.5(22.5) | 61.0(8.5) | 51.1(24.3) | |
| Postintervention action, mean (SD) | 62.5(26.9) | 80.4(5.6) | 76.8(11.6) | |
| Change (post-pre), mean (SD) | 5.9(7.9) | 19.4(7.2) | 25.7(13.4) | .04 |
| Control (sites = 5) | Electronic/direct (sites=7) | |||
|---|---|---|---|---|
| Change (post-pre), mean (SD) | 5.9(7.9) | 22.1(9.9) | .01 | |
⁎ P value is a Kruskal-Wallis test. |
Discussion
Our results suggest that electronic detailing of lipid-treatment information along with direct detailing translate into modest improvements in lipid management among high-risk diabetic patients in a community-based primary care setting. Coronary heart disease remains the principal cause of premature morbidity and mortality in the United States among diabetic patients. Hypercholesterolemia is a significant coronary heart disease risk factor in this population, and with the advent of statins, it is relatively straightforward to treat. However, despite this ease of treatment, recent data indicate that only a minority of diabetic patients are being optimally treated, even in those diabetic patients with coexistent coronary heart disease.11 In the third National Health and Nutrition Survey 1999 to 2000, more than half of diabetic patients had total cholesterol values greater than 200 mg/dL.12 Our study suggests that an opportunity exists to bridge this gap, between compelling scientific evidence that favors aggressive lipid-lowering and community practice, through a simple intervention. The intervention is based on communication through either electronic or direct provider detailing by local opinion leaders. Of note in a very recent study, mailing educational materials to patients with hypertension did not yield a significant increase in blood pressure control.13
Changing physician behavior to expedite the application of recent therapeutic advances into everyday clinical practice is a challenging task. The ability to positively change physician’s practices could have a profound effect on the quality of health care. In the past, clinical practice guidelines by themselves have had a limited role to alter the way physicians practice or improve patient outcomes.14 The publication of landmark randomized controlled trials is also clearly not sufficient to ensure timely adoption of the new evidence into medical practice.15 Although the traditional treatment focus for diabetic patients has been in the area of glycemic control, there is growing evidence that improvements in lipid levels provide greater reductions in coronary heart disease risk.16 This message is critically important because patients with type 2 diabetes possess a coronary heart disease risk that is fourfold greater than the general population.17 Moreover, aggressive dyslipidemia management in diabetic patients has been shown to be as cost-effective as in nondiabetic patients with overt coronary heart disease.18
In our study, an electronic-based intervention using both facsimiles and e-mail favorably impacted lipid evaluation and management. Improvements in care process were similar to that achieved by direct face-to-face detailing. This is somewhat surprising given the one-way nature of communication inherent to electronic detailing in contrast with direct detailing, which is an interactive process delivered face-to-face. However, the efficiency advantages of a successful technology-based intervention are hard to ignore. Electronic communications have the potential to more widely disseminate best practices and are more amenable to adaptation as population-based interventions in diabetes care. This could ensure that the cardiovascular health care needs of type 2 diabetic patients are more optimally met.
We acknowledge potential limitations to our study. Data quality are limited by the chart abstraction process. We were unable to determine basic demographics such as race and smoking status on up to 40% of the population. Our stratified randomization resulted in an unequal number of patients in each of the three groups, decreasing our statistical power to adequately discern modest differences in process of care indicators. In particular, small differences between electronic and direct detailing were noted but need to be interpreted cautiously. It is tempting to infer that electronic detailing was equally effective compared with direct detailing given the lack of statistically significant difference. However, larger studies are required to confirm this finding given the established benefits and widespread adoption of direct detailing.19 Another limitation is that our randomization resulted in sizeable differences in the insurance and racial mix between intervention groups. Although we attempted to control for this variation in the analysis, differences in outcomes could still be attributed to unmeasured factors associated with race and insurance status. These limitations notwithstanding, the positive effects of our intervention are probably underestimated because we notified all practice sites (including control groups) of the general purpose of the study initiative.
Our results contrast with recent data from Reiber et al20 demonstrating that even ambitious computerized detailing programs may have no effect on achieved lipid values and other important process of care measures in type 2 diabetes. This may reflect the inability of overextended primary care providers to add new tasks to an already full schedule. Unless an intervention or process of care alteration reduces provider workload, there is little likelihood of long-term adherence. Because our intervention was simple and occurred only quarterly it may have been more acceptable to clinicians in a busy practice environment. In addition, Reiber et al’s study did not target the provider-patient interaction, in which a constructive dialogue may be essential for optimizing the process of care. Our intervention emphasized the importance of an interactive discussion with patients, which may have enhanced its effectiveness.
Relatively little information is available to clarify the reasons why patients are inadequately screened for dyslipidemia or fail to receive lipid-lowering agents.21 Previously, physicians have indicated that they may not have been aware of the recommended LDL-cholesterol goals for patients with coronary heart disease or coronary heart disease risk equivalents such as diabetes mellitus.22 These observations emphasize the potential importance of physician education efforts such as those described in this study.23 Despite all the advances of medicine, health care providers often do not screen or treat patients appropriately during health care visits. Such behaviors have been referred to as clinical inertia.24 Almost 20% of these high-risk patients never had a lipid level measured despite being seen at least twice during the study period. Thus, a substantial quality improvement opportunity exists.
Conclusion
A simple educational intervention seems to positively influence provider behavior in the area of lipid management in diabetes mellitus. Both electronic and direct detailing seem to be viable approaches. Future studies to determine optimal educational components that facilitate appropriate provider actions to initiate or intensify lipid treatment seem warranted given the burgeoning population of diabetic patients at risk for coronary heart disease morbidity and mortality.
Acknowledgments
The authors thank Nancy Gifford for her efforts implementing the initiative. We are also grateful to Adriana Padgett for administrative assistance.
References
- Impact of diabetes on mortality after the first myocardial infarction . Diabetes Care . 1998;21(1):69–75
- . Mortality from coronary heart disease in subjects with type 2 diabetes and in nondiabetic subjects with and without prior myocardial infarction . N Engl J Med . 1998;339(4):229–234
- Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in AdultsExecutive summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) . JAMA . 2001;285(19):2486–2497
- . American College of Physicians Pharmacologic lipid-lowering therapy in type 2 diabetes mellitus (background paper for the American College of Physicians) . Ann Intern Med . 2004;140(8):650–658
- . Lipid intervention trials in diabetes . Diabetes Care . 2000;23(Suppl 2):B49–B53
- . Lack of improvement in the treatment of hyperlipidemia among patients with type 2 diabetes . Am J Med . 2003;114(5):377–382
- . Diabetes and decline in heart disease mortality in US adults . JAMA . 1999;281(14):1291–1297
- . MRC/BHF Heart Protection Study of cholesterol lowering with simvastatin in 20 536 high-risk individuals (a randomised placebo-controlled trial) . Lancet. . 2002;360:7–22
- . Cholesterol lowering with simvastatin improves prognosis of diabetic patients with coronary heart disease. A subgroup analysis of the Scandinavian Simvastatin Survival Study (4S) . Diabetes Care . 1997;20(4):614–620
- . Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge . Clin Chem . 1972;18(6):499–502
- . Cardiovascular risk factor control among veterans with diabetes (the ambulatory care quality improvement project) . Diabetes Care. . 2004;27(Suppl 2):B33–B38
- . Poor control of risk factors for vascular disease among adults with previously diagnosed diabetes . JAMA . 2004;291(3):335–342
- . Impact of educational mailing on the blood pressure of primary care patients with mild hypertension . J Gen Intern Med . 2004;19:925–930
- Why don’t physicians follow clinical practice guidelines? A framework for improvement . JAMA . 1999;282(15):1458–1465
- . Synergy between publication and promotion (comparing adoption of new evidence in Canada and the United States) . Am J Med . 2003;115(6):467–472
- . Multifactorial intervention and cardiovascular disease in patients with type 2 diabetes . N Engl J Med . 2003;348(5):383–393
- . Increased incidence of coronary atherosclerosis in type 2 diabetes mellitus (mechanisms and management) . Ann Intern Med . 2003;139(10):824–834
- . How cost-effective is the treatment of dyslipidemia in patients with diabetes but without cardiovascular disease? . Diabetes Care . 2001;24(1):45–50
- . Principles of educational outreach (‘academic detailing’) to improve clinical decision-making . JAMA . 1990;263(4):549–556
- . Diabetes quality improvement in Department of Veterans Affairs Ambulatory Care Clinics (a group-randomized clinical trial) . Diabetes Care. . 2004;27(Suppl 2):B61–B68
- . Use of aspirin, beta-blockers, and lipid-lowering medications before recurrent acute myocardial infarction (Missed opportunities for prevention?) . Arch Intern Med . 1999;159(6):561–567
- . Primary care practice adherence to National Cholesterol Education Program guidelines for patients with coronary heart disease . Arch Intern Med . 1998;158(11):1238–1244
- . A community-wide survey of physician practices and attitudes toward cholesterol management in patients with recent acute myocardial infarction . Arch Intern Med . 2002;162(7):797–804
- Clinical inertia . Ann Intern Med . 2001;135(9):825–834
Financial support was received from Colorado Trust and COPIC.
PII: S0002-9343(05)00632-7
doi:10.1016/j.amjmed.2005.07.038
© 2005 Elsevier Inc. All rights reserved.
Volume 118, Issue 12 , Pages 1414.e13-1414.e19, December 2005
