Secondary Prevention following CABG: Findings of a National Randomized Controlled Trial and Sustained Society-Led Incorporation into Practice

Abstract

Background

Despite evidence supporting the use of aspirin, B-blockers, ACE inhibitors, and lipid-lowering therapies in eligible patients, adoption of these secondary prevention measures following coronary artery bypass grafting (CABG) has been inconsistent. We sought to rigorously test on a national scale whether low-intensity continuous quality improvement (CQI) interventions can be used to speed secondary prevention adherence following CABG.

Methods and Results

A total of 458 hospitals participating in the Society of Thoracic Surgeons National Cardiac Database and treating 361,328 patients undergoing isolated CABG were randomized to either a control or an intervention group. The intervention group received CQI materials designed to influence the prescription of the secondary prevention medications at discharge. The primary outcome measure was discharge prescription rates of the targeted secondary prevention medications at intervention vs. control sites, assessed by measuring pre-intervention and post-intervention site differences. Pre-randomization treatment patterns and baseline data were similar in the control (N=234) and treatment (N=224) groups. Individual medication use as well as composite adherence increased over 24 months in both groups, with a markedly more rapid rate of adherence uptake among the intervention hospitals and a statistically significant therapy hazard ratio in the intervention vs. control group for all 4 secondary prevention medications.

Conclusions

Provider-led, low-intensity CQI efforts can improve the adoption of care processes into national practice within the context of a medical specialty society infrastructure. The findings of the present trial have led to the incorporation of study outcome metrics into a medical society rating system for ongoing quality improvement.

Continuous quality improvement (CQI) in its present form emerged in the early 1980’s from the business theories of W. Edwards Deming,¹ Joseph M. Juran,² and Philip B. Crosby.³ Each stressed the importance of process measurement and feedback in any mechanism designed to measure and manage quality. Coronary artery bypass grafting (CABG) has been at the forefront of provider measurement and feedback systems. Specifically the Society of Thoracic Surgeons’ national cardiac database has been used as a platform for promoting better care and outcomes of CABG for two decades. A prior study has found that the STS infrastructure could be used to promote CQI in perioperative surgical issues.⁴ Yet it is unclear whether this system could be used to also impact pre-discharge secondary medication choices on a national scale.

The benefits of secondary prevention in patients with coronary artery disease has been established, including following CABG.⁵ Unless contraindications exist, medications at discharge after CABG should include aspirin (ASA),⁶ β-blocker,⁷ lipid-lowering therapy,⁸ and angiotensin-converting enzyme inhibitor (ACEI).⁹ The American Heart Association (AHA), American College of Cardiology (ACC), and Society of Thoracic Surgeons (STS) have published guidelines specifically directed toward these and other secondary prevention measures for CABG patients.^10,11 However, the implementation of secondary prevention therapies following CABG have lagged behind percutaneous catheter interventions.¹²

The goal of this Agency for Healthcare Research and Quality (AHRQ) grant was to evaluate whether we could successfully engage cardiac surgeons via an initiative to improve adherence to AHA/ACC/STS secondary prevention guidelines on a national scale. The design of this CQI intervention was a cluster randomized trial, randomizing sites to control or intervention.

Methods

The STS NCD

The Society of Thoracic Surgeons (STS) National Cardiac Database (NCD) was begun in 1989 and has since evolved into one of the largest specialty-specific clinical data registries in the world. The STS NCD currently houses data from over 950 participants, representing just fewer than 90% of the cardiac surgery providers in the United States, with data on more than 3.6 million procedures. In collaboration with the Duke Clinical Research Institute’s outcomes research group, the STS has developed mortality, morbidity, and length-of-stay risk models for CABG and other major cardiac procedures for adults.^13,14 Modifications aimed to expand the potential of the STS NCD to facilitate quality improvement efforts have been documented elsewhere.¹⁵

Completeness of the NCD data has been compared with data from a Centers for Medicare and Medicaid Services diagnosis–related group dataset for CABG, and no evidence for underreporting or misrepresenting of cases or events was found in the STS data.¹⁶ The quality of the data has been further assessed in a regional independent chart abstraction study, which documented a 96.2% correlation between submitted and re-abstracted data elements.¹⁶

Intervention

The intervention was a low-intensity educational effort, directed at a pre-determined local opinion leader functioning as a quality champion at each individual site. Sites received educational information designed to influence the prescription of four medications at hospital discharge: ASA, β-blockers, lipid lowering agents, and ACE inhibitor (or angiotensin receptor blocker [ARB]). Site-specific feedback reports highlighting sites’ use of these four pharmacologic agents were generated every six months, along with standardized care orders, care reminders, a ‘Call to Action’ letter on STS letterhead, and periodic newsletters. The site-specific performance data was illustrated against regional, national, and national “best practice” benchmarks. In addition, patient activation materials, including patient educational materials that stress the importance of secondary prevention medications and other lifestyle modification interventions, were supplied for review prior to hospital discharge. Finally, patients and their physician were given a discharge ‘flight-plan’ check-list of evidence-based discharge medications to be considered. Each site documented whether or not the intervention material was received and by whom.

Outcome Measures

The primary endpoint of this trial was a composite measure of site performance rate for post-CABG prevention of cardiovascular disease by prescription at discharge of ASA, β-blocker, lipid-lowering agent, and ACEI. Each patient contributed up to four records, each one corresponding to an associated measure opportunity. The opportunity outcomes were coded as dichotomous responses indicating whether or not the patient was discharged with the particular secondary prevention measure. Each record included an indicator for the associated measure. Patient-specific, site-specific, and study-specific characteristics were incorporated as well as the patient’s date of admission, enabling a time trend for uptake of secondary prevention to be derived.

Two additional outcomes were also identified. First, “Overall Adherence Score” was defined as the number of therapies given (at both the patient and site level) divided by the total number of eligible opportunities. Second, an “All vs. Not All” outcome (i.e. “no missed opportunities”) took a value of 1 if a patient had every medication for which they were eligible, and a value of 0 if the patient had less than every medication. The “All vs. Not All” performance score is the proportion of patients having every medication for which they were eligible.

Randomization

Prior to the trial, all NCD participant sites were surveyed regarding CABG care processes. All NCD sites were informed that they may periodically receive supplemental educational reports in addition to the standard site-specific semiannual reports. Participants were specifically not told of the current study design, nor that other NCD participants might be receiving interventions different from the ones they received. Hospitals in close geographic proximity or with shared surgeons were randomized as clusters. Participant sites were stratified by yearly CABG volume prior to randomization as CABG processes and outcomes may be influenced by procedural volumes at a given site. Clusters were then paired, such that each pair was similar in terms of geography and CABG volume, and randomized within pairs so that one cluster received the intervention and the other received the control assignment. The Duke University institutional review board served as the multiple projects assurance entity for this study for the STS, and determined that informed consent was not required.

Statistical Analyses

Demographic patient-level and hospital-level characteristics were compared between the control and intervention arms to evaluate the adequacy of randomization. A weighted two-sample t-test was used to compare the overall difference in improvement in usage rate of secondary prevention medications among control participants versus among treatment participants. The usage rate was defined as the percentage of eligible patients who did receive the eligible medication.

Time trend analyses were performed with a conditional logistic model at the patient level, conditioning on the randomized pairs, to examine whether the rate of change in use of therapies differed between the intervention and control groups. The conditional logistic model accounts for the pairing of clusters and implicitly adjusts for differences in volume and geography. A significant interaction between period of time and treatment can be explained as the difference in slope between treatment and control participants.

A further subgroup analysis was performed on the change in the percentage of use of secondary prevention medications, examining the interaction between intervention group and site of CABG, academic versus nonacademic center site characteristic, use of process measure at baseline, and order for discharge by surgeon or non-surgeon. For volume and measure performance, sites were categorized as low, medium, or high according to whether they were in the first, second, or third tertile for the variable of interest. All analyses were performed using SAS version 9.1.3 (SAS Institute; Cary, NC).

Sample Size

The power for these analyses was conservatively calculated using sites as the unit of analysis since sites were also the unit of randomization. The mean and standard deviation percentage of discharge ACEI use in the 2003 year STS database retrieval of data were respectively 39.9% and 18.6%, across all sites. A conservative estimate was made for a total of 450 sites entering the two study arms, and an increase of 5 percentage points in use of the secondary prevention measure ACEI over time in the control arm was assumed. Thus, the probability was 81% that the study would detect a treatment difference at a two-sided 0.05 significance level if the true difference between treatment and control arms is 5 percentage points from baseline to end-of-study use of ACEI at discharge (i.e., 45% use of ACEI in the intervention group vs. 50% in the controlarm at study end, a difference in improvement of 5 percentage points between groups). The standard deviations among sites for use of ASA (7.30%), β-blocker (9.84%), and lipid lowering agents (12.14%) were each considerably lower than for ACEI; hence, assuming the same level of difference, fewer sites were needed for these analyses to detect the same magnitude of difference as for the ACEI measure.

Results

As figure 1 shows, a total of 491 sites with active cardiac surgery programs were considered for inclusion in this study based upon ongoing participation in the NCD—these sites had accrued 377,658 patients over the previous seven NCD reporting intervals. The sites were distributed geographically within the continental United States. After excluding sites with greater than 5% of patient records missing all discharge medications or greater than 10% of patients missing any discharge medications, we randomized 234 control sites and 224 intervention sites in this trial. From these 458 sites, a total of 361,328 patients who underwent isolated CABG between July 1, 2002 and December 31, 2005 and were discharged alive to home from the hospital were included.

Figure 1.

Flow of sites through trial

Table 1 shows demographic patient-level variables for the two trial arms, averaged first within sites and then across sites. As expected following randomization, baseline clinical characteristics at the site level were similar between the Control and Intervention groups. Baseline use of aspirin, β –Blocker, and lipid lowering agents were slightly greater among the control hospitals while baseline use of ACEI was slightly greater among the intervention hospitals. The All vs. Not All adherence outcome was slightly higher at baseline in the control group (49.4% vs. 46.3%). As demonstrated in Figures 2a and 2b, there was a wide variance in hospitals’ baseline All vs. Not All performance scores (Interquartile Range [IQR] 39.2% to 56.1%), as well as in overall adherence scores (IQR 71.3% to 81.6%).

Table 1.

Baseline Patient and Hospital Characteristics by Randomized Group^*

	Control Hospitals (N = 234)	Intervention Hospitals (N = 224)
Age	63.8 ± 1.9	63.7 ± 1.8
Male	75.0 ± 5.2	75.2 ± 5.0
Body mass index	29.3 ± 0.73	29.3 ± 0.69
Diabetes	34.5 ± 6.3	34.2 ± 5.0
Hypertension	75.3 ± 6.7	74.8 ± 6.4
Hypercholesterolemia	72.5 ± 11.9	72.3 ± 10.4
Smoker	61.2 ± 8.9	62.3 ± 7.7
Chronic lung disease	17.9 ± 11.5	19.4 ± 10.7
Peripheral vascular disease	14.4 ± 6.6	14.0 ± 5.3
Previous stroke	5.9 ± 2.3	5.7 ± 2.3
Renal failure	4.5 ± 2.9	4.1 ± 2.3
Prior PCI	22.7 ± 7.4	3.4 ± 5.8
Prior cardiac surgery	7.0 ± 7.1	6.4 ± 3.8
Prior MI	43.4 ± 9.6	44.8 ± 8.7
Arrhythmia	7.9 ± 4.1	7.8 ± 3.3
3-Vessel disease	73.6 ± 9.8	74.6 ± 7.8
Ejection fraction	51.1 ± 4.1	51.7 ± 3.7
Elective procedure	50.4 ± 21.8	51.4 ± 22.3
Discharge meds
Aspirin	91.0 ± 9.5	89.6 ± 9.4
β -Blockers	81.7 ± 13.3	78.7 ± 14.6
ACEI	39.8 ± 16.7	40.0 ± 16.0
Lipid lowering	76.2 ± 15.9	72.8 ± 16.0
All vs. Not All	49.4 ± 14.2	46.3 ± 13.7

^*All values are mean percentages +/− standard deviation, except where noted PCI=percutaneous coronary intervention; MI=myocardial infarction; ACEI=angiotensin converting enzyme inhibitor.

Figure 2.

Histograms demonstrating the wide variance in both A, baseline hospital All vs. Not All performance scores (assigned a value of 1 if a patient had every medication for which they were eligible and a value of 0 if patients had less than every medication) and B, overall hospital adherence scores (number of therapies given divided by the number of eligible opportunities).

Table 2 shows the improvements over the intervention period for each of the four secondary prevention medications by control and intervention hospitals. The positive change in adherence score for each of the measures was greater among the intervention hospitals. The difference between the mean change in adherence score between the two study groups reached statistical significance for β –Blocker (9.7% vs. 12.2%, p=0.032), ACEI (6.4% vs. 13.1%, p<0.001), and lipid lowering agent (13.1% vs. 15.7%, p=0.017). The All vs. Not All outcome was also improved significantly more among the intervention hospitals (12.1% vs. 16.7%, p=0.001).

Table 2.

Improvement in Secondary Prevention Medication Adherence Scores^*

	Control Mean Δ (N = 234)	Intervention Mean Δ (N = 224)	p-value
Discharge meds
Aspirin	2.9	4.2	0.255
β -Blockers	9.7	12.2	0.032
ACEI	6.4	13.1	<0.001
Lipid Lowering	13.1	15.7	0.017
All vs. Not All	12.1	16.7	0.001

^*All values are percentages, except where noted

P-values refer to between group changes.

ACEI= angiotensin converting enzyme inhibitor

Figures 3a and 3b show the time trends for the All vs. Not All and Adherence Score outcomes over the study period. A conditional logistic model at the patient level examined whether the rate of change in use of therapies differed between the control and intervention hospitals. Analysis of the composite treatment metric shows the change in the outcome ‘All vs. Not All’ was significant, depicted in Figure 3a. The Hazard Ratio per one six-month period increase was 1.1361 (for intervention hospitals) vs. 1.1190 (for control hospitals), p < 0.001 for different slopes. Similarly, the change in recommended adherence was significant over time (Figure 3b). The Hazard Ratio per one six-month period increase was 1.0833 (for Secondary Prevention Intervention) vs. 1.0710 (for Control), p < 0.001 for different slopes for the recommended population.

Figure 3.

Plot for change in adherence to recommended secondary prevention (SP) medications. A, change in adherence to All vs. Not All performance scores recommended SP medications. B, change in overall hospital adherence scores for recommended secondary prevention medications.

Table 3 shows the change in adherence score for subgroups based upon baseline site volume, on whether or not the site was an academic center, and on whether the surgeon or another allied health provider was responsible for the discharge order. Low site volume at baseline was associated with a greater improvement in adherence score in the treatment group (mean change 5.42% vs. 9.60%, p = 0.019). Overall, the non-academic centers had a higher change in adherence (mean change 6.69% vs. 9.96%, p =0.048), and if the discharge order was made by non-surgeons the change in adherence was significantly greater in the treatment arm compared to control (mean change 5.06% vs. 9.64%, p = 0.002).

Table 3.

Change in Overall Adherence Score by Randomized Group^*

	Control Mean Δ (N = 234)	Intervention Mean Δ (N = 224)	p-value for interaction
Overall	6.84	9.78	0.165
Order for discharge			0.079
Surgeon	7.96	10.10	0.835
Other	5.06	9.64	0.002
Site volume			0.093
Low	5.42	9.60	0.019
Medium	7.76	10.44	0.240
High	7.42	8.86	0.718
Academic center			0.017
No	6.69	9.96	0.048
Yes	8.42	7.19	0.060

^*All values are percentages except where noted

Discussion

The hospitalization following invasive procedures provides a valuable “teachable moment” during which secondary prevention measures can be effectively implemented.¹⁷ The present trial demonstrates the ability of a low-level CQI intervention to engage surgeons and other providers in the use of secondary prevention measures at this crucial time for the coronary artery disease patient. This report documents a significant impact of the STS NCD CQI program on important secondary prevention measures on a national scale over a rapid time frame. In fact, local CQI teams embraced the concept and execution of secondary prevention following CABG to a degree that has changed national practice of care.¹⁸

This trial validates the provider-led CQI model developed by the STS and DCRI. The results, with p<0.001 for a difference in rate of adoption of secondary prevention between the intervention and control groups, were achieved against a background of ongoing baseline improvement in adherence to the chosen secondary prevention measures. Moreover, the present trial demonstrates that a surgical society CQI program can successfully accelerate the real-world application of best-practice measures directed beyond the surgical procedure itself. The application by surgeons of measures for secondary prevention of atherosclerotic progression was a relatively complicated CQI measure, requiring thought and coordination at multiple levels of care. This CQI effort focused not on more traditional short-term surgical processes or outcomes, but on an effect to potentially impact the long-term benefit of the revascularization procedure. Importantly, this focus on care processes surrounding preventative medications supports the potential applicability of the present CQI model to other provider networks, societies, or the like.

The secondary prevention metrics used in this trial have since been incorporated into the STS quality score system for rating performance of participating hospitals. Participants may be awarded one, two, or three stars based on a composite measure of four quality domains: avoidance of mortality, avoidance of morbidity, use of the internal thoracic artery, and adherence to recommended secondary prevention measures. STS NCD participants receive their quality scores with semi-annual reports detailing their performance compared with national aggregate data for each of the quality domains. This incorporation of secondary prevention metrics with continuous feedback to clinical centers underscores the importance of the STS NCD in providing a sustainable structure and straightforward mechanism for incorporating effective CQI interventions into general practice. Adherence to recommended discharge medications among 144,526 CABG patients from 733 STS hospitals from January 1 through December 31, 2007 shows 95.7%, 90.3%, and 88.7% adherence for discharge ASA/antiplatelet, β –Blocker, and lipid-lowering agent respectively.¹⁹ These findings are suggestive of sustained quality improvement following completion of the trial.

The present findings are consistent with other studies indicating that adoption of best practices may largely be locally-driven^4,20 and that local providers respond to contemporary benchmarks.²¹ In this trial, individual sites and local leaders were allowed to determine how best to implement changes in their practices without specifically required CQI tools. This allowed active, as opposed to passive, involvement in the CQI process by the local providers. Many previously reported successes with CQI in medicine have involved either high-intensity and/or site-specific interventions.^22,23 However, the present study employed a low-intensity CQI effort, which resulted in significant and rapid improvements relative to other CQI efforts.²⁴ The present CQI effort hinged on the local leaders and on empowerment of individual sites to implement their own changes. Site-specific feedback with aggregated benchmark data from peer facilities provided motivation for goal-oriented improvements. Thus, by garnering the support of local quality champions, encouraging active individualized CQI changes, and adding a measure of “peer pressure”, the ingredients were in place for a marked increase in secondary prevention adherence among the trial’s intervention group hospitals.

Success of the low-intensity CQI intervention is likely referable in part to baseline CQI expertise already in place at individual sites, expertise gained from previous internal or external quality improvement initiatives. Other CQI studies have encountered difficulty deciphering intervention effects against baseline changes in quality measures.²⁵ A key methodological consideration in this study is that without randomization and a group of control hospitals, the effect of the CQI interventions may have been grossly overestimated and trial results confounded by both nationwide improvements in secondary prevention adherence and regression towards the mean among poor performers. The observed improvements among control arm sites in the present trial are significant. These improvements may be attributed to cross-talk between sites regarding information contained in the intervention-group materials if clinicians compared clinical practices and results at regional and national meetings, as all were blinded to the trial design. Control arm improvements may also reflect use by these hospitals of site-specific data included in routine ongoing STS NCD reports.

All contributing members of the STS NCD are voluntary and whether a general focus on process improvement is greater at participating sites compared to non-STS NCD participating centers is unknown. The present study validates an established platform for achieving meaningful and sustained improvements in quality of care processes, but it remains for government and third-party organizations to capitalize on the success of this type of effort in order to expand the approach to other areas of medicine. Importantly, the form of low-intensity intervention herein represents an approach which may be easily replicated by other provider groups.

Limitations

First, longer term adherence measures evaluating discharge medications were not available; as such, we were unable to assess longer term outcome differences. Furthermore, the outcomes assessed were processes of care and not direct metrics of morbidity and mortality, although the association between secondary prevention and improved outcomes has been shown.⁵ Second, this was a single intervention and it remains unclear whether a more or less intensive intervention would have been equally effective. Further, differences in baseline CQI expertise between sites were not assessed and, despite randomization, confounding cannot be excluded. Finally, we did not assess the CQI infrastructure costs, prohibiting a cost effectiveness analysis. However, the modest trial costs combined with the individual site cost of STS NCD participation compare favorably with prior regional quality improvement efforts in CABG.²⁶

Conclusions

This national randomized controlled trial demonstrates that a professional society CQI program can incrementally speed adoption of secondary prevention therapies. Surgeons and other providers were successfully engaged in the CQI process with reproducible, low-intensity interventions. The findings of the present trial have led to the incorporation of study outcome metrics into a medical society rating system for ongoing quality improvement. Further investigation is needed to determine whether these interventions result in continued quality measure adherence and, ultimately, improved patient outcomes.