This was a retrospective cohort study of patients undergoing first-time, isolated CABG at the East Carolina Heart Institute between July 2002 and May 2011. Demographic data, comorbid conditions, coronary artery disease (CAD) severity, and surgical data were collected at the time of surgery. Patients who were and were not discharged on β-blockers were compared. Only black and white patients were included to minimize the potential for residual confounding (~1% other races). Racial identity was self-reported. Emergent cases were considered a different population due to clinical instability and were excluded in our analysis (n = 98).

A history of β-blocker use at discharge was the exposure variable of interest. Mortality was defined as any cause of death postoperatively. CAD was defined as at least 50% stenosis and confirmed by angiography before surgery.

The East Carolina Heart Institute is a population-based tertiary referral heart hospital located in the center of eastern North Carolina, a rural region with a large black population (18). The institute is the largest stand-alone hospital devoted to cardiovascular care in the state of North Carolina, with an emphasis on reducing the unequal burden of cardiovascular disease in this region. The center provides care to patients that predominantly live and remain within a 150-mile radius of the medical center.

The primary sources of data were the Society of Thoracic Surgeons Adult Cardiac Surgery Database linked with the electronic medical record at the Brody School of Medicine. The National Death Index was used to obtain death dates for patients lost to follow-up and also used to validate death information captured in our electronic medical record. Linkage with the National Death Index was based on a multiple criteria, deterministic matching algorithm, which included a patient’s social security number (19). In our database, <5% of validated deaths failed to correctly match with the National Death Index. Beginning in 2012, the use of social security numbers as a patient identifier was proscribed within our university system in compliance with §205(r) of the Social Security Act (20). Information on β-blocker use at discharge was not collected in our database prior to 2002.

Categorical variables were reported as frequency and percentage while continuous variables were reported as mean ± standard deviation, median, and range. Follow-up time was measured from the date of surgery to the date of death or censoring. Survival probabilities were computed using the Kaplan–Meier product-limit method and stratified by β-blocker status. The log-rank test was used to compare survival between patients with and without discharge β-blockers. Cox proportional hazard regression models were used to compute hazard ratios (HR) and 95% confidence intervals (CI) for long-term mortality. The main model was adjusted for variables that have been previously reported to be associated with cardiovascular-related mortality and included the following covariates: age, sex, race, hypertension, CAD severity, heart failure, and prior stroke. The post hoc addition of other variables into the model was performed in a pairwise fashion. P-for-interaction between race and discharge β-blocker use was computed using a likelihood ratio test. The test statistic of Grambsch and Therneau was used to check the proportional hazards assumption (21). Temporality during the study period was assessed by stratifying the analysis by two time periods (e.g., 2002–2006, 2007–2011). There were no missing values for the variables used in this analysis. Statistical significance for categorical variables was tested using Fisher’s exact test and the Deuchler–Wilcoxon procedure for continuous variables. Statistical significance was defined as p < 0.05. SAS Version 9.3 (Cary, NC, USA) was used for all analyses.

Our study is strengthened by its large racially dichotomous sample size and long-term follow-up. Furthermore, we were able to accurately determine time of death using a combination of the National Death Index and our comprehensive electronic medical record.

Data regarding socioeconomic position, education, and income were not collected and these factors potentially influenced survival (26). Alcohol consumption also was not recorded in our dataset, but this variable has not been associated with increased morbidity and mortality after CABG (27). We did not have information on long-term β-blocker use. Adherence to medication among study participants may have changed. In general, β-blocker adherence has been shown to be relatively stable after CABG and any non-adherence to β-blockers after discharge likely would have biased results toward the null, decreasing the probability of finding an association of β-blocker use and long-term survival (28). However, a statistically significant result was observed in our study despite potential non-adherence to β-blocker therapies.

Physicians may have prescribed β-blockers to patients perceived to have better prognoses, potentially introducing selection bias into the analysis. However, the association of discharge β-blocker use and lower mortality was preserved after excluding patients who died within the first 30 days of surgery (black: adjusted HR = 0.42, 95% CI = 0.28–0.61; white: adjusted HR = 0.71, 95% CI = 0.56–0.90). Presumably, the accuracy of selecting patients who will survive diminishes following the postoperative period. While β-blocker therapy is relatively contraindicated in patients with chronic obstructive pulmonary disease and may have resulted in selection bias, only a small number of these patients were not discharged on β-blockers and this percentage was similar between races. β-Blocker use also is relatively contraindicated in heart failure patients with depressed ejection fraction (< 0.35). However, few patients with depressed ejection fraction who did not receive β-blockers were in our dataset (black, 2.5%; white, 2.2%). Excluding these patients did not substantively alter our findings (black, adjusted HR = 0.29, 95% CI = 0.20–0.43; white, adjusted HR = 0.48, 95% CI = 0.38–0.60).

Patients in this study were recruited over a relatively long period (~10 years), during which practice methods and clinical care may have changed. However, adjusting for study period had little impact on our findings. Similarly, the status of several variables in our analysis may have changed over time (e.g., medications and postoperative complications). We did not adjust for these variables in a time-dependent manner due to their potential to be in the causal pathway. Cause of death is not recorded in the National Death Index and β-blocker use may have been unrelated to their mortality. Although we adjusted for known clinically relevant variables, we acknowledge that other unmeasured factors may have resulted in residual confounding. Additionally, the data reported are from a single center and may reflect a chance finding specific to our institution and region.

Multivariable Cox regression models were used to adjust for confounding because of potential “non-collapsibility bias” inherent to propensity scores based on log-linear models and the possible decrease in power due to incomplete matching (29). Additionally, a random forest algorithm (e.g., machine learning) was not used for matching. This algorithm potentially introduces misspecification bias into the propensity score model due to the “black box” nature of the algorithm that obscures the etiologic relationship between predictors and outcome (30, 31).

In conclusion, β-blocker use at discharge was associated with a survival advantage among black patients after CABG and the magnitude of the effect was similar to that observed among white patients. In contrast with prior studies that have reported β-blocker use is less effective in black patients, we did not observe a similar association at our institution among black CABG patients discharged on β-blockers. Additional studies are needed to better understand the underlying mechanisms of our results and to appropriately target future treatment strategies among CABG patients.