Racial Differences in Heart Failure Outcomes: Evidence from Telemonitoring to Improve Heart Failure Outcomes (Tele-HF) Trial

Abstract

Objectives—

To determine if there are racial differences in patient-reported health status as well as mortality and rehospitalization after hospitalization for heart failure (HF).

Background—

Little is known about whether racial differences exist in patient-reported outcomes after a HF hospitalization.

Methods—

We analyzed data from 1427 patients (636 non-Hispanic African Americans, 45%; 791 non-Hispanic whites, 55%) enrolled in the Telemonitoring to Improve Heart Failure Outcomes (Tele-HF) Trial. Health status was measured with the Kansas City Cardiomyopathy Questionnaire (KCCQ) at baseline, 3 and 6 months. Generalized linear mixed models and propensity score methods were used to adjust for clustering within sites and differences between races.

Results—

While black patients reported better adjusted health status at baseline (black vs. white difference in KCCQ summary score: 6.22, 95% confidence interval [CI]: 2.98–9.46, P<0.001), after adjusting for patient demographics, comorbidities, clinical laboratory values, and baseline KCCQ score, we detected no significant racial differences in patient-reported health status at 3 (black vs. white difference in KCCQ score: 2.28, 95% CI: [−0.84, 5.41], P=0.15) or 6 months (black vs. white difference in KCCQ score: 1.91, 95% CI: [−1.31, 5.13], P=0.24).

Conclusions—

Compared with whites, black patients with HF had better patient-reported health status shortly after a HF admission, but not at 3 or 6 months. Our study fails to show that black patients are disadvantaged with regard to health status after a HF hospitalization.

Racial disparities in cardiovascular disease have long been recognized. However, little is known about whether racial differences exist in patient-reported outcomes after a heart failure (HF) hospitalization. Despite pharmacologic and technical advances in the care of patients with HF, racial disparities in HF care and outcomes remain an enormous public health concern.(1–4) Although HF is disproportionately present among African Americans,(5–9) it is unclear whether there are racial differences in patient-reported health status outcomes after HF hospitalization.

Prior studies, using Medicare data, of older patients discharged after a hospitalization for HF have indicated that, compared with white patients, African American patients have higher readmission rates(1,2) and lower mortality.(1) However, these studies did not include younger patients and focused on mortality and readmission, although health status (patients’ symptoms, function and quality of life) is a primary concern for patients. Illuminating racial disparities in outcomes is important because HF affects black individuals at a younger age than white individuals. It thus remains unknown whether racial differences exist in mortality and readmission in a younger HF population and whether there are racial differences in patient- reported health status after a HF hospitalization.

To address these gaps in knowledge, we sought to compare a comprehensive set of HF outcomes including health status, readmission, and mortality in non-Hispanic white and non- Hispanic black patients enrolled in a multicenter, randomized controlled trial, Telemonitoring to Improve Heart Failure Outcomes (Tele-HF). A previous study reported that telemonitoring did not improve health outcomes among patients recently hospitalized for HF using Tele-HF dataset.(10) Our main goal was to determine whether there are short-term and long-term racial differences with respect to white and black patients’ health status, mortality, and readmission rates.

Methods

Study Population

The primary data source was the Tele-HF trial, which has been previously described.(10,11) In brief, the Tele-HF trial was funded from the National Heart, Lung, and Blood Institute (5 R01 HL080228) as a multicenter, randomized, controlled trial to assess whether daily, remote telemonitoring of symptoms and body weight would improve HF outcomes. The Yale University School of Medicine and each participating site (see the Appendix) approved the trial protocol. An independent Data and Safety Monitoring Board was appointed to monitor adherence to the protocol, to evaluate the recruitment and retention of patients, and to assess the quality of the data and the safety of the telemonitoring intervention.

Patients were recruited from 2006 through 2009 at 33 participating sites across the U.S. Candidates considered for inclusion in the trial were patients hospitalized for HF in the previous 30 days. The exclusion criteria were residence in a nursing home, low expected probability of survival for the next 6 months, inability to stand on a scale, severe cognitive impairment,(12) and a planned hospitalization for a procedure. There were no significant differences between the telemonitoring group and the usual care group in mortality and readmission rates.(10) After excluding 226 patients who self-reported Hispanic or “other” race from the study, our analyses included 1427 patients with HF of the 1653 Tele-HF patients.

Variables and Outcomes

Self-reported race information was recorded when a patient was enrolled. The primary outcome measures for this study were the Kansas City Cardiomyopathy Questionnaire (KCCQ) overall summary scores (which are derived from the physical function, symptom, social function, and quality of life domains; range: 0–100 with higher scores reflecting better health status)(13) at baseline, 3 and 6 months). Scores range from 0–100, where higher scores indicate better health status (e.g. fewer symptoms, less function impairment and higher quality of life). A centralized call center was used to administer the KCCQ shortly after discharge (baseline) and again at 3 and 6 months. Patient readmission and mortality within 180 days of enrollment were analyzed as secondary outcomes.

Statistical Analyses

We compared patient demographics, socioeconomic status, and medical history as well as clinical characteristics by race using means and standard deviations or medians and interquartile range for skewed data and percentages for categorical variables. Wilcoxon’s rank sum test for continuous variables and the Pearson χ2 test/Fisher’s exact test for categorical variables were used to examine statistical significance of observed differences.

Multivariable linear regression analyses using generalized linear mixed models (GLMM) were performed to examine the relationship between race (black vs. white) and the overall summary score of KCCQ across time (baseline, 3 and 6 months) between these two racial groups. The GLMM models included a random intercept to account for clustering within site and to adjust the covariance structure to account for repeated observations on subjects. Missing data at follow-up were assumed to be missing at random (MAR) after empirical data checking. Maximum likelihood estimation method was used in all GLMM models. We first identified an overall model predicting KCCQ summary scores across all three time intervals by identifying those variables which were either clinically important or statistically significant (P<0.05). Because black and white patients differ in numerous demographic, socio-economic and clinical characteristics, we sought to balance the groups by creating a propensity score to be black, as has been done in prior analyses.(14) The propensity to be black was derived from a nonparsimonious logistic regression on all available variables at each sequential step. We estimated the propensity score and then included it as a covariate along with other covariates related to the outcome in the final fully-adjusted model predicting black vs. white differences in KCCQ’s overall summary scores across time.(15) The model was estimated sequentially. First the unadjusted model was fitted, and then we sequentially adjusted for demographics, insurance payer, medical history, and lab values each time fitting a new model. It is suggested that regression adjustment for propensity score has great balancing properties.(16)

Likewise, multivariable logistic regression analyses using GLMMs were conducted to evaluate the association between race (black vs. white) and readmission and mortality rates (180- day). Models accounted for the clustering of patients within site through the use of random intercepts. Similar model building techniques, including sequential propensity score method and regression method of adjusting for propensity score, were used to test for racial differences for each outcome.

In addition, to ensure that detected racial difference in any outcome measure did not depend on the group assignment (telemonitoring vs. usual care), an interaction term between race (black vs. white) and group (telemonitoring vs. usual care) was added to each full model for readmission and mortality measures and a three way interaction term (i.e., race, group, and time) was added to the KCCQ model. Treatment assignment was not found to be statistically significant in any of these models and therefore was not included in the final models.

All statistical analyses were performed using SAS version 9.3 software (SAS Institute, Cary, NC). A 2-tailed significant level of 0.05 was used for all tests. The authors are responsible for the integrity of the data in the study.

Results

There were 1427 white and black patients with HF (636 non-Hispanic African Americans, 45%; 791 non-Hispanic whites, 55%) from 33 participating hospitals in the Tele-HF trial. The mean age was 61.5±15.2 (SD) years and 42.5% were women. There were significant racial differences in patient demographics, socioeconomic status, medical history, risk factors, and laboratory values (Table 1).

Table 1.

Demographic, Socioeconomic Status, Medical History, Risk Factors, and Laboratory Values at Enrollment by Race

Variables	Non-Hispanic White (n=791)	Non-Hispanic Black (n=636)	P Value
Demographics
Age, years
Mean (SD)	67.2 (13.7)	54.3 (13.9)	<.001
Median (IQR)	68.0 (20.0)	54.0 (18.0)
Female	327 (41.3)	280 (44.0)	0.31
Payer
Commercial/PPO	265 (33.5)	147 (23.1)	<.001
Medicare	495 (62.6)	211 (33.2)	<.001
Medicaid	72 (9.1)	160 (25.2)	<.001
HMO	49 (6.2)	43 (6.8)	0.67
VA	11 (1.4)	6 (0.9)	0.44
Self-Pay	66 (8.3)	117 (18.4)	<.001
Other	32 (4.1)	16 (2.5)	0.11
Unknown	21 (2.7)	16 (2.5)	0.87
Socioeconomic Status
Annual household income < $10 000 Number (%)	89 (11.3)	184 (28.9)	<.001
High school graduation Number (%)	567 (71.7)	382 (60.1)	<.001
Medical History
CAD/MI/IC	500 (63.2)	237 (37.3)	<.001
Hypercholesterolemia	466 (58.9)	301 (47.3)	<.001
Hypertension	568 (71.8)	542 (85.2)	<.001
Liver disease	9 (1.1)	20 (3.1)	0.008
Renal failure	194 (24.5)	181 (28.5)	0.09
Cardiac resynchronization therapy	56 (7.1)	26 (4.1)	0.02
Peripheral vascular disease	104 (13.2)	46 (7.2)	<.001
Permanent pacemaker	132 (16.7)	57 (9.0)	<.0001
Chronic pulmonary disease	201 (25.4)	117 (18.4)	0.002
Diabetes	383 (48.4)	284 (44.7)	0.16
Cerebrovascular disease	76 (9.6)	56 (8.8)	0.60
Risk Factors
Current Smoker	47 (5.9)	81 (12.7)	<.001
BMI
Mean (SD)	26.2 (10.4)	27.4 (12.1)
Median (IQR)	26.1 (13.3)	26.6 (17.2)
Systolic Blood Pressure (mmHg)
Mean (SD)	117.2 (19.8)	126.0 (24.6)
Median (IQR)	116.0 (28.0)	124.0 (35.0)
Diastolic Blood Pressure (mmHg)
Mean (SD)	66.7 (11.9)	75.7 (14.7)	<.001
Median (IQR)	66.0 (15.0)	75.0 (21.0)	<.001
NYHA Functional Classification			0.13
I	46 (5.8)	37 (5.8)
II	309 (39.1)	211 (33.2)
III	389 (49.2)	342 (53.8)
IV	47 (5.9)	46 (7.2)
Low cognition: Folstein score ≤ 24(12)	50 (6.3)	75 (11.8)	<.001
Laboratory Values at Enrollment
Low LVEF (<40%)	486 (61.4)	499 (78.5)	<.001
Baseline KCCQ Summary Score
Mean (SD)	59 (24)	60 (25)	0.69

Values in parentheses are percentages unless otherwise indicated. P values are based on Pearson Chi-square test or Fisher’s exact test when cell counts are less than 5 (for categorical variables) or Wilcoxon’s rank sum test (for continuous variables). SD, standard deviation; IQR, interquartile range; PPO, preferred provider organization; HMO, health maintenance organization; VA, veterans affairs; CAD, coronary artery disease; MI, myocardial infarction; IC, ischemic cardiomyopathy; BMI, body mass index; NYHA, New York Heart Association; LVEF: left ventricular ejection fraction; KCCQ, Kansas City Cardiomyopathy Questionnaire.

Compared with white patients, black patients were much younger, more likely to be in the Medicaid and self-pay categories, had lower annual household income and lower high school graduation rates, were less likely to have a medical history of coronary artery disease (CAD)/myocardial infarction (MI)/ischemic cardiomyopathy (IC). Blacks were more likely to have a medical history of hypertension, higher systolic and diastolic blood pressure, cognitive impairment (Folstein score(12) ≤ 24) and reduced left ventricular ejection fraction (LVEF<40%).

Primary Outcomes

After accounting for clustering within site and correlation among repeated measurements, black patients with HF reported better health status at baseline, as measured by the KCCQ overall summary score, (black vs. white score difference, 3.97; 95% confidence interval [CI], 1.00–6.95, P=0.009). However, difference in the unadjusted KCCQ summary score was not observed at 3 months (black vs. white score difference, 1.17; 95% CI, −1.93–4.27, P=0.46) or 6 months (black vs. white score difference, 0.71; 95% CI, −2.48–3.91, P=0.66). After further adjustment for all other patient level factors, together with the propensity score, the racial difference in the overall summary score of KCCQ at baseline still persisted (black vs. white score difference, 6.22; 95% CI, 2.98–9.46, P<0.001) (Table 2). To further understand this difference we performed a post hoc analysis of propensity score adjusted KCCQ baseline sub-scores to identify which domains might be most influential. We found that there were significant racial differences in almost all domains of the baseline KCCQ scores except self-efficacy and symptom stability (Table 3). By contrast, no significant racial differences were detected in fully adjusted models at 3 months (black vs. white score difference, 2.28; 95% CI, −0.84–5.41, P=0.15) or 6 months (adjusted black vs. white score difference, 1.91; 95% CI, −1.31–5.13, P=0.24). Importantly, both racial groups experienced notable improvement in mean quality of life scores during the follow-up period (Table 2, Appendix Table 1, and Figure 1) ranging from 58.4 to 68.9 points for whites and 64.6 to 70.9 for blacks (p<0.01 for interaction of race and time).

Table 2.

Estimated Black vs. White Differences in the Overall Summary Score of KCCQ

Models	Estimate	95% Confidence Interval	P Value
Time: at Enrollment
Unadjusted Model	3.97	1.00 – 6.95	0.009
Model 1: adjusted for demographics	5.82	2.73 – 8.92	<.001
Model 2: Model 1 + adjusted for payer	6.07	2.94 – 9.19	<.001
Model 3: Model 2 + adjusted for medical history	5.93	2.73 – 9.14	<.001
Model 4: Model 3 + adjusted for lab values	6.22	2.98 – 9.46	<.001
Time: 3 Months after Enrollment
Unadjusted Model	1.17	−1.93 – 4.27	0.46
Model 1: adjusted for demographics	2.98	−0.21 – 6.17	0.07
Model 2: Model 1 + adjusted for payer	2.86	−0.33 – 6.04	0.08
Model 3: Model 2 + adjusted for medical history	2.29	−0.84 – 5.41	0.15
Model 4: Model 3 + adjusted for lab values	2.28	−0.84 – 5.41	0.15
Time: 6 Months after Enrollment
Unadjusted Model	0.71	−2.48 – 3.91	0.66
Model 1: adjusted for demographics	2.54	−0.74 – 5.83	0.13
Model 2: Model 1 + adjusted for payer	2.42	−0.86 – 5.70	0.15
Model 3: Model 2 + adjusted for medical history	1.88	−1.34 – 5.09	0.25
Model 4: Model 3 + adjusted for lab values	1.91	−1.31 – 5.13	0.24

Table 3.

Propensity Score Adjusted Baseline KCCQ Sub-scores

KCCQ Sub-score	Non-Hispanic White (n=791)	Non-Hispanic Black (n=636)	(n=636)
Physical Limitation Sub-score
Mean (SE)	67.7 (1.6)	73.0 (1.8)	0.01
Symptom Stability Sub-score
Mean (SE)	73.7 (1.1)	77.0 (1.4)	0.06
Symptom Frequency Sub-score
Mean (SE)	55.8 (1.6)	62.4 (1.8)	0.001
Symptom Burden Sub-score
Mean (SE)	64.3 (1.6)	71.8 (1.8)	<0.001
Total Symptom Score Sub-score
Mean (SE)	60.1 (1.5)	67.2 (1.8)	<0.001
Self-Efficacy Sub-score
Mean (SE)	84.5 (1.1)	83.8 (1.3)	0.65
Quality of Life Sub-score
Mean (SE)	52.5 (1.4)	58.4 (1.6)	0.002
Social Limitation Sub-score
Mean (SE)	53.8 (2.2)	59.9 (2.5)	0.02

^*Propensity score: the propensity to be black is derived from a nonparsimonious logistic regression on available demographic, socioeconomic and clinical characteristics variables.

Figure 1. Plot for Estimated Differences in KCCQ Summary Score from Final Model*.

*Final model adjusted for patients’ demographics, payer, medical history, and laboratory values

Secondary Outcomes

We observed no significant racial differences in unadjusted 30-day (white: 18.3%, 95% CI: 15.7% - 21.2%; black: 16.4%, 95% CI: 13.6% - 19.5%; P=0.33) or 180-day (white: 50.4%, 95% CI: 46.9% - 54.0%; black: 47.8%, 95% CI: 43.9% - 51.8%; P=0.32) all-cause readmission and 30-day mortality (white: 2.3%, 95% CI: 1.4% - 3.6%; black: 1.3%, 95% CI: 0.5% - 2.5%; P=0.15). Significant racial difference occurred in unadjusted 180-day mortality rates (white: 13.4%, 95% CI: 11.1% - 16.0%; black: 9.0%, 95% CI: 6.9% - 11.5%; P=0.01) using χ2 test. However, such racial differences were no longer statistically significant after fully adjusting for propensity score to be black, patient demographics, payer, socioeconomic status, medical history, risk factors, and laboratory values (adjusted odds ratio [AOR], 0.85; 95% CI, 0.52–1.37, P=0.49). Likewise, fully adjusted 180-day all-cause readmission rates were similar (AOR, 0.92; 95% CI, 0.70–1.21, P=0.53) (Table 4). Due to a low number of events we were unable to calculate propensity score adjusted models for 30-day mortality and readmission outcomes.

Table 4.

Using Sequential Propensity Score Adjustment of Race (Black vs. White) to Predict 180-day Readmission and Mortality Rates

Models	AOR*	95% Confidence Interval	P Value
180-Day Readmission
Unadjusted Model	0.87	0.70 – 1.09	0.23
Model 1: adjusted for demographics	0.88	0.69 – 1.11	0.28
Model 2: Model 1 + adjusted for payer	0.89	0.70 – 1.14	0.36
Model 3: Model 2 + adjusted for socioeconomic status*	0.83	0.64 – 1.06	0.14
Model 4: Model 3 + adjusted for medical history	0.86	0.66 – 1.13	0.27
Model 5: Model 4 + adjusted for laboratory values	0.92	0.70 – 1.21	0.53
180-Day Mortality
Unadjusted Model	0.63	0.43 – 0.91	0.01
Model 1: adjusted for demographics	0.84	0.56 – 1.25	0.39
Model 2: Model 1 + adjusted for payer	0.79	0.53 – 1.20	0.27
Model 3: Model 2 + adjusted for socioeconomic status*	0.71	0.47 – 1.09	0.12
Model 4: Model 3 + adjusted for medical history	0.81	0.51 – 1.29	0.38
Model 5: Model 4 + adjusted for laboratory values	0.85	0.52 – 1.37	0.49

AOR: adjusted odds ratio; socioeconomic status included annual household income < $10000 and high school level education.

Discussion

We found that, as compared with whites, black patients with HF had better self-reported health status overall and in several domains (including physical limitation, symptom frequency, symptom burden, total symptom score, quality of life, and social limitation) early after discharge. However, such racial advantages were attenuated and no longer statistically significant at 3 or 6 months, although the health status of both racial groups improved during follow-up. Moreover, we did not detect significant racial differences in 180-day readmission and mortality rates from the fully adjusted models.

Although there is a pressing need to report patient-centered outcomes, including patient’s health status,(17–21) to the best of our knowledge, our study is one of the first studies to describe patients’ health status (e.g., symptoms, function, and quality of life) by race for patients recently discharged from the hospital after a HF exacerbation. To quantify patients’ health status, we used the KCCQ questionnaire which is a valid, sensitive, disease-specific health status measure for patients with HF. Compared with other health-related quality of life instruments for HF, the KCCQ questionnaire is more sensitive to clinical changes and thus an ideal tool to demonstrate improvements in health status.(22) Furthermore, unlike previous research using large administrative datasets to examine racial disparities in HF,(1,2) our analyses were based on data collected from a multicenter randomized controlled trial (Tele-HF) with detailed clinical information.(10,11)

Our findings that black patients with HF had higher summary score of KCCQ at baseline differed from what another study recently reported.(23) Using the Heart Failure-A Controlled Trial Investigating Outcomes in Exercise TraiNing (HF-ACTION) trial database (details of the trial design were published previously(24,25)), which included 2,175 patients with HF (black: n=749, 34%; white: n=1426, 66%), investigators showed that the unadjusted baseline KCCQ overall score was higher in white patients with HF than in black patients (white vs. black: 69 vs. 66, P<0.001).(23) There are several possible explanations for the different findings between the two studies. First and foremost, we focused on hospitalized HF patients enrolled in the Tele-HF trial whereas HF-ACTION trial examined stable outpatients with diagnosis of HF. As expected, white patients in our study were older (67 vs. 62), tended to have had a history of myocardial infarction (63% vs. 51%), and were more likely to have hypertension (72% vs. 51%). In addition, the percentage of black patients was much higher in our study (45% vs. 34%). And, a recent study using American Heart Association Get With The Guidelines – HF (GWTH-HF) national registry reported an even lower percentage of non-Hispanic black patients with HF (23%).(26) Moreover, we further limited to non-Hispanic white and non-Hispanic black patients to get a more homogeneous study population. Finally, we cannot rule out the possibility that black patients with HF admitted to hospitals might have lower admission severity than whites (e.g., risk factors that were not captured by the Tele-HF trail may have contributed to differential severity) , which has been reported previously.(27–29) Nonclinical factors such as being less well-insured and educated, inadequate outpatient follow-up, poor social support, and nonadherence with medications or diet are possible explanations for blacks being hospitalized at an earlier stage of HF.(27) Also, this might reflect the racial differences in response to the HF treatment due to phenotype differences across racial groups. (30,31)

We also found that there were no significant racial differences in readmission and mortality rates. These findings contrast with the results from two prior studies using a national Medicare fee-for-service sample. Rathore et al reported that black Medicare patients hospitalized with HF had slightly higher risk-adjusted rates of 1-year readmission (relative risk <RR>, 1.09; 95% CI, 1.06–1.13) but had lower mortality rates up to 1 year (30-day mortality RR, 0.78; 95% CI, 0.68–0.91; 1-year mortality RR, 0.93; 95% CI, 0.88–0.98).(1) Joynt et al showed that black Medicare patients with HF were more likely to get readmitted within 30 days (adjusted odds ratio <AOR>, 1.04; 95% CI, 1.03–1.06).(2) However, the 95% confidence intervals in our study covered the point estimates reported from these prior papers, therefore our findings are not significantly inconsistent. Importantly, the effect size in the Joynt study was quite small, indicating that the odds of readmission were only 4% greater for blacks versus whites. In addition, our study differed from these two studies with respect to the patient population and study design. While prior studies used national Medicare fee-for-service recipients (which were elderly patients) with HF to study racial differences in mortality and readmission, our analyses was based on a randomized controlled trial (our patients were overall much younger) to study racial disparities in HF outcomes. As shown previously, using different types of datasets (large administrative datasets vs. randomized controlled trial database with detailed clinical and health status information) could arrive at different conclusions.(32) Furthermore, we examined 180-day mortality and readmission after hospitalization for HF whereas Rathore et al examined 1-year mortality and 1-year readmission and Joynt et al reported 30-day readmission. It is very likely that studying different outcome measures and adjusting for sufficient clinical risks (e.g., LVEF) in our study might lead to dissimilar findings. This also suggests that future studies are needed to confirm whether the length of follow-up period can play a critically important role in evaluating racial differences in HF outcomes.

There are several clinical implications of our findings from this study. First, our results highlight how race is strongly associated with many features of patients with HF (e.g., demographics, baseline characteristics, clinical profiles, responses to treatment, etc) although we did not observe significant racial differences in the outcomes during the follow-up. This suggests the importance of clinicians being aware of the ways that race is linked to the manifestation of HF and the need for them to provide personalized HF treatment for different subgroups. As the Tele-HF database did not include detailed clinical information for the recent hospitalization with HF, future studies are needed to improve our understanding of the mechanism of the race in the presentation and the outcomes of HF. Second, our findings of higher KCCQ scores at baseline and very similar KCCQ scores during follow-up supports the use of this disease-specific, valid and reliable, self-administered KCCQ questionnaire at different times to monitor clinical changes and better predict medium- and long-term outcomes in HF.(33) Third, we cannot rule out the possibility of “self-report bias” (i.e., selective revealing or suppression of information by specific groups) in the Tele-HF database. Literature has consistently demonstrated that self-reports are at risk of reporting bias when assessing quality of care (34) and health status.(35) In addition, a prior study reported that there were black vs. white disparities in health literacy among patients with HF. (36)

Our study should be interpreted in the context of the following potential limitations. First, the Tele-HF trial had specific inclusion and exclusion criteria so that our findings might not be generalizable to the actual patient population. We excluded patients with HF who resided in a nursing home, or had low expected probability of survival for the next 6 months, or were unable to stand on a scale, or had severe cognitive impairment, or had a planned hospitalization for a procedure. In other words, the patients included in the study might be systematically less severe. Moreover, black patients were much younger than their white counterparts in the study. Given that quality of life and outcomes for patients with HF could vary across age groups,(37) it is possible that our regression models did not account for residual confounding due to age or other factors. While we adjusted for many patient level variables of prognostic importance, residual confounding may still have influenced our findings. Second, Tele-HF’s follow-up period was 180 day after enrollment. Therefore, we were unable to evaluate the racial differences in longer terms. Third, it is possible that this study was unable to detect very modest racial differences in these outcomes. Although we had non-significant findings, we found that the confidence intervals encompassed ranges that included clinically meaningful differences. It is possible that modest racial differences could be detected with a high-powered study. Fourth, there might be racial differences with regard to the prior hospitalization for HF (e.g., number of complications, length of stay, etc) and such differences might explain some findings in this study. Unfortunately, we did not have detailed clinical information about the prior HF hospitalization and thus were unable to accounts for such differences in our analyses. Fifth, as we performed multiple statistical tests between black and white patients with HF, we cannot completely rule out the possibility of a false positive finding. However, the extremely low p value (<0.001) from testing racial difference in the KCCQ summary score at enrollment suggests that this is very unlikely.

In summary, we found no differences in patient-reported health status, mortality or readmission outcomes by race in a multicenter randomized controlled trial of HF patients. We found that compared with whites, black HF patients had higher self-reported health status early after hospitalization, but we failed to detect racial differences in health status at 3 and 6 months. No significant racial differences in 180-day readmission and mortality rates were detected after adjusting for clustering at sites and patient factors. These findings suggest that racial disparities in patients with HF are less likely to be evident after they are hospitalized for treatment.

Abbreviations List

HF

heart failure

Tele-HF

Telemonitoring to Improve Heart Failure Outcomes trial

KCCQ

Kansas City Cardiomyopathy Questionnaire

GLMM

generalized linear mixed models

MAR

missing at random

CAD

coronary artery disease

MI

myocardial infarction

IC

ischemic cardiomyopathy

LVEF

left ventricular ejection fraction

HF-ACTION

Heart Failure-A Controlled Trial Investigating Outcomes in Exercise TraiNing trial

GWTH-HF

Get With The Guidelines – HF

Appendix Table 1. Overall Model Predicting KCCQ’s Overall Summary Score

Models	Estimate	95% Confidence Interval	P Value
Demographics
Age≥65 (vs. Age<65)	6.74	4.10 – 9.37	<.001
Female (vs. Male)	−4.58	−6.93 - -2.23	<.001
Black (vs. White)	5.33	2.35 – 8.31	<.001
Time: 6 months	11.57	9.56 – 13.57	<.001
Time: 3 months	9.74	8.03 – 11.44	<.001
Black X Time: 6 months	−3.15	−6.23 - -0.07	0.045
Black X Time: 3 months	−2.79	−5.42 - -0.16	0.037
Telemonitoring group (vs. Control group)	3.71	1.53 – 5.89	<.001
Medical History & Risk Factors
Chronic pulmonary disease	−7.23	−9.99 - -4.47	<.001
Liver disease	−10.18	−18.28 - -2.08	0.01
Peripheral vascular disease	−3.54	−7.36 – 0.27	0.07
Prior AICD implant	−3.73	−6.55 - -0.91	0.01
Prior MI	−3.35	−6.06 - -0.65	0.02
Sleep apnea	−4.32	−7.32 - -1.32	0.005
NYHA functional classification II/III vs. I	−11.21	−16.25 - -6.17	<.001
NYHA functional classification IV vs. I	−14.96	−21.71 - -8.21	<.001
CAD	−0.14	−2.77 – 2.48	0.92
Hypercholesterolemia	−0.03	−2.37 – 2.32	0.98
Illicit drug use	−4.30	−11.18 – 2.58	0.22
Permanent pacemaker	−2.35	−5.69 – 1.00	0.17
Tumor	−0.86	−5.72 – 4.00	0.73
Laboratory Values
Hemoglobin	0.99	0.38 – 1.61	0.002
Hemoglobin value unknown/missing	10.99	2.13 – 19.84	0.02
LVEF<40 (vs. LVEF>40)	0.99	−1.64 – 3.62	0.46