Effects of socioeconomic status on clinical outcomes with ventricular assist devices

Abstract

Background

Lower socioeconomic status (SES) is a known risk factor for worse outcomes after major cardiovascular interventions. Furthermore, individuals with lower SES face barriers to evaluation for advanced heart failure therapies, including ventricular assist device (VAD) implantation.

Hypothesis

Examination of the effects of individual determinants of SES on VAD outcomes will show similar survival benefit in patients with lower compared with higher SES.

Methods

All VAD implants at the University of Florida from January 2008 to December 2015 were reviewed. Patient‐level determinants of SES included place of residence, education level, marital status, insurance status, and financial resources stratified by percent federal poverty level. Survival or transplantation at 1 year, 30‐day readmission, implant length of stay (LOS), and an aggregate of VAD‐related complications were assessed in univariate fashion and multivariable regression modeling.

Results

A total of 111 patients were included (mean age at the time of implant 57.6 years, 82.8% men). More than half received destination therapy. At 1 year, 78.3% were alive on device support or had undergone successful transplantation. There were no differences in survival, 30‐day readmission, or aggregate VAD complications by the SES category. Although patients with lower levels of education had longer LOS in univariate analysis, on multivariable ordinal regression modeling, this relationship was no longer seen.

Conclusions

Patients with lower SES receive the same survival benefit from VAD implantation and are not more likely to have 30‐day readmissions, complications of device support, or prolonged implant LOS. Therefore, VAD implantation should not be withheld based on these parameters alone.

1. INTRODUCTION

Heart failure (HF) affects more than 6.5 million people in the U.S. and remains the leading cause of hospitalization for patients >65 years of age.1 Despite medical advances, hospitalization and readmission rates for HF remain high.2 Furthermore, healthcare costs for HF hospitalizations are estimated to be >$32 billion annually, and this is expected to continue to grow.3 As the number of patients with congestive HF increases, there remains a greater shortage of donor organs and also a growing number of patients identified as poor candidates for transplantation. In patients with New York Heart Association Class IV HF, continuous‐flow left ventricular assist devices (LVADs) remain an increasingly available option and provide patients with improved functional capacity and quality of life.4 Regardless of implant strategy, whether bridge to transplant (BTT) or destination therapy (DT), rates of LVAD implantation seemed to have plateaued, despite continued improvements in survival.5 This apparent lack of widespread adoption of life‐saving technology may be related to minimal improvements with adverse event rates and how these impact hospital utilization and cost effectiveness of LVAD implantation.5, 6, 7, 8 Predictors of outcome in LVAD recipients, therefore, remain of interest, with increasing focus on how to deliver greater quality of care at better value for both patients and payers.

Among HF predictors, lower socioeconomic status (SES) has been strongly associated with increased incidence of HF and readmission rates.9 Education, a SES indicator, has been recognized in multiple investigations as a predictor of HF hospitalization rates.10 Elderly patients with a lower SES have a higher risk of 1‐year mortality as well as hospital readmission within 1 year of discharge for HF, compared with those with higher SES.11 Studies further demonstrate that lower education level is associated with greater 30‐day readmission rates after cardiac surgery.12 Fewer studies, however, address predictors of outcome for patients receiving LVADs. Early investigations uncovered disparities in access to LVADs, and, more recently, lower income has been associated with shorter times to readmission after implantation.13, 14 Also, while an analysis of the United Network for Organ Sharing database demonstrated no effect of SES on waitlist death or delisting with patients implanted with a BTT LVAD or in device‐related complication rates, the assessment of SES was based on neighborhood and census‐derived measures and not patient‐specific data.15 Our study focuses on furthering the understanding of patient‐level markers of SES and its effect on outcomes after LVAD implantation in a mixed population including both BTT and DT strategies.

2. METHODS

This study was a retrospective cohort study at the University of Florida, with the approval of the local Institutional Review Board. All LVAD placements between January 1, 2008 and December 31, 2015 were reviewed. Only those patients who received a continuous‐flow LVAD completed a psychosocial assessment prior to implantation, continued to follow at our institution postimplant, and were at least 1 year removed from implant at time of analysis were included. The psychosocial assessments were performed by a multidisciplinary team including a transplant and ventricular assist device (VAD) dedicated licensed medical social worker, clinical psychologist, and neuropsychologist. Parameters of SES included place of residence, marital status, level of education, financial resources, insurance status, substance use history, and medication adherence. Income was stratified by federal poverty level (FPL) using 2015 poverty guidelines (https://aspe.hhs.gov/2015-poverty-guidelines). Place of residence was divided into rural, micropolitan, and metropolitan, determined by the most recent U.S. Census data for each patient's residential address at time of implant. Rural was defined as a population of <10 000; micropolitan, 10 000 to 50 000; and metropolitan, a population of >50 000. Additional variables analyzed include age, sex, race, body mass index, comorbid conditions, implant strategy, Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) classification, survival status, infectious complications, bleeding, stroke, thrombus, status of readmission in 30 days, and length of initial admission. By institutional policy, all patients demonstrated abstinence from any recreational drug, tobacco and alcohol use for a period of 3 to 6 months prior to implantation. The primary outcomes were survival on device support or transplantation at 1 year (composite), implant length of stay (LOS), hospital readmission within 30 days of implant, and an aggregate of LVAD complications comprised of bleeding requiring transfusion more than 72 hours after implant, stroke, pump thrombosis, or device‐related infection requiring intervention.

Data were initially summarized by computing means and proportions with their standard errors. Univariate associations between two categorical variables were analyzed by use of either a χ ² or Fisher exact test. As age was continuous, its univariate association with each of the binary outcomes, that is, survival on device support or transplantation, readmission within 30 days and aggregated VAD complications were summarized using the two‐sample student t tests, and a nonparametric Wilcoxon rank‐sum (Mann‐Whitney U) test. As an inclusion criterion was continued follow‐up at our institution, by definition, there were no patients lost to follow‐up, and therefore no patient data were censored in our survival analysis. A descriptive analysis between age and the LOS response variable was performed using the nonparametric Kruskal‐Wallis rank test as LOS was categorical with three levels. Additionally, univariate logistic and ordinal regression models were fit for binary responses and naturally ordered LOS response, respectively. Covariates that indicated to be potential predictors of the outcome at the univariate models with a P value ≤0.05 were then adjusted for in multivariable robust logistic or ordinal regression models appropriately depending on the response variable whether binary or ordinal, while adjusting for potential confounding effects such as age, race and gender. For the ordinal logistic regression models, the proportional odds assumption (ie, the parallel regression assumption) was tested using a likelihood ratio test and the Brant test procedures provided in Stata software (Stata version 15). (StataCorp 2017, Stata Statistical Software: Release 15; StataCorp LLC, College Station, Texas).

3. RESULTS

During the study period, a total of 1051 patients were evaluated for advanced therapies of HF and ultimately 144 durable LVADs were implanted. Of those denied implantation at our center, 593 were found to be at the extremes of the HF spectrum (either “too sick” or “too well”) precluding implantation, 298 were noted to have a psychosocial barrier to advanced therapies, and 16 declined implantation. Of the 144 who received an LVAD, 33 of these implanted patients were excluded from analysis, of whom 3 did not receive follow‐up at our institution. The remaining 30 exclusions were unable to have preimplant income level stratified by FPL with the available documentation. A total of 111 patients met inclusion criteria and were included in the analysis. The mean age at time of implant was 57.6 years (range 19‐80), of which 92 were male (82.8%). Sixty‐one of the patients received LVADs as DT (54.9%), and 87 of the patients were alive or transplanted at 1 year (78.3%). Three patients received a HeartWare device (2.7%), 107 patients received a HeartMate II (96.3%), and 1 patient received a HeartMate III device (0.09%) (Table 1). Neither age, sex, body mass index (BMI) nor implant strategy (DT vs BTT) were found to significantly impact the primary outcomes (Table 2). INTERMACS classification (profiles 1 and 2 vs 3‐7) at the time of implant was associated with an increased risk of a prolonged LOS (>21 days, odds ratio [OR] = 2.04, P = 0.046) (Table 2).

Table 1.

Baseline characteristics

	Number (N = 111)	Percentage of total participants, %
Sex
Male	92	82.8
Female	19	17.2
Race
White	82	73.9
Black	19	17.1
Hispanic	5	4.5
All other race	5	4.5
Income
<100% FPL	13	11.7
100%‐200% FPL	45	40.5
>200% FPL	53	47.7
Insurance status
Medicare	33	29.7
Medicaid or Medicare + Medicaid	17	15.3
Private	61	54.9
Education
Less than high school	12	10.8
High school graduate	36	32.4
Some higher education	63	56.7
Geographic location
Micropolitan	38	34.2
Metropolitan	50	45.0
Rural	23	20.7
Implant strategy
Bridge to therapy	50	45.0
Destination therapy	61	54.9
INTERMACS classification at time of implant
Profiles 1‐2	52	46.8
Profiles 3‐7	59	53.2
Survival
<12 month	24	21.7
≥12 month	87	78.3
Length of stay
<14 days	35	31.5
14‐21 days	32	28.8
>21 days	44	39.6
Complications
Aggregate (bleeding, infection, CVA, or thrombus)	66	59.46
Readmitted within 30 days	16	14.4

CVA, cerebrovascular accident; FPL, federal poverty level; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support.

Table 2.

Univariate analysis of outcomes

	1‐year survival		Readmission within 30 days		Length of stay		Aggregate VAD complications
	Odds ratio	Confidence interval	Odds ratio	Confidence interval	Odds ratio	Confidence interval	Odds ratio	Confidence interval
Female gender	6	0.76‐47.47	2.62	0.79‐8.73	0.81	0.31‐2.1	0.93	0.34‐2.52
BMI	1.4	0.38‐5.3	1.2	0.25‐6.64	0.9	0.33‐2.5	1.06	0.34‐3.3
Race	0.50	0.19‐1.3	0.61	0.16‐2.32	1.18	0.54‐2.56	0.45	0.19‐1.05
Income	0.23	0.03‐1.95	0.84	0.15‐4.6	0.86	0.29‐2.6	1.2	0.36‐4.08
Insurance status	0.83	0.33‐2.1	0.6	0.2‐1.71	1.18	0.6‐2.4	0.96	0.45‐2.1
Education	1.42	0.33‐6.03	0.83	0.16‐4.45	0.30	0.09‐0.99 *	0.44	0.19‐1.8
Location	1.78	0.42‐7.52	3.8	0.64‐22.6	1.04	0.41‐2.66	1.17	0.41‐3.31
Marital status	0.37	0.05‐2.6	5.7	0.52‐62.2	0.92	0.24‐3.41	1.96	0.47‐8.11
Diabetes	0.92	0.36‐2.27	1.31	0.45‐3.8	0.68	0.33‐1.35	1.14	0.54‐2.44
HTN	1.6	0.67‐4.23	1.28	0.41‐3.99	0.89	0.44‐1.79	1.13	0.51‐2.49
COPD	0.67	0.12‐3.69	2.57	0.45‐14.5	1.26	0.25‐6.27	1.76	0.32‐9.5
HLD	1.55	0.63‐3.86	1.67	0.53‐5.18	0.59	0.27‐1.13	0.77	0.35‐1.68
Renal disease	0.72	0.29‐1.78	2.01	0.677‐5.99	0.83	0.42‐1.66	0.92	0.43‐1.97
PVD	1		0.83	0.09‐7.3	2.28	0.52‐9.98	2.15	0.41‐11.2
CAD	2.8	0.61‐13.3	2.42	0.73‐7.98	1.79	0.73‐4.43	2.3	0.78‐7.02
Implant strategy	0.67	0.26‐1.7	0.59	0.21‐1.71	1.9	0.95‐3.86	1.29	0.61‐2.78
Stroke	0.81	0.15‐4.32	0.83	0.09–7.3	1.9	0.51‐7.12	2.15	0.41–11.2
INTERMACS	0.48	0.18‐1.23	0.66	0.23‐1.91	2.04	1.01‐4.1 **	0.71	0.33‐1.52

BMI, body mass index; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; HLD, hyperlipidemia; HTN, hypertension; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; NS, non‐significant; PVD, peripheral vascular disease; VAD, ventricular assist device.

All other P values = NS.

^* P = 0.049.

^** P = 0.046.

Outcomes by SES are also reported in Table 2. There were no significant differences found for income level (FPL > 200% vs all others) or insurance status (private insurance vs all others) related to 1‐year survival, 30‐day readmission, LOS, or aggregate VAD complications. The same was true when examining race, patient location, and marital status. Likewise, there were no significant differences in education level (some higher education vs all others) with regard to survival, 30‐day readmission, or aggregate VAD complications. However, there was a noted association between LOS and education demonstrating that those with higher levels of education were at a lower risk of prolonged LOS (>21d, OR 0.30, P = 0.049). However, when further examining this association in a multivariable logistic regression model while controlling for age, race and gender, only INTERMACS classification was found to be predictive of a prolonged LOS (Table 3).

Table 3.

Multivariable logistic regression for the outcome length of stay >21 days

Category	Odds ratio	Confidence interval	P value
Age	0.99	0.96‐1.01	0.525
Gender	0.74	0.28‐1.99	0.552
Race	0.76	0.32‐1.84	0.558
Education	0.42	0.11‐1.61	0.209
INTERMACS	2.16	1.01‐4.64	0.047

INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support. Bold indicates statistically significant results.

4. DISCUSSION

Our study demonstrates that the survival benefit of VAD implantation spans across SES strata, in a mixed, real‐world population. And although our survival at 1 year is somewhat lower than that reported in the most recent INTERMACS report,5 and indeed lower than our contemporary experience, this may be due to our analysis crossing different eras of VAD implantation, that nearly half our cohort was INTERMACS profile 1 or 2, and that over 50% were implanted as DT. Nevertheless, this finding is similar to prior investigations demonstrating that Medicaid beneficiaries and individuals of lower median household income did not have worse survival after VAD implantation.14 These single‐center data were followed by an analysis of the United Network for Organ Sharing (UNOS) database wherein lower SES did not increase the composite of waitlist death or delisting for VAD recipients awaiting transplantation.15 It should be noted, however, that a slightly more contemporary analysis of the UNOS database did indeed demonstrate an increase in wait list mortality in Medicaid recepients.16 While our data confirm some these earlier findings, importantly our study does not rely on population‐based estimates of SES, such as US Census‐based median household income or neighborhood and Zip‐code‐based assessments of education as utilized previously. Rather, only patient‐level adjudicators of SES were analyzed, potentially providing a more robust fidelity of results. Even the relatively novel Agency for Healthcare Research and Quality SES index, a tool which incorporates multiple variables for determining levels of SES does so based on US Census data and therefore may not represent the same degree of granularity as individual, patient‐level adjudicators of SES such as those employed here.

Importantly, our study demonstrates that those with lower SES were not more likely to be readmitted or have a VAD‐related complication. This is in contrast to Smith et al14 who noted that individuals with lower median household income were more likely to be admitted after LVAD implantation. This discrepancy may be explained by the relative small numbers in both analyses, and the aforementioned difference in population‐based vs individual‐patient‐based assessment of SES, as we found no difference in outcomes based on percent of FPL. Nevertheless, the relationship between income and patient outcomes is likely complex and requires additional investigation to determine to what extent it may be a marker for health literacy.

A unique finding in our study demonstrates that those individuals with an education level beyond high school were 70% less likely to have an implant LOS >21 days compared with those with a high school education or less when examined in a univariate fashion. While this relationship did not remain in our multivariable analyses, it does perhaps provide a hypothesis to be tested in a larger study. We selected 21 days for our analysis as several studies have noted the median implant LOS to be 20 to 21 days.17, 18, 19 The need to focus on VAD implant LOS is underscored by the close relationship between LOS and total hospital cost, demonstrated across conditions20 and more recently determined to be the main driver of costs during the implant hospitalization with implant LOS explaining 77% of the cost variance at one center.21 Improving cost effectiveness of this life‐saving therapy will require continued refinement, not only in outpatient management, but also patient selection and identifying higher risk features. While this has traditionally centered on assessing the risk of mortality and right ventricular failure, expanding this risk assessment beyond traditional medical parameters, while challenging, may be of great utility in identifying novel variables that may serve as potential targets for quality improvement and cost containment.

The limitations of this study include its retrospective nature and that we are restricted to single‐center data. Our small sample size certainly constrains our ability to detect differences between groups. Therefore, in an attempt to overcome this limitation, insurance status, education, and income categories were dichotomized as outlined above. Additionally, patients who did not undergo a complete preimplantation psychosocial assessment, including income‐level stratification, may have excluded more urgent cases requiring expedited evaluations, and therefore may have unintentionally excluded patients with more severe illness. Furthermore, without an assessment of all patients who underwent evaluation for, but ultimately did not receive an LVAD, across strata of SES there may be a component of ascertainment bias. Further study of the 298 patients who did not receive an LVAD for various psychosocial barriers to advanced therapies would help to overcome this bias; however, many of these individuals did not have complete evaluations and after denial for LVAD implantation, a fair number did not follow at our institution rendering outcomes in this population difficult to assess. Beyond this, while LOS was utilized as an outcome, without precise knowledge of the specific barriers to discharge in each case, identifying strategies to improve this metric is challenging and therefore a limitation to the present investigation. That our study population was rather homogeneous, specifically in regards to gender and race, and with limited numbers of those at the most extreme range of poverty is a limitation and may have blunted our ability to detect differences in outcomes. The same can be said of our limited population of Medicaid LVAD recipients. Finally, that our study spans across several eras of LVAD implantation, surgical technique and medical management strategies, which have evolved present a possible confounder which, due to our limited patient population, we are unable to account for.

5. CONCLUSIONS

Our data suggests that patients should not be denied LVAD implantation solely based on their lower levels of SES. With potentially equivalent rates of survival as well as readmissions and complications postimplant, these individuals may receive similar benefits from implantation without an added marker of risk. Those with lower levels of education, however, may be at risk for longer implant LOS. While this may be a marker for a lower health literacy, it may suggest that this population would potentially benefit from greater targeted pre‐ and postimplant education and an altered management strategy. Further study is warranted to both refine estimates of preimplant risk, with particular attention to psychosocial risk and how to best determine SES, and thereafter create interventions based on those estimates of risk, to grow the field of mechanical circulatory support in an equitable and cost‐effective fashion.

CONFLICTS OF INTEREST

J.M.A. has served as a consultant for Thoratec Inc. C.T.K. has received research grant support from Thoratec Inc. and HeartWare.

Ahmed MM, Magar SM Jr, Jeng EI, et al. Effects of socioeconomic status on clinical outcomes with ventricular assist devices. Clin Cardiol. 2018;41:1463–1467. 10.1002/clc.23070