Association between psychosocial factors and left ventricular assist device implant outcomes: A systematic review

Abstract

Background:

Psychosocial assessment is a core component of the multidisciplinary evaluation for left ventricular assist device (LVAD) implantation. The degree to which psychosocial conditions are considered a contraindication to LVAD implantation continues to be debated. This systematic review examines modifiable psychosocial factors as predictors of outcomes in patients undergoing LVAD implantation.

Methods:

We conducted a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline. The search resulted in 2,509 articles. After deduplication, abstract and full-text review, 20 relevant articles were identified.

Results:

Included studies evaluated socioeconomic status (n = 6), caregiver characteristics (n = 6), non-adherence (n = 6), substance use (n = 13), and psychiatric disorder (n = 8). The most commonly measured outcomes were all-cause death, readmission rate, and adverse events. Studies varied widely in definition of each psychosocial factor and selected outcomes. No psychosocial factor was consistently associated with a specific outcome in all studies. Socioeconomic status was generally not associated with outcomes. Non-adherence, psychiatric disorder, and substance use were associated with higher risks of mortality, adverse events, and/or readmission. Findings on caregiver characteristics were mixed.

Conclusion:

Of the psychosocial factors studied, non-adherence, psychiatric disorder, and substance use were the most consistently associated with an increased risk of mortality, readmission, and/or adverse events. Heterogeneity in research methodology and study quality across studies precludes firm conclusions regarding the impact of psychosocial factors on long-term patient outcomes. The results of this review reveal a need for adequately powered studies that use uniform definitions of psychosocial factors to clarify relationships between these factors and outcomes after LVAD implantation.

Introduction

Psychosocial factors are the leading barrier to receiving a left ventricular assist device (LVAD) for patients with advanced heart failure (HF) [1]. According to a recent multisite study, approximately 20% of patients undergoing evaluation for LVAD do not receive this life-saving therapy due to psychosocial concerns [1]. Psychosocial factors are typically evaluated as a part of the psychosocial assessment by social worker and/or a mental health clinician (e.g., psychiatrist, psychologist) with expertise in these subject matter [2]. The psychosocial assessment is a core component of the multidisciplinary assessment process for patients being evaluated for LVAD and covers multiple domains including active repeated non-adherence to medications or other medical recommendations, alcohol or drug misuse, active psychiatric disorders that impact adherence without mitigating factors, lack of sufficient social support, and lack of cognitive or physical capacity to operate the LVAD safely [2, 3]. Yet, whether these psychosocial conditions should be considered a contraindication to LVAD implantation remain unclear due to limited availability of high-quality studies on this subject.

In 2014 Bruce et al. conducted a systematic review of psychosocial factors and their impact on LVAD outcomes [4]. Their review consisted of only five studies lacking in methodologic rigor, which prevented the authors from drawing firm conclusions. In that review, psychosocial factors were analyzed only as broad categories (e.g., psychological functioning, social functioning, etc.), limiting the study’s generalizability.

Unlike heart transplants, LVADs are not a scarce resource requiring the same utilitarian approach to recipient selection [5–7]. Nevertheless, LVADs are costly to implant, resource-intensive to maintain, associated with psychological distress and limited improvements in quality of life in recipients and caregivers, particularly those who struggle to adapt to the new therapy [8–12]. Therefore, it is crucial to establish the relationship between the psychosocial factors and outcomes of LVAD implantation (i.e., morbidity, mortality) to inform evidence-based guidelines and optimize patient selection for LVAD therapy.

This systematic review builds upon the previous work by Bruce et al. by taking a more nuanced approach to psychosocial categories and focusing on psychosocial factors that are considered modifiable (e.g., mental illness, social support, substance use) and, as a result, could have direct impact on clinical care and policymaking.

Materials and Methods

This systematic review was performed following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Figure 1). The authors collaborated with two research librarians to develop search terms for the following databases: PubMed, Embase, PsycINFO, Web of Science and Cochrane Library. Search strategy included three sets of terms: LVAD, psychosocial factors, and outcome. The search strategy used and search dates can be found in the Supplementary Table 1. All databases were searched using the same search strategy again during revision of this manuscript to ensure that any recent papers were included. The updated search date and results are included in Supplementary Table 1 and noted in the PRISMA flow diagram in Figure 1.

Figure 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Flow Diagram

* 3 additional studies were included from citations

** 2 additional study was included from an updated literature search

The inclusion criteria for this systematic review were: 1) English language, 2) peer-reviewed, 3) adult human subjects ≥ 18 years, 4) documented pre-LVAD implant psychosocial assessment, and 5) documented post LVAD-implant medical or surgical adverse events.

Studies that included pulsatile-flow LVAD devices were excluded, as these devices have been discontinued in favor of continuous-flow models due to higher rates of adverse events (e.g., stroke, device failure) [13, 14]. Inclusion of studies examining pulsatile-flow LVAD would limit this review’s generalizability to clinical practice. Studies that began prior to the approval of the first continuous-flow LVAD by the US Food and Drug Administration (April 2008; HeartMate II) [15] but did not specify device type (i.e., pulsatile or continuous-flow) were excluded. Studies that included non-LVAD mechanical circulatory devices (e.g., biventricular assist device) were also excluded. The subject of nonmodifiable risk factors (e.g., age, race, sex assigned at birth) is beyond the scope of this review and studies exclusively reporting on these factors were excluded. Studies only examining in-hospital outcomes (e.g., length of stay) were excluded to focus the scope of review. Case reports, abstracts, book chapters, professional consensus, or guidelines were also excluded.

Inclusion and exclusion decisions for all the studies were made by two of the reviewers (SB or JL, and JJC). Title and abstract were reviewed to exclude irrelevant publications (e.g., pediatric cohort, conference poster). The reviewers then reviewed the full text of studies that were considered relevant. In the case of reviewer disagreement, all three reviewers reviewed the text and established consensus. Information extracted from each study included pre-implant psychosocial factors, mortality and/or morbidity outcome, and time(s) to follow-up. Quality in Prognosis Studies (QUIPS) [16] was utilized for risk of bias assessment for the included studies and the rovis tool [17] was utilized to generate a summary table Supplementary Fig.1. All three reviewers completed QUIPS assessment for each study and established consensus. A statistician (PW) assisted with the assessment of bias in statistical analysis and reporting. QUIPS does not produce an overall quality score for a study. It assesses for six types of bias and if a study scores high on any of the risks, it is recommended to exclude the study from final data synthesis. Therefore, we defined any study with one or more domains scored as high risk as being at high risk of bias (ROB).

PROSPERO registration was not possible as only COVID-related studies were accepted at the start of this review.

Results

Literature search identified 2,509 records. After de-duplication (n = 221), 2253 records were excluded based on title and abstract review. The remaining 35 articles underwent full text review to determine eligibility and 15 of these articles screened in. Three additional articles that met review inclusion and exclusion criteria were identified from the references of these texts. Two additional articles were included after an updated literature search as described in Methods. In total, 20 unique articles were included in this systematic review Supplementary Table 2. All studies except one [18] adjusted for covariates Supplementary Table 2 and results reported here are derived from multivariate analyses that account for demographic (e.g., age, gender) and clinical (e.g., comorbidity, medication) covariates unless otherwise specified. Due to a wide variety of psychosocial factors and outcomes measured, meta-analysis was not feasible.

Nine of the 20 included studies scored high in at least one type of bias and are therefore considered high ROB Supplemental Fig.2 [18–23]. No study scored low on all six types of bias. The most common reason for risk of bias was unclear definition and measurement of psychosocial factors.

1. Socioeconomic Status

Six studies¹³⁻¹⁸ examined the relationship between SES and outcomes (Table 1). Most studies found that SES did not predict mortality, readmission, or adverse events, with some exceptions as described below. Of these six studies, one, whose authors describe it as hypothesis generating, had high ROB due to poor definition of prognostic factors and lack of adjustment for confounders [18].

Table 1.

Socioeconomic Factors

Study	Definition	Outcome association with socioeconomic factor
Insurance Type
Khatana (2021)	Medicaid Medicare Commercial	Mortality:  No difference by insurance type Adverse events:  Medicaid < Commercial insurance  Medicare > Commercial insurance
Clemons (2020)	Medicaid Medicare Private	Mortality:  No difference by Medicaid and Medicare Time to first admission:  No difference by insurance type
Ahmed (2018)	Medicaid Medicare Medicaid + Medicare Private	Mortality, readmission, adverse event:  No difference by private and all other insurance type
Kaiser (2019)*	Government Private None	Readmission:  No difference by insurance type
Smith (2014)	Medicaid Non-Medicaid	Mortality, readmission:  No difference by insurance type
Income
Ibarra (2021)	Cut-off for high- vs low-income level at 200% FPL	Mortality, readmission:  No difference by income level
Ahmed (2018)	Low: <100% FPL Medium: 100–200% FPL High: > 200% FPL	Mortality, readmission, adverse event:  No difference by income level
Smith (2014)	Low: <$38,370 Medium: $38,371 to <$56,890 High: >$56,891	Mortality  No difference by MHI level Readmission  Medium MHI < low MHI No difference between high and low MHI
Other
Ibarra (2021)	ADI	Mortality, readmission:  No difference by ADI
Clemons (2020)	nSES calculated using AHRQ SES index and reported as tertiles	Mortality, readmission: No difference by nSES level

Abbreviations: ADI, area deprivation index; AHRQ Agency for Healthcare Research and Quality; FPL, federal poverty line; MHI, Median Household Income; nSES, neighborhood-level socioeconomic status; SES, socioeconomic status.

^*Denotes high risk of bias

1.1. Health insurance type

Five studies examined the association between health insurance type and outcome [18, 24–27], one of which had high ROB [18]. Studies varied in categorizing health insurance type and status. All but one study [18] examined the effect of Medicaid and Medicare separately. Three studies compared Medicaid, Medicare and private insurance [24, 26, 27], while one study examined the effect of having Medicaid compared to all other types of insurance [25]. Only one study included uninsured status [18].

No study found an association between insurance type and mortality [24–27] or readmission rate [18, 25–27]. Two studies reported conflicting results on adverse events: in one, insurance type (i.e., private vs. others) was not associated with rates of adverse events [26] but another reported that, compared to private insurance, Medicaid recipients had a lower adjusted incidence of adverse events (IRR = 0.88, 95% CI: 0.78–0.99), whereas Medicare recipients were at an increased risk of adverse events (IRR = 1.16, 95% CI, 1.03–1.0) [24]. The latter study was larger, based on multisite data, and had a lower ROB than the one with negative findings.

1.2. Income

As with insurance type, studies varied in their measurement of income level. One study used patient zip code to determine median household income (MHI) [25], while the other two reported using the federal poverty line (FPL) [26, 28]. None of the three studies reporting on income were considered to have high ROB, and the study reporting MHI had the lowest ROB.

Despite varying definitions across studies, no association between income level and either mortality [25, 26, 28] or adverse events [26] was reported. Both studies using PFL reported no difference in readmission rates by income level [26, 28]; however, the study using MHI (and with the lowest ROB) found that medium MHI ($38,371 to $56,890) was associated with a lower rate of readmission than low MHI (< $38,370) (HR = 0.4, 95% CI: 0.24–0.65) [25].

1.3. Aggregate markers of SES

Two studies used composite SES indices: neighborhood SES (nSES) [27] and area deprivation index (ADI) [28]. nSES is calculated using the Agency for Healthcare Research and Quality (AHRQ) SES index that considers the following variables: mean number of 1 person or more per room, the median value of owner-occupied dwelling, the percentage unemployed, percentage living below the poverty level, the median household income, the percentage 25 years or older with a bachelor’s degree or higher, and the percentage 25 years or older with less than a 12th-grade education [29]. ADI is another measure of neighborhood disadvantage that is composed of 17 education, employment, housing, and poverty-related measures [30].

Neither study had high ROB. Both studies failed to find an association between nSES or ADI with mortality or readmission rates. Neither study reported on adverse events.

2. Caregiver Characteristics

Six studies examined caregiver characteristics in association with post-implant outcomes (Table 2) [12, 19, 20, 31, 32]. One study developed an assessment tool to analyze specific caregiver characteristics [19] whereas the other studies broadly assessed support availability [12, 20, 31, 32].

Table 2.

Caregiver characteristics and non-adherence

Study	Assessment tool/definition	Outcome association with caregiving or non-adherence
Caregiver characteristics
Wang (In press)*	INTERMACS: Limited social support	Readmission: Not associated
Maukel (2023)*	INTERMACS: Limited social support	Readmission, adverse events: not associated
Olt (2023)	SIPAT items VI – VII	Mortality: Not associated Readmission, device-related:  Associated with being divorced
Dew (2020)	m-PACT item IV-1	Adverse events:  No difference by family and support system stability, means, or physical ability/availability.
DeFilippis (2020)*	INTERMACS: Limited social support	Mortality: Not associated Readmission: Associated (device-related infection)
Bruce (2017)*	33-item tool developed by the study authors	Lower mortality:  Associated with caregivers who:   1. Understand the severity of patient’s illness and medical options.   2. Identified backup caregiver plan.   3. Can provide logistical support.   4. Is an adult child living within 50 miles   5. Is an extended family member Higher mortality:   Associated with living alone Readmission, adverse events: No difference by caregiver characteristic
Non-adherence
Wang (In press)*	INTERMACS: Repeated nonadherence	Readmission: Not associated
Maukel (2023)*	INTERMACS: Repeated noncompliance	Readmission: Not associated Adverse events: Bleeding and neurologic dysfunction (male only)
Olt (2023)	SIPAT item A-IV	Mortality, readmission:  Not associated
Dew (2021)	m-PACT item IV-2	Adverse events: Associated with greater degree of non-adherence
DeFilippis (2020)*	INTERMACS: Noncompliance	Mortality:  Not associated with non-adherence Readmission, adverse events:  Associated with non-adherence
Kirklin (2018)*	History of repeated noncompliance	Pump-thrombosis: Associated with non-adherence

Abbreviations: INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; m-PACT, Psychosocial Assessment of Candidates for Transplantation modified for mechanical circulatory support; SIPAT, Stanford Integrated Psychosocial Assessment for Transplant

^*Denotes high risk of bias

The study that examined specific caregiver characteristics was notable for having an especially high ROB (i.e., 3 of 6 domains rated as high ROB) although it provided the most nuanced investigation of these associations [19]. It found the following to be associated with a lower risk of 1-year mortality: 1) understanding of patient medical condition (HR = 0.21, 95% CI: 0.08–0.73); 2) presence of backup caregiver plan (HR = 0.27, 95% CI: 0.09–0.83); 3) logistical support capability (HR = 0.42, 95% CI: 0.18–0.97), 4) presence of an adult child within 50 miles (HR = 0.28, 95% CI: 0.09–0.91), and 5) availability of extended family member to help (HR = 0.41, 95% CI: 0.18–0.91). Conversely, living alone was associated with higher risk of 1-year mortality (HR = 3.15, 95% CI: 1.07–9.29). Two other studies with comparatively lower ROB failed to identify an association between caregiver availability and mortality [20, 32]. Of the four studies that investigated caregiver availability as a predictor of hospital readmission, the high ROB study was one of three that failed to find a significant association [12, 19, 33]. Predictors of readmission included divorced status (HR = 1.13, 95% CI: 1.04–1.24) and limited social support (HR = 1.13, 95% CI: 1.04–1.24) [20, 32].

Of the two studies that investigated caregiver availability in relation to adverse events, two high ROB studies failed to find an association [19, 33], whereas another study with high ROB, although in fewer domains, reported that limited social support predicted device-related infection (HR = 1.44, 95% CI: 1.22–1.69) [20].

3. Non-adherence

Four studies with high ROB [12, 20, 33, 34] and two additional studies with lower ROB [31, 32] reported on non-adherence (Table 2). There was no evidence supporting an association between mortality and non-adherence [20, 32]. Three studies differed on the significance of association between readmission and non-adherence [12, 20, 32, 33] with only one high ROB study reporting significant findings (HR = 1.13, 95% CI 1.04–1.24) [20].

Various types of adverse events were associated with non-adherence in all four studies that reported on this domain [20, 31, 33, 34]: device-related infections (HR = 1.58, 95% CI 1.31–1.90), [20] device thrombosis (HR = 1.29, 95% CI 1.03–1.62), [20] stroke (HR = 1.38, 95% CI 1.11–1.71), [20] gastrointestinal bleeding (HR = 1.40, 95% CI 1.17–1.67), [20] cardiac arrhythmia (HR = 1.41, 95% CI, 1.12–1.78), [31] and device malfunction (HR = 1.52, 95% CI, 1.10–2.08) [31]. In a study that stratified the results by sex [33], individual adverse events were significantly associated with repeated non-adherence only in male patients (Bleeding: HR=1.31, 95% CI 1.11–1.54; neurologic dysfunction: HR=1.23, 95% CI, 1.01–1.50) [33].

4. Substance Use

Thirteen studies evaluated substance use as a pre-implant psychosocial factor (Table 3) [12, 18, 20, 22, 23, 31–38]. Studies varied in definition of substance use. Only one study clearly defined substance use using the Diagnostic and Statistical Manual of Mental Disorders (DSM), 4^th Edition [38]. The remaining studies did not operationalize substance use, abuse, or dependence. Studies at times specified which substance was studied (e.g., alcohol, tobacco) [20, 22, 23, 31, 32, 34–37] and at other times combined the use of several substances into a single category (e.g., narcotic dependence, substance or drug use/abuse) [12, 18, 20, 23, 31, 32, 38]. No study examined cannabis use as a standalone category. No study included cannabis use. The lack of clear definition of substance use was a common reason that six of these papers scored as having high ROB [12, 18, 20, 22, 23, 34].

Table 3.

Substance Use

Study	Terminology	Outcome association with substance use by type
Tobacco
Olt (2023)	SIPAT item D XVIII	Mortality, readmission: no difference by smoking status
Maukel (2023)*	INTERMACS: History of smoking	Readmission, adverse events: associated
Combs (2020)	Never, former, or current smoker	Mortality: no difference by smoking status Readmission: Current > never Sepsis: Current < former, never Other adverse events: Current > former
Imamura (2020)	Never or current smoker (analysis by sex)	Male Mortality: no difference by smoking status Readmission: Current > never Adverse events: no difference by smoking status Female: Mortality, readmission, adverse events: no difference by smoking status
Kutyifa (2020)*	smoking at baseline	Readmission: not associated
Gupta (2019)	Previous smoking	Readmission: not associated
Kirklin (2015)*	Current smoker	Pump thrombosis: associated
Alcohol
Olt (2023)	SIPAT item D XIV-V	Mortality, readmission: not associated
Maukel (2023)*	INTERMACS: History alcohol abuse	Readmission, adverse events: not associated when stratified by sex (M/F)
DeFilippis (2020)*	INTERMACS: Alcohol abuse	Mortality: not associated Readmission, adverse events: associated
Mullan (2020)*	INTERMACS: Alcohol abuse	Mortality: not associated Device infection, readmission: associated
Narcotic
Mullan (2020)*	Narcotic dependence	Mortality: Associated Readmission, adverse events: not associated
General
Wang (In press)*	INTERMACS: Substance abuse	Readmission: associated
Olt (2023)	SIPAT items D XVI - XVII	Mortality, readmission, adverse events: not associated
Maukel (2023)*	INTERMACS: History drug abuse	Readmission: not associated Adverse events: associated with pump thrombosis (male only)
Mullan (2020)*	Drug Use	Mortality: not associated Readmission, adverse events: associated
DeFilippis (2020)*	Drug use	Mortality: not associated Adverse events, readmission: associated
Dew (2020)	m-PACT item III-2	Device malfunction: associated with a greater degree of substance use
Kaiser (2019)*	Substance abuse	Readmission: not associated
Cogswell (2014)	Substance abuse (DSM-IV)	Mortality: associated Drive-line infection: associated

Abbreviations: DSM-IV, Diagnostic and Statistical Manual of Mental Disorders, 4th Edition; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; m-PACT, Psychosocial Assessment of Candidates for Transplantation modified for mechanical circulatory support; SIPAT, Stanford Integrated Psychosocial Assessment for Transplant.

^*Denotes high risk of bias

4.1. Tobacco

Seven papers examined the association between smoking tobacco and post-LVAD implantation mortality, morbidity, complications, and readmission [22, 32–37]. No study quantified the extent of pre-implant tobacco exposure or continued use post-implant. The pre-implant tobacco exposure was generally categorized as active smoking at the time of implant and/or smoking in the past. Active smoking at the time of implant was the most frequent exposure type studied with its point prevalence ranging from 5–36%. One study examined the differences in the impact of smoking by sex assigned at birth and reported results for each sex separately [37]. Three studies score high on the ROB assessment [22, 33, 34].

Mortality was not associated with exposure to smoking, present or past [32, 36, 37]. Findings on readmission varied; a mix of high and lower ROB papers found significant [33, 36, 37] and non-significant [22, 32, 35] positive associations between smoking and readmission.

Three studies varied in their findings on how adverse events are associated with tobacco smoking. Kirklin and colleagues examined pump thrombosis exclusively as its primary outcome and found active smoking at the time of implant to be associated with increased risk of pump thrombosis (HR = 1.49, p <0.02) [34]. The other two studies which analyzed a nearly identical data set, demonstrated that the risks of gastrointestinal bleeding, stroke, driveline infection and hemolysis were greater in active smokers compared to those who never smoked [36] but these findings were no longer significant when stratified by sex [37]. Of note, the stratification of results by sex in the latter study limited its power with a very low number of smokers included in the analysis (male =25, female = 9) [37]. Lastly, contrary to expectations, one study reported that smoking was associated with a lower risk of post-implant sepsis (p < 0.05). [36]

4.2. Alcohol

Three studies examined alcohol misuse and its association with post-implant complications [20, 23, 32]. Two studies presented results of INTERMACS data analysis of the same study period (2008–2017) with results differing only on associations with some of the adverse events [20, 23]. Both were scored as having high ROB largely due to lack of standardized definition of “alcohol abuse” for the INTERMACS database. The other study utilized the Stanford Integrated Psychosocial Assessment for Transplant (SIPAT) to assess for alcohol use disorder [32] and had significantly lower ROB. The INTERMACS-based studies estimate prevalence of alcohol misuse to about 7% of the over 15,000 participants analyzed [20, 23]. Unfortunately, the single-site study with a clearer definition of alcohol misuse, does not report the prevalence. Like the studies on tobacco smoking, no study reported on post-implant alcohol consumption.

All three studies found no significant association between alcohol misuse and mortality rate. The INTERMACS-based studies were consistent in their findings of a significant association between alcohol misuse and readmission (HR = 1.1, 95% CI 1.1–1.2 [23]; HR = 1.11, 95% CI 1.03–1.19 [20]), but not when stratified by sex [33]. A single-center study of lower ROB also did not find any significant associations [32].

The three INTERMACS-based studies reported inconsistent findings on adverse events. [20, 23, 33]. Two of the studies reported a significant association between alcohol abuse and device-related infection (HR = 1.3, 95% CI 1.2–1.5 [23]; HR = 1.26, 95% CI 1.09–1.45 [20]), and one [20] found a significant link with gastrointestinal bleeding (HR = 1.22, 95% CI 1.08–1.38). When stratified by sex, these associations were no longer significant [33].

4.2. Opioids and Other Substances

Eight studies examined LVAD outcomes associated with general substance use, opioid dependence, or other substances [18, 20, 23, 31, 32, 38]. One of the studies used the term “narcotic dependence” to describe opioid dependence as this is the language used in INTERMACS, the source of their study data [23]. The INTERMACS-based studies additionally reported on “drug use” or “history of drug abuse” which lacks operationalized definition [12, 20, 23]. Other studies used the term “substance abuse” either undefined [18] or defined by validated psychosocial assessment tools [31, 32] or the DSM-IV [38]. Four papers were scored as high ROB due to lack of clear definitions [12, 18, 20, 23].

Narcotic dependence, but not substance/drug use, was associated with increased risk of mortality (HR = 1.9; 95% CI 1.2–3.0) [23] across all studies. However, it should be noted that the other INTERMACS studies did not analyze the effect of narcotic dependence, citing high rate of missingness [12, 20].

Substance abuse/drug use was not associated with mortality in all studies [12, 20, 23, 32] except one (HR =5.6, 95% CI 1.9–16.3) [38]. The study with the positive finding was a smaller, single-site study but methodologically was robust in using a matched cohort design and DSM-IV to diagnose a disorder.

Findings on readmission were mixed. Three studies analyzing INTERMACS data (similar study period, slightly different cohort) found a significant association with drug use (HR=1.1, CI 95% 1.1–1.2 [23]; HR=1.2, 95% CI 1.03–1.22 [12]; HR=1.09, 95% CI 1.01–1.17 [20]); when stratified by sex, the results were no longer significant [33]. Other studies also did not find evidence supporting the association [18, 31, 32]. The studies with negative finding had overall lower ROB.

Of all the adverse events reported, device-related infection was consistently associated with substance abuse/drug use across studies (HR = 1.24, 95% CI 1.07–1.44 [20]; HR=1.5, 95% CI 1.4–1.7 [23]; rate ratio = 5.43, 95%CI 1.3–31.8 [38], except when results were stratified by sex [33].

5. Depression and Other Psychiatric Symptoms

Eight studies investigated associations between psychiatric symptoms and LVAD outcome (Table 4) [12, 18, 20, 21, 23, 31–33]. The most studied psychiatric symptom was depression followed by catch-all categories: “other major psychiatric diagnosis,” “other mental health issues,” “psychopathology.” Except for the two utilizing the SIPAT or the Psychosocial Assessment of Candidates for Transplantation Modified for Mechanical Circulatory Support (m-PACT) [31, 32], other studies lacked a clear definition of psychiatric symptoms and/or diagnoses, rendering them as high ROB [12, 18, 20, 23, 33].

Table 4.

Psychiatric Symptoms

Study	Terminology	Outcome association with psychiatric disorders by type
Depression
Maukel (2023)*	INTERMACS: Severe depression	Readmission: not associated Adverse events: associated (device infection – male only; pump thrombosis – male and female)
Olt (2023)	SIPAT item C-IXa	Mortality, readmission: not associated
DeFilippis (2020)*	INTERMACS: Severe depression	Mortality: not associated GI bleeding, readmission: associated
Mullan (2020)*	INTERMACS: Severe depression	Mortality: not associated Adverse events, readmission: associated
Kaiser (2019)*	Depression	Readmission: associated
Gordon (2013)*	Depression	Device-related infection: associated
Anxiety
Olt (2023)	SIPAT item C-IXb	Mortality, readmission: not associated
Kaiser (2019)*	Anxiety	Readmission: associated
Personality disorder
Olt (2023)	SIPAT item C-XI	Mortality, readmission: not associated
Coping/family psychiatric history
Olt (2023)	SIPAT item C-XIII	Mortality, readmission: not associated
Other
Wang (In press)*	INTERMACS: Major psychiatric disorders	Readmission: associated
Maukel (2023)*	INTERMACS: Other major psychiatric diagnosis	Readmission: associated (female only) Adverse events: associated – infection (male only); thrombosis (female only); neurologic dysfunction (male only)
Olt (2023)	SIPAT item C-IX (mood, anxiety, psychosis and others)	Mortality: not associated Readmissions: Associated with presence of psychopathology
Dew (2021)	m-PACT item II:  Severe chronic psychopathology  Moderate personality/coping and adjustment issues Well-adjusted	Adverse events: associated with greater degree of psychopathology
DeFilippis (2020)*	INTERMACS: Other major psychiatric diagnosis	Mortality, adverse events, readmission: associated
Mullan (2020)*	INTERMACS: Other major psychiatric diagnosis	Mortality, adverse events, readmission: associated
Kaiser (2019)*	Other mental health issues	Readmission: not associated

Abbreviations: GI, gastrointestinal; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; m-PACT, Psychosocial Assessment of Candidates for Transplantation modified for mechanical circulatory support

^*Denotes high risk of bias

Studies that reported “depression” [18, 21] and “severe depression” [20, 23, 33] did not provide a definition for these terms. One study used the SIPAT [32], whose item IXa is scored based on symptoms of clinical depression determined by clinical assessment or validated depression rating scales [39]. No study found depression symptoms to be associated with mortality risk [20, 23, 32]. Studies that used the term “severe depression” found a significant association between depression and readmission rate (HR=1.3, 95% CI 1.2–1.9 [23]; HR=1.8, 95% CI 1.06–1.33 [20]); however, results were no longer significant upon stratification by sex [33]. The study using SIPAT to denote presence and severity of clinical depression did not find any significant links between depression readmission either [32]. The three INTERMACS-based studies examined the association between “severe depression” and adverse events [20, 23, 33]. Despite sharing very similar study methods for the same data set, the three studies did not agree on significant findings for any of the adverse events.

Seven studies examined the association between presence of other psychopathologies and outcomes [12, 18, 20, 23, 31–33]. The vagueness of this category renders comparison of results across studies even more difficult than other psychosocial factors. Mortality, readmission and adverse events were inconsistently associated with psychopathology across these studies (Table 4). One study with lower ROB examined adverse events by phase (i.e., early versus late post-implantation) and found almost all patients experienced an adverse event by the end of study period but patients with psychosocial factors were more likely to have it earlier in the post-implantation phase [31]. In this study, greater severity of mental health problems was associated with a higher cumulative risk of adverse events (HR=1.32, 95% CI 1.08–1.60).

Data on other psychiatric conditions (i.e., anxiety, personality disorder, coping) was too limited to draw any conclusions on their association with post-implant outcomes.

Discussion

In the era of growing LVAD therapy for treatment of heart failure, understanding factors that impact post-implant outcomes is crucial to the informed consent process as well as to optimizing outcomes. This systematic review includes 20 studies encompassing a wide range of psychosocial factors and provides mixed evidence for each factor. Importantly, although this review includes potentially modifiable psychosocial factors, included studies only report associations; therefore, causality cannot be inferred.

Overall, the results suggest absent or weak associations between psychosocial factors and outcomes included in this review. Exceptions are the significant associations of non-adherence with adverse events, substance use with adverse events and readmission, and depressive symptoms with adverse events and readmission. While one study [19] reported significant findings between specific caregiver characteristics and outcomes, it scored high risk in three domains of bias on the QUIPS. Therefore, in absence of other studies to replicate its findings, its results should be interpreted with caution.

It stands to reason that non-adherence would be associated with higher risk of adverse events. Medication non-adherence would increase the risk of LVAD complications and worsening underlying medical condition, while appointment and laboratory testing non-adherence could result in inadequate medication/LVAD management and/or a delay in recognition of complications. To effectively mitigate the risk of adverse outcomes associated with non-adherence, understanding the reasons for non-adherence is important. [40–42] Systems-level barriers to non-adherence (e.g., transportation, cost, etc.) could be overcome through resource provision (e.g., subsidized transportation, medication vouchers), while more individual reasons (e.g., personal or cultural experiences) require a more nuanced approach. Similarly, it is important to distinguish between intentional (e.g., disregard for clinician) and unintentional non-adherence (e.g., lack of access to transportation) as each type has different intervention implications (i.e., behavioral- versus resource-focused interventions) [42]. Out of the four studies on non-adherence included in this systematic review, only one study makes a distinction between intentional and unintentional non-adherence using the m-PACT scale [31]. Future studies on this subject should examine specific, objectively measurable components of non-adherence not reliant on self-report to guide risk stratification as well as intervention strategies.

The potential mechanism for the increased risk of adverse outcome associated with substance use is less clear. The findings are difficult to interpret due to the heterogeneity in definition of substance use, lack of specificity in substance type, and absent record of amount and frequency used. Studies using INTEMACS data use the outdated term “narcotic,” which is not only unclear in definition but also stigmatizing given its connotations of illegality. Additionally, these observational studies do not indicate whether substances of misuse may directly cause physiological effects leading to complications, or whether behavioral or other psychosocial factors associated with substance use (i.e., drug-induced mood disorder, neglecting self-care) might result in specific outcomes. Further, none of the studies that investigated substances defined use patterns (e.g., duration of use and abstinence, amount, frequency, or use post-implant), so these results are unable to inform risk stratification in a graded fashion.

The studies examining the association between psychiatric symptoms and LVAD outcomes share similar limitations. Strikingly, no included study reported a definition for the psychiatric disorder being assessed. Whereas “depression” was the most frequently studied mental health condition, it is unclear in each study whether this refers to major depressive disorder, self-reported depressive mood at the time of assessment, the results of a screening questionnaire (e.g., Patient Health Questionnaire), or otherwise. It is unclear how to apply these findings to individual patients. Likewise, reporting all other non-depressive psychiatric disorders as a single entity renders use of the results for risk stratification impossible. The study by Bruce et al. on caregiver characteristics avoids this problem by clearly defining the domain of social support into specific subcategories. [19] Although the large quantity of missing data in many of the categories raises concern for bias, this study nevertheless provides a blueprint for future studies on studying psychosocial factors.

The results of this systematic review is consistent with that of Bruce et al. in 2014 [4] in its conclusion that available data is insufficient to “identify definitive evidence-based psychosocial contraindications” due to lack of robust, high-quality data that is harmonized across studies and amenable to menta-analysis. Although the number of studies included in systematic review has grown from five to 20, the call for greater clarity in definition of psychosocial factors remain unanswered.

Several limitations were common across included studies. First, all studies excluded patients who were evaluated for but not offered LVAD therapy. Some of the exclusions may be secondary to psychosocial factors, which would mean included subjects in these studies represent a less psychosocially complex population, introducing the risk of selection bias. Second, no study examined the effect of combining specific psychosocial factors on outcome. While one psychosocial factor itself may not significantly be associated with increased risk of mortality or morbidity, the presence of certain combinations of factors may be associated with differences in risk, including potential protective effects. Additionally, the studies include various types of continuous-flow LVAD, including ones that are no longer on the market, such as the HVAD HeartWare. Devices vary in their proclivity to malfunction or adverse events, which introduces another variable in to interpreting medical outcome. However, many studies that included more than one type of device did not account for device type. Lastly, risk of bias assessment revealed that the quality of the studies ranged from moderate to low.

Conclusions

The presence and strength of evidence for the link between psychosocial factors and LVAD outcome remain limited. Of the factors studied, non-adherence, substance use, and depressive symptoms appear to have the most consistent association with adverse outcome. The results of this systematic review cautions clinicians against considering any of these risk factors as absolute contraindications to LVAD implantation. Standardized definitions of psychosocial factors across LVAD centers will be important to enhance the quality of data that can inform clinical practice and future studies on interventions for these modifiable factors.