Impact of Substance Use Disorder on Midterm Mortality After Valve Surgery for Endocarditis

Abstract

Background.

Fueled by the burgeoning opioid epidemic, valve surgeries for substance use disorder-related infective endocarditis (SUD-IE) are increasing. The impact of substance use disorder on postvalve replacement morbidity needs further investigation.

Methods.

We queried The Society of Thoracic Surgeons Adult Cardiac Surgery Database for all valve surgeries for infective endocarditis at Tufts Medical Center (2002-2016) and collected demographic and disease-related data, including timing of mortality subclassified as short-term (<6 months including operative), midterm (6 months to 5 years), and extended-term (>5 years). Patients with documentation of illicit drug use before the operation were considered to have SUD-IE. Deaths were confirmed through review of medical record and matching with the Massachusetts Vital Statistics Database. We performed univariate and multivariate proportional hazard regressions examining the impact of substance use disorder mortality in people who received a valve replacement.

Results.

In the cohort of 228 patients, 80 (35%) had SUD-IE. Eight-six people (38%) died, and overall mortality was higher in people with SUD-IE compared with non–SUD-IE (48% vs 32%, P = .025). SUD-IE was associated with a higher risk of overall mortality (adjusted hazard ratio, 2.41; 95% confidence interval, 1.38-4.20; P = .002). Although the difference between short-term or extended-term mortality was not significant, SUD-IE was associated with increased frequency of midterm mortality (53% vs 31%, P = .003).

Conclusions.

Our data reflect high rates of postvalve surgery morbidity and mortality in people with SUD-IE at a tertiary care center. The midterm postoperative period is a vulnerable period for people with SUD-IE worthy of further investigation.

Despite advances in surgical techniques and antibiotics, infective endocarditis is associated with significant morbidity and mortality.^1,2 One risk factor for developing infective endocarditis is the injection of drugs such as opiates, opioids (eg, heroin, fentanyl), cocaine, or methamphetamines. Parallel with the burgeoning epidemic of nonprescription opioid use, there are increasing numbers of people with infective endocarditis associated with injection drugs use (referred to from this point on as substance use disorder related-infective endocarditis [SUD-IE]) who are being hospitalized.^3-5

With increasing numbers of people with SUD-IE admitted to the hospital and updated guidelines for the surgical management of infective endocarditis suggesting criteria for early surgical intervention and improved outcomes,⁶ the number of valve surgeries on people with SUD-IE is increasing.^5,7,8 Literature about the impact of substance use disorder on postvalve replacement mortality after the postoperative period is mixed. A meta-analysis by Hall and colleagues⁹ found no difference in 30-day mortality between people with SUD-IE and people with non–SUD-IE; however, 2 studies published recently showed lower inpatient mortality in people with SUD-IE.^5,10 Studies to date have discordant results on association with substance use disorder with short-, mid-, and extended-term outcomes.^7,11-14 A better understanding of the timing of postoperative mortality in people with SUD-IE compared with people with non–SUD-IE is urgently needed to proactively investigate ways to improve postsurgical outcomes.

Recognizing discordant results in the literature, we performed a retrospective cohort analysis of all patients who underwent valve surgery for infective endocarditis at our institution between 2002 and 2016 with the goal comparing mortality and timing of mortality in people with SUD-IE to people with endocarditis unrelated to substance use disorder.

Patients and Methods

Study Design

We queried The Society of Thoracic Surgeons Adult Cardiac Surgery Database (STS-ACSD) for all patients who underwent cardiac surgery with a listed diagnosis of infective endocarditis between January 2002 and December 2016. Institutional medical records were retrospectively reviewed to evaluate patient eligibility by excluding any patients who underwent surgical procedures that did not include repair or replacement of a valve affected with infective endocarditis. The Tufts Medical Center Institutional Review Board reviewed and approved the study, including a waiver for patient consent.

Variables Collected and Definitions of Variables

Demographic, disease-related and perioperative data collected included patient age, sex, history of risk factors, immunologic disorders, New York Heart Association Functional Classification, comorbidities, medications, blood culture results, transthoracic or transesophageal echocardiogram results, or both, operative details, and postoperative course, complications, and hospital length of stay. Endocarditis was considered to be “active” if the patient demonstrated any ongoing fevers, positive blood cultures, or septic emboli before the operation. Immunologic disorders included systemic steroid therapy, antirejection medications, and chemotherapy. The type of immunologic disorder was not specified when present. Postoperative arrhythmia included atrial fibrillation, postoperative dysrhythmia needing a pacemaker, and cardiac arrest.

The STS ACSD has a variable “Illicit Drug Use.” All medical records were reviewed to confirm that this variable was assigned appropriately to people who injected nonprescription injection drugs (eg, heroin, cocaine, methamphetamines), including review of toxicology screens and notes documenting self-reported or clinician concern for injection drug use. No addiction consultation service was available during the study period, and data about available psychiatric consultations in the electronic medical record were limited, so this information was not collected.

Recurrent infective endocarditis was defined as evidence of reinfection deemed to meet the criteria for infective endocarditis, including the presence of positive blood cultures and an echocardiogram confirming vegetation.

Outcome

We were interested in mortality, subdivided into (1) short-term mortality, (2) midterm mortality, and (3) extended-term mortality. We defined short-term mortality as occurring within 6 months of the operation (inclusive of operative mortality), midterm mortality was between 6 months and 5 years after the operation, and extended-term mortality was greater than 5 years from the operation. Definitions of short-, mid-, and extended-term were based on discussion with surgical colleagues and review of the cardiac surgery literature. If people underwent more than 1 valve operation, the time to mortality was based on the date of the first valve operation. Mortality was determined through (1) medical record review and (2) query of the Massachusetts Vital Statistics Database, updated with in-state mortality data through December 31, 2016.

We reviewed each of the discharge summaries for cause of death of the people who died. If the person did not die in the hospital, the final discharge summary was reviewed to further investigate whether during the final encounter at the hospital, the patient had endocarditis on the last encounter at our hospital before death. If the person did not die in the hospital, causes of death were classified by Automatic Classification of Medical Entry death codes.

Statistical Analysis

Data are described as medians with interquartile ranges for continuous variables and as frequencies with percentages for categorical variables. Wilcoxon rank sum tests were used to assess statistically significant differences between continuous and binary variables, and χ² tests were used for comparing categorical variables. Fisher exact tests were used for comparison of categorical variables with expected cell counts of less than 5 in any cell. P values of .05 or less were considered statistically significant. Kaplan-Meier survival curves and the logrank test were used to determine differences in overall mortality between SUD-IE and non–SUD-IE patients. Univariable and multivariable Cox proportional hazards regression models were used to identify predictors of time to overall mortality. Statistical calculations were performed using SAS 9.4 software (SAS Institute, Inc, Cary, NC).

Results

The final cohort represented 228 unique patients who underwent valve surgery for infective endocarditis and 246 surgeries (18 people had replacement valve surgery). Baseline characteristics are summarized in Table 1. Eighty patients (35%) had SUD-IE, and the SUD-IE group was significantly younger (38 [29, 48] years vs 60 [50, 69] years, P < .001). More common in the SUD-IE group were Staphylococcus aureus (49% vs 22%, P < .001) and tricuspid (38% vs 8%, P < .001) endocarditis. A greater proportion of patients in the SUD-IE group had depression (40% vs 14% P < .001) and used alcohol (21% vs 7%, P = .001), but fewer were immunosuppressed (5% vs 14%, P = .046). There was no difference between groups in the proportion of nonelective surgery, proportion of preoperative heart failure, or mean days from admission to surgery; however, preoperative arrhythmia was less likely in the SUD-IE group (4% vs 32%, P < .001).

Table 1.

Baseline Demographics and Culture Results

Characteristic	All Patients (N = 228)	SUD-IE (n = 80)	Non–SUD-IE (n = 148)	P Value
Age at time of surgery, y	53 (39, 63)	38 (29, 48)	60 (50, 69)	<.001
Male	154 (68)	56 (70)	98 (66)	.561
Type of admission				.248
Transfer from another institution	172 (75)	65 (81)	107 (72)
From emergency department	31 (14)	7 (9)	24 (15)
Scheduled admission	25 (11)	8 (10)	17 (11)
Type of infected endocarditis				.366
Active	199 (87)	72 (90)	127 (86)
Treated	29 (13)	8 (10)	21 (14)
Culture microorganism				<.001
Staphylococcus aureus	71 (31)	39 (49)	32 (22)
Non-Staphylococcus aureus	134 (59)	38 (48)	96 (65)
Streptococcus species	65 (29)	17 (21)	48 (32)
Enterococcus species	28 (12)	8 (10)	20 (14)
Coagulase-negative Staphylococcus	15 (7)	2 (3)	13 (9)
Fungal	7 (3)	3 (4)	4 (3)
Other	19 (8)	8 (10)	11 (7)
Unknown	23 (10)	3 (4)	20 (14)
Comorbid conditions
Depression	52 (23)	32 (40)	20 (14)	<.001
Alcohol	27 (12)	17 (21)	10 (7)	.001
Immunosuppression	24 (11)	4 (5)	20 (14)	.046
History of previous heart failure	39 (17)	10 (13)	29 (20)	.176
Valves with endocarditis
Aortic valve	123 (54)	34 (43)	89 (60)	.011
Prosthetic aortic valve	23 (10)	4 (5)	19 (13)	.061
Mitral valve	105 (46)	28 (35)	77 (52)	.014
Prosthetic mitral valve	5 (2)	0 (0)	5 (3)	.097
Tricuspid valve, all native	42 (18)	30 (38)	12 (8)	<.001
Pulmonary valve, all prosthetic	1 (.4)	1 (1)	0 (0)	.174
Multiple valves	33 (14)	8 (10)	25 (17)	.159
Preoperative NYHA classification				.661
I	5 (2)	2 (3)	3 (2)
II	63 (28)	26 (33)	37 (25)
III	82 (36)	27 (34)	55 (37)
IV	78 (34)	25 (30)	53 (36)
Preoperative arrhythmia	51 (22)	3 (4)	48 (32)	<.001
Nonelective surgery	173 (76%)	65 (81%)	108 (73%)	.164
Index surgery was first cardiac surgery at Tufts	186 (82)	71 (89)	115 (78)	.064
Interval from admission to surgery, d	4 (1, 7)	3 (1, 8)	4 (1, 6)	.618

Data are presented as n (%) or as median (interquartile range).

NYHA, New York Heart Association; SUD-IE, substance use disorder–related infective endocarditis.

Endocarditis after valve replacement occurred in 17 people, was more common in people with SUD-IE (15% vs 3%, P = .002), and repeat valve replacement for endocarditis was more likely to happen in people with SUD-IE (14% vs 3%, P = .003). People with SUD-IE had less time between valve replacements compared with non–SUD-IE (240 [143, 402] days vs 304 [19, 840] days, P = .002).

Eighty-six patients (38%) died, with more deaths in the SUD-IE group than in the non–SUD-IE group (48% vs 32%, P = .025; Table 2); of these, 29 (34%) died at our medical center. Among the 86 deaths, 36% were short-term mortality, 41% had midterm mortality, and 23% had extended-term mortality. Differences in short-term mortality or extended-term mortality were not statistically significant; however, people with SUD-IE were more likely than people with non–SUD-IE to die in the midterm period (53% vs 31%, P < .003). Recurrent endocarditis, overdose, liver failure, and unknown/other were more likely to be the cause of death in people with SUD-IE.

Table 2.

Perioperative and Postoperative Outcomes

Outcomes	All Patients (N = 228)	SUD-IE (n = 80)	Non–SUD-IE (n = 148)	P Value
Operative mortalitya	16 (7)	3 (4)	13 (9)	.157
Postoperative complications	143 (63)	44 (55)	99 (67)	.077
Reoperation for cardiac indication	23 (10)	8 (10)	15 (10)	.974
Strokea	10 (4)	3 (4)	7 (5)	.731
Prolonged ventilation	80 (35)	24 (30)	56 (38)	.238
Pneumonia	27 (12)	10 (13)	17 (11)	.822
New need for hemodialysis	20 (9)	5 (6)	15 (10)	.323
Acute limb ischemiaa	7 (3)	4 (5)	3 (2)	.215
Sepsisa	10 (4)	5 (6)	5 (3)	.331
Arrhythmia	78 (34)	28 (35)	50 (34)	.854
Readmission within 30 days	35 (15)	11 (14)	24 (16)	.623
Endocarditis after valve replacementa	17 (7)	12 (15)	5 (3)	.002
Redo valve surgery completed
For any reason	18 (8)	11 (14)	7 (5)	.016
For valve endocarditis	15 (7)	11 (14)	4 (3)	.003
Time between valve operations, d	240 (121, 471)	240 (143, 402)	304 (19, 840)	.002b
Mortality in people who had redo valve surgery	6 (7)	5 (13)	1 (2)	.047
Mortality
Short-term (<6 months)a	31 (36)	10 (26)	21 (44)	.116
Midterm (6 months-5 years)	35 (41)	20 (53)	15 (31)	.003
Extended-term (>5 years)a	20 (23)	8 (21)	12 (25)	.799
Overall mortality	86 (38)	38 (48)	48 (32)	.025
Causes of death
Cardiovascular	28 (33)	12 (32)	16 (33)	.400
Recurrent endocarditisa	16 (19)	11 (29)	5 (10)	.029
Cancera	8 (9)	2 (5)	6 (13)	.716
Infection (not involving heart)a	4 (5)	1 (3)	3 (6)	>.99
Overdosea	4 (5)	4 (11)	0	.014
Renal failurea	4 (5)	0	4 (8)	.300
Liver failurea	3 (3)	3 (8)	0	.042
Other/unknowna,c	18 (21)	11 (29)	7 (15)	.021

^aIndicates use of the Fisher test

^bIndicates use of the t test

^cCauses of death in 2 people were unknown, and 16 died of other causes, including diabetes, diverticulitis, endocrine, trauma, and other vascular.

SUD-IE, substance use disorder–related infective endocarditis.

Substance use disorder was the strongest predictor of mortality (adjusted hazard ratio [HR], 2.41; 95% confidence interval [CI], 1.38-4.20; Table 3). Older age (adjusted HR, 1.004; 95% CI, 1.003-1.04) and New York Heart Association class III/IV (adjusted HR, 1.98; 95% CI, 1.12-3.49) were also associated with increased risk of mortality. The Kaplan-Meier survival probability curve for SUD-IE patients was significantly less than for non–SUD-IE patients (P = .003, Figure 1).

Table 3.

Crude and Adjusted Hazard Ratios for All-Cause Mortality

	Crude Model			Adjusted Model
Characteristic	Hazard Ratio	95% CI	P Value	Hazard Ratio	95% CI	P Value
Substance use disorder	1.59	1.04-2.43	.034	2.41	1.38-4.20	.002
Age at surgery	1.01	0.996-1.02	.149	1.022	1.004-1.04	.016
Staphylococcus aureus	1.79	1.17-2.75	.008	1.58	0.998-2.51	.051
Female	1.09	0.70-1.70	.706	0.98	0.61-1.56	.920
Surgery status
Elective	Reference
Emergent/urgent	1.29	0.77-2.18	.330	1.03	0.60-1.74	.925
NYHA class
I and II	Reference
III and IV	2.16	1.26-3.72	.005	1.98	1.12-3.49	.018
Race
White	Reference			Reference
Nonwhite	1.42	0.80-2.52	.228	1.46	0.81-2.65	.211
Immunocompromised
No	Reference			Reference
Yes	1.46	0.77-2.74	.245	1.46	0.76-2.81	.253

CI, confidence interval; NYHA, New York Heart Association.

Figure 1.

Survival after valve replacement in people with substance use disorder (SUD)-related infective endocarditis vs people with non-SUD-related infective endocarditis.

Comment

Cardiothoracic surgeons have long respected infective endocarditis as a serious condition, and substance use disorder adds another layer to the complexity of managing postoperative valvular heart disease. In our study, almost half of patients with SUD-IE died after valve repair, and more than one-third of people with SUD-IE died from drug overdose or endocarditis. Despite the average 20-year younger age of people with SUD-IE, they had a 240% increased risk of mortality compared with people with non–SUD-IE.

The reasons why 6 months to 5 years postvalve surgery represents a distinct period of increased disparities in mortality is likely multifactorial and certainly deserves further investigation. Posthospitalization care that occurs in the short-term period after discharge often comes with medical care, treatment of pain, and linkage to stable housing at residential nursing facilities, which may explain the absence of differences in short-term outcomes compared with non–SUD-IE. In the midterm period, people with SUD-IE may return to unstable housing conditions, limited access to transportation, and other barriers preventing consistent engagement in addiction-related services—synergizing into increased relapse of underlying substance use disorder, recurrent endocarditis, and death. Homelessness and poverty are inextricably related to poor outcomes related to overdose mortality,¹⁵ and housing instability has emerged as a major stressor for people with SUD-IE.^16,17 One potential reason no statistical difference was found in extended-term mortality in the 2 groups is because people with SUD-IE who have remained stable in recovery for 5 years after valve repair have secured stable housing and engaged in sustainable treatment, with decreased risk of relapse or death. In addition, extended-term mortality rates in non–SUD-IE may be relatively elevated because they are an average of 20 years older than people with SUD-IE.

When to offer a valve replacement is a topic of debate, especially if the patient is someone who has substance use disorder and prosthetic-valve endocarditis.^18,19 High-rates of mortality after repeat valve replacement are concerning, and with limited time and resources as well as increased operative mortality with subsequent valve surgeries, some surgeons argue for “a futility threshold in an effort to balance our obligations to individual patients versus society.”^20-22

From another perspective, clinicians argue that substance use disorder is a chronic illness with potential for relapse, similar to people with uncontrolled diabetes,²³ and that policies prohibiting repeat valve surgery do not take into consideration length of previous sobriety or conditions that may have led to recurrent use of drugs. Several centers, including our own, have started using multidisciplinary meetings with surgeons, infectious diseases clinicians, addiction specialists, case workers, and nurses to discuss and decide case-by-case about medical and surgical options for patients.^16,24 This provides an opportunity for surgeons to express concerns about the technical difficulty and increased operative risk from redo surgeries and allows addiction specialists to discuss the patient’s insight into their illness and the probability of the patient engaging in addiction-related care after hospitalization.

A systematic way to improve outcomes and reduce the need for reoperation is addressing addiction during the hospitalization with the assistance of addiction psychiatry or addiction medicine specialists.^25,26 There is increasing evidence that patients hospitalized for SUD-IE are interested in engaging with addiction specialists to initiate treatment in the hospital,²⁶ and patients will follow-up after hospitalization for continued treatment.^27,28 Addressing substance use disorder in the hospital reduces the frequency of patients with substance use disorder leaving against medical advice, postdischarge addiction severity, readmission rates, and mortality.^29-31

Despite increasing evidence that it decreases readmission and mortality, implementation of addiction care in people hospitalized with severe infections has lagged.^32-34 One study performed in Boston during similar years to our cohort found that 24% of people with SUD-IE had an addiction consultation during hospitalization.³⁴ The suboptimal outcomes of people with SUD-IE in our study are likely at least partly due to limited opioid use disorder management during and after hospitalization, and with introduction of an addiction consult service in August of 2018, our goal is to improve outcomes.

Limitations

Our study has several notable limitations. First, any classification of substance use disorder is heavily prone to misclassification, particularly when drug use needs to be captured in the medical record and disclosed by the patient or discovered by the clinician. There are likely cases of endocarditis related to substance use disorder that are misclassified as non-SUD-IE.

Our study design may have missed deaths or redo surgeries if they occurred outside of the institution or were not available in the vital statistics database.

The scope of the original database did not include details about history of addiction, such as history of or concurrent medication for opioid use disorder, or any documentation of attempts for posthospitalization linkage into addiction care. We also do not have any data about linkage and participation of patients with addiction-related care after hospitalization. Given that this study was performed during period of time and in the same city as previous work showing low addiction medicine consultation rates,³⁴ it is likely our hospital had similarly suboptimal addiction management in the hospital. With the addition of a formal addiction psychiatry consult service in 2018, we look forward to a second round of analysis examining the impact on mortality of addiction management during and after hospitalization.

Lastly, because approximately two-thirds of people died while not hospitalized, we could not confirm cause of death. Notably, when people died in the hospital and we were able to compare Automatic Classification of Medical Entry cause of death to review of medical record, only 3 patients (10%) had incorrect causes of death, a reminder that vital statistics death classifications are less reliable sources of data than medical records data. Future prospective studies can overcome this hurdle by securing patient consent for request and review of outside hospital records to confirm cause of death.

Conclusion

Our findings indicate that people with SUD-IE have similar short- and extended-term mortality but increased midterm mortality. Because recurrent endocarditis was a common cause of death in people with SUD-IE, the factors leading to recurrent drug use, especially in the midterm period, should be investigated further to improve outcomes for people with SUD-IE.