Introduction

Infective endocarditis (IE) is a life-threatening multisystem disease that is commonly defined as a bacterial infection of native or prosthetic heart valves, the endocardial surface, or a cardiac device[1]. IE has been associated with prolonged hospitalization, significant morbidity, and a high mortality [1,2,3,4]. There is a large variation in the estimated crude incidence of IE, reported to be between 1.5 and 11.6 cases per 100,000 people from 1990 to 2010 [5]. However, there is a relative lack of data from low- and middle-income regions which can significantly affect the overall incidence of IE. A nationwide study from the United States (US) reported nearly half a million IE-related hospitalizations from 2000 to 2011 and a steady increase in its incidence [6,7,8,9,10,11]. The short-term mortality related to infective endocarditis has been estimated to be 10–24% [11]. Since there is significant variation among different regions with regards to epidemiology of IE, hence it is important to report local data on patient characteristics and outcomes.

Trends in populations at risk, the responsible pathogens, and their antimicrobial resistance patterns have been dynamic over the past few decades [12]. Previously, IE was primarily associated with rheumatic heart disease (RHD) and congenital heart disease (CHD) [13]. However, in the twenty-first century, increasing age, intracardiac devices, prosthetic heart valves, hemodialysis access, central venous catheters, immunosuppressive therapy, and intravenous drug use (IVDU) are some of the strongest risk factors [14,15,16,17,18,19,20,21,22]. There has also been a shift in the most common pathogens causing IE. While Streptococcus species were the most common causative agents in the previous century, there is strong evidence that Staphylococcus aureus (S. aureus) has prevailed since then [6, 12, 22,23,24]. This shift carries significant consequences, as S. aureus IE has been associated with a higher likelihood of requiring surgical treatment and overall worse outcomes [24,25,26,27]. The emergence of resistant pathogens, especially methicillin-resistant Staphylococcus aureus (MRSA), pose additional therapeutic challenges with data suggesting worse outcomes [2, 4, 28, 29]. The Bronx is one of the most diverse areas in the United States according to Census Bureau. Nearly a third of its residents immigrated from low- or middle-income countries and thus, potentially carry traditional risk factors of IE such as RHD. In addition, the Bronx has been one of the most socio-economically disadvantaged regions in the US and ranks last among all 62 counties of New York state in health outcomes, quality of life and important health and socioeconomic factors[30]. These characteristics may be associated with higher burden of endemic RHD [31]. The Bronx has also been suffering greatly from the opioid epidemic[32, 33], allowing IVDU to play a central role as a risk factor in the local IE cases. However, there is paucity of data on the epidemiology, microbiology, and outcomes of patients with IE in the Bronx.

Our primary objective was to retrospectively identify the epidemiological, microbiological, and clinical characteristics of patients who were admitted with infective endocarditis in a public hospital in the Bronx, New York from January 2010 to September 2020. Our secondary objectives were to investigate for differences in these characteristics over time and identify possible predictors of in-hospital death.

Methods

Study design and patient population

This retrospective cohort study was conducted at Jacobi Medical Center, a tertiary care hospital in the Bronx, New York that is part of the public health system of New York City. All adult patients who were hospitalized with a primary diagnosis of new-onset IE between January 1st, 2010 and September 30th, 2020 were included. ICD-9 (421.0, 421.1, 421.9, 112.81, 036.42, 098.84, 115.04, 115.14, 115.94, 424.90, 424.91, 424.99) and ICD-10 diagnostic codes (I33.0, I33.9, I38, I39) were utilized to identify possible eligible patients. We excluded patients who met one or both of the following exclusion criteria: (i) age < 18 years, (ii) the diagnosis of IE was not of new onset. The included patients were followed until discharge, transfer to another institution, or death. The study was approved by the institutional review board (IRB) of Albert Einstein College of Medicine with a waiver of informed consent (IRB number 2020–12072).

Data extraction

Two pairs of researchers (WaL, ArT and WeL, LP) reviewed all electronic medical records (EMR) independently in a pre-defined data extraction sheet. Discrepancies were resolved by reaching consensus via discussion. The documentation of the index admission from emergency medicine providers, inpatient providers, consultants, nurses, therapists, social workers; and laboratory and imaging data were reviewed. Post-discharge notes were also reviewed when available and as needed. Documentation from past visits and the search engine of the EMR were also utilized as needed.

The extracted data included baseline demographic information [age, sex, race/ethnicity, residence status (community-based, skilled nursing facility resident or homeless/shelter resident)], clinical characteristics [body mass index (BMI), history of IVDU, hypertension, diabetes, hyperlipidemia, coronary artery disease (CAD), peripheral artery disease (PAD), heart failure, RHD or CHD, presence of cardiac or non-cardiac prosthetic device or material, recent cardiac operation, heart transplantation, valvular abnormalities, atrial fibrillation, chronic obstructive pulmonary disease (COPD), obstructive sleep apnea (OSA), pulmonary embolism (PE) or deep vein thrombosis (DVT), chronic kidney disease (CKD), end-stage renal disease/ESRD (on hemodialysis), active malignancy, malignancy in remission, colorectal cancer, liver cirrhosis, human immunodeficiency virus infection (HIV) or acquired immunodeficiency syndrome (AIDS), recent hospitalization (within three months from index admission), placement of a central venous catheter (within 3 months from the index admission)], pertinent home medications (immunosuppressive agents), clinical and laboratory data on presentation [fever, new or worse murmur, white cell count (WBC), hemoglobin (Hgb), platelet count (PLT), creatinine, blood urea nitrogen (BUN), aspartate transaminase (AST), alanine transaminase (ALT), international normalized ratio (INR), partial thromboplastin time (PTT), albumin, total bilirubin, C-reactive protein (CRP), complement C3, complement C4, rheumatoid factor (RF), antinuclear antibodies (ANA), syphilis tests, and blood cultures with antibiotic sensitivity], echocardiographic data, need for invasive mechanical ventilation (IMV), development of circulatory shock, need for intensive care unit (ICU) admission, occurrence of acute decompensated heart failure (ADHF), myocardial infarction (MI), or advanced heart block, need for surgical treatment, occurrence of septic embolization, meningitis, intracranial hemorrhage, or acute kidney injury (AKI), length of stay (LOS), and discharge status (death, transfer to another institution, discharge to a rehabilitation center, community, or shelter). IE cases were classified as possible or definite per modified Duke’s criteria. Possible IE cases were reviewed post hoc by two independent attending physicians (LP and MG) to adjudicate that the diagnoses of IE were clinically appropriate.

The data were processed and analyzed without any personal identifiers to maintain patient confidentiality as per Health Insurance Portability and Accountability Act (HIPAA). The STROBE statement guidelines for reported observational studies was followed.

Outcomes and statistical analysis

Patients were classified as two groups based on the time of the index hospitalization: (i) January 1st, 2010 to December 31st, 2015 (cohort 2010–2015) and (ii) January 1st, 2016 to September 30th, 2020 (cohort 2016–2020). The primary endpoint was in-hospital mortality. Secondary endpoints included need for IMV, need for ICU admission, occurrence of ADHF, MI, advanced heart block, or circulatory shock, need for surgical treatment, and occurrence of septic embolization, meningitis, intracranial hemorrhage, or AKI, and LOS). Patients that were transferred to other hospitals for higher level of care were excluded from the mortality analysis. Continuous data are presented as median with interquartile range (IQR) and categorical data as absolute and relative frequencies. The ANOVA test was used to compare the continuous variables, while chi-square was used for discrete variables. A logistic regression model was used to identify baseline variables associated with in-hospital mortality. To build a multivariate model, we used the following method: model 1: all the variables with significant univariate associations (p value ≤ 0.05), and model 2: all variables with univariate associations at the verge of statistical significance (p value ≤ 0.1). Results of logistic regression are given as the odds ratio (OR) with the 95% confidence interval (CI). The threshold of statistical significance was p ≤ 0.05. All analyses were performed using STATA software (version 14·1; STATA Corporation, College Station, TX, USA).

Results

Baseline characteristics, clinical and laboratory findings

399 charts were identified by the utilized diagnostic codes. 24 charts were duplicates and 193 patients did not have new-onset infective endocarditis during index hospitalization. Therefore, 182 patients were included in this analysis [female sex: 70 (38.5%), median age: 54 years (IQR 44–67), non-Hispanic Black: 69 (37.9%)]. 79 patients (43.4%) were considered to have possible IE and 103 (56.6%) definite IE per modified Duke criteria; all 182 patients were diagnosed with and treated for IE. 135 patients (75.8%) were community-based, 8.2% were homeless, and 16% were SNF residents. The median BMI was 24 kg/m2 (IQR 21–28). 46 patients (25.3%) reported IVDU. Hypertension, diabetes mellitus, and hyperlipidemia were the most common comorbidities, prevalent in 52.8%, 32.4% and 26.9% of our patients, respectively. 14.3% of our patients were on hemodialysis and 29.1% had a recent hospitalization for any cause. 10.4% had a history of infective endocarditis and five patients (2.8%) had RHD. Five patients (2.8%) had a cardiac device. 86 patients (47.2%) were included in the cohort 2010–2015 and 96 patients (52.8%) in the cohort 2016–2020. There were significant differences between the two cohorts in the mean age (59.58 vs. 50.47 years, p < 0.001), mean BMI (27.03 vs 23.9, p = 0.001), hypertension (69.8% vs. 87.4%, p < 0.001), CAD (26.7% vs 14.6%, p = 0.042), COPD (10.5% vs. 2.1%, p = 0.018), hemodialysis (19.8% vs. 9.4%, p = 0.045), IVDU (16.3% vs. 33.3%, p = 0.008), active malignancy (9.3% vs. 2.1%, p = 0.033), and recent hospitalization (44.2% vs. 15.6%, p < 0.001). The detailed baseline patient characteristics are presented in Table 1.

Table 1 Baseline patient characteristics
Full size table

Fever was observed in 105 (58%) and new or worse murmur in 39 (21.4%) patients. The median WBC on presentation was 11.6 (IQR 7.7–16.0) 103/μL and median CRP concentration was 116.6 (63.7–135.0) mg/dL. Among patients in whom RF and ANA were measured, they were positive in 13/24 (54.2%) and 10/40 (25%) of the patients, respectively. 153 patients (84.1%) had positive blood cultures and the median number of positive sets was 2 (IQR 2–4). 99 out of the 153 patients with positive blood cultures (64.7%) were found to have one or more pathogen(s) with some resistance. The clinical and laboratory data are presented in Table 2.

Table 2 Clinical and laboratory data
Full size table

Echocardiographic and microbiological findings

Transesophageal echocardiogram (TEE) was performed in 102 patients (55.6%), while only transthoracic echocardiogram (TTE) was performed in 81 patients (44.4%). The mitral valve was involved in 32.4% of patients followed by aortic valve (19.8%) and tricuspid valve (18.1%). The pulmonary valve was involved in only 2.2% of patients. Multivalvular disease was observed in 6.6% of patients and no vegetation was detected in 20.9% of patients. There were significant differences in the cardiac valves involved between the cohort 2010–2015 and 2016–2020 (mitral valve: 41.9% vs. 24%, aortic valve: 19.8% vs. 19.8%, tricuspid valve: 10.5 vs. 25%, pulmonary valve: 1.2% vs. 3.1%, multivalvular disease: 1.2 vs. 11.5%, no vegetation: 25.6% vs. 17.7%; p = 0.001). The data on echocardiographic findings and involved valves are presented in Table 3.

Table 3 Echocardiographic data and valvular involvement
Full size table

153 patients (84.1%) had positive monomicrobial blood cultures, 29 patients (16%) had culture-negative endocarditis and 18 patients (9.9%) had polymicrobial endocarditis. The rate of possible IE per modified Duke criteria in patients with culture-negative IE was remarkably higher compared to the overall rate of possible IE in our cohort (89.7% vs. 43.4%). In contrast, the rate of definite IE in patients with positive blood cultures was significantly higher compared to the overall rate of definite IE in our cohort (65.4% vs. 56.9%). Similarly, the rate of possible IE was somewhat higher in the patient group that underwent only TTE compared to the overall cohort (49.4% vs. 43.4%). These findings might indicate that the higher rate of possible IE in our cohort was mainly driven by the culture-negative endocarditis cases and partially by those IE cases where only TTE was performed. However, it should be emphasized that all possible IE cases were adjudicated as true clinical IE by two attending-level reviewers. Gram-positive bacteria were isolated in 139 of 153 patients (90.9%) with positive monomicrobial blood cultures, followed by gram-negative bacteria (6.5%) and fungi (2.6%). Staphylococci were isolated in 87 of 139 patients (62.6%) with gram-positive bacteremia, followed by Streptococci (26.6%) and Enterococci (8.6%). S. aureus was isolated in 79.3% of cases secondary to staphylococcal infection (69/87) and in the 45.1% (69/153) of all patients with monomicrobial IE. Coagulase-negative Staphylococci (CNS) were isolated in 20.7% of cases secondary to staphylococcal infection (18/87) and in the 11.8% (18/153) of all patients with monomicrobial IE. S. Epidermidis was the most common CNS (66.7% of CNS cases, 7.8% of all monomicrobial IE cases). Streptococcus mitis was the most common isolated Streptococcus species (43.2% of streptococcal cases, 10.5% of all monomicrobial IE cases). Enterococcus faecalis was the most common isolated enterococcus species (66.7% of the Enterococcus cases, 5.2% of all monomicrobial IE cases). Klebsiella pneumoniae was the most common pathogen in gram-negative endocarditis (6 out 10 cases). All fungal endocarditis cases were secondary to candida species. The detailed microbiological data are presented in Table 4.

Table 4 Microorganism data
Full size table

MRSA was the isolated pathogen in about half of the cases secondary to S. aureus (34/69) and 22.2% of all monomicrobial IE cases. Vancomycin-resistant Enterococcus (VRE) was found in one of 12 enterococcal cases. Extended-spectrum b-lactamases (ESBL) strains were detected in 50% of IE secondary to Klebsiella pneumoniae (three out of six).

Outcomes

164 patients (90.1%) were diagnosed with native valve IE, 15 patients (8.2%) with prosthetic valve IE, and 3 patients (1.7%) with both native and prosthetic valves IE. IMV and admission to ICU were required in 59 (32.4%) and 101 (55.8%), respectively. 40 patients (22%) required surgical treatment. Only one patient received oral antibiotics as an inpatient and was discharged to community in a stable condition. The in-hospital mortality was 18.1%. The mortality was significantly higher in the cohort 2010–2015 compared to the cohort 2016–2020 (22.1% vs 14.6%). Two patients (1.1%) were transferred to another institution for higher level of care and 83 (45.6%), 61 (33.5%), and 3 (1.7%) patients were discharged to community, rehabilitation center, and shelter, respectively. The median LOS was 9 (IQR 11–33) days after excluding the two patients, who were transferred to another hospital. The detailed presentation of outcomes is presented in Table 5. We performed an additional analysis classifying our patients in two groups: native vs. prosthetic/device endocarditis. There was no significant difference in in-hospital death rate (18.29% vs. 16.67%, p = 0.865) or in any other outcome between the two groups. This analysis is presented in Supplementary Table 1. The in-hospital mortality was 21.8% in patients who did not require surgery and 2.5% in patients who required surgery (p = 0.009). Patients who decided to stop antibiotics prematurely (18.1% of patients; average antibiotic treatment duration: 13.3 days; inpatient death rate: 63.6%) had significantly higher in-hospital mortality as compared to patients who completed therapy in the hospital (18.7% of patients; average duration of antibiotic treatment: 34.5 days; inpatient death rate: 14.7%), and patients who were discharged to complete antibiotic therapy as outpatient (60.4% of patients; average inpatient antibiotic treatment duration: 35.7 days; inpatient death rate: 5.5%) (p = 0.001).

Table 5 Outcomes
Full size table

Logistic regression analysis – predictors of in-hospital death

The univariate associations with in-hospital mortality were examined for all the baseline demographic and clinical characteristics, as well as selected characteristics related to IE. Increasing age (analyzed per 10 years), SNF residence, atrial fibrillation, COPD, and active malignancy were found to have a significant univariate association with in-hospital death, while IVDU and MRSA IE were at the verge of statistical significance (Table 6).

Table 6 Logistic regression analysis for the outcome of in-hospital mortality
Full size table

In the multivariable analysis, MRSA IE (OR: 3.71; 95% CI 1.31–10.55; p = 0.014 in model 2) was the only variable found to have significant association with in-hospital death, while active malignancy (OR: 3.55; 95% CI 0.88–14.37; p = 0.076 in model 1) and atrial fibrillation (OR: 2.72; 95% CI 0.86–8.65; p = 0.090 in model 2) were at the verge of statistical significance. The logistic regression analysis is presented in Table 6.

Discussion

Our study describes the baseline characteristics, clinical features, microbiological data, and outcomes of patients hospitalized with IE in a public hospital of the city of New York in the past decade. This is the first study evaluating the epidemiology of IE in the Bronx, New York. The main findings can be summarized as following: (1) S. aureus followed by streptococci and enterococci were the most common identified pathogens, (2) MRSA accounted for about half of the S. aureus IE cases, (3) the incidence of IE in patients with IVDU increased over time and the median age decreased, (4) the overall in-hospital mortality was 18.1% and was higher in 2010–2015 compared to 2016–2020, and (5) MRSA IE was found to be associated with a higher likelihood for in-hospital death.

S. aureus was the most common microbe identified in our population accounting for about 45% of cases. Similarly, S. aureus was the most prevalent pathogen in recent cohorts of patients with IE in Australia (45–53%)[34,35,36], Israel (33.3%) [37], Norway (31.4%) [38], Japan (27.2%) [39], Italy (27%) [40], and Taiwan (25.7%) [41]. On the other hand, recent observational studies from low- and middle-income countries demonstrated lower prevalence of S. aureus and revealed that streptococci largely remain the most common causative agents in these regions [21, 42,43,44,45]. The existing literature has repeatedly demonstrated increasing prevalence of staphylococcal IE particularly in the western world [6, 11, 34, 46,47,48,49]. Increased use of central venous catheters, prosthetic valves, and cardiac devices, along with the performance of higher number of invasive procedures than before seem to be responsible for this change in the epidemiology of the disease over time [14, 50]. It should be emphasized that patients with S. aureus IE are more likely to require longer hospitalization, surgical treatment and are associated with higher likelihood for large vegetations, prosthetic valve involvement, abscess, recurrence, readmission, and in-hospital death [44, 51,52,53,54,55,56].

MRSA accounted for about half of the S. aureus cases in our cohort. Although S. aureus has a well-known propensity to develop resistance to methicillin and MRSA strains have been a universally recognized problem [57, 58], such a high prevalence of MRSA IE is deeply concerning. Recent observational studies from four different continents demonstrated that MRSA accounted for 19% to 40.9% of S. aureus IE and bacteremia cases [39, 59,60,61] which is significantly lower compared to our findings. The high rates of IVDU (25.3%), recent hospitalization (29.1%), and SNF residence (16%) in our population are the most likely explanations of higher MRSA IE rates since all these characteristics are known risk factors for MRSA infections [62,63,64,65].

Several differences were observed comparing the early cohort (2010–2015) to the late cohort (2016–2020). Most notably, the rate of patients with IVDU was more than double in 2016–2020 compared to 2010–2015. This remarkable difference likely drove the other observed differences, namely higher median age and prevalence of chronic diseases in 2010–2015 and significantly higher rates of right-sided and multivalvular IE in 2016–2020 [27, 66,67,68]. It is likely that the US opioid epidemic [69, 70], which peaked in the middle of the past decade, is associated with this epidemiological and clinical shift. Similar trends over time have been observed across the country [71] and Sweden [66].

The overall in-hospital mortality was 18.1% in our study, which is in line with the high in-hospital death rates observed in other recent cohorts and indicates that IE remains a life-threatening condition requiring immediate and special attention [4, 39, 45, 72,73,74,75]. A significant decrease in the in-hospital mortality from 22.1% to 14.6% was noted comparing 2010–2015 to 2016–2020, which can be explained at least partially by the higher representation of patients with IVDU in the late cohort, who typically are of younger age and have less comorbidities [71, 76, 77]. MRSA IE was the only predictor of higher likelihood for in-hospital death identified in the multivariate analysis. The latter finding is in accordance with the results of a large systematic review that included 17,563 patients and showed that MRSA IE was associated with a 95% higher likelihood for death compared to MSSA IE [29].

One of the key strengths of the current study is that the study population represents underserved and economically disadvantaged minorities; thus, revealing the epidemiologic characteristics and outcomes of infective endocarditis in this usually underreported and underrepresented population in clinical research. Moreover, we report our findings on the epidemiology and outcomes of infective endocarditis in the Bronx, where there is a paucity of relevant data. Additionally, two pairs of researchers independently and blindly collected data which reduces errors and bias. On the other hand, our study has several limitations. First, this is a single-center study performed in a hospital that does not have significant cardiac surgery capabilities, therefore our findings cannot be easily generalized and the outcomes could have been affected by this limitation. Second, the relatively low rates of performed TEE might have posed diagnostic limitations and affected the in-hospital outcomes. Third, this was a real-world study with a retrospective design utilizing the electronic medical records, which is suboptimal compared to a prospective study that would allow for a more accurate follow-up assessment and would additionally provide information on long-term outcomes.

In conclusion, in this cohort of patients hospitalized with IE in the past decade in a public hospital in the Bronx, New York, we found that S. aureus was the most common causative agent and MRSA accounted for nearly half of the S. aureus IE cases. The incidence of IE in patients with IVDU increased over time, while the median age decreased. The in-hospital death rate was higher in 2010–2015 compared to 2016–2020. MRSA IE was found to be associated with a higher likelihood for in-hospital death. Data from more local institutions are needed to better characterize the changing epidemiology of IE at a regional level and tailor our preventative and therapeutic approaches accordingly.