Author + information
- Received August 6, 2017
- Revision received September 28, 2017
- Accepted October 2, 2017
- Published online November 27, 2017.
- Carmen Olmos, MD, PhDa,∗ (, )
- Isidre Vilacosta, MD, PhDa,
- Cristina Fernández-Pérez, MD, PhDb,c,
- José L. Bernal, MDc,d,
- Carlos Ferrera, MD, PhDa,
- Daniel García-Arribas, MDa,
- Carlos N. Pérez-García, MDa,
- J. Alberto San Román, MD, PhDe,
- Luis Maroto, MD, PhDa,
- Carlos Macaya, MD, PhDa and
- Francisco J. Elola, MD, PhDc
- aCardiovascular Institute, Hospital Clínico San Carlos, Madrid, Spain
- bDepartment of Preventive Medicine, Instituto de Investigación Sanitaria San Carlos, Universidad Complutense de Madrid, Madrid, Spain
- cFundación Instituto para la Mejora de la Asistencia Sanitaria, Madrid, Spain
- dManagement Control Department, Hospital Universitario 12 de Octubre, Madrid, Spain
- eCardiology Department, Instituto de Ciencias del Corazón, Centro de Investigación Biomédica en Red de Enfermedades Cardiovasculares (CIBERCV), Valladolid, Spain
- ↵∗Address for correspondence:
Dr. Carmen Olmos, Instituto Cardiovascular, Hospital Clínico San Carlos, Prof. Martín Lagos s/n, 28040 Madrid, Spain.
Background Little information exists regarding population-based epidemiological changes in infective endocarditis (IE) in Europe.
Objectives This study sought to analyze temporal trends in IE in Spain from 2003 to 2014.
Methods This retrospective, population-based, temporal trend study analyzed the incidence, epidemiological and clinical characteristics, and outcome of all patients discharged from hospitals included in the Spanish National Health System with a diagnosis of IE, from January 2003 to December 2014.
Results Overall, 16,867 episodes of IE were identified during the study period, 66.3% in men. The rate of IE significantly increased, from 2.72 in 2003 to 3.49 per 100,000 person-years in 2014, and this rise was higher among older adults. The most frequent microorganisms were staphylococci (28.7%), followed by streptococci (20.4%) and enterococci (13.1%). Twenty-three percent of patients underwent cardiac surgery. The in-hospital mortality rate was 20.4%. Throughout the study period, the proportion of patients with previously known heart valve disease and diabetes mellitus significantly increased, whereas the prevalence of intravenous drug use decreased. Regarding microorganisms, Staphylococcus aureus and streptococci slightly declined, whereas coagulase-negative staphylococci and enterococci consistently increased over the years. In-hospital complications and cardiac surgery rates significantly increased across the years. The risk-adjusted in-hospital mortality rate diminished (0.2% per year) during the study period.
Conclusions The incidence of IE episodes significantly increased over the decade of the study period, particularly among older adults. Relevant changes in clinical and microbiological profile included older patients with more comorbidity and a rise in enterococci and coagulase-negative staphylococcal infections. Adjusted mortality rates slightly declined over the study period.
Most changes in the epidemiological profile of infective endocarditis (IE) described in recent years, including a rise in older adult patients with degenerative valve disease, prosthetic valves, and intracardiac device infections, a decline of patients with rheumatic heart disease, and an increasing incidence of staphylococcal and enterococcal infections, came from studies performed in tertiary care referral centers, with a high percentage of transferred patients, cardiac surgical facilities, and a high volume of cases (1–7). However, this information may be influenced by selection and referral bias (8–10). Population-based studies are more reliable in capturing the complete picture of the disease and provide more representative information of IE epidemiology. Regrettably, well-planned epidemiological studies are scarce, and most countries are not represented (10–12). In addition, the few studies available showed contradictory results regarding IE incidence changes over the years, as well as other relevant clinical and microbiological information (10,11,13–18).
The aims of our study were to provide population-based epidemiological data of IE in Spain from 2003 to 2014 and to evaluate temporal trends in IE incidence, clinical and microbiological characteristics, and outcome by using the Spanish National Health Service (SNHS) database.
Study design and data source
We performed a retrospective temporal trend study, using data obtained from national in-patient hospital activity from January 2003 to December 2014. All patients admitted to hospitals included in the SNHS database have anonymous standard data recorded, corresponding to the Spanish Minimum Basic Dataset, which includes both demographic and clinical information. Statistics comprise age, sex, length of stay, type of hospital information, primary discharge diagnosis, up to 13 secondary diagnoses, and 20 procedures performed during hospitalization, all coded according to the International Classification of Diseases-9th Revision-Clinical Modification (ICD-9-CM) coding system (19).
We identified all episodes with a primary or secondary ICD-9-CM discharge diagnosis of IE (421.0, 421.1, 421.9, and 424.99). All episodes of IE for a single patient during the study period were included in the study.
By searching individual patient-level data (date of birth, sex, and dates of discharge and admission), we were able to identify those patients admitted to 1 hospital and then transferred to another center for further treatment (8.76% of the total study population). To avoid overcounting, these patients were considered to have a single IE episode with a single hospitalization and were paired with the final institution.
Clinical information, previous comorbidities (renal insufficiency or diabetes, history of intravenous drug use, previously known heart valve disease, previous prosthetic valve replacement), and evolution during hospitalization (heart failure, embolic events, septic shock, renal failure, cardiac surgery) were extracted from the ICD-9-CM codes reported for each episode. The microorganisms involved were obtained by searching for the diagnostic codes 0.41.x. Records of in-hospital death and 30-day readmission for any cause were also obtained.
The incidence rates of IE are expressed as the number of episodes per 100,000 person-years, assuming that the entire population was at risk. The denominators (total population and age- and sex-specific groups) were obtained from the nationwide census.
Continuous variables are presented as mean ± SD or median and interquartile range, and categorical variables are expressed as frequencies and percentages.
To assess differences in clinical, epidemiological, and microbiological variables across time, a univariate logistic regression analysis was performed considering 2003 as the reference year. The presence of temporal trends across the time period with regard to the incidence of IE and risk-adjusted mortality was evaluated by means of a Poisson regression model. Incidence risk ratios (IRRs) and 95% confidence intervals (CIs) are represented. In addition, direct standardization was performed to account for changes in age and sex in the Spanish population throughout the study period.
To analyze the influence of different variables in in-hospital mortality, a multivariable logistic model was created. Variables with p values <0.10 and considered clinically relevant were included in a multivariable logistic regression analysis. The final model was built by means of stepwise forward selection and backward elimination technique. The significance levels for selection and elimination were p < 0.05 and p ≥ 0.10, respectively.
To evaluate the potential impact of underlying differences among health care facilities on patients’ outcomes, a second model (multilevel model) that incorporated a random hospital-specific intercept was included along with the selected patient-level epidemiological and clinical characteristics of the logistic regression model. By using this method, it was possible to obtain risk-adjusted mortality rates (ratio of the predicted number of deaths to the expected number of deaths, multiplied by the unadjusted mortality rate). The predicted mortality rate was calculated by applying the estimated parameters of the regression model for in-hospital mortality to each patient according to the patient’s characteristics (presence of risk factors) and adding the hospital-specific intercept. The estimated hospital-specific intercept was added to the sum of the estimated regression coefficients multiplied by the patient’s characteristics. The results were then transformed and summed over all patients attributed to a hospital to obtain a predicted value. The expected mortality rate was obtained in the same way, but instead of the hospital-specific intercept, a common intercept using all hospitals in the sample was added. The results were then transformed and summed over all patients in the hospital to obtain an expected value (20). Discrimination of both models (logistic and multilevel) was assessed with receiver-operating characteristics (ROC) curves. The multilevel model was used to obtain risk-adjusted mortality to compare outcomes among different hospitals and across years.
To analyze the relationship between risk-adjusted mortality and hospital volume (number of patients with IE treated per year), a univariate k-mean clustering method was used, thus achieving the maximum intragroup density and the maximum intergroup disparity. The cutoff point obtained was 13 episodes per center. This analysis was performed only for episodes diagnosed in 2014, to avoid potential errors because there is a temporal trend to reduce risk-adjusted mortality. The association between volume and adjusted-risk mortality was tested with the Pearson correlation coefficient and a linear regression model.
All tests were 2-sided, and differences were considered statistically significant at p values <0.05. Statistical analysis was performed with Stata software version 13.0 (Stata Corp., College Station, Texas).
The present study was exempt from additional review by the ethical investigation committee because all data were de-identified.
A total of 16,867 episodes were identified during the study period (2003 to 2014). The mean age was 63.8 ± 17.5 years, and 11,184 (66.3%) patients were male. Demographic, clinical, and microbiological characteristics are represented in Table 1, and in-hospital evolution is shown in Table 2.
Regarding comorbidities, 16.5% of patients were diabetic, 18% had valvular prostheses, 8.9% had rheumatic valve disease, and 37.4% had nonrheumatic valvulopathy. The most frequent group of involved microorganisms were staphylococci (28.7%; 17.1% Staphylococcus aureus and 12.2% coagulase-negative staphylococci), followed by streptococci (20.4%) and enterococci (13.1%).
With regard to in-hospital evolution and complications, 5.5% of patients had septic shock, 16.2% had an embolic phenomenon, 29.2% had acute renal insufficiency, and 25.3% had heart failure. In the total cohort, cardiac surgery was performed in 23% of episodes, although this percentage was higher in patients treated in high-volume referral centers (35.5%). The median length of hospital stay was 26 days, and 20.4% of patients died during hospitalization.
Temporal trends over the study period
The number of IE episodes significantly increased over the study period, at 2% per year (IRR: 1.02; CI 95%: 1.02 to 1.03), from 2.72 in 2003 to 3.49 per 100,000 person-years in 2014 (Central Illustration), particularly among older adults (Table 1, Online Table 1). In fact, IE incidence in the study population in young patients (≤60 years) decreased over the years. Although episodes during the study period were more than twice as frequent in male than in female patients (IRR: 2.37; 95% CI: 2.30 to 2.45), no differences between the sexes were found as time progressed.
After adjusting for age and sex, direct standardized incidence of IE also increased from 2.90 cases per 100,000 person-years in 2003 to 3.34 cases per 100,000 person-years in 2014, (IRR: 1.02; CI: 1.00 to 1.05; p = 0.042). Over time, the proportion of patients with cardiac implantable electronic devices, previously known heart valve disease, and prosthetic valves significantly increased, as well as the proportion of patients with diabetes mellitus, whereas episodes of IE in intravenous drug users decreased. No significant changes were found in patients with rheumatic valve disease (Table 1). Main in-hospital complications (embolisms, acute renal failure, septic shock, heart failure) increased over the years, as well as the proportion of patients who underwent cardiac surgery during the IE episode (Table 2), whereas the in-hospital mortality rate adjusted by risk decreased linearly by 0.2% per year (IRR: 0.998; CI 95%: 0.997 to 0.999; p = 0.043).
Regarding microorganisms, we found a slight decline in the incidence of infections with S. aureus and streptococci in the total cohort, whereas the incidence of enterococcal and coagulase-negative staphylococcal infection consistently increased over the years (Table 1, Figure 1).
In patients without previously known heart valve disease, the incidence of S. aureus infections also decreased over the years, whereas the incidence of streptococcal infections in patients with previously known native valve disease remained stable (Online Figure 1).
Prognostic factors associated with in-hospital mortality
Variables found to be independently associated with in-hospital mortality in the multivariable analysis were as follows: increasing age; female sex; stroke; peripheral embolism; acute renal insufficiency; heart failure; S. aureus infection; and septic shock (Table 3). Combining these variables in a multivariable logistic model, we obtained an area under the ROC curve of 0.764 (95% CI: 0.755 to 0.773). When we included intercenter variability in a multilevel model (center odds ratio: 1.46), the area under the ROC curve significantly increased to 0.786 (95% CI: 0.777 to 0.794).
Relationship between hospital volume and in-hospital mortality
Hospitals with higher volume (≥13 episodes/year) had higher crude mortality rates and higher variability among centers, compared with low-volume hospitals. Once mortality was adjusted by risk, the difference was reduced (Table 4, Figure 2). When analyzing exclusively centers with cardiac surgery, a significant negative correlation between volume and risk-adjusted mortality (r = −0.279; p < 0.001) was found.
In the present population-based study, we analyzed the epidemiological changes in IE in Spain throughout a period of 12 years (2003 to 2014). We found an increase in the incidence of IE over the study period, and this increase was accompanied by changes in the clinical and microbiological profile of IE episodes. Regarding predisposing factors, the proportion of intravenous drug users significantly decreased, whereas the prevalence of diabetes among patients with IE episodes increased. Concerning the microbiological profile, staphylococci are still the most frequent microorganisms, as reported in other series (1,7,10,17), but in the last years of the study the proportion of S. aureus progressively declined, and the incidence of infections with enterococci and coagulase-negative staphylococci regularly and significantly increased. Finally, the proportion of patients who underwent surgery progressively increased, and risk-adjusted in-hospital mortality rate slightly declined in during the study period.
Several population-based studies have also reported an increase on IE incidence rates over time. A study in Denmark evaluated 5,486 incident cases along 18 years (1994 to 2011) and found a substantial rise in IE incidence from 3.93 to 7.55 per 100,000 person-years, an increase that was more evident among male and old patients (16).
Other studies from European countries have also confirmed this trend (13,15). An Italian study included 1,873 patients from the Veneto region (2000 to 2008) and reported an increase from 4.1 to 4.9 cases per 100,000 person-years (15). In the same way, a study from the United Kingdom that aimed to investigate the effect of National Institute for Health and Care Excellence 2008 recommendations regarding IE prophylaxis on the incidence of IE in England analyzed 19,804 patients (2000 to 2013) and found a significant increase of 0.11 cases per 10 million persons per month after 2008 (13).
Conversely, some community-based survey studies did not find significant changes in the incidence of IE. The Rochester Epidemiology Project has evaluated the profile of patients with IE from Olmsted County, Minnesota during different time periods (18,21,22). The most recent study, published in 2015 (18), found an increase of IE incidence rates in older patients, although the overall incidence did not change. Survey-based studies from several regions of France (17,23,24) did not identify changes in IE incidence rates over 3 periods of time (1991, 1999, 2008). However, both types of studies were limited by the small numbers of patients included in the surveys, restricted to particular regions and, in the case of French studies, precise cross-sectional periods of time. Finally, 2 population-based studies in United States did not detect any increase in the incidence of IE (14,25). In fact, in 1 study a decline in hospitalization rates for endocarditis from 2006 to 2010 was documented (14).
Most population-based studies, despite including a large number of episodes, fail in having relevant clinical and microbiological information (11,13,15,16) that could explain changes in incidence and mortality rates. In the present study, we were able to extract significant clinical information. Several predisposing comorbidities were analyzed, and microbiological data were reported in 70% of IE episodes.
Changes in the microbiological profile of IE episodes agree with those reported in survey-based studies (17,18) and tertiary referral center studies (1,5,7). In our study, staphylococci were the most prevalent microorganisms, and S. aureus continues to be the most frequent cause of IE. However, we found a significant increase in the incidence of coagulase-negative infections, as well as enterococcal IE. The rise of these 2 groups of microorganisms was previously described in smaller works (4,10,17), and it is probably related to an increase in the prevalence of health care–related interventional procedures (intravenous lines placement, catheterization, hemodialysis) in recent years.
Finally, we found a small decrease in the incidence of streptococcal infections, although we are not able to analyze whether there was any change in the incidence of oral viridans streptococci because this group of microorganisms does not have a specific ICD code. For this reason, in our study we included codes 041.00, 041.01, 041.02, 041.03, 041.05, and 041.09, which comprised infections caused by group A, B, C, and G streptococci, as well as those streptococcal infections codified as “other” or “unspecified.” However, because of the well-known microbiological profile of IE, it is probable that most microorganisms included under these aforementioned streptococcal codes were in fact viridans group streptococci.
Regarding comorbidities, as reported in previous studies (7,10,17), we found an increase in the proportion of patients with diabetes mellitus, along with a higher the prevalence of cardiac implantable electronic devices and cardiac prostheses. This rise in comorbidities and predisposing conditions is probably related to population aging and explains why the group of older adult patients with IE, particularly those more than 80 years of age, gradually increased over the years.
In our study, 23% of patients underwent cardiac surgery during hospitalization. This percentage is similar to those reported in population-based studies from the United States and Italy (15,21,22,25), as well as some tertiary referral centers (26), but it is significantly lower than percentages described in the majority of studies from tertiary care centers (1,7,17,24,27–30). This disparity is most likely the result of referral bias because patients transferred to referral hospitals are probably sicker and have more complications during the course of the disease that require urgent surgery.
In this sense, 2 prospective cohort studies that analyzed differences comparing patients with IE who were or were not transferred to tertiary care centers found that transferred patients tended to have a more complicated course and underwent surgery much more frequently than patients who were not transferred (9,31). In accord with these 2 studies, in the present study those patients treated in high-volume referral centers underwent surgery more frequently than did patients treated in smaller hospitals (10.1% vs. 35.5%).
The same reasoning applies to explain why in-hospital mortality rates and lengths of hospital stay are higher in high-volume centers. Patients in low-volume hospitals who are not transferred frequently have an uncomplicated course, in particular without the need for cardiac surgery, and have low in-hospital mortality rates (1,31). Conversely, in referral hospitals, patients are sicker and undergo surgery more frequently. The reason that adjusted mortality rates were higher in these hospitals probably reflects an imperfect adjustment (i.e., surgery and periannular complications were not considered in the model). However, it was found that, among hospitals with cardiac surgery, adjusted mortality rates were lower in those centers with the highest volume of IE episodes per year.
Finally, in-hospital mortality rates in our study (20%) were in accord with those reported in updated registries. Moreover, we found a slight but significant decrease in in-hospital mortality rates throughout the study period. The most recent French study reported an in-hospital mortality rate of 21% (17), and the already mentioned population-based study from the Veneto region had a rate of 18.5% (15).
Regarding U.S. studies, the most recent report from Olmsted County, Minnesota had a 6-month mortality rate of 29% (18). The Medicare registry found a strikingly low in-hospital mortality rate (9% to 11%), although the 6-month mortality rate was quite high (28% to 31%). This discrepancy between in-hospital and 6-month mortality rates was probably the result of short lengths of stay in diagnosing hospitals and a high incidence of patients’ discharge to intermediate care facilities and home health care (14).
First, because of the design and data source used, the accuracy of the data relies on adequate hospital coding, and this also affects IE diagnosis. As with any population-based study, it is prone to underestimation of the true number of cases and disease misclassification. However, given the specific features and relevance of this disease, incorrect coding of patients is less likely than in other diseases. Nevertheless, certainly some episodes may have been missed, and others may have been incorrectly reported as IE. In addition, we could not differentiate between definite and possible IE episodes. Conversely, because of the population-based, nationwide design and the characteristics of the SNHS, which provides access to health care for almost the entire population of Spain, this study provides information that it is not affected by referral or selection bias.
Second, we lack relevant information such as the type of acquisition (nosocomial, community acquired) and diagnostic and therapeutic procedures, and we were not able to evaluate previous hospitalizations that could have provided information regarding interventional procedures or hospital admissions.
In addition, microbiological information was missed in 30% of cases. We cannot be sure whether these episodes were culture-negative IE or episodes in which the causative agent was simply not recorded. Finally, we do not have data from private hospitals, which represent a small percentage of centers in Spain. Thus, incidence rates of IE have been probably underestimated.
The incidence of IE in Spain increased during the study period, particularly among older adults. The proportion of patients who underwent surgery increased, whereas the risk-adjusted in-hospital mortality rate slightly decreased. In addition, the clinical and microbiological characteristics of IE changed over the years, including the presence of older patients with more comorbidities and a rise in enterococcal and coagulase-negative staphylococcal infections.
COMPETENCY IN MEDICAL KNOWLEDGE: The incidence of IE rose in Spain from 2003 to 2014, particularly among older patients with comorbidities. Staphylococci remain the most frequent microorganisms, but in the last years S.aureus progressively declined, whereas enterococal and coagulase-negative species are more commonly encountered pathogens. The proportion of patients undergoing surgery increased, whereas the risk-adjusted mortality rate decreased slightly.
TRANSLATIONAL OUTLOOK: More effective strategies, focused on prevention of health care-related infections, are needed to reduce the morbidity and mortality associated with IE.
The authors thank the Spanish Ministry of Health, Social Services, and Equality for providing the Spanish Minimum Basic Dataset, with special gratitude to the General Directorate of Public Health, Quality, and Innovation.
The present study was partly financed by an unconditional grant from the Fundación Interhospitalaria para la Investigación Cardiovascular (Interhospital Foundation for Cardiovascular Research), Madrid, Spain. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- International Classification of Diseases-9th Revision-Clinical Modification
- infective endocarditis
- incidence risk ratio
- receiver-operating characteristic
- Spanish National Health Service
- Received August 6, 2017.
- Revision received September 28, 2017.
- Accepted October 2, 2017.
- 2017 American College of Cardiology Foundation
- López J.,
- Revilla A.,
- Vilacosta I.,
- et al.
- Sevilla T.,
- López J.,
- Gómez I.,
- et al.
- Slipczuk L.,
- Codolosa J.N.,
- Davila C.D.,
- et al.
- Bikdeli B.,
- Wang Y.,
- Kim N.,
- Desai M.M.,
- Quagliarello V.,
- Krumholz H.M.
- Duval X.,
- Delahaye F.,
- Alla F.,
- et al.
- DeSimone D.C.,
- Tleyjeh I.M.,
- Correa de Sa D.D.,
- et al.
- ↵National Center for Health Statistics. Centers for Disease Control and Prevention. ICD-9-CM Guidelines, Conversion Table, and Addenda. Classification of Diseases, Functioning, and Disability. Available at: https://www.cdc.gov/nchs/icd/icd9cm_addenda_guidelines.htm. Accessed April 14, 2017.
- ↵Suter LG, Wang C, Araas M, et al. Yale New Haven Health Services Corporation/Center for Outcomes Research & Evaluation (YNHHSC/CORE) Hospital-Level 30-Day All-Cause Mortality Following Coronary Artery Bypass Graft Surgery Updated Measure Methodology Report Centers for Medicare & Medicaid Services (CMS). Available at: https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Downloads/Coronary-Artery-Bypass-Graft-CABG-Mortality.zip. Accessed March 9, 2017.
- Toyoda N.,
- Chikwe J.,
- Itagaki S.,
- Gelijns A.,
- Adams D.H.,
- Egorova N.N.
- Tleyjeh I.M.,
- Ghomrawi H.M.,
- Steckelberg J.M.,
- et al.
- Olmos C.,
- Vilacosta I.,
- Habib G.,
- et al.
- Tornos P.,
- Iung B.,
- Permanyer-Miralda G.,
- et al.
- Habib G.,
- Lancellotti P.,
- Antunes M.J.,
- et al.
- Fernández-Hidalgo N.,
- Almirante B.,
- Tornos P.,
- et al.