Over the past decades, the epidemiology of IE has changed due to changes in demographic characteristics and risk factors. Researches from developed countries showed that the incidence of IE had increased in the past decades . Cresti et al  found the incidence was 4.6/100,000 person-years with a significant linear increase between 1998 and 2014. Keller et al  found that the incidence of IE in Germany was 11.6/100,000 person-years between 2005 and 2015 and the incidence increased continuously during the study period, especially in the last five years. However, in our study, we found that the incidence of IE was between 0.33 and 0.72 patients per 1000 admissions and the incidence remained stable during the period, which was lower in comparison with the developed countries (approximately 1-1.3 patients per 1000 hospital admissions) [4, 7] .The difference in incidence of IE from different studies may be related to the geographical location of the study, the time period selected, and the difference in diagnostic techniques of different institutions.
In our study, we found that the mean age at the time of IE episodes were younger than those reported from developed countries, although the mean age increased during this period. The proportion of old patients with IE increased gradually in our study period, and this shown the same trend as the developed countries [7, 9].Erichsen et al  found that the incidence increased substantially for elderly patients between 1994 and 2011, with the highest incidence rate of 3.38 for patients more than 80 years old at IE onset. In Oliver et al’ s report , 49% of IE patients were over 65 years old and 11.2% were over 80 years old. The changes in the age of IE onset may owing to the aging population. And as a result of the older onset age, patients with IE in the latter part of the study period were more likely to have comorbidities compared with patients in the earlier part of the study period.
We found that CHD had become the most common underlying heart disease for patients with IE in our study. RHD was once the most common underlying heart disease in the 1990s according to Chao et al’ s report . But the proportion of RHD gradually decreased in our study. The results from other medical institutions in China also found the same change on underlying heart disease in patients with IE. Previous reports from developed countries also presented that the proportion of RHD had dropped obviously, and DVD, PVE and implantable electronic devices related IE had gradually increased and replaced RHD as the leading heart disease [1, 4, 8] .This difference may contributed to the large number of patients with CHD and the relative low proportion of patients with prior cardiac surgery. According to the published reports, the risk of developing into IE increased in patients with CHD [14–16]. Darren et al  found that the IE risk exceeded 100 times in patients with ACHD compared to that of the general population and 2.5 times that of children with CHD [16, 17].The proportion of patients with CHD in our study was as high as 25.8%.
TEE was performed in only 13% patients with IE in our study, as most patients diagnosed after TTE did not perform TEE routinely. Based on a review of the literature published from China on IE, the rate was similarly but it approached to other studies in our region [13, 18, 19]. But this rate was significantly lower than that reported by other studies from developed countries with 74%-100%[20, 21].However, according to the 2015 European Association of Cardiology guidelines on IE, for TTE positive patients, TEE should be performed to exclude perivalvular complications. The low utilization rate of TEE could directly resulted in the low detection rate of perivalvular complications such as valvular perforation and perivalvular abscess in our study. They were very important data for the cardiac surgeon. Besides, this could cause the missed diagnoses of IE in patients with unobvious valvular lesions or with basic valvular lesions. This underlined that our research institution still needs to strengthen the learning of the value of TEE in IE diagnosis, strengthen the study of the latest IE guidelines and improve the use of TEE in the diagnoses of IE to reduce missed diagnoses.
The positive rate of blood culture in our study was lower compared with other studies from developed countries, where the rate of a positive blood culture varied from 83–96%[2, 8, 11, 22] .However, the positive culture rate was approximate to other studies from China varied from 38.5–70.1%[13, 18, 23].This finding could be related to the use of antibiotics before blood culture, because in our study the proportion of patients with prior antibiotics use was as high as 81.7%. As a tertiary hospital, most patients in our hospital had been referred from other medical institutions. They usually had a history of antibiotic use prior to the blood cultures. Besides, the blood cultures of the HACEK group, serology tests for mycoplasma, bartonella, legionella and the polymerase chain reaction (PCR) were not performed in our hospital. All these factors resulted in the low microbiology detection rate. We suggested that patients with suspected bloodstream infection should receive blood cultures routinely before the use of antibiotics to rule out or confirm the diagnosis of bloodstream infection.
In our study, streptococci was the main pathogen accounting for 24.1%, followed by staphylococci accounting for 18.9%, which was the same as the report in the 1990s from China . This was similar to other developing countries [5, 6] but different from developed countries . For IE in developed countries, the proportion of staphylococci increased gradually and became the main pathogenic bacteria [4, 6].The increase in staphylococcus IE was mainly due to the high incidence of intravenous drug addicts, hemodialysis patients and elderly patients with comorbidities [3, 4, 24]. However, in our study, the proportion of patients with intravenous drug addicts, hemodialysis patients and octogenarians was lower compared with developed countries. Besides, streptococcus IE was the main for community origin IE according to the previous reports, and most of patients in our study were community origin IE .
According to published studies, older age, prosthetic valve IE, heart failure, septic shock staphylococcus aureus, and large vegetation were predictors of poor outcome in patients with IE [1, 4]. In our study, we also found that hemodialysis and pulmonary arterial hypertension were independent risk factors of in-hospital mortality. The risk of in-hospital mortality in patients with Hemodialysis increased by 4.697 times and in patients with pulmonary arterial hypertension, the risk of in-hospital mortality increased by 5.308 times.
There are several limitations in our study. First, as a retrospective study, there exists information bias, and we were unable to obtain information on patient exposure to dental procedures and the use of antibiotic prophylaxis. Second, the study used a single-center cohort in a referal teaching hospital with possible selection bias that could not represent the entire Chinese condition. Finally, as a nonrandomised study, there were associated limitations and selection bias affecting comparisons between the in-hospital outcomes. Therefore, we suggest multiple-center prospective cohort studies performed in our region.