Proposal for the use of echocardiography in bloodstream infections due to different streptococcal species

This is a sub-study to a population-based study including adult patients with monospecies streptococcal BSIs admitted from January 1, 2008 to December 31, 2017 in the Capital Region of Denmark (1.5 million people) [9]. The full method has been described in detail in the main study. In addition to the main study exclusion of cases with unavailable species identification, the current sub-study also excluded rare species accounting for < 0.5% of BSIs. Using Danish nationwide registries [13, 14] we identified IE cases with hospital admissions with primary and secondary ICD-10 codes (I33, I38 and I39.8) and IE was considered associated to the streptococcal BSI, if the positive blood cultures occurred during the IE admission or up to 30 days prior. Since detailed patient chart data including data on echocardiograms are not available in the registries the estimated IE prevalence is based on the assumption that when the hospital doctors diagnose a patient with IE the patient actually has IE and in the same way, when a patient is not diagnosed with IE it is because the patient does not have IE. To reduce the number of false positive IE diagnoses, we required IE cases to have a hospital admission of minimum 14 days unless they died during the first 14 days of admission because this algorithm have been validated with a positive predictive value (PPV) above 90% for the diagnosis of IE in the Danish registries [15, 16]. See Supplemental Table 1 for ICD-, procedure- and ATC-codes used in this study.

Based on the crude estimated IE prevalence, each streptococcal species was assigned to a risk group: low (< 3%), moderate (3–10%), high (10–30%), and very high (> 30%). After grouping the streptococcal species in risk groups, the IE prevalence was further assessed in subgroups according to number of positive blood culture bottles and presence of conventional IE risk factors (prosthetic valve, previous IE, native valve disease and cardiac device) [17, 18]. From these subgroups we decided to use the following cut-offs for the proposed use of TTE and TOE: (i) IE prevalence > 10%, TTE and are recommended (ii) IE prevalence of 3–10%, TTE is recommended (iii) IE prevalence < 3%, “wait & see” is recommended, meaning that no initial echocardiography is recommended unless there is a strong clinical suspicion of IE.

From a total of 7122 streptococcal BSI cases we identified 6506 eligible cases with monospecies streptococcal BSIs in 6224 unique patients, of which 236 patients had more than one BSI case with a median of 332 days [IQR 102–869] between the first and the second episode. After excluding rare species (causing < 0.5% of BSIs) we included 6393 cases with monospecies streptococcal BSIs of which 451 were diagnosed with IE (7.1%) (Fig. 1). Based on the IE prevalence streptococcal species were classified as low-risk, moderate-risk, high-risk and very high-risk streptococci as described in the main study [9]. There were 3230 BSI cases (51%) in the low-risk group, 2106 BSI cases (33%) in the moderate-risk group, 556 BSI cases (9%) in the high-risk group, and 501 BSI cases (8%) in the very high-risk group (Fig. 1). In the total cohort the mean age was 68.1 years (SD 16.2) and 52.8% were men. Baseline characteristics according to risk groups are shown in Table 1. Cases in the very high-risk streptococci group were older and had the highest rate of native valve disease (13.2%), prosthetic valve (12.4%), previous IE (4.4%), and cardiac device (9.2%) (Table 1).

Figure 2 shows the crude prevalence of IE according to streptococcal species and the multivariable logistic regression adjusted for age, sex, and IE risk factors (≥ 3 positive blood culture bottles, native valve disease, prosthetic valve, previous IE and cardiac device). In the adjusted analysis including a slightly smaller cohort than in the main study, the overall risk pattern remained stable with progressively higher associated risk of IE in the species with the higher IE prevalence (Fig. 2).

The IE prevalence was calculated for each streptococcal risk group and is presented relative to the number of positive blood culture bottles, and IE risk factors (Table 2). Low-risk streptococci (S. pneumoniae, S. pyogenes, and S. intermedius) had an IE prevalence of 0.5–1.5%, unless they had ≥3 positive blood culture bottles and at least one further risk factor in which case IE prevalence was 6.9% [95% CI: 3.4–12.3%]. Moderate-risk streptococci (S. agalactiae, S. anginosus, S. constellatus, S. dysgalactiae, S. salivarius, and S. thermophilus) had an IE prevalence of 1.5% [95% CI: 0.9–2.4%] if they neither had a risk factor nor ≥3 positive blood culture bottles, but an IE prevalence of 3–10% if they had either ≥3 positive blood culture bottles (IE prevalence 9.6% [95% CI: 7.5–12.1%]) or a risk factor (IE prevalence 8.7% [95% CI: 4.4–15.1%]). Moderate-risk streptococci presenting with both ≥3 positive blood culture bottles and a risk factor had an IE prevalence of 31.1% [95% CI: 23.3–39.7%]. High-risk streptococci (S. mitis/oralis, S. parasanguinis, and G. adiacens) had an IE prevalence of 3.6% [95% CI: 1.9–6.0%]) if they neither had a risk factor nor ≥3 positive blood culture bottles, but an IE prevalence of 14.0% [95% CI: 5.3–27.9%] if they presented a risk factor and < 3 positive blood culture bottles and an IE prevalence of 48.4% [95% CI: 40.4–56.5%] if they had ≥3 positive blood culture bottles. Finally, very high-risk streptococci (S. gordonii, S. gallolyticus (formerly S. bovis), S. mutans, and S. sanguinis) had a very high baseline IE prevalence of 35.7% [95% CI: 31.5–40.1%] (Table 2).

Based on the estimated IE prevalence in BSIs with different streptococcal species and taking the number of positive blood culture bottles and IE risk factors into account we designed a flowchart for the use of TTE and TOE (Fig. 3). Using the specified IE prevalence cut-offs, we propose the following strategy: “wait & see” (IE prevalence < 3%), TTE (IE prevalence 3–10%), and TTE plus TOE (IE prevalence > 10%). In this way, low-risk streptococci are directed to “wait & see”, unless they have ≥3 positive blood culture bottles and an IE risk factor in which case they are led to TTE. Moderate-risk streptococci with < 3 positive blood culture bottles and no risk factors are guided to “wait & see”, whereas a TTE is recommended if they have either ≥3 positive blood culture bottles or a risk factor. However, moderate-risk streptococci presenting with both ≥3 positive blood culture bottles and a risk factor are recommended a TTE plus TOE. High-risk streptococci are recommended a TTE if they neither have a risk factor nor ≥3 positive blood culture bottles, but a TTE plus TOE if they have either ≥3 positive blood culture bottles or a risk factor. Finally, very high-risk streptococci are recommended a TTE plus TOE due to a very high baseline IE prevalence.

The flowchart is created on retrospective data from a partly register-based setup, which naturally introduces limitations. Firstly, the diagnosis of IE was based on ICD-10 discharge codes with the inherent risk of misclassification bias. However, the applied method using ICD-10 codes in the Danish registries has been validated with a positive predictive value for the diagnosis of IE of 90% [15]. Since the negative predictive value of the IE diagnosis in the Danish registries has not been investigated, we cannot accurately estimate the amount of overlooked IE patients that were misclassified (false negatives) by not receiving an ICD-10 IE diagnosis. Secondly, since data on TTE and TOE are not available, the flowchart was created retrospectively without knowledge of the use of echocardiography in the different groups. Since echocardiography plays a central role in diagnosing IE it would have been relevant to know if some species were less often examined with echocardiography thereby increasing the probability of ascertainment bias and missed IE cases. Having that in mind, the IE prevalence should be interpreted as conservative estimates. At the same time, the missing data on echocardiograms introduces a risk of circular deduction where the species more typically examined by echocardiography also becomes the species where IE is diagnosed most often. In other words that the current clinical selection of patients for echocardiography are already performed in a manner parallel to the suggested flow chart without collective consciousness of the specific selection process. In this way the suggested flowchart may just be the first actual description of the performed clinical practice applied by IE experts. The only way to answer this question is to perform a prospective screening study with systematic echocardiography in all patients with streptococcal BSI. Thirdly, all positive BC bottles occurring within 14 days of the first positive BC were considered part of same BSI episode without the possibility to distinguish persistent bacteraemia from several positive BCs obtained in a single initial BC sampling. Fourth, the present study was performed in a specific geographical setting in Scandinavia, which may not represent the prevalence of IE according to streptococcal species in other regions of the world. Finally, it is important to emphasize that the proposed flowchart should not stand alone and the clinical evaluation of the patient is still central in the decision making as stated in the international guidelines [10, 11].