Sources of Primary Bloodstream Infections in Internal Medicine Patients – a Cohort Study

Objectives: To describe the sources of bloodstream infections (BSIs) in internal-medicine patients, on admission and during hospitalization, and to determine the proportion of BSIs in which no secondary cause could be de�ned (i.e. primary-BSI). Methods: We analyzed all BSIs at the internal-medicine wards of the two campuses of the Hadassah Hebrew-University Medical Center, during 2017-2018. We de�ned the BSI source of each event (secondary, Central-line associated BSI (CLABSI) or primary non-CLABSI) and compared BSIs present on admission (POA) to hospital acquired (HA). Results: There were 595 patient-unique BSI events, 316 (53.1%) POA-BSI and 279 (46.9%) HA-BSI. Overall, 309 (51.9%) were secondary, 194 (32.6%) primary non-CLABSI and 92 (15.5%) CLABSI. Primary non-CLABSI in the POA-BSI group was 20.6% vs. 46.2% in the HA-BSI group (p=0.001). The length of hospital stay (LOS) from culture to discharge of the HA-BSI group was longer than in the POA-BSI group (mean LOS, 19 days vs. 13.6 days, p=0.01) and the mortality rate was higher (48.7% vs. 19%, p=0.001). Staphylococcus aureus was more common in primary non-CLABSI than in CLABSI and secondary BSI (29.5%, 12.8% and 16.2%, respectively). Interpretations: The proportion of primary non-CLABSI among HA-BSI events is very high (46.2%). The absence of any plausible source for these BSIs, and the fact that most patients in Internal-medicine wards have peripheral lines, suggests that the peripheral catheter is a probable source for primary non-CLABSIs. Measures to prevent peripheral line associated BSI (PLABSI), similar to those implemented successfully for the prevention of CLABSI, should be considered.


Introduction
Bloodstream infection (BSI) is a common diagnosis in hospitalized patients causing signi cant morbidity and mortality (1)(2). Its incidence rate has increased from 7. Hospital acquired BSIs have been extensively studied among patients in intensive-care units, (4,5) especially central-line associated BSI (CLABSI) (6); however, there is a research gap regarding BSIs among patients hospitalized in other departments, especially internal medicine wards. Previous studies among hospitalized patients described the common pathogens causing BSI and their susceptibility pattern but did not relate to the possible source of the BSI (7)(8)(9).
BSI is usually secondary to an infection at another body site (e.g. urinary tract infection, intra-abdominal infection etc.). When there is no clear secondary source of infection, the term primary BSI is used. When the patient has primary BSI in the presence of a central line, the BSI is attributed to the line and de ned as CLABSI. If a patient with primary BSI has no central line, the BSI is de ned as primary non-CLABSI (10).
The common infection complicating peripheral venous catheters (PVC) is thrombophlebitis while BSIs secondary to the presence of PVCs are considered a rare complication (11)(12)(13). Nevertheless, given the large proportion of primary non-CLABSI among hospitalized patients with BSI, this assumption needs to be re-examined. If the peripheral line can be appointed as a signi cant cause of BSI, then preventive measures should be taken in order to minimize it, similar to CLABSI prevention (14).
In our Medical Center, we perform routine surveillance of BSIs among patients hospitalized in the medical wards. We have noticed that in many cases of HA-BSI a secondary source cannot be de ned; in most cases there is no central line either. In order to further investigate this group, we initiated a two-year study of BSIs in the medical wards.

Design
We preset a cohort study among patients admitted to internal medicine wards.

Setting
The Hadassah-Hebrew University Medical Center consists of two academic hospitals with 1075 in-patient beds, Jerusalem's largest. The Hadassah Ein-Kerem Campus is a 775-bed, academic tertiary care hospital with all facilities,. The Mount-Scopus Campus is a 300-bed, community hospital. In total, there are six internal medicine departments with 170 beds.
A waiver of informed consent was granted by the institutional research ethics committee, as this was considered a quality control initiative, and there were no patients identi ed or treatments given.

Blood cultures
Blood cultures are obtained when BSI is suspected, either in the emergency department or during hospitalization, at the discretion of the attending physician. It is recommended to obtain two sets of BCs when BSI is suspected, but this is not always carried out. Each BC set includes both aerobic and anaerobic bottles. Bottles are used according to the manufacturer's instructions. Blood cultures are usually obtained from peripheral venipuncture following skin prep with 70% isopropyl alcohol or 0.5% chlorhexidine in 70% alcohol.

Microbiologic Methods
Becton-Dickinson BACTEC Plus aerobic and anaerobic bottles (Becton, Dickinson & co., Sparks, MD, USA) were used for all blood cultures and were processed using the BACTEC 9240 system. Bacteria were identi ed using standard microbiological techniques (15).

Data collection and Analysis of Bloodstream infection events
The study population included patients aged 18 years or above, hospitalized in an internal medicine department, from January 1, 2017 to December 31, 2018, and had at least one positive blood culture, upon admission or during hospitalization. Analysis of positive BCs was done at the end of each month. Data was obtained from the Information Systems Division using a dedicated computerized system, including demographics, admission date, department, culture date and the pathogen. Clinical data (i.e. past medical history, vital signs, physical examination, laboratory) and clinical course (i.e. length of hospitalization, intensive-care unit admission, mortality) were further collected from the electronic medical chart.
An infectious disease specialist assessed the data: The rst step in every BC analysis was to exclude contaminants according to Centers for Disease Control surveillance guidelines (10,16); Blood cultures with true pathogens were further analyzed. Each BSI was classi ed either as POA-BSI if the culture was obtained within two calendar days since admission, or as HA-BSI if the culture was obtained later. When more than one positive BC were related to the same clinical event they were counted as one BSI. If a patient had more than one BSI event during the study period, only the rst was included in the analysis.
Every BSI was further categorized to one of three groups: 1) secondary BSI, a BSI that is thought to be seeded from a speci c infection at another body site (unique bacteria, e.g. Brucella species, which are endemic in Israel (17), were classi ed as secondary BSI); 2) CLABSI, a BSI that is not secondary and the patient has a central line for at least two days prior to the date of the positive BC; 3) primary non-CLABSI, a BSI that is not secondary and the patient doesn't have a central line (3). In the secondary BSI group the source of infection was de ned according to CDC de nitions (e.g. catheter-associated UTI, intraabdominal infection etc.).

Statistical analysis
Categorical variable distributions are presented with absolute count and percentages, continuous variables with mean and standard deviation. We compared variable distribution between clinical events that were de ned as POA and HA. Additionally, we compared variable distribution between secondary BSI, CLABSI and primary non-CLABSI events. Categorical variables were compared with the chi-square test and continuous variables with the Student t-test and ANOVA (between three groups). We considered pvalues below 0.05 statistically signi cant. In order to assess the association of BSI group (CLABSI, primary non-CLABSI and secondary infection) and community/hospital source with in-hospital mortality, we generated a logistic regression model including age and sex variables as well as co-morbidities, which were unevenly distributed between the groups of interest. The associations are described with odds ratio and a 95% con dence interval.

Results
During the two-year period (2017-2018), there were 2084 positive blood cultures among patients hospitalized at the internal medicine wards ( Figure 1). After excluding contaminants and recurrent BSI events in the same patient, 595 patient-unique events were included in the nal analysis. Overall, 309 (51.9%) were secondary, 194 (32.6%) primary non-CLABSI and 92 (15.5%) CLABSI. The classi cation into POA-BSI and HA-BSI, and the source of the BSI, are presented in the Figure 1.
Baseline and clinical patients' characteristics according to the timing of the event are presented in Table   1. The patients with HA-BSI were older than those with POA-BSI (mean age 71.3 vs. 66.6, p=0.001) and the proportion of nursing home residence among them was higher (25.1% vs. 16.8%, p=0.027). The clinical characteristics of both groups were generally similar, with only minor differences. The length of hospital stay (LOS) from culture to discharge in the HA-BSI group was longer than in the POA-BSI group (mean LOS, 19 days vs. 13.6 days, p=0.01) and the mortality rate was higher (48.7% vs. 19%, p=0.001).
This latter nding remained signi cant in the multivariate model, controlling for age, sex, hemodialysis, chronic obstructive pulmonary disease and nursing home residency, where we found that HA-BSIs were associated with an increased OR of mortality of 3.76 (95%CI, 2.46 -5.74) compared to POA-BSIs. The source of BSI differed signi cantly between the two groups: among the POA-BSI group 67% were secondary BSIs and 20.6% were primary non-CLABSIs. Among the HA-BSI group 34.4% were secondary BSI and 46.2% were primary non-CLABSI (p=0.001) ( Figure 1, Table 1).
We compared patients' characteristics and clinical outcomes according to the source of the BSI (Table 2).
Chronic renal failure and the need for dialysis were more common among patients with CLABSI than in patients with primary non-CLABSI or secondary BSI (p=0.001). In addition, in-hospital death rate was signi cantly higher in the CLABSI and primary non-CLABSI groups (43.5% and 39.7% respectively) than in the secondary BSI group (25.6%, p=0.001). However, in the multi-variate analysis the difference in mortality stayed signi cant only for CLABSI. Compared to secondary BSI, CLABSI was associated with an increase OR of mortality of 2.1 (95% CI, 1.1 to 3.8) while the OR of mortality among primary non-CLABSI was 1.2 (95% CI, 0.78-1.79).  Table 3. The most common secondary sources were urinary tract infections (44.1% and 26.0% respectively) followed by skin and soft tissue infections (13.6% and 18.8% respectively) and lower respiratory tract infections (7.5% and 27.1% respectively). Unique bacteria (i.e. Brucella sp., Salmonella enterica) were found only in the POA-BSI group. The bacteria isolated according to the timing of the event (POA or HA) and according to the source of infection are presented in Table 4. In POA events, the most common bacteria belonged to Enterobacteriaceae followed by Staphylococcus aureus, Enterococcus species and Pseudomonas aeruginosa. In HA events the ndings were quite similar: Enterobacteriaceae were the most common bacteria followed by S. aureus (Table 4a). Among CLABSI events, Gram-negative bacilli account for most events followed by S. aureus and coagulase-negative Staphylococcus. In contrast, in primary non-CLABSI, S. aureus was much more dominant and accounted for about a third of cases (Table 4b).   20.6% respectively, p=0.001) (Figure 1, Table 1).
True BSI should always warrant clinicians to look for its source, in order to treat the patient properly. In clinical practice, many times such a source cannot be de ned. Several reasons can be given for that, such as lack of appropriate workup or previous antibiotic treatment. In addition, adherence to the strict CDC criteria when analyzing a BSI event can sometimes diminish the ability to classify events as secondary.
We conducted a rigorous evaluation in each BSI event in order to identify the likely source. The nding of lack of a source for HA-BSI in almost half of the cases requires explanation.
Over the last two decades, attention has been focused on central venous lines as an important and common source of BSI. The same mechanisms leading to BSI secondary to the presence of a central venous catheter (i.e. bacterial invasion through or around the catheter) (18), may apply also to PVCs. Though peripheral lines are placed for a shorter duration than central lines (19), PVCs are much more prevalent in the hospital setting. Practically, almost every hospitalized patient in the internal medicine ward has a PVC (20). These lines are usually inserted and maintained without the strict precautions applied for central lines (11). Hence, it is reasonable to conclude that PVCs can be an important source of primary non-CLABSI.
The peripheral line is an obvious cause for primary non-CLABSI when thrombophlebitis is present (13,(21)(22)(23), however in our study, thrombophlebitis was found only in 3/129 (2.3%) cases of primary non-CLABSI. Lack of local signs in primary non-CLABSI is not surprising, as patients with CLABSI also usually have no local signs of infection. We propose that when there is a BSI without a secondary cause, in the presence of a peripheral line (and without a central line), peripheral-line associated BSI (PLABSI) is the presumed diagnosis. As in the CDC guidelines for the diagnosis of CLABSI, the PLABSI de nition is also not relied on tip of the line cultures (i.e. catheter-related BSI), since these are not feasible in everyday clinical practice. In our opinion, PLABSI is the commonest hospital acquired cause for primary non-CLABSI, and in most cases without local signs.
In a systematic review by Mermel (24) in most studies included, PLABSI was diagnosed by local signs of infection (i.e. thrombophlebitis). According to our ndings, those events of PLABSI with local signs probably constitute only a small part of PLABSI cases.
Once the concept of PLABSI is de ned, it should lead us to take appropriate preventive measures. There is plenty of information regarding CLABSI prevention, on insertion of the catheter and during its use. It is reasonable to assume that the proven practices for the prevention of CLABSI (25) can be effective as well for PLABSI prevention. (26-29). Prevention of BSI as a complication of a PVC through appropriate handling of insertion and maintenance of such lines could prevent a considerable number of deaths from HA-BSIs (30). Moreover, when there is a BSI and no secondary cause can be identi ed, it may be prudent to replace the peripheral line.
In our study, both in POA and HA, Gram-negative bacteria were the commonest pathogens identi ed followed by Staphylococcus aureus. The same prevalence was found in primary non-CLABSI (39.1% and 29.5%, respectively) (Table 4b).
We found worse outcomes of HA-BSI vs. POA-BSI, with a prolonged length of stay since the positive culture (21.4 days vs. 14.4 days, p=0.01) and signi cantly higher mortality rate (48.7% vs. 19.0%; OR 3.76, 95%CI, 2.46 -5.74) ( Table 1). In our cohort, the patients with HA-BSI were older and had more comorbidities compared to the patients with POA-BSI; this might explain their worse outcomes.
Comparing the clinical outcomes according to the BSI source (Table 2), the mortality rates in CLABSI and primary non-CLABSI were similar and signi cantly higher than that in secondary BSI (43.5%, 39.7% and 25.6% respectively, p=0.001). However, in multi-variate analysis, only in CLABSI the mortality rate was signi cantly higher than in secondary BSI (OR 2.1, 95% CI, 1.1 to 3.8).
Our study has some limitations. We conducted analyses at the end of each month and not immediately on the date of the positive BC. This delay could have led to missing of some information. However, we conducted a thorough investigation in each case and the perspective nature of the analysis allowed us a better understanding of the disease course. Additionally, we could not record the presence of a peripheral line at the time of BSI, yet it is rare to nd a patient in internal medicine wards without a line, as is known also from the literature (12). Finally, we did not prove microbiologically that the peripheral line is the cause of BSI in primary non-CLABSI. This would have been di cult to perform and was not part of this study. For the same reason, the CDC surveillance guidelines use the de nition of CLABSI rather than the microbiologically proven catheter-related BSI (10).
In conclusion, we found that a large proportion (46.2%) of HA-BSI among internal medicine patients do not have a secondary cause. We suggest that the source for many of these BSIs is the peripheral line. Measures to prevent PLABSI, similar to those implemented for the prevention of CLABSI, should be applied.

Declarations
Ethics approval and consent to participate -A waiver of informed consent was granted by the Hadassah hospital research ethics committee (HMO-12-0460), as this was considered a quality control initiative, and there were no patients identi ed or treatments given.
All methods were carried all methods were performed in accordance with the relevant guidelines and regulations.
The study was not experimental.
The Hadassah institutional research ethics committee granted waiver from the need for informed consent.
Hadassah institutional research ethics committee approved all the study methods.

Consent for publication -not applicable
Availability of data and materials -the datasets generated and/or analyzed during the current study are not publicly available due hospital policy but are available from the corresponding author on reasonable request.