Impact of Continuation of Parenteral Nutrition on Outcomes of Patients with Blood Stream Infections

Abstract

Background:

The impact of continuing parenteral nutrition (PN) in patients who develop blood stream infections (BSI) while receiving PN is largely unknown.

Patients and Methods:

Adult patients admitted to a large academic center over three consecutive years and seven months who had a positive blood culture while receiving PN were included in the study. The cohort was divided into those who had PN continued (PN-c) or discontinued (PN-dc) after the positive culture. We evaluated the effect of continuing PN on clinical outcomes by comparing a composite outcome of recurrent BSI, severe sepsis/septic shock, and death within 30 days between the two groups using a propensity score-weighting regression analysis.

Results:

Of 154 patients included in the study, approximately 70% of whom were surgical patients, 65 (42%) had PN discontinued whereas 89 (58%) had PN continued. Cohort characteristics were similar between the two groups including the Pitt bacteremia score and source control. There were more cases of candidemia (18% vs 6%, p = 0.03) and more cases of intra-abdominal infections (IAI; 42% vs 25%, p = 0.02) in the PN-c group compared with the PN-dc group. The most common sites of infection were endovascular and IAI in both groups. The median duration of bacteremia for both groups was one day. After applying propensity score weighting, the composite outcome of recurrent BSI, severe sepsis/septic shock, and death within 30 days was similar between the PN-dc and PN-c groups (43% and 49%, respectively; p = 0.61).

Conclusions:

Continuing PN in patients with bacteremia or candidemia was not associated with worse clinical outcomes.

Parenteral nutrition (PN) provides an alternative source of nutrition for patients unable to eat or tolerate enteral feeding [1]. Some have speculated that a higher risk of infection, including blood stream infections (BSI), associated with PN may be related to higher blood glucose levels and caloric intake [2] and the fact that PN is often administered through central venous catheters, which are independently more likely to cause BSI than peripheral venous access [3]. Several studies have identified factors associated with increased risk of BSI during PN treatment [2], however, there are limited data on the impact of continuing PN on outcomes of patients with established BSI. Blood stream infections can occur during PN treatment as a result of infection of the vascular catheter through which PN is infused (catheter-related blood stream infection; CR-BSI), another source of infection (e.g., intra-abdominal infection), or less frequently because of contamination of the PN solution [4,5]. At our hospital, we have observed variable practices regarding continuation of PN in patients diagnosed with a BSI (bacteremia or fungemia), with some discontinuing PN only if it is a CR-BSI, others discontinuing only if it is fungemia, and still others who do not discontinue at all. We hypothesized that continuation of PN after a positive blood culture would not worsen clinical outcomes.

Patients and Methods

Study population

All patients ≥18 years old who were admitted to The Johns Hopkins Hospital (JHH) between January 26, 2016, and August 20, 2019, and had a positive blood culture while receiving a course of PN were included in the study. Only the first PN-BSI episode was included for a given hospitalization. The cohort was divided into two groups: PN-continued (PN-c), which included patients who had PN continued for more than 72 hours (treatment interruptions of <48 hours within the first two days of the positive blood culture were counted as continued) and PN-discontinued (PN-dc), which included patients who had PN discontinued within 48 hours of a positive blood culture and for >72 hours. This study was approved by the Johns Hopkins Medicine Institutional Review Board.

Data collection and outcomes

Cohort characteristics, blood culture results, comorbidities, and source of infection were collected for all patients. Source control was defined as the removal of any infected hardware or devices, drainage of infected fluid collections, or resolution of biliary or urinary obstruction. Recurrence of BSI was defined as a positive blood culture for the same or a different organism after 72 hours of a negative blood culture or at least seven days from the last positive blood culture. Sepsis and septic shock were determined by chart review as documented in progress notes using standard definitions. Immunosuppression was defined as active malignancy, chemotherapy within the previous six months, hematologic stem-cell transplantation (HSCT) or solid organ transplantation (SOT) within the previous 12 months, CD4 < 200 cells/mm³, ≥10 mg/d of corticosteroids for longer than two weeks, or immunomodulatory therapy. The primary outcome was a composite outcome that included recurrent BSI, sepsis or septic shock, or death, all within 30 days of the first positive blood culture that occurred during PN.

Statistical analysis

Categorical and continuous parameters were compared using the χ² and the Wilcoxon rank sum test as appropriate. To balance differences between the two groups, we used inverse probability of treatment weighting [6]. Using multivariable logistic regression, propensity scores were created for each patient with the dependent parameter being a binary outcome of PN treatment (PN-c or PN-dc). Covariables used for generating propensity scores were selected a priori based on clinical relevance and possible confounding effect and included age, gender, immunosuppression status, source of infection, organism causing BSI (bacterial versus yeast), source control, Pitt bacteremia score on day one of BSI, and intensive care unit status on day one of BSI. Patients who were in the PN-dc group were weighted by the inverse of the propensity score, and those who received continued treatment (PN-c) were weighted by the inverse of 1 minus the propensity score. The Pitt bacteremia score is a severity of acute illness index and predictor of mortality risk in patients with BSIs (both bacterial and yeast infections). It ranges from 0 to 14 points with a score ≥4 commonly used as an indicator of critical illness and increased risk of death. It includes five parameters: hypothermia, hypotension, respiratory failure, cardiac arrest, and altered mental status [7,8]. Weights were stabilized to increase precision by reducing the influence of extreme weights. Baseline characteristics were considered balanced if standardized difference values were <10%.

In the final analysis, odds ratios (ORs) and 95% confidence intervals (CIs) for the composite outcome were estimated using weighted regression, adjusting for parameters with standardized differences >10%. A two-tailed p value ≤0.05 was deemed statistically significant (STATA version 16.0; StataCorp, College Station, TX).

Results

During the study period, there were 154 patients who developed a positive blood culture while receiving PN. Of these patients, 65 (42%) had their PN discontinued (PN-dc) and 89 (58%) had their PN treatment continued (PN-c). Age, gender, type of PN received, intensive care unit (ICU) location at time of positive blood cultures, clinical service, duration of bacteremia, median Pitt bacteremia score, and source control were all similar between the two groups (Table 1). Overall, 51% of blood cultures grew gram-positive organisms, 30% grew gram-negative organisms, 20% grew yeast, 10% grew anaerobes, and only 5.2% of blood cultures were polymicrobial. Of the 20 patients whose blood cultures grew yeast, only one grew a non-Candida species (Rhodotorula mucilaginosa). Fungemia was more common in the PN-c group (16/89; 18%) compared with the PN-dc group (4/65; 6%; p = 0.03). More patients in the PN-c group had intra-abdominal infections (39/89; 42%) than in the PN-dc group (15/65; 25%; p = 0.02). The proportion of patients with central-vascular catheter infection who had source control was similar in both groups (67% of patients in the PN-c group and 75% in the PN-dc group, p = 0.49). The median duration of PN from the first positive blood culture in the PN-c group was 6.8 days (interquartile range [IQR] 3.9, 10.9); for patients with fungemia, it was 3.8 days (IQR 2.27, 7.72).

Table 1.

Cohort Characteristics

Cohort characteristics, n = patient-hospital admissions	PN-dc	PN-c	p
Cohort characteristics, n = patient-hospital admissions	n = 65	n = 89	p
Age, median, (IQR)	62 (47, 72)	60 (47, 70)	0.95
Female gender, (%)	31 (48)	50 (56)	0.38
Service, (%)			0.93
Medicine	20 (31)	30 (34)
Surgery	45 (69)	59 (66)
ICU location at time of BSI, (%)	16 (25)	13 (15)	0.11
Pitt bacteremia score, median, (IQR)	1 (0, 2)	0 (0, 2)	0.09
Type of PN, (%)			0.21
TPN	37 (57)	43 (48)
Peripheral PN	28 (43)	46 (52)
Length of hospital stay in days, median, (IQR)	32 (13, 63)	29 (16, 63)	0.54
Source of bacteremia, (%)
Endovascular^a	36 (55)	38 (43)	0.12
Blood culture contaminant	9 (14)	13 (15)	0.89
Intra-abdominal	15 (23)	39 (44)	< 0.01
Lungs	6 (9)	2 (2)	0.21
Urinary tract	2 (3)	2 (2)	0.72
Skin, soft tissue, and bone	1 (2)	1 (1)	—0.94
Not specified	3 (5)	4 (4)	—0.94
Source control, (%)	14 (22)	29 (33)	0.16
Median duration of bacteremia in days, (IQR)	1 (1, 1)	1 (1, 1)	—
Pathogen on blood culture, (%)			0.03
Bacteria	61 (94)	73 (82)
Yeast	4 (6)	16 (18)
Comorbidities, (%)
Active malignancy	18 (28)	39 (44)	0.04
Chemotherapy	7 (11)	13 (15)	0.53
Immunosuppressant	5 (8)	5 (6)	0.57
Solid organ transplant	1 (2)	6 (7)	0.12
HIV (CD4 < 200 cells/mm³)	3 (5)	2 (2)	0.41

Includes vascular-catheter infections.

PN-c = parenteral nutrition continued; PN-dc = parenteral nutrition discontinued; IQR = interquartile range; ICU = intensive care unit; BSI = blood stream infection; PN = parenteral nutrition; TPN = total parenteral nutrition; HIV = human immunodeficiency virus.

In the weighted cohort, the primary outcome, which was a composite of recurrent BSI, sepsis or septic shock and all-cause mortality within 30 days, was similar between the PN-dc and PN-c and groups (OR 0.82; 95% CI, 0.38–1.75; Table 2). More specifically, 4% of patients in both groups had recurrent BSI (OR 1.18; 95% CI, 0.27–5.14); 33% and 34% of patients in the PN-dc and PN-c groups, respectively, met criteria for sepsis/septic shock (OR 0.97; 95% CI, 0.44–2.11); and death was observed in 27% and 24% of patients in the PN-dc and PN-c groups, respectively (OR 1.22%; 95% CI, 0.53–2.77).

Table 2.

Outcomes for the Weighted Cohort

Outcome, (%)	PN-dc	PN-c	OR (95% CI) adjusted	p
Outcome, (%)	n = 64	n = 87	OR (95% CI) adjusted	p
Recurrence of BSI	3 (4)	4 (4)	1.18 (0.27–5.14)	0.82
30-day sepsis	21 (33)	31 (34)	0.97 (0.44–2.11)	0.94
All-cause 30-day mortality	17 (27)	22 (24)	1.22 (0.53–2.77)	0.63
Composite outcome	27 (43)	44 (49)	0.82 (0.38–1.75)	0.61

PN-dc = parenteral nutrition discontinued; PN-c = parenteral nutrition continued; OR = odds ratio; CI = confidence interval; BSI = bloodstream infection.

Discussion

In this observational propensity score-weighted cohort of 154 patients who developed a BSI while receiving PN, there was no difference in the composite outcome of recurrent BSI, subsequent sepsis/septic shock, or death within 30 days of the first positive blood culture regardless of whether patients had their PN treatment continued or discontinued, after adjusting for gender, age, immunosuppression status, type of organism causing the BSI, source of infection, source control, location in ICU, clinical service, and Pitt bacteremia score at the time of first positive blood culture.

Few studies have addressed the impact of continuing PN on clinical outcomes of patients with BSI. One study evaluated continuation of total parenteral nutrition (TPN) in 245 patients with concomitant positive blood cultures at an academic hospital [9]. The authors found that continuation of TPN was independently associated with longer hospitalization (44.6 ± 32.3 days vs 28.2 ± 18.5 days; p < 0.001), although mortality was similar between patients who had TPN continued and those who had TPN discontinued. The authors hypothesized that by discontinuing TPN, providers were prompted to consider oral/enteral nutrition at an earlier point, thereby speeding recovery. The authors did not comment on source control for underlying infections.

In vitro studies have shown that lipid emulsion induces candidal germination and biofilm formation on vascular catheters material [10]. However, large human studies have not found a difference in infection rates between patients receiving intravenous fatty emulsions (IVFE) and those not receiving IVFE [2]. Our cohort included 20 cases of yeast infection; most of these patients had their PN treatment continued. Although we did not see a difference in outcomes between the groups based on causative organism (yeast vs bacterial), studies with a larger sample of yeast cases are needed to confirm our observations.

Our study has some limitations. It was conducted in a single center, however, patients from both medicine and surgery services were evaluated, providing patient and provider diversity. Our estimates of severe sepsis and septic shock relied on provider documentation and may be underestimated. Treatment allocation was not randomized because of the retrospective nature of the study, which could introduce indication bias, however, we investigated outcomes through a propensity score-weighted regression model, adjusting for severity of illness and other clinical characteristics that may impact outcome results. Documentation of the reasons why PN was discontinued was not recorded, and patients in the discontinued group may have been less ill than those who continued PN. Severity of illness, however, was included in the propensity score-weighting regression analysis.

Conclusions

Our study suggests that continuing parenteral nutrition in the presence of bacteremia or candidemia is a safe practice. Adequate source control is important to ensure good outcomes.

Footnotes

Acknowledgments

The authors thank Eili Klein, PhD for assistance with data extraction.

Funding Information

No funding was received.

Author Disclosure Statement

The authors have no competing financial interests.

References

Guenter

, Boullata

, Ayers

, et al. Standardized competencies for parenteral nutrition prescribing: The American Society for Parenteral and Enteral Nutrition Model. Nutr Clin Pract, 2015; 30:570–576.

McCleary

, Tajchman

. Parenteral Nutrition and infection risk in the intensive care unit: A practical guide for the bedside clinician. Nutr Clin Pract, 2016; 31:476–489.

Mermel

. Short-term peripheral venous catheter-related bloodstream infections: A systematic review. Clin Infect Dis, 2017; 65:1757–1762.

Gupta

, Hocevar

, Moulton-Meissner

, et al. Outbreak of Serratia marcescens bloodstream infections in patients receiving parenteral nutrition prepared by a compounding pharmacy. Clin Infect Dis, 2014; 59:1–8.

Solomon

, Khabbaz

, Parker

, et al. An outbreak of Candida parapsilosis bloodstream infections in patients receiving parenteral nutrition. J Infect Dis, 1984; 149:98–102.

Amoah

, Stuart

, Cosgrove

, et al. Comparing Propensity score methods versus traditional regression analysis for the evaluation of observational data: A case study evaluating the treatment of gram-negative bloodstream infections. Clin Infect Dis [Epub ahead of print; DOI: 10.1093/cid/ciaa169].

Rhee

, Kwon

, Ki

, et al. Scoring systems for prediction of mortality in patients with intensive care unit-acquired sepsis: A comparison of the Pitt bacteremia score and the Acute Physiology and Chronic Health Evaluation II scoring systems. Shock, 2009; 31:146–150.

Henderson

, Luterbach

, Cober

, et al. The Pitt Bacteremia Score predicts mortality in nonbacteremic infections. Clin Infect Dis, 2020; 70:1826–1833.

Patel

, Liu

, Paul

, Belisle

. Longer hospitalization of patients with positive blood cultures receiving total parenteral nutrition. Surg Infect, 2014; 15:227–232.

10.

Swindell

, Lattif

, Chandra

, et al. Parenteral lipid emulsion induces germination of Candida albicans and increases biofilm formation on medical catheter surfaces. J Infect Dis, 2009; 200:473–480.