Thrombocytopenia and Bloodstream Infection: Incidence and Implication on Length of Stay in the Pediatric Intensive Care Unit

Abstract

The incidence and prognosis of thrombocytopenia in critically ill patients with bloodstream infection (BSI) is not well delineated in the pediatric intensive care unit (PICU) setting. We assessed these variables in our PICU and sought to determine whether thrombocytopenia could serve as a prognostic marker for length of stay (LOS). The study was conducted at the medical PICU of a university hospital, on all critically ill pediatric patients consecutively admitted during a 3-year period. Patient surveillance and data collection have been used to identify the risk factors during the study period. The main outcomes were BSI incidence and implication on morbidity and LOS. Data from 2,349 PICU patients was analyzed. The overall incidence of BSI was 3.9% (93/2,349). Overall, 85 of 93 patients (91.4%) with BSI survived and 8 patients died (8.6% mortality rate). The overall incidence of thrombocytopenia among these 93 patients was 54.8% (51/93) and 100% (8/8) for the nonsurvivors. Out of the 85 survivors, 27 thrombocytopenic patients were hospitalized for >14 days versus 14 of nonthrombocytopenic patients ( p  = 0.007). Thrombocytopenia was associated with borderline significance with an increased LOS (adjusted odds ratio = 3.00, 95% confidence interval: 0.93–9.71, p  = 0.066). Thrombocytopenia is common in critically ill pediatric patients with BSI and constitutes a simple and readily available risk marker for PICU LOS.

Introduction

Thrombocytopenia is one of the most common laboratory-detected abnormalities in pediatric intensive care unit (PICU) patients. The main factors that contribute to thrombocytopenia in the PICU are impaired platelet production, increased platelet consumption or destruction, and sequestration of platelets in the spleen or at the endothelial level ¹ _. Drug-induced thrombocytopenia, hemophagocytosis, bleeding, and hemodilution are also major factors. ² Thrombocytopenia has been well described in adults with bloodstream infection (BSI), ^{3 4 5 6 7} which is known to alter the clinical course, laboratory findings, and hemodynamic parameters of affected patients. BSI account for approximately 50% of all thrombocytopenia cases in the severely ill, and the severity of sepsis correlates with the decrease in platelet count. ⁶ Data on the incidence and prognosis of thrombocytopenia in critically ill pediatric patients with BSI are sparse. Furthermore, nearly all adult studies analyzing thrombocytopenia as a prognostic marker in critically ill patients found an inverse relationship between the platelet count and risks for a prolonged intensive care unit LOS and mortality. ⁸ Furthermore, many studies stress the importance of infections related to the duration of hospital stay and the associated costs involved with prolonged LOS. ⁹

Therefore, we sought to determine the prevalence of thrombocytopenia in pediatric critically ill patients with BSI and its implication on LOS.

Methods

Study Population

The population of this retrospective cohort study comprised pediatric patients with a microbiologically documented BSI who were admitted to the PICU at the Safra Children's Hospital at Sheba Medical Center in Israel between January 2014 and December 2016. Patients with bone marrow suppression, regardless of the etiology, were excluded. Those with more than one episode of sepsis or who were later readmitted for recurrent sepsis were included only once in our analysis.

The protocol for this study was approved by the ethics committee of our institution and conforms to the provisions of the Declaration of Helsinki (2016).

Data Collection

The primary outcome of this study was the incidence of thrombocytopenia among PICU patients with BSI. The secondary aim was to determine whether thrombocytopenia is a reliable predictor of morbidity and LOS in critically ill pediatric patients with BSI. BSI was defined as isolation of a true pathogen in at least one blood culture. Twenty-three different pathogens were isolated and classified as either gram-positive, gram-negative, or fungi. We required at least two positive blood cultures, drawn on separate occasions, to confirm cases of a gram-positive infection with coagulase-negative staphylococci due to its relatively low virulence.

A hospital-wide surveillance system was used to search for PICU patients with BSI. Data on each study patient were ret from the hospital's electronic patient database and included sociodemographic characteristics, diagnosis on admission, comorbidities, date of BSI, pathogen identified in blood culture, laboratory results, required supportive therapy (to include length of mechanical ventilation, and vasoactive-inotropic score [VIS]), LOS, and survival.

The day that the positive blood culture drawn was documented as day 0. Thrombocytopenia was defined as a platelet count ≤150 × 10 ⁹ /L (as defined in the literature). Each sample that was suspected of low platelet count was examined twice at least once during the index hospitalization, which was defined as the period of time of 2 days prior to BSI until 2 days after BSI (day −2, −1, 0, +1, and +2).

Statistical Analysis

Demographic and clinical characteristics—including survival status, thrombocytopenia, and LOS—were analyzed. Qualitative variables were presented as frequencies and percentages, and they were compared between groups with the Chi-square test and Fisher's exact test. Quantitative variables were presented as medians and interquartile ranges (IQRs), and between group comparisons took place with Student's t -test or the Wilcoxon's rank sum test. A multivariable logistic regression model was fitted for an LOS of at least 15 days compared with one of less than 15 days. The model compared the thrombocytopenic and nonthrombocytopenic groups while adjusting for demographic and clinical variables. The goodness of fit of the model was evaluated with the Hosmer and Lemeshow goodness-of-fit test and a receiver operating curve. All tests conducted were two-tailed, and p -values below 0.05 were considered statistically significant. The data were analyzed with SAS software version 9.4 (SAS Institute, Cary, North Carolina, United States).

Results

During the 3-year study period, 2,349 pediatric patients were admitted to our PICU, of whom 93 (3.9%) had a confirmed BSI and were eligible for study inclusion. A total of 51 BSI patients (54.8%) had thrombocytopenia and 42 BSI patients (45.2%) did not ( Fig. 1 ).

Fig. 1.

Flow chart of study cohort.

Overall, 85 patients (50 males and 35 females) survived to discharge (91.4%). The median age at admission was 11.6 years (IQR: 2.5–53.4), and the median weight was 8.0 Kg (IQR: 4.0–17.8). Eight patients (four males and four females) died (8.6%), all of which had thrombocytopenia. Their median age was 2.6 months (IQR: 1.9–42.2), and their median weight was 6.7 Kg (IQR: 4.0–13.5). Among the patients that survived, 50 (58.8%) were mechanically ventilated with a median PaO ₂ /FIO ₂ ratio of 200 (IQR: 124–310). In comparison, all nonsurviving patients required mechanical ventilation with a median PaO ₂ /FIO ₂ ratio of 106 (IQR: 78–247) and also required hemodynamic support, with a median VIS of 29 (IQR: 14–40).

A total of 85 positive cultures were identified in surviving patients. Gram-positive microorganisms were most frequently isolated (45%, n  = 38), followed by gram-negative (38%, n  = 32) and fungal (17%, n  = 15) microorganisms. In comparison, there were significant differences in the type of bacterial and fungal organisms found between those patients with and without thrombocytopenia ( p  = 0.007). Precisely, 21 thrombocytopenic patients (48.8%) had gram-negative BSI compared with 11 nonthrombocytopenic patients (26.2%), whereas gram-positive BSI was more prominent in the nonthrombocytopenic group 26 (61.9%) versus 12 (27.9%) in the thrombocytopenic group. The presence of a fungal BSI among thrombocytopenic patients was twice that compared with nonthrombocytopenic patients ( n =10 [23.3%] vs. n  = 5 [11.9%]).

Comparison of inflammatory biomarker findings between thrombocytopenic patients and nonthrombocytopenic patients revealed several differences. The main difference was in the C-reactive protein (CRP) levels, which was significantly different between the groups with a mean of 130 mg/L (95% confidence interval [CI]: 85.0–210.0) for the thrombocytopenic group versus a mean of 89.5 mg/L (95% CI: 19.0–177.0) for the nonthrombocytopenic group ( p  = 0.035). The white blood count (WBC) revealed a trend in the opposite direction, where thrombocytopenic patients demonstrated a mean of 11.9 K/μL (95% CI: 5.5–20.5) for the thrombocytopenic group versus a mean of 17.2 K/μL (95% CI: 12.3–23.1) for the nonthrombocytopenic group ( p  = 0.0059).

Otherwise, respiratory failure with mechanical ventilation ( p  = 0.038), cardiovascular failure with VIS ( p  = 0.0004), platelet transfusion ( p  = 0.001), and elevated international normalized ratio (INR; p  = 0.0006) were more evident in thrombocytopenic patients as demonstrated in Table 1 . Sex ( p  = 0.312), age ( p  = 0.58), and PaO ₂ /FiO ₂ ( p  = 0.48) were insignificant between both groups.

Table 1. Demographic and clinical characteristics of the surviving patients with and without thrombocytopenia.

Variables	Thrombocytopenia	Non-thrombocytopneia	P -value
Sex (%)			0.312
Male	23 (53.5)	27 (64.3)
Female	20 (46.5)	15 (35.7)
Platelet transfusion (%)			0.001
Yes	10 (23.3)	0 (0)
No	33 (76.7)	42 (100)
Microorganisms (%)			0.007
Gram-negative	21 (48.8)	11 (26.2)
Gram-positive	12 (27.9)	26 (61.9)
Fungi	10 (23.3)	5 (11.9)
Mechanical ventilation (%)			0.038
Yes	20 (47.6)	30 (69.8)
No	22 (52.4)	13 (30.2)
Weight (kg)	7.0 (3.4–17.1)	9.3 (4.4–18.1)	0.238
Age (mo)	9.9 (2.5–45.9)	13.9 (2.3–60.6)	0.580
WBC (K/μL)	11.9 (5.5–20.5)	17.2 (12.3–23.1)	0.0059
CRP (mg/L)	130 (85.0–210.0)	89.5 (19.0–177.0)	0.035
VIS	1.0 (0–10)	0.0 (0.0–0.0)	0.0004
PaO 2 /FiO 2 ratio	180.0 (122.0–321.5)	244.0 (176.0–300.0)	0.484
INR	1.4 (1.2–1.6)	1.1 (1.1–1.35)	0.0006

Abbreviations: CRP, C-reactive protein; INR, international normalized ratio; VIS, vasoactive inotropic score; WBC, white blood count.

Note: Values are given as frequencies (percentages) for qualitative variables and median (interquartile range) for quantitative variables.

In univariate analysis, we used LOS as a dependent variable to reflect morbidity. LOS was subdivided into two groups of 0 to 14 and >14 days. Patients with thrombocytopenia required significantly longer hospitalization where 65.9% (27/43) required >14 days of LOS versus 36.4% of them (16/43) who required only 0 to 14 days of LOS ( p  = 0.007). Again, VIS ( p  = 0.0004) and respiratory support (mechanical ventilation) ( p ≤ 0.0001) was greater in patients with thrombocytopenia. There was no significant difference between the groups in terms of sex, platelet transfusions, microorganisms, age, weight, WBC, and CRP findings as seen in Table 2 .

Table 2. Univariate analysis of prognostic factors associated with length of stay.

Variables	Length of stay		p -Value
	0–14 d ( n  = 44)	>14 d ( n  = 41)
Sex (%)			0.41
Male	24 (54.6)	26 (63.4)
Female	20 (45.4)	15 (36.6)
Thrombocytopenia (%)			0.007
Yes	16 (36.4)	27 (65.9)
No	28 (63.6)	14 (34.1)
Platelet transfusion (%)			0.51
Yes	4 (9.1)	6 (14.6)
No	40 (90.9)	35 (85.4)
Microorganisms (%)			0.16
Gram-negative	14 (31.8)	18 (43.9)
Gram-positive	24 (54.6)	14 (34.2)
Fungi (%)	6 (13.6)	9 (22.0)
Mechanical ventilation (%)			<0.0001
Yes	14 (31.8)	36 (87.8)
No	30 (68.2)	5 (12.2)
Age (mo)	14.1 (4.2–62.4)	7.3 (2.3–51.6)	0.45
Weight (kg)	9.6 (4.3–18.2)	6.9 (3.6–16.1)	0.42
WBC (K/μL)	15.6 (10.4–22.2)	13.4 (9.0–22.6)	0.63
CRP	99.5 (49.0–177.0)	129.0 (47.0–206.0)	0.42
VIS	0.0 (0.0–0.0)	1.0 (0.0–10.0)	0.0004

Abbreviations: CRP, C-reactive protein; VIS, vasoactive inotropic score; WBC, white blood count.

Note: Values are given as frequencies (percentages) for qualitative variables and median (interquartile range) for quantitative variables.

A multivariable analysis was also performed to examine LOS as a dependent variable to reflect morbidity ( Table 3 ). The logistic multivariable model included thrombocytopenia, mechanical ventilation, type of microorganism, sex, and age. Thrombocytopenia emerged as a potential prognostic variable for increasing the LOS (odds ratio [OR]: 3.0, 95% CI: 0.93–9.71, p  = 0.066). Mechanical ventilation was also a significant independent variable associated with a prolonged LOS (OR: 14.82, 95% CI: 4.52–48.65, p  < 0.0001). Neither sex ( p  = 0.193), age ( p  = 0.93), gram positive ( p  = 0.776), nor gram negative ( p  = 0.825) organism was associated with LOS. VIS was not included in the multivariable analysis because it was found to be in closed correlation with thrombocytopenia.

Table 3. Multivariate analysis independent predictors of length of stay.

Variables	OR	95% CI	p -Value
Thrombocytopenia
No	1.00	Ref
Yes	3.00	0.93–9.71	0.066
Ventilated
No	1.00	Ref
Yes	14.82	4.52–48.65	<0.0001
Microorganisms
Fungi	1.00	Ref
Gram-positive	0.79	0.16–3.87	0.776
Gram-negative	0.84	0.18–3.98	0.825
Sex
Male	1.00	Ref
Female	0.47	0.15–1.47	0.193
Age
Increase per month	1.003	0.99–1.01	0.525

Abbreviations: CI, confidence interval; OR, odds ratio.

Discussion

This study looks at the association between thrombocytopenia and BSI in patients hospitalized in PICU. BSI was confirmed in 93 out of a total 2349 (3.9%) PICU admissions of whom 85 patients survived and 8 died. All the deceased patients developed thrombocytopenia. Significantly higher mortality rates had been previously reported among thrombocytopenic patients in general ICUs, which is consistent with our findings. Even though thrombocytopenia is often multifactorial, BSI was likely a contributing factor as seen in prior studies. ^{3 7 10 11}

Thrombocytopenia, whether present at the onset of a BSI or developing later during the clinical course, is associated with nonfavorable outcomes. ^{10 11} In this study, we found that thrombocytopenia was associated with increased LOS, this finding was statistically significant in the univariate analysis ( p  = 0.007), and of borderline significance ( p  = 0.066) in the multivariate analysis. These findings are similar to multiple adult studies that have shown that the development of thrombocytopenia is associated with increased LOS during an ICU admission. ¹ In addition, LOS in an ICU is independently related to in-hospital mortality. ^{12 13} Our findings are also consistent with previous pediatric studies examining the effect of thrombocytopenia. Krishnan et al ¹⁴ reported that the development of thrombocytopenia is predictive of an extended PICU LOS. Similarly, Agrawal et al ¹⁵ found that pediatric patients requiring cardiopulmonary resuscitation or with evidence of more severe disease such as presence of circulatory shock, coagulopathy, or sepsis have higher risk of developing thrombocytopenia. ¹⁵ Yilmaz et al ¹⁶ showed that thrombocytopenia could be related to mortality and act as an indicator of poor prognosis in the PICU. Lastly, Kaur et al ¹⁷ concluded that thrombocytopenia is commonly associated with sepsis and linked to a higher mortality rate in comparison to nonthrombocytopenic patients. Therefore, thrombocytopenia-associated risk factors should be closely monitored by physicians of critically ill children due to its association with poor prognosis, prolonged length of stay, and mortality.

As days in an intensive care unit are substantially more costly than those associated with general ward care due to increased resource utilization and labor intensity, the occurrence of adverse events occurring during hospitalization add additional burden to the health care system. Among adverse events, new onset conditions such as hospital-acquired infections, add to an escalation in use of hospital resources, and costs associated with necessary hospital treatment of such adverse events. As a result, reduction in LOS would likely reduce hospital costs. ^{18 19}

This study showed that all the deceased patients were thrombocytopenic, required hemodynamic support and were mechanically ventilated. Among the patients that survived, more than half (58.8%) were mechanically ventilated. Kaier et al ¹⁸ showed that mechanical ventilation is associated with a substantial increase in the daily costs of ICU care. Therefore, thrombocytopenia itself, and in association with mechanical ventilation, is a possible prognostic variable for prolonged PICU LOS and costs.

Another interesting finding was a significant association between the type of the bacteria (gram-positive versus gram negative) and the presence of thrombocytopenia. Earlier studies showed conflicting evidence on the association between type of bacteria (based on gram stain) and thrombocytopenia. ^{20 21}

Thrombocytopenia in critically ill patients, although often multifactorial, is often related to the presence BSI raising the possibility of BSI to serve as an early warning sign. The specific mechanisms of linking BSI to thrombocytopenia are not fully understood; however, prior work suggests that they may include disseminated intravascular coagulation, immune-mediated platelet destruction, and hemophagocytic histiocytosis. ²²

Therefore, given the data as presented above, thrombocytopenia may serve as a marker for unfavorable outcomes such as increased PICU LOS, PICU mortality, and associated hospital costs. However, we advise its use as complementary to the clinical course and not as an independent marker or treatment goal per se. Nevertheless, thrombocytopenia may assist clinicians in patient management as a platelet count is a simple routine measure that acts as a readily available tool for evaluating prognosis in critically ill patients with BSI. Since patients with sepsis-related thrombocytopenia are more critically ill than nonthrombocytopenic patients, the platelet count should be part of the early and ongoing evaluation of the severity of illness in septic patients. Furthermore, the platelet count is commonly utilized in scoring systems, acting as a component of early warning scoring systems used for adult and pediatric patients alike. ²³ The assessment of platelet count is inexpensive, easy to perform, and a common routine practice that facilitates its role as an early warning device to alert physicians as to the presence of thrombocytopenia among patients with sepsis. ²⁴ Nevertheless, the exact pathogenesis of thrombocytopenia is still poorly understood and warrants further research. Ideally, a prospective multicenter study would able to stratify the occurrence of thrombocytopenia in pediatric sepsis and to identify risk factors in more detail.

Conclusion

In conclusion, thrombocytopenia is common in critically ill pediatric patients who present with BSI and is significantly associated with an increase in the PICU LOS. Thrombocytopenia may serve as a marker of mortality and morbidity risk in the PICU.

A platelet count, which is a routine measure used in PICU, may help clinicians identify potential “high-risk” patients. Further studies are needed to better define the precise influence of BSI and thrombocytopenia on morbidity and mortality among the critically ill patients in the PICU.

Limitations

This is a historical cohort and data were extracted retrospectively; therefore, severity scores were not part of the recorded data and mortality data for the equivalent patients without BSI is not available. Another limitation of the study is relatively low number of included patients. Finally, data regarding the platelet indexes were not available.

Funding Statement

Funding None.