Introduction

Sepsis with septic shock is the major cause of death in hospitals and a major reason for admission to the intensive care unit (ICU; Evans et al., 2021). According to the World Health Organization, the global incidence of sepsis is 189 per 10,000 patients, with approximately 20% of patients with sepsis dying and more than half of patients with sepsis admitted directly to the ICU (World Health Organization, 2020). In the United States, medical expenses related to sepsis management total approximately 16–25 billion USD annually (Rodríguez et al., 2021). As patients with sepsis often encounter issues related to aging and chronic diseases, their medical expenses and risk of mortality increase dramatically with length of stay (Fathi et al., 2019). Furthermore, mortality and ICU stay ≥ 3 days are both recognized as ICU adverse events (Evans et al., 2021) and may be used as clinical criteria for identifying patients with sepsis. The early identification and evaluation of risks of early interventions in patients with sepsis have become important.

Sepsis is characterized by highly variable and nonspecific signs and symptoms, making its diagnosis and evaluation of severity very complicated (Villegas & Moore, 2018). The mortality rate for sepsis and septic shock in surgical patients is approximately 36% (Vincent et al., 2019), which is higher than the 12%–18% rate observed in other medical patients (Fernando et al., 2018). Different from other patients, monocyte dysfunction, which puts patients at a high risk of death and secondary infections, has been found to be present in patients with sepsis during the first 3–5 days after surgery (Baëhl et al., 2016). Intraoperative blood loss volume also affects T-lymphocyte proliferation (Albertsmeier et al., 2015). Moreover, degree of T-lymphocyte proliferation severity is positively associated with the complexity of the surgical procedure performed, as patients undergoing surgery cannot effectively defend themselves against bacteria and endotoxins and may develop organ failure or death (Albertsmeier et al., 2015, 2017). Furthermore, broad-spectrum antibiotics are commonly prescribed to treat or prevent infections in surgical patients, which further complicates the identification, assessment, and monitoring of septic-related changes (Martínez et al., 2020). Sepsis assessment tools specific to surgical patients have been rarely reported in the literature (Posadas-Calleja et al., 2018). However, early identification and monitoring of septic condition severity in surgical patients are imperative to initiate timely treatment and reduce mortality (Kim & Park, 2019).

Assessment Tools

Over the past two decades, several assessment tools for sepsis have been developed that quantify sepsis severity and predict mortality risk and prognosis as an aid in allocating medical resources (Evans et al., 2021). However, these tools are lengthy and difficult to apply, making them difficult to use regularly and repeatedly in clinical practice. The most commonly used tools with reasonable numbers of assessment items include the sequential organ failure assessment (SOFA); quick SOFA (qSOFA); and predisposition, infection/injury, response, and organ dysfunction (PIRO).

SOFA, developed by the European Society of Intensive Care Medicine, is a widely used tool consisting of six systemic variables: ratio of partial pressure of oxygen to fraction of inspiration O2 (PaO2/FiO2), mean arterial pressure, administration of vasopressors, bilirubin, platelets, creatinine, and Glasgow Coma Scale (GCS; Vincent et al., 1996). Higher scores indicate a higher risk of mortality. However, this tool does not consider host factors such as age and comorbidities, which are also known to significantly affect mortality in patients with sepsis (Evans et al., 2021).

qSOFA was developed in 2016 using only three criteria: GCS, blood pressure, and respiration rate. Although qSOFA does not require laboratory data and may be used quickly and repeatedly (Finkelsztein et al., 2017), its sensitivity in predicting mortality has been found to be lower than that of the other assessment tools (Finkelsztein et al., 2017). In addition, the feasibility of using this tool with patients undergoing surgery with multiple organ failure has yet to be confirmed.

PIRO, developed by Levy et al. (2003), consists of four dimensions. Because of its theoretical nature, criteria weightings may differ for different patient groups. Thus, it is not an assessment tool standardized for clinical practice. PIRO has been applied to patients undergoing surgery with intra-abdominal sepsis. Posadas-Calleja et al. (2018) found that eight factors, including age, comorbid condition, white blood count, body temperature (BT), systolic blood pressure, GCS, creatinine, and PaO2/FiO2, were significantly related to mortality. PIRO has been shown to be a good predictor of mortality and severity in surgical patients, with results suggesting this tool may be applicable to other surgical intensive patients.

Several recent studies have shown that, in the host defense mechanism, BT is not a good predictor of mortality in infectious patients (Schuttevaer et al., 2019), as BT in patients with infections is often within the normal range. Thus, normal BT may lead to the mis-assessment of disease severity, delaying the initiation of antibiotics and increasing mortality risk (Inghammar & Sunden-Cullberg, 2020). According to O'Grady et al. (2008), patients with infections, normal BT, and positive blood culture tend to experience higher levels of disease severity. The cutoff BT is 36°C in current assessment tools. Although hypothermia is often diagnosed in the late phase of sepsis (Ostadi et al., 2019), it is not a valid indicator for the early detection of sepsis. Furthermore, it is difficult to determine the actual influence of antipyretics and antibiotics on BT (Schuttevaer et al., 2022). Hence, using BT as a predictor of mortality in patients with infections should be done cautiously.

SOFA and Acute Physiology and Chronic Health Evaluation II both use thrombocytopenia as an important parameter for predicting mortality in critical patients. Thrombocytopenia is often defined as a platelet count of < 150 × 103/μl (Claushuis et al., 2016). The incidence of thrombocytopenia in patients with sepsis is approximately 70% and is even higher in patients requiring ICU treatment (Giustozzi et al., 2021). Platelets are inflammatory and coagulation elements that function in hemostasis and immune mechanisms (Giustozzi et al., 2021). The large quantities of inflammatory and bioactive molecules released by platelets help control and modulate the immune cells (Ribeiro et al., 2019). During sepsis, platelets inhibit inflammation and promote tissue repair in a receptor- and organ-dependent manner (Ribeiro et al., 2019). Moreover, patients with thrombocytopenia have a higher risk of surgical bleeding and organ dysfunction (Assinger et al., 2019), which leads to higher risks of hospital mortality and increased length of stay (Lyons et al., 2018). Detecting thrombocytopenia in a timely manner and initiating treatment early can effectively improve the rate of survival in patients with sepsis (Giustozzi et al., 2021).

In light of the above, this study was designed to create a modified Posadas-Calleja PIRO (mPIRO) that (a) includes a count of platelets and (b) eliminates consideration of BT. Furthermore, because adverse events have rarely been assessed in previous studies and are now considered a key outcome in patients with sepsis (Evans et al., 2021), the predictive accuracy of mPIRO in terms of mortality and adverse events in a sample of patients with surgical sepsis was assessed and compared against the results obtained by SOFA, qSOFA, and PIRO.

Methods Study Design and Setting

This retrospective, observational study was conducted using a review of the electronic medical records of a surgical ICU (SICU) in southern Taiwan from January 2016 to December 2017. The inclusion criteria were as follows: age > 20 years, severe sepsis or septic shock diagnosis, surgery, and SICU admission. Patients who had been diagnosed with secondary infection or readmitted to the SICU were excluded. Patients who had stayed in the SICU for < 8 hours were also excluded because detecting early changes in either their physiological response or organ dysfunction would be difficult. Patients with altered consciousness before admission were excluded because it would be difficult to identify the impact of sepsis. Of the 2,055 medical records reviewed, 103 met the inclusion criteria. The study design flow is shown in Figure 1.

Figure 1:

Enrollment Process Flowchart

Measurements

A three-part checklist for data collection was created, including demographic variables, assessment tools, and outcome indicators. Demographic data included age, gender, diagnosis at ICU admission, and source of infection. The parameters of SOFA, qSOFA, PIRO, and mPIRO and their respective scoring methods are depicted in Table 1. Outcome indicators included ICU length of stay, mortality, and adverse events. Adverse events were defined as ICU death, a stay greater than 3 days, or both and were identified by researchers based on length-of-stay and mortality data.

Table 1 - Comparison of Assessment Tool Scoring Systems Assessment Tool Score SOFA PIRO mPIRO qSOFA PaO2/FiO2 (mmHg)  ≥ 400 0 0 (≥ 400) 0 (≥ 400) 0 (RR < 22)  < 400 1 1 (< 400) 1 (< 400) 1 (RR ≥ 22)  < 300 2  < 200 3  < 100 4 Mean arterial pressure (MAP; mmHg)  No hypotension 0 0 (≥ 70) 0 (≥ 70) 0 (SBP > 100)  MAP < 70 1 1 (< 70) 1 (< 70) 1 (SBP ≤ 100)  Dopamine ≤ 5 (μg/kg/min) or dobutamine of any dose 2  Dopamine > 5 or norepinephrine ≤ 0.1 3  Dopamine > 15 or norepinephrine > 0.1 4 Bilirubin (mg/dl)  < 1.2 0  1.2–1.9 1  2.0–5.9 2  6.0–11.9 3  ≥ 12 4 Platelet (×103/μl)  ≥ 150 0 0 (≥ 150)  < 150 1 1 (< 150)  < 100 2  < 50 3  < 20 4 Creatinine (mg/dl)  < 1.2 0 0 0  1.2–1.9 1 1 (≥ 1.2) 1 (≥ 1.2)  2.0–3.4 2  3.5–4.9 3  ≥ 5.0 4 Glasgow Coma Scale  15 0 0 0 0  13–14 1 1 (< 15) 1 (< 15) 1 (< 15)  10–12 2  6–9 3  < 6 4 Age (years)  ≤ 65 0 0  > 65 1 1 Comorbid conditions  None 0 0  Chronic health points 1 1 WBC count (/μl)  ≥ 4,000 0 0  < 4,000 1 1 Body temperature (°C)  ≥ 36 0  < 36 1 Total score 0–24 0–8 0–8 0–3

Note. SOFA = sequential organ failure assessment; PIRO = predisposition, infection/injury, response, and organ dysfunction; mPIRO = modified PIRO; qSOFA = quick SOFA; PaO2 = partial pressure of oxygen; FiO2 = fraction of inspiration oxygen; RR = respiration rate; SBP = systolic blood pressure; WBC = white blood count.

Data Collection Procedures

The protocol for this study was approved by the institutional review board of the affiliated institution (KMUHIRB-E[I]-20190020). Electronic medical records were retrieved from the electronic database and hospital information system in the medical center. All of the parameters with extreme values within the first 8 hours of ICU admission were collected and recorded in an Excel file.

Statistical Analysis

Baseline characteristic data were analyzed using descriptive statistics. We used the area under the receiver operating characteristic curve (AUC) to determine the discrimination among the scores. Discrimination is the ability of a score to distinguish between survivors and nonsurvivors. Pairwise comparisons between the AUC of each score in the one-tailed t test were performed based on the method described by Hanley and McNeil (1983). The optimal cutoff point was decided for each assessment tool based on Youden's index. The values of sensitivity, specificity, positive likelihood ratio (LR+), and negative likelihood ratio (LR−) were also calculated based on the optimal cutoff point in predicting mortality and adverse events in patients with surgical sepsis. IBM SPSS Statistics 22.0 (IBM Inc., Armonk, NY, USA) was used for all statistical analyses, and significance was set at p < .05.

Results

The characteristics and outcomes are presented in Table 2. Most were men (64.1%), and the mean age was 66.56 (SD = 12.87) years. Forty-eight (46.6%) patients were diagnosed with sepsis, and 55 (53.4%) had septic shock. The most frequent infection site was the gastrointestinal tract (47.6%), followed by the skin or soft tissue (22.3%). No significant differences were found between survivors and nonsurvivors in terms of age (p = .333), gender (p = .431), diagnosis (p = .063), or infection source (p = .439). The means and standard deviations of the scores for SOFA, qSOFA, PIRO, and mPIRO were 6.7 (4.57), 1.7 (1.03), 3.21 (1.61), and 3.26 (1.74), respectively. The scores of the four assessment tools for predicting mortality all differed significantly for survivors and nonsurvivors (p < .001). The overall mortality rate in the ICU was 20.4%, and the rate of adverse event occurrence (mortality or length of stay in the ICU ≥ 3 days) was 85.4%.

Table 2 - Participant Characteristics and Outcomes (N = 103) Variable All (N = 103) Survivors (n = 82) Nonsurvivors (n = 21) p n % n % n % Age (years; M and SD) 66.56 12.87 65.94 12.84 69 12.96 .333 Gender .431  Male 66 64.1 51 62.2 15 71.4  Female 37 35.9 31 37.8 6 28.6 Diagnosis .063  Sepsis 48 46.6 42 51.2 6 28.6  Septic shock 55 53.4 40 48.8 15 71.4 Disease severity (M and SD)  SOFA 6.7 4.57 6.16 4.2 10.95 3.85 < .001  qSOFA 1.7 1.03 1.51 0.97 2.43 0.93 < .001  PIRO 3.21 1.61 2.85 1.52 4.62 1.12 < .001  mPIRO 3.26 1.74 2.87 1.63 4.81 1.21 < .001 Source of infection .439  Respiratory tract 7 6.8 6 7.3 1 4.8  Gastrointestinal tract 49 47.6 36 43.9 13 61.9  Hepatobiliary tract 4 3.9 4 4.9 0 0.0  Urinary tract 6 5.8 6 7.3 0 0.0  Skin or soft tissue 23 22.3 20 24.4 3 14.3  Catheter-related bloodstream infection 2 1.9 1 1.2 1 4.8  ≥ 2 infection sources 12 11.7 9 11.0 3 14.3 Outcome  ICU mortality 21 20.4 0 0 21 100 < .001  ICU LOS (days; M and SD) 12.38 11.97 10.0 9.1 21.62 16.73 .005  ICU LOS ≥ 3 days 85 82.5 67 81.7 18 85.7 .666  Sepsis adverse events 88 85.4 67 81.7 21 100.0 .034

Note. SOFA = sequential organ failure assessment score; qSOFA = quick SOFA; PIRO = predisposition, infection/injury, response, and organ dysfunction; mPIRO = modified PIRO; ≥2 infection sources = patients who suspect or show two or above two infection sources; ICU = intensive care unit; LOS = length of stay.

In this study, the optimal cutoff points for predicting mortality in mPIRO, SOFA, qSOFA, and PIRO were 4, 6, 3, and 4, respectively. The optimal cutoff points for predicting adverse events in mPIRO, SOFA, qSOFA, and PIRO were 3, 4, 2, and 4, respectively. On the basis of the optimal cutoff point, the LR+ and LR− were calculated for ruling in and ruling out mortality and adverse events (Table 3). Of these four assessment tools, mPIRO was the most reliable in terms of predictive results.

Table 3 - Sensitivity and Specificity at Different Thresholds for Predicting Mortality and Adverse Event in Four Assessment Tools for Patients With Surgical Sepsis Threshold Mortality SEN SPE PPV NPV LR+ LR− YI Death
(n = 21) Survivors
(n = 82) Total SOFA ≥ 6 20 (a) 37 (c) 57 0.95 0.55 0.35 0.98 2.11 0.09 0.50 SOFA < 6 1 (b) 45 (d) 46 qSOFA ≥ 3 14 (a) 15 (c) 29 0.67 0.82 0.48 0.91 3.64 0.41 0.48 qSOFA < 3 7 (b) 67 (d) 74 PIRO ≥ 4 17 (a) 26 (c) 43 0.81 0.68 0.40 0.93 2.55 0.28 0.49 PIRO < 4 4 (b) 56 (d) 60 mPIRO ≥ 4 19 (a) 23 (c) 42 0.91 0.72 0.45 0.97 3.23 0.13 0.62 mPIRO < 4 2 (b) 59 (d) 61 Threshold Adverse Event SEN SPE PPV NPV LR+ LR− YI Yes
(n = 88)