Accurate prognostic indicators for patients’ survival in intensive care unit (ICU) are important and helpful to guide early decision-making. The Acute Physiological and Chronic Health Evaluation (APACHE) II score is a disease severity classification system designed to measure the severity of disease for adult patients admitted to the ICU (1). The Sequential Organ Failure Assessment (SOFA) score indicates organ dysfunction and is also widely used as a severity measure for critically ill patients (2, 3). However, rapid calculation of these scores is both difficult and complicated because when patients arrive in the ICU, they have unstable vital signs and need urgent management. Therefore, we designed a more rapid, simple new risk scoring system using the delta neutrophil index (DNI) and thrombotic microangiopathy (TMA) score which are easily obtained from the complete blood cell count by an automated blood cell analyzer.
The DNI reflects the fraction of circulating immature granulocytes. This index identified by an automated blood cell analyzer as the difference between the leukocyte subfraction (determined by the cytochemical myeloperoxidase reaction) and the leukocyte subfraction (determined using nuclear lobularity assays by the reflected light beam). The DNI is significantly associated with the diagnosis of true bacteremia, and the severity and prognosis of sepsis (4–8).
The TMA score has developed to detect thrombocytopenia-associated multiple organ failure (TAMOF) in critically ill patients, associated with a variety of diseases, including sepsis, hemolytic uremic syndrome, thrombotic thrombocytopenic purpura, autoimmune disease, malignant hypertension, vascular rejection, and endothelial damage due to drug toxicity (9, 10). This score is composed of the following: red cell distribution width (RDW), hemoglobin distribution width (HDW), microcytes, hyperchromic red cells, and platelet count. Yoo et al. (11) reported that early detection of TAMOF by red blood cell (RBC) parameters and a volume/hemoglobin concentration (V/HC) cytogram with an automated hematology analyzer improves clinical outcome of TAMOF in critically ill patients.
In this study, we evaluated a new risk scoring system using DNI and TMA scores, to determine whether this would be a simpler, more rapid means to predict mortality in critically ill patients.
MATERIALS AND METHODS
Study design and patients
This was a retrospective study performed in the medical ICU at Yonsei University College of Medicine from June 2015 to February 2016. A total of 291 patients were admitted to our medical ICU during this period. We excluded two patients without DNI or TMA score measurements. After these exclusions, 289 patients were enrolled in the study. The patients were divided into two sets: the training set (n = 232), for the development of the new risk scoring model and evaluation of the predictive accuracy of the scoring system, and the test set (n = 57), for internal validation (Fig. 1). Training set and test set were constructed by random sampling with 8:2 hold out validation method. Demographic, clinical, and biochemical data at the time of ICU admission were recorded.
The protocol was approved by Severance hospital institutional review board (IRB No. 4-2016-0951). The need for informed consent was waived given the retrospective nature of the study.
We collected data on demographic indicators, preexisting comorbidities, hemodynamic factors, other clinical information (including the presence of acute renal failure, acute respiratory distress syndrome, bacteremia, primary focus of infection, and the use of mechanical ventilation), and clinical outcomes (28-day all-cause mortality).
The laboratory tests were performed within 24 h of ICU admission. To assess the severity of illness and prognosis of the patients, the APACHE II score and SOFA score were evaluated on admission to the ICU.
Assessment of DNI and TMA score
DNI was assessed by an automated blood cell analyzer (ADVIA 120; Siemens, Forchheim, Germany). This hematologic analyzer is flow cytometry-based and analyzes white blood cell by both a myeloperoxidase channel and a lobularity/nuclear density channel. The formula for calculating DNI is as follows: DNI = (neutrophil subfraction + eosinophil subfraction measured in the myeloperoxidase channel) – (polymorphonuclear subfraction measured in the nuclear lobularity channel) (4, 5).
The TMA score was assigned a point value from 0 to 5, with one point for each of the following: RDW more than 15%, HDW more than 3.2%, microcytes more than 0.4%, hyperchromic red cells more than 1.9%, platelet count less than 140 × 109/L. All these RBC parameters were measured by an automated blood cell analyzer (ADVIA 120; Siemens, Forchheim, Germany). The percent microcytes and percent hyperchromic red cells were calculated from the V/HC cytogram. The percent microcytes indicates the percent of RBCs smaller than 60 fL, and the percent hyperchromic red cells indicates the percent of RBCs with more than 41 g/dL of hemoglobin (11). The TMA score was automatically calculated and reported.
Categorical variables are presented as numbers and percentages; continuous variables are presented as the median and interquartile range (IQR). Baseline characteristics of the groups were compared using Student t test or the Mann–Whitney U test for continuous variables and the chi-square test or Fisher exact test for categorical variables. We evaluated 28-day all-cause mortality as the study endpoint.
In the training set, a multivariate Cox proportional-hazards model was used to determine the contribution of variables, including age, sex, DNI, and TMA score. We included age and sex to make a simple formula that is easy to apply, although there are no statistical significances in univariate analysis. The predictive accuracy of this new scoring model (age + sex + DNI + TMA score) was examined by calculating the Harrell's C index. Using a bootstrap method, we compared this new scoring model with the APACHE II score and SOFA score. Finally, to develop a new practical prognostic score (which we designated the “simplified mortality score” [SMS]), we assigned to each risk factor a weighted point proportional to its β-coefficient (rounded to the nearest integer). An SMS was then calculated for each patient and the Contal and O’Quigely's method was used to find the cutoff value of SMS in patients at low, intermediate, and high risk of death. The population was divided into three categories: patients at low, intermediate, and high risk of death. Survival was estimated using the Kaplan–Meier method, and differences in survival between groups were assessed using the log-rank test. For internal validation, Harrell's C index and the bootstrap method for comparison between SMS, APACHE II, and SOFA score were used in the test set. A two-sided P <0.05 was considered to indicate statistical significance. Statistical analyses were performed using R statistical software, version 3.2.4 (the R Foundation for Statistical Computing, Vienna, Austria)
Baseline characteristics of the study population
We enrolled 289 patients admitted to our medical ICU: 178 (61.2%) men and 111 (38.1%) women. The mean age was 66.1 years (range, 19–92 years). The 28-day mortality was 31.1% (n = 90); these patients were classified as nonsurvivors. The other patients were defined as survivors. Table 1 shows the comparison of demographic, clinical, and laboratory parameters between nonsurvivors and survivors. The nonsurvivors had higher APACHE II (P < 0.001) and SOFA (P < 0.001) scores, and a higher prevalence of liver disease, chronic kidney disease, and cancer than survivors. Furthermore, nonsurvivors had a higher prevalence of acute renal failure and a higher rate of blood culture positivity than survivors; however, the primary focus of infection did not differ. Nonsurvivors had a higher heart rate (P = 0.006), but a lower median platelet count (P < 0.001), serum albumin level (P = 0.002), and PaO2/FiO2 ratio (P = 0.005) than survivors. The DNI (P = 0.005) and TMA (P = 0.034) scores of nonsurvivors were significantly higher than those of survivors, but no significant differences were observed in other blood biomarkers, such as procalcitonin levels and C-reactive protein (CRP) levels (Table 1).
New risk scoring model development and comparison for training set
To develop a new risk scoring model, we performed a multivariate Cox proportional-hazards analysis with data from the training set (Table 2). Variables included in this analysis were DNI, TMA score, age, and sex. DNI and TMA score were found to be of prognostic significance by univariate analysis, and age and sex are basic clinical parameters. We also performed a multivariate Cox proportional-hazards analysis with APACHE II and SOFA scores for comparison with the new risk-scoring model. The multivariate Cox proportional hazards analysis revealed that increased DNI (hazard ratio [HR], 1.031; 95% confidence interval [CI], 1.015–1.047; P < 0.001), APACHE II score (HR, 1.076; 95% CI, 1.047–1.107; P < 0.001), and SOFA score (HR, 1.219; 95% CI, 1.147–1.295; P < 0.001) were independent risk factors for 28-day mortality. In contrast, age, sex, and TMA score were not (Table 2). To evaluate predictive accuracy, we calculated the Harrell's C index of the new risk scoring model, APACHE II score, and SOFA score. Harrell's C index of the new risk scoring model (0.660) was slightly inferior to SOFA score (0.720) but not inferior to the APACHE II score (0.664). Moreover, comparisons between the new risk scoring model and APACHE II score (Harrell's C index difference, 0.031; 95% CI, −0.045 to 0.117) or SOFA score (Harrell's C index difference, −0.025; 95% CI, −0.095 to 0.045) were performed and did not show significant differences. Therefore, we determined that it would be reasonable to develop a new practical prognostic score, named the “simplified mortality score” (SMS), using this new risk scoring model.
To develop the SMS, we assigned each prognostic variable a weighted point proportional to its β-coefficient value, rounded to the nearest integer (Table 2). Finally, SMS was defined by the following formula: [age + 11 if a male patient + (2 × DNI) + (61 if TMA score is 1, 76 if TMA score is 2, 74 if TMA score is 3, 26 if TMA score is 4, 99 if TMA score is 5)]. An SMS was then calculated for each patient; these values ranged from 63 to 284. The nonsurvivors had a higher median SMS than survivors (167.7 [IQR: 144.6–187.5] vs. 153.0 [IQR: 127.2–170.4]; P < 0.001), and the Harrell's C index for the SMS was 0.660. The populations were divided into three subgroups on the basis of the SMS: a low risk group (0–120), an intermediate-risk group (121–180), and a high-risk group (≥181). Kaplan–Meier curves were generated for 28-day mortality based on these risk groups. The log-rank test results indicated that the SMS was an independent prognostic marke8r in critically ill patients (Fig. 2).
Internal validation using the test set
Data from the test set patients (n = 57) were analyzed for internal validation. Harrell's C index was calculated, and comparisons between the new risk scoring model and the APACHE II score, SOFA score, or SMS were performed (Table 3). In the training set, Harrell's C index for the new risk scoring model was 0.581. That was inferior to that of the SOFA score (0.676), but not inferior to that of the APACHE II score (0.564), similar to the results of the training set. Harrell's C index for the SMS was 0.579. Comparisons between the new risk scoring model and the APACHE II score (Harrell's C index difference, 0.037; 95% CI, −0.371 to 0.463), SOFA score (Harrell's C index difference, −0.047; 95% CI, −0.462 to 0.403), or SMS (Harrell's C index difference, 0.001; 95% CI, −0.078 to 0.014) were performed and did not show significant differences (Table 3). The Harrell's C index of the new risk scoring model and SMS was lower than those obtained in the training set. However, this tendency was observed for the APACHE II and SOFA scores, too. Therefore, the results were reliable for internal validation.
Severe sepsis and septic shock are the most important causes of death among patients in ICU, and high morbidity and mortality from severe sepsis and septic shock remain unresolved problems (12, 13). With infection or systemic inflammation, the number of immature granulocytes increase in the peripheral circulation. However, measurement of immature granulocytes is difficult in clinical practice because manual counting is time consuming, requires a trained clinical pathologist, and is not accurate (6, 7). Recently, DNI was developed as an indicator of immature granulocytes; it is calculated by an automated blood cell analyzer (4). Because the complete blood count is routinely evaluated in patients admitted to the ICU, DNI can be easily obtained. In many previous studies, DNI has been reported to be associated with the severity of sepsis, sepsis-related mortality, true bacteremia, disseminated intravascular coagulation, and the differential diagnosis of infection with noninfectious critical illness (4, 5, 8, 14–17).
Furthermore, some recent studies were conducted for evaluation of predictive accuracy of new predictive markers made by combination of DNI with other biomarkers in critically ill patients. Kim et al. (18) reported that the combination of DNI and procalcitonin (the sum of these two markers) had higher area under the curve than that of each biomarker alone (such as DNI, procalcitonin, and CRP) to predict the severity of sepsis and survival. Hwang et al. (19) reported that the DNI/albumin ratio on day 1 defined as the DNI divided by the albumin level on admission to the emergency department, and on the peak day were associated with 28-day mortality in patients with severe sepsis receiving early goal directed therapy. In the present study, we designed a novel prognostic scoring system by combining DNI with the TMA score.
In our hospital, the TMA score is automatically calculated and reported by an automated hematology analyzer when conducting a complete blood cell count. No additional laboratory tests are needed. Hence, it is a very simple, fast marker, as is the DNI. The TMA score is designed for early and easy detection of organ failure based on the concept of TAMOF. TAMOF results from microvascular endothelial cell injury with thrombocytopenia, hemolytic anemia, thrombosis, and tissue ischemia (20). An increased percentage of microcytic hyperchromic cells with anisocytosis and anisochromia indicate the presence of schistocytes; a good automated screening marker for TAMOF. Using these characteristics of schistocytes, Yoo et al. (11) reported that they can establish an early diagnosis of TAMOF using RBC parameters and a V/HC cytogram with an automated hematology analyzer. In critically ill patients, transplantation, cardiopulmonary bypass, autoimmune disease, infection, cancer, and exposure to radiation and medications can result in systemic endothelial injury (11). In previous studies, patients with TAMOF were demonstrated to have an unfavorable prognosis and high mortality in the absence of early and appropriate treatment (21–23). Therefore, early detection of TAMOF is very important for patients admitted to ICU, to institute early treatment.
To the best of our knowledge, this is the first study to evaluate a new risk scoring system using DNI, TMA score, and basic clinical parameters (age and sex) as a prognostic marker for critically ill patients. This SMS could provide faster information and a simple means to predict mortality. This is the most important strength of the SMS. Furthermore, the predictive accuracy of the SMS is acceptable and is not inferior to APACHE II score. The SMS assigned risk appropriately compared with the APACHE II or SOFA score.
Our study has several limitations. First, it was a small, single-center, retrospective cohort study, making the findings difficult to generalize. Although we assessed internal validity of the SMS, external validation in an independent data sample and on a larger scale by means of multicenter, prospective validation studies, is required. Second, the HR did not represent a linear increase according to TMA score in the multivariate Cox proportional-hazards analysis (Table 2). We suggest that these results are due to the small population sample. Further investigation of a large population sample is desirable. Finally, the TMA score is not currently widely known or used; hence, the cutoff value of each category was decided in our institution. Further validation of the TMA score is needed.
We developed a new risk scoring system (SMS) using DNI and TMA score, both of which are easily obtained from complete blood cell count, and age and sex, which are basic clinical parameters. SMS is a more rapid, simple prognostic score to predict 28-day mortality and to stratify risk group than either the APACHE II or SOFA scores. However, an external validation study, using a larger sample, is needed in the near future.
We thank all the patients who participated in this study, and also thank all the healthcare professionals at Severance Hospital, especially those from the Department of Pulmonology and the ICU.
1. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II
: a severity of disease classification system. Crit Care Med
2. Peres Bota D, Melot C, Lopes Ferreira F, Nguyen Ba V, Vincent JL. The Multiple Organ Dysfunction Score (MODS) versus the Sequential Organ Failure Assessment
) score in outcome prediction. Intensive Care Med
3. Minne L, Abu-Hanna A, de Jonge E. Evaluation of SOFA
-based models for predicting mortality
in the ICU
: A systematic review. Crit Care
4. Nahm CH, Choi JW, Lee J. Delta neutrophil index
in automated immature granulocyte counts for assessing disease severity of patients with sepsis. Ann Clin Lab Sci
5. Park BH, Kang YA, Park MS, Jung WJ, Lee SH, Lee SK, Kim SY, Kim SK, Chang J, Jung JY, et al. Delta neutrophil index
as an early marker of disease severity in critically ill
patients with sepsis. BMC Infect Dis
6. Kim HW, Ku S, Jeong SJ, Jin SJ, Han SH, Choi JY, Kim JM, Song YG. Delta neutrophil index
: could it predict mortality
in patients with bacteraemia? Scand J Infect Dis
7. Seok Y, Choi JR, Kim J, Kim YK, Lee J, Song J, Kim SJ, Lee KA. Delta neutrophil index
: a promising diagnostic and prognostic marker for sepsis. Shock
8. Lee CH, Kim J, Park Y, Park YC, Kim Y, Yoon KJ, Uh Y, Lee KA. Delta neutrophil index
discriminates true bacteremia from blood culture contamination. Clin Chim Acta
9. Benz K, Amann K. Pathological aspects of membranoproliferative glomerulonephritis (MPGN) and haemolytic uraemic syndrome (HUS)/thrombocytic thrombopenic purpura (TTP). Thromb Haemost
10. Zipfel PF. Thrombotic microangiopathies as a prime example for translational medicine. Thromb Haemost
11. Yoo JH, Lee J, Roh KH, Kim HO, Song JW, Choi JR, Kim YK, Lee KA. Rapid identification of thrombocytopenia-associated multiple organ failure
using red blood cell
parameters and a volume/hemoglobin concentration
cytogram. Yonsei Med J
12. Angus DC, Linde-Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med
13. Martin GS, Mannino DM, Eaton S, Moss M. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med
14. Park HJ, Ha YJ, Pyo JY, Park YB, Lee SK, Lee SW. Delta neutrophil index
as an early marker for differential diagnosis of adult-onset Still's disease and sepsis. Yonsei Med J
15. Shin DH, Kim EJ, Kim SJ, Park JY, Oh J. Delta neutrophil index
as a marker for differential diagnosis between acute graft pyelonephritis and acute graft rejection. PLoS One
16. Lim TS, Kim BK, Lee JW, Lee YK, Chang S, Kim SU, Kim DY, Ahn SH, Han KH, Chon CY, et al. Use of the delta neutrophil index
as a prognostic factor of mortality
in patients with spontaneous bacterial peritonitis: implications of a simple and useful marker. PLoS One
17. Kim HW, Yoon JH, Jin SJ, Kim SB, Ku NS, Jeong SJ, Han SH, Choi JY, Kim JM, Song YG. Delta neutrophil index
as a prognostic marker of early mortality
in gram negative bacteremia. Infect Chemother
18. Kim H, Kim Y, Lee HK, Kim KH, Yeo CD. Comparison of the delta neutrophil index
with procalcitonin and C-reactive protein
in sepsis. Clin Lab
19. Hwang YJ, Chung SP, Park YS, Chung HS, Lee HS, Park JW, Lee JW, Hong JH, You JS, Park I. Newly designed delta neutrophil index
-to-serum albumin ratio prognosis
of early mortality
in severe sepsis. Am J Emerg Med
20. George JN, Nester CM. Syndromes of thrombotic microangiopathy
. N Engl J Med
21. Moake JL. Thrombotic microangiopathies. N Engl J Med
22. Carcillo JA, Kellum JA. Is there a role for plasmapheresis/plasma exchange therapy in septic shock, MODS, and thrombocytopenia-associated multiple organ failure
? We still do not know—but perhaps we are closer. Intensive Care Med
23. Sadler JE, Moake JL, Miyata T, George JN. Recent advances in thrombotic thrombocytopenic purpura. Hematology Am Soc Hematol Educ Program