Quantitative Assessment of Significant Hematological Parameters Deviation from Healthy State for Sepsis Clinical Monitoring

Summary

This study proposes a novel quantitative metric based on the Euclidean distance of three significant hematological parameters (Hb, Lymph%, PDW) from a healthy state to assess sepsis severity. Results demonstrate a strong correlation with SOFA scores, offering a simple, objective tool for monitoring sepsis using routine blood tests.

Abstract

Sepsis, a life-threatening condition, requires improved diagnostics and severity assessment. Based on long-term clinical observation, Hemoglobin (Hb), lymphocyte percentage (Lymph%), and platelet distribution width (PDW) change significantly during sepsis. This study aimed to evaluate these three Complete Blood Count (CBC) parameters for distinguishing sepsis patients from healthy individuals and to develop a novel quantitative metric based on their combined deviation from a healthy state to assess sepsis severity. This retrospective case-control study included 496 sepsis patients (diagnosed according to Sepsis-3 criteria upon admission) and 1136 healthy controls. Initial Hb, Lymph%, and PDW were z-score standardized. The healthy cohort's 3D centroid was calculated and each sepsis patient's Euclidean distance from this centroid was computed. Group differences and correlation with SOFA scores were analyzed. 3D scatter plots visualized these relationships. Hb, Lymph%, and PDW differed significantly between sepsis patients and controls (p < 0.001), with sepsis showing lower Hb/Lymph% and higher PDW. The calculated Euclidean distance from the healthy centroid was significantly greater in sepsis patients (p < 0.001). This distance strongly correlated with SOFA scores (r = 0.45, p < 0.001) and was larger in patients with higher disease severity. 3D visualization confirmed group separation, with color intensity reflecting this distance and correlating with increasing deviation from the healthy centroid. The Euclidean distance of key CBC parameters from a healthy cohort's centroid effectively quantifies deviation from a healthy state and correlates with disease severity. This approach using routine CBC data offers a promising, simple tool for sepsis severity assessment and monitoring.

Introduction

Sepsis, as a systemic inflammatory response syndrome caused by infection, can lead to severe organ dysfunction and remains one of the leading causes of death in intensive care units worldwide^1,2. Despite advances in diagnostic and treatment technologies in recent years, sepsis mortality rates remain high at 20%-40%³. Timely identification and accurate assessment of sepsis severity are crucial for clinical decision-making and improving prognosis⁴.

Currently, sepsis diagnosis and severity assessment primarily rely on scoring systems such as the Sequential Organ Failure Assessment (SOFA) and Acute Physiology and Chronic Health Evaluation II (APACHE II)^5,6. Although these systems are widely applied in clinical practice, they have limitations including complex calculations, time-consuming assessments, requirements for multiple physiological indicators and biochemical test results, and a certain degree of subjectivity and inconsistency in clinical practice^7,8. Implementation of these scoring systems poses significant challenges, especially in primary healthcare institutions and resource-limited areas⁹. Therefore, developing a simple, reliable, and easily implementable sepsis assessment tool has important clinical significance.

Hematological indicators, as the most basic clinical examination items, feature rapid detection, low cost, and wide availability^10,11. Clinical observations have found that multiple hematological indicators in sepsis patients undergo significant changes, with hemoglobin (Hb), lymphocyte percentage (Lymph%), and platelet distribution width (PDW) showing particularly notable alterations^12,13. Decreased hemoglobin reflects sepsis-related anemia, reduced lymphocyte percentage indicates immune function suppression, while increased platelet distribution width is associated with inflammatory response and coagulation dysfunction^14,15. Previous studies^12,13,14,15 have confirmed that these indicators individually correlate with sepsis severity and prognosis to some extent, but there is a lack of systematic methods for integrated assessment of multiple indicators¹⁶.

In recent years, multidimensional data analysis methods have seen increasing applications in disease assessment¹⁷. Euclidean distance, as a geometric measurement method for quantifying differences between samples, can be used to evaluate the degree of deviation of patient indicators from normal reference values. Applying this concept to multidimensional clinical data analysis may provide new insights for disease severity assessment. However, there is currently a lack of quantitative sepsis assessment research based on the multidimensional spatial characteristics of hematological indicators^18,19.

Based on the above background, this study aims to: evaluate the efficacy of hemoglobin, lymphocyte percentage, and platelet distribution width in distinguishing sepsis patients from healthy populations; establish a standardized multidimensional space based on these three indicators and calculate the Euclidean distance of patient samples from the healthy center point; analyze the correlation between this distance and sepsis severity (SOFA score); explore the application value of this new assessment method based on routine hematological examinations in clinical monitoring of sepsis. This method provides a new objective and convenient tool for early identification, severity assessment, and dynamic monitoring of sepsis, particularly suitable for clinical practice in healthcare institutions at various levels¹⁰.

Protocol

This study was a retrospective case-control investigation approved by the Ethics Committee of Sunsimiao Hospital, Beijing University of Chinese Medicine (approval number: SSMYY-KYPJ-2024-065). All participants or their legal representatives signed informed consent forms.

Study design and participants
The study included 496 patients diagnosed with sepsis (according to Sepsis-3 criteria³) and 1136 healthy controls from the health examination center during the period from January 2024 to March 2025. Exclusion criteria for the observation group included: (1) age < 18 years; (2) pregnant women; (3) primary hematological diseases; (4) cancer patients; (5) blood transfusion treatment prior to admission; (6) recent immunosuppressive therapy or immunodeficiency; (7) incomplete clinical data. Daily monitoring included morning blood draws (06:00-08:00) for CBC analysis, with results available within 12 h. SOFA scores were calculated by trained intensivists using standardized criteria. Patients were monitored until clinical resolution, discharge, death, or a 30-day maximum follow-up period.

Laboratory testing and data collection
Demographic data (age, gender, height, weight, etc.) and clinical information were collected for all included subjects. CBC samples were tested using a hematology analyzer, serum samples were tested using a clinical analyzer, and coagulation samples were tested using a hematology analyzer. All blood samples were required to be tested within the specified testing time. All tests were performed by trained laboratory technicians following standard operating procedures. The laboratory regularly conducted internal and external quality control.

For sepsis patients, the main diagnosis, infection site, comorbidities, Sequential Organ Failure Assessment (SOFA) score, Acute Physiology and Chronic Health Evaluation II (APACHE II), and laboratory examination results at admission were recorded. Based on the SOFA score at admission, sepsis patients were divided into three groups: mild (SOFA ≤ 6 points), moderate (7 ≤ SOFA ≤ 12 points), and severe (SOFA ≥ 13 points). For a subset of patients, laboratory examination results and SOFA scores were collected daily during their hospitalization for continuous tracking analysis, with an observation period of 30 days or until the patient was discharged or died. Samples from the healthy control group were collected from health examination center participants confirmed to have no acute or chronic diseases.

Data analysis and visualization
Standardization, processing, and significance analysis
To eliminate dimensional differences between different indicators, Z-score standardization was performed on all sample Hb, Lymph%, and PDW data:

Data_norm, mu, sigma] = zscore([H;P])

where H is the healthy control group data matrix, P is the sepsis patient group data matrix, and mu and sigma are the mean and standard deviation, respectively. After standardization, the data conformed to a standard normal distribution with a mean of 0 and a standard deviation of 1.

Significance visualization plots (Figure 1) were created for the significant indicators Hb, Lymph%, and PDW, which intuitively show the differences between healthy samples and sepsis patients in each indicator dimension. The actual significance of quantitative analysis between indicators is shown in Table 1. Descriptive statistics (mean, standard deviation, median, quartiles, etc.) were calculated, independent sample t-tests were used to compare differences between groups, and Cohen's d effect size was calculated to evaluate the significance of differences.

Healthy center point calculation and Euclidean distance measurement
Based on the standardized data from the healthy control group, the coordinates of its center point in the three-dimensional feature space were calculated. Subsequently, the Euclidean distance from each sepsis patient sample to the healthy center point was calculated to quantify the degree of deviation of the patient's hematological indicators from the healthy state, as an objective assessment indicator of disease severity.

Three-dimensional visualization method
Three-dimensional scatter plots (Figure 2) were created to display the distribution of the two groups of samples in the feature space. They use a color gradient (Jet color scheme) to represent sepsis severity, where dark blue represents mild sepsis, red represents severe sepsis, and the healthy control group is marked with green +.

Clinical validation of long cohort samples
To evaluate the applicability of this method on new samples from long-cohort (followed for 3-6 months post-admission) patients, a predictive validation process was designed. For a typical long cohort sample of sepsis, standardized processing was performed using the previously saved mean (mu) and standard deviation (sigma) from the regularization process, and then the Euclidean distance from the healthy center point was calculated. Based on the calculated distance value, combined with the pre-established correspondence between distance and disease severity, the sepsis severity of new samples was assessed. Through continuous tracking, time series comparison curves of Euclidean distance and SOFA score (Figure 3) were drawn to analyze the consistency and correlation of changes in the two trends.

Results

Significance analysis
Based on long-term clinical practice, it has been observed that as sepsis progresses, patients show significantly decreased hemoglobin levels and lymphocyte percentages, along with significantly increased platelet distribution width^12,16. Through analysis of these three key hematological indicators -- hemoglobin, lymphocyte percentage, and platelet distribution width -- significant differ...

Discussion

New method for sepsis assessment based on key hematological indicators
This study proposes a new method for assessing sepsis severity based on three key hematological indicators (hemoglobin, lymphocyte percentage, and platelet distribution width). This method quantifies the degree of deviation of hematological indicators from the healthy state by calculating the Euclidean distance of patient samples from the center point of the healthy population, providing a new assessment tool for clinical monito...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
MATLAB	MathWorks	2023B	Computing and visualization
The Multivariate State Colorful Scatter Plot	Intelligent  Entropy	Multivariate State V1.0	Beijing Intelligent Entropy Science & Technology Co Ltd. Modeling for multivariate State