Sepsis Prediction From Complete Blood Count Components

Abstract

Background. Sepsis is a clinical syndrome that defines human body reaction to infection by microbial pathogens. Sepsis is a major healthcare problem around the world with significant mortality. Timely and correct management improves outcome. Initial empirical therapy is started before definitive microbiological pathogen identification. Complete blood count is a major baseline laboratory investigation for a suspected case. Multiple parameters are analyzed and reported in the complete blood count analysis that estimate human body immune response to microbial infection. However, but for white cell count elevation, correlated information from majority of other measurements remain tangled up with minimal input as regards to diagnosis, stratification and management of sepsis. Statistical analysis of complete blood count in sepsis may reveal further clinical relevance of other components that is may be additive to diagnosis and management leading to improved outcome in sepsis. Broad objectives; To measure the degree of relationship among complete blood count changes that occur in sepsis. Study design and site; This study was a retrospective study using laboratory measurements recorded from patients treated for sepsis in Beth Israel hospital. Materials and methods; This study used secondary data from medical information mart for intensive care (MIMIC III) database. This database contains de- identified clinical information of ICU patients admitted over ten years in Beth Israel hospital. Data cleaning and imputation was done using R software. Multivariate analysis was carried out using stata analytical software. Multiple linear Regression modelling; the main model had white cell count as response variable and differential counts and other complete blood count values as multivariate variables. Data management; After approval, data was acquired from MIMIC III database management and stored in data files in the computer that was used for analysis. Data files were password protected and no manual coping was done. Expected main outcome measure/ utility of the study; Main outcome measures were correlation of laboratory measurements from the initial complete blood count investigations. These correlated variables were used in prediction model for sepsis in addition to elevated white cell count. Results; Mean cell volume, mean cell hemoglobin and red cell distribution width were the highest correlated variables with the principal factor explaining the variance in sepsis. They also loaded highly in discriminating elevated white cell count from low and normal counts. Multivariate predictive model using these variables had a cronbach’s alpha of 0.4. Conclusion; Mean cell volume, mean cell hemoglobin and red cell distribution width may be used in predictive diagnosis of sepsis in addition to white cell changes.