Predictive Score Model of Clinical Outcomes Sepsis in Intensive Care Unit Tertier Referral Hospital of Eastern Indonesia

Abstract

AIM: This study aimed to design a predictive score model of clinical outcome sepsis and bacterial profiles of blood and sputum cultures in the intensive care unit (ICU) of a tertiary referral hospital. METHODS: An observational retrospective study was conducted in 2017–2020 using medical record data in the ICU of Dr. Soetomo Hospital as tertiary referral hospital. The predictor of sepsis prognosis was Acute Physiology and Chronic Health Evaluation II (APACHE II), blood and sputum culture results, procalcitonin (PCT) levels, and antimicrobial resistance in blood and sputum cultures. The model was prepared by logistic regression analysis and receiver operating characteristic (ROC) curves. RESULTS: Data from 355 subjects showed that predictor score was APACHE II, blood and sputum culture results; besides PCT levels were found to contribute significantly to predictive score of sepsis clinical output (p<0.05), while the predictor test of antimicrobial resistance in blood and sputum cultures was not significant to predictive score of sepsis clinical output (p > 0.05). The resulting scores to predict sepsis clinical outcomes include PCT level >2 ng/mL (1.61), APACHE score >20 (1), sputum culture as true pathogen (1.1), and blood culture as true pathogen (1.35). When the total score ≥3, the patient will die, while when the score <3, the patient will survive. ROC curves analysis obtained area under curve 0.859 (p < 0.05) which indicates that the equation is statistically significant in predicting the sepsis clinical outcome. Probability scores and death outcomes indicate that the higher the predictive score, the higher the probability of dying, with a score >3 the probability of dying is above 95.27%, whereas if the score is 5, the probability of dying is above 99%. The bacterial profile of blood cultures leading to mortality is predominately Gram-positive (34.4%), consisting of coagulase-negative Staphylococcus (22.9%), and Staphylococcus aureus (4.3%), while Gram-negative is only 14.7%, which consists of Enterobacteriaceae group (8.7%), Acinetobacter baumannii (4%), polymicrobial infection (2%), Burkholderia cepacia (0.8%), and Pseudomonas aeruginosa (0.4%). Sputum culture profile of patients with sepsis who died in the ICU of a tertiary referral RSUD Soetomo is dominated by Gram-negative, namely, A. baumannii (22.1%), Enterobacteriaceae group (20.6%), P. aeruginosa (11.1%), while Gram-positive is S. aureus (22.9%). CONCLUSION: The predictive score model for sepsis clinical outcomes in the ICU of a tertiary referral hospitals can be used as a basis for determining of patient management and the profile of the bacteria that causes sepsis that results in death.