Clinical and microbiological characterization of sepsis and evaluation of sepsis scores

Background Despite the necessity of early recognition for an optimal outcome, sepsis often remains unrecognized. Available tools for early recognition are rarely evaluated in low- and middle-income countries. In this study, we analyzed the spectrum, treatment and outcome of sepsis at an Ethiopian tertiary hospital and evaluated recommended sepsis scores. Methods Patients with an infection and >= 2 SIRS criteria were screened for sepsis by SOFA scoring. From septic patients, socioeconomic and clinical data as well as blood cultures were collected and they were followed until discharge or death; 28-day mortality was determined. Results In 170 patients with sepsis, the overall mortality rate was 29.4%. The recognition rate by treating physicians after initial clinical assessment was low (12.4%). Increased risk of mortality was significantly associated with level of SOFA and qSOFA score, Gram-negative bacteremia (in comparison to Gram-positive bacteremia; 42.9 versus 16.7%), and antimicrobial regimen including ceftriaxone (35.7% versus 19.2%) or metronidazole (43.8% versus 25.0%), but not with an increased respiratory rate (>= 22/min) or decreased systolic blood pressure (<= 100mmHg). In Gram-negative isolates, extended antimicrobial resistance with expression of extended-spectrum beta-lactamase and carbapenemase genes was common. Among adult patients, sensitivity and specificity of qSOFA score for detection of sepsis were 54.3% and 66.7%, respectively. Conclusion Sepsis is commonly unrecognized and associated with high mortality, showing the need for reliable and easy-applicable tools to support early recognition. The established sepsis scores were either of limited applicability (SOFA) or, as in the case of qSOFA, were significantly impaired in their sensitivity and specificity, demonstrating the need for further evaluation and adaptation to local settings. Regional factors like malaria endemicity and HIV prevalence might influence the performance of different scores. Ineffective empirical treatment due to antimicrobial resistance is common and associated with mortality. Local antimicrobial resistance statistics are needed for guidance of calculated antimicrobial therapy to support reduction of sepsis mortality.