Methods: A prospective study enrolling adult patients (18 years and older) diagnosed with sepsis at two tertiary healthcare facilities in Rwanda - University Teaching Hospital of Kigali (CHUK) and King Faisal Hospital (KFH), Kigali, Rwanda. Patient enrollment began on October 1st ,2014. Clinical, laboratory and management data were collected to assess for predictors of mortality. Rapid acute physiology score (RAPS), rapid emergency medicine score (REMS) and mortality in emergency department sepsis (MEDS) scores were calculated.
Results: 172 patients have been enrolled so far in our study. 31 individuals were lost to follow-up. 38 (26.9 %) and 103 patients (73.1 %) were from KFH and CHUK respectively. 57 enrollees were female (40.4 %). 90 patients (63.8 %) were below 50 years of age. Overall 28-day mortality was 29.1%. On preliminary univariate analysis, we found that independent predictors of mortality were hypoxia (oxygen saturation below 90 % vs. above 90%, OR 6.89 95%CI 3.06-1.52 p<0.001), stage of sepsis on admission (severe sepsis vs. sepsis OR 10.71, 95%CI 3.46-33.14, p<0.001), late diagnosis of sepsis (7 hours post presentation vs 1 hour, OR 73.33, 95%CI 15.03-357.73, P<0.001), and appropriateness of antibiotic received based on blood culture results (inappropriate vs appropriate antibiotics OR 185.25, 95 % CI 21.83-1572.12, p<0.001). These factors remain significant on multivariate analyses. Volume of IV fluids received early in disease course was positively correlated with survival (p=0.033). We found that the MEDS score was the most reliable prognostication tool for 28-day mortality with an area under the curve (AUC) of 0.84.
Conclusion: Hypoxia, stage of sepsis on presentation, use of inappropriate antibiotics and late recognition of sepsis were associated with higher mortality while volume of IV fluid administration was positively correlated with improved survival among our patient cohort. Of prognostication systems feasible for use in our practice setting, the MEDS score was found to be most reliable in predicting mortality.
O. Ogbuagu, None
T. Walker, None