Mortality associated with third-generation cephalosporin resistance in Enterobacteriaceae bloodstream infections at one South African hospital.
Dramowski A., Aiken AM., Rehman AM., Snyman Y., Reuter S., Grundmann H., Scott JAG., de Kraker MEA., Whitelaw A.
ObjectivesEnterobacteriaceae are common pathogens causing bloodstream infection (BSI) in sub-Saharan Africa and frequently express third-generation cephalosporin (3GC) resistance; however, the impact of 3GC resistance on clinical outcomes is rarely studied.MethodsWe conducted a single-site prospective cohort study at Tygerberg Hospital, Cape Town, South Africa to examine the feasibility of measuring impacts of 3GC resistance in Enterobacteriaceae BSI. We included patients with 3GC-susceptible and 3GC-resistant BSIs and matched each BSI patient to two uninfected patients. We determined the concordance of initial antibiotic treatment with the corresponding isolate's susceptibility profile. We performed exploratory impact analysis using multivariable regression models.ResultsBetween 1 June 2017 and 31 January 2018, we matched 177 Enterobacteriaceae BSI patients to 347 uninfected patients. Among these BSIs, 35% were phenotypically 3GC resistant. Parameters describing clinical comorbidity showed strong associations with mortality. We found that 18% of 3GC-R and 3% of 3GC-S BSI patient received non-concordant initial therapy. In multivariable Cox regression, we found a mortality impact over their matched patients for both 3GC-R (cause-specific hazard ratio 23.77; 95% CI 5.12-110.3) and 3GC-S (HR 7.49; 95%CI 3.08-18.19) BSI. There was a nonsignificant ratio of these ratios (HR 3.18; 95% CI 0.54-18.70), limited by the small sample size.ConclusionThis form of impact estimation was feasible in one hospital in South Africa where 3GC-R status was associated with non-concordant initial antibiotic treatment. There was a possible increase in mortality among individuals with 3GC-resistant Enterobacteriaceae, but with broad confidence intervals. These analytical approaches could be applied to larger datasets to improve precision of estimates.